22.12.2012 Views

Praca Dyplomowa - Photogrammetry and Remote Sensing - AGH

Praca Dyplomowa - Photogrammetry and Remote Sensing - AGH

Praca Dyplomowa - Photogrammetry and Remote Sensing - AGH

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Praca</strong> <strong>Dyplomowa</strong><br />

Temat pracy: Badanie metod dotyczących tworzenia szczegółowych modeli<br />

cyfrowych rzeźb i innych obiektów<br />

Title of work: Investigation on methods for making detailed digital models<br />

of sculptures <strong>and</strong> other artefacts<br />

Oppgave tittel: Undersøkelse av metoder for å lage detaljerte digitale<br />

modeller av skulpturer og <strong>and</strong>re objekter<br />

Nazwisko i imię: Małgorzata Aulejtner<br />

Kierunek studiów: Geodezja i Kartografia<br />

Specjalność: Geoinformacja, Fotogrametria i Teledetekcja<br />

Ocena:……………<br />

Akademia Górniczo-Hutnicza<br />

Im. Stanisława Staszica w Krakowie<br />

Wydział Geodezji Górniczej<br />

i Inżynierii Środowiska<br />

Katedra Geoinforamcji, Fotogrametrii<br />

i Teledetekcji Środowiska<br />

Recenzent: Opiekun Pracy:<br />

Dr inż. Sławomir Mikrut dr hab. inż. Regina Tokarczyk<br />

Oświadczam, świadomy(a) odpowiedzialności karnej za poświadczenie nieprawdy, że<br />

niniejszą pracę dyplomową wykonałem(am) osobiście i samodzielnie i że nie<br />

korzystałem(am) ze źródeł innych niż wymienione w pracy.<br />

...........................................<br />

czytelny podpis autora pracy<br />

Kraków 2011


Streszczenie pracy :<br />

Różnego rodzaju rzeźby i inne artefakty dorobku kulturalnego są narażone na zniszczenia,<br />

które czasem trudno uniknąć. Tworzy się więc nowe repliki tych obiektów w celu ochrony<br />

tego dziedzictwa. Aby uniknąć dodatkowych zniszczeń w czasie wykonywania tych kopii,<br />

odchodzi się od dawnych metod – wykonywanych metodą odlewów. Stosuje się więc tutaj<br />

metody jak trójwymiarowe skanowanie za pomocą skanera lub metody fotogrametryczne, w<br />

których nie trzeba dotykać w żaden sposób obiektu dzięki czemu ewentualne zniszczenia są<br />

minimalizowane. Trójwymiarowe digitalne modele utrzymane z nowszych metod mogą być<br />

użyte następnie do utworzenia realnej repliki przy zastosowaniu specjalistycznych drukarek<br />

trójwymiarowych, które “wyrzeżbią” dany obiekt.<br />

Celem tej pracy było badanie bezdotykowych, nowszych metod tworzenia szczegółowych<br />

modeli 3D na przykładzie rzeźby z katedry Nidaros w Trondheim (Norwegia). Głównym<br />

priorytetem było uzyskanie jak najbardziej dokładnego modelu, przede wszystkim aby<br />

uzyskać jak najwięcej szczegółów. W pracy mieści się więc porównanie metod<br />

fotogrametrycznych, skaningu a także narzędzi do modelowania 3D.<br />

Do uzyskania modelu z dużą ilością małych szczegółów rzeźby użytej w projekcie potrzebna<br />

jest duża chmura punktów. Trzeba zwrócić uwagę, iż rzeźba nie była płaska, a dość okrągła,<br />

złożona z różnych powierzchni z wieloma małymi obiektami. Uzyskana chmura punktów<br />

powinna być więc tutaj bardzo dokładna z wysoką rozdzielczością. Nie mniej jednak, takie<br />

parametry powodują iż niemalże każdy program potrzebuje dość dużo czasu aby obliczyć tak<br />

dużo danych, czy to z metody fotogrametrycznej czy też skaningu. Głównie nie jest to jednak<br />

spowodowane słabymi parametrami używanego komputera, ale ograniczeniami programów.<br />

Na rynku dostępne są różne skanery, są jednak i takie które potrafią uzyskać bardzo gęstą<br />

chmurę punktów a dużą dokładnością. Zazwyczaj, jeżeli skaner jest stworzony do skanowania<br />

mały obiektów z bliska a rozdzielczość i dokładność rośnie, są one coraz mniejsze w<br />

rozmiarze. Im większa dokładność tym mniejszy zasięg również, tak więc dla mniejszych<br />

obiektów – mniejsza odległość skanowania jest zalecana. Metody fotogrametryczne również<br />

mogą dać dobre wyniki i dobrą rozdzielczość, jednak tylko w momencie gdy materiał<br />

fotografowanego obiektu nie świeci się, nie odbija światła ani nie jest ze szkła.<br />

W projekcie zostały wykorzystane następujące fotogrametryczne programy: PhotoModeler<br />

Scanner i Topcon ImageMaster. Są to dość podobne programy, gdzie obydwa potrafią<br />

stworzyć powierzchnie trójwymiarowe używając do tego zdjęć zrobionych kamerą<br />

niemetryczną. Obydwie firmy zapewniają również kalibrację takiej kamery. Resultat z tych<br />

programów jest dość dobry, jednak powierzchnia rzeźby świeciła się nieco i było widać<br />

refleksy, odbicie światła w niektórych miejscach na zdjęciu, przez co matching i generacja<br />

punktów 3D nie działała w tym wypadku najlepiej. Powyższe problemy resultowały na<br />

modelu w ten sposób, iż widoczny był szum, źle obliczone punkty – odstrzelone od modelu.<br />

Ponadto, czas potrzebny do obliczenia i wygenerowania DSM był dość długi i to przede<br />

wszystkim powinno zostać wzięte pod uwagę przy dalszym ulepszaniu programów.<br />

Oprócz wymienionych programów, aplikacja Autodesk – Project Photofly 2.0 była również<br />

testowana. Program jest w stanie wygenerować teksturowany mesh – powierzchnię TIN 3D.<br />

Jest to program freewere, dostępny na stronie Autodesk do ściągnięcia i jest bardzo łatwy w<br />

obsłudze – użytkownicy bez znajomości fotogrametrii będą w stanie bez problemu wykonać<br />

model. Jednakże, program nie jest w stanie wykonać tak dokładnego modelu jak wspomniane<br />

ii


wcześniej aplikacje. Nie mniej jednak, w wielu sytuacjach może to wystarczyć i dać dużo<br />

możliwości.<br />

Ze skaningu w projekcie został użyty skaner Konica Minolta VI-910, ale także bardzo<br />

ciekawy, mały, przenośny skaner Artec MHT 3D. Obydwa skanery mają dość krótki zasięg,<br />

Artec ma jednak mniejszy. Artec używa światła żarówki zamiast lasera jak Konica Minolta.<br />

Artec również wydaje się być lepszy w zastosowaniu do mniejszych obiektów, charakteryzuje<br />

się też lepszą dokładnością i rozdzielczością. Używając tych skanerów, programy które daje<br />

producent zostały użyte do pozyskania danych 3D. Jednakże, na rynku jest też dostępny<br />

program jak Geomagic, który może zostać użyty do post-processingu, do pracy na surowej<br />

chmurze punktów pozyskanej czy to z fotogrametrii czy też skaninngu. Polygon Editing Tool<br />

od Konici Minolty wydaje się nie działać najlepiej z nowym Windowsem 7 wersji 64-bitowej,<br />

jednak program od Artec jest w stanie wykorzystać komputer znacznie lepiej – można ustawić<br />

ilość rdzeniów procesora do obliczeń.<br />

Najlepszy model został osiągnięty przy użyciu skanera Artec MHT, który potrafi pozyskać<br />

dane z rozdzielczością do 0,5mm i dokładnością aż do 0,1mm. Jeżeli chodzi o końcowy<br />

model uzyskany w programie Artec, jest on dość dobry. Jednak, zdaniem autora, program<br />

Geomagic spisał się lepiej, gdzie jest więcej kontroli nad manipulacją modelu. Wygładzanie<br />

może tu zostać zastosowane z “uwzględnieniem krzywych”, gdzie model przy tych krzywych<br />

nie jest aż tak bardzo wygładzany i potrzebne linie – ostre krawędzie zostaną lepiej widoczne<br />

na modelu.<br />

Ostateczny model jest wyeksportowany w pliku 3D PDF a także w innych dodatkowych<br />

plikach, które można obejrzeć na komputerze.<br />

iii


Summary:<br />

Sculptures <strong>and</strong> other artifacts are exposed to damages <strong>and</strong> it is hard to avoid them. For<br />

cultural heritage preservation new replicas of these objects should be done. For this purpose<br />

the best solution is to use non-touching methods, like photogrammetry <strong>and</strong> three-dimensional<br />

scanning, which will not do any harm to the object. Then 3D digital model may be used <strong>and</strong><br />

by applying of 3D printers the real copy may be done<br />

The purpose of the report was to investigate methods on making detailed digital models on<br />

example of a sculpture from Nidaros Cathedral. The goal was to get as accurate – detailed<br />

model as possible. There is comparison on photogrammetric, 3D scanning methods <strong>and</strong><br />

available 3D modelling tools.<br />

To make a very detailed digital model of round, complex, <strong>and</strong> small objects – very big point<br />

cloud is needed. The point cloud has to be accurate with high resolution. To make such a<br />

point cloud from images any software needs a lot of time for computation. Also, computation<br />

of big amount of points in the data from scanning takes a lot of time. This is not because of<br />

low PC, but because of software limitations. Some scanners are able to make more dense<br />

point cloud with better accuracy than others. With higher resolution <strong>and</strong> better accuracy<br />

scanners become smaller, <strong>and</strong> also distance range is decreasing. Though, for small object<br />

smaller distance is required. <strong>Photogrammetry</strong> may produce also very good results, if only the<br />

material is non-reflective/shiny, or glass.<br />

In the project photogrammetric software has been used, such as PhotoModeler Scanner <strong>and</strong><br />

Topcon ImageMaster, where both are able to create 3D surface from photographs, taken by<br />

non-metric camera. Both softwares provide camera calibration. The results at the end were<br />

quite good, but the sculpture is quite shiny <strong>and</strong> reflects the light, <strong>and</strong> this situation is not<br />

required for image matching <strong>and</strong> creating 3D points. Therefore, the model results with a lot of<br />

noise. Also, the time needed for DSM creating was very long <strong>and</strong> this part still needs some<br />

improvement. Free software from Autodesk – Project Photofly 2.0 was also tested. The<br />

application creates mesh <strong>and</strong> may be used by non-experienced users. Nevertheless, the results<br />

are not so accurate or the resolution is lower, it still may be used in most common situations.<br />

From laser scanning, Konica Minolta VI-910 was used for data acquisition, but also very<br />

interesting h<strong>and</strong>held small scanner from Artec was used. Both are created for small range<br />

distances; nevertheless Artec MHT seems to be better for smaller objects. It has better<br />

accuracy <strong>and</strong> higher resolution. In both cases, softwares recommended by the companies were<br />

used for modelling. Geomagic software was also used for post processing afterwards. Polygon<br />

Editing Tool from Konica Minolta seemed not working properly with new Windows 7 x64<br />

bits, but Artec was even able to use most of the processors cores for computations.<br />

The best model was achieved by applying small structured scanner from Artec with 0,5mm of<br />

resolution <strong>and</strong> accuracy up to 0,1mm. The model created using Artec software works good,<br />

but in the author‟s opinion Geomagic software works better, where smoothing may be done in<br />

the respect to curves.<br />

The final model is exported in 3D PDF file <strong>and</strong> other additional files<br />

iv


NORWEGIAN UNIVERSITY OF SCIENCE AND TECHNOLOGY<br />

DEPARTMENT OF CIVIL AND TRANSPORT ENGINEERING<br />

Report Title:<br />

Investigation on methods for making detailed digital models<br />

of sculptures <strong>and</strong> other artefacts<br />

Undersøkelse av metoder for å lage detaljerte digitale<br />

modeller av skulpturer og <strong>and</strong>re objekter<br />

Badanie metod dotyczących tworzenia szczegółowych<br />

modeli cyfrowych rzeźb i innych obiektów<br />

Name:<br />

Professor in charge/supervisor:<br />

Other external professional contacts/supervisors:<br />

Date: 25.07.2011<br />

Number of pages (incl.<br />

appendices): 90<br />

Master Project Work<br />

Thesis X<br />

Małgorzata Aulejtner<br />

Professor Knut Ragnar Holm<br />

Ph.D. Regina Tokarczyk<br />

Abstract:<br />

Digital modelling of real objects, like sculptures for cultural heritage preservation became a<br />

modern non-touching method. New methods are being developed. There are already existing<br />

methods like photogrammetry <strong>and</strong> relatively newer three-dimensional scanning. In this work<br />

comparison of these methods in connection with the software has been done. The main goal was<br />

to get detailed a model, which would represents the object as real one as best as possible. The<br />

model here is needed to make a real replica afterwards, by applying three-dimensional printer.<br />

The main limitation of the examinations was time period. Dada acquisition was not a problem,<br />

however the amount of data, <strong>and</strong> amount of generated point clouds itself was sometimes<br />

problematic. There were also some limitations of software – every computational step was taking<br />

too much time. Even though the used computer was very fast, the older software was not able to<br />

take full advantage of the hardware. It was also hard to set a proper time index to the quality of<br />

details. The best model was needed, the high resolution was required, what affects that time for<br />

processing increases. It was especially visible when it comes to the photogrammetric part,<br />

because the scanning method was more user-friendly. Geomagic software was also used for<br />

processing the point clouds. All results are compared, <strong>and</strong> the best model was made by using<br />

Artec MHT 3D Scanner in connection with Geomagic Qualify software. The scanner is capable<br />

to acquire data with 0,5mm resolution with 0,1mm accuracy. <strong>Photogrammetry</strong> methods may also<br />

give good results, though in this case, as the material of the sculpture was reflecting the light, it<br />

did not. The model resulted with big amount of noise.<br />

The final model is exported in 3D PDF file <strong>and</strong> other additional files.<br />

Keywords:<br />

1. <strong>Photogrammetry</strong><br />

2. Camera Calibration<br />

3. Three-dimensional scanning<br />

4. Three-dimensional digital modelling<br />

5. Visualisation<br />

6. Comparison<br />

_________________________________________<br />

(Signature)


iii


PREFACE AND ACKNOWLEDGMENTS<br />

This master thesis has been written for The Restoration Workshop of Nidaros Cathedral<br />

(Trondheim) request. I had opportunity to investigate various methods <strong>and</strong> instruments for<br />

three-dimensional modelling, where replica of a sculpture from the Cathedral was object for<br />

tests. I was able to work <strong>and</strong> test available instruments <strong>and</strong> software at the NTNU – The<br />

Norwegian University of Science <strong>and</strong> Technology in Trondheim, Norway. During the year as<br />

exchange student from <strong>AGH</strong> – The University of Science <strong>and</strong> Technology in Kraków, Pol<strong>and</strong>,<br />

I had participated in a few measurement projects, which were based on digital<br />

photogrammetric <strong>and</strong> scanning methods.<br />

Thanks to Kristin Bjørlykke, working at The Restoration Workshop of Nidaros Cathedral, for<br />

borrowing the replica (sculpture of Archbishop) from the Cathedral to work on it. I got<br />

helpful information about the problems, which they had met before <strong>and</strong> historical background<br />

about the Archbishop itself.<br />

I am also very grateful to Odd Erik Mjørlund, manager in Geoplan 3D - company in Oslo, for<br />

giving introduction <strong>and</strong> opportunity to test Artec MHT 3D Scanner.<br />

I would like to express my gratitude to Ph.D. Regina Tokarczyk, working at <strong>AGH</strong> at the<br />

Department of Geoinformation, <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong> of Environment for<br />

supervising the thesis in Pol<strong>and</strong>, while I was studying abroad.<br />

Since the beginning of this thesis, there has been a continuous communication with Professor<br />

Knut Ragnar Holm, working at NTNU, at the Department of Civil <strong>and</strong> Transport Engineering,<br />

to whom I would like to express my grateful acknowledgments. I am thankful for contacts to<br />

people who were helpful to try more methods.<br />

v


SUMMARY<br />

Sculptures <strong>and</strong> other artifacts are exposed to damages <strong>and</strong> it is hard to avoid them. For<br />

cultural heritage preservation new replicas of these objects should be done. For this purpose<br />

the best solution is to use non-touching methods, like photogrammetry <strong>and</strong> three-dimensional<br />

scanning, which will not do any harm to the object. Then 3D digital model may be used <strong>and</strong><br />

by applying of 3D printers the real copy may be done<br />

The purpose of the report was to investigate methods on making detailed digital models on<br />

example of a sculpture from Nidaros Cathedral. The goal was to get as accurate – detailed<br />

model as possible. There is comparison on photogrammetric, 3D scanning methods <strong>and</strong><br />

available 3D modelling tools.<br />

To make a very detailed digital model of round, complex, <strong>and</strong> small objects – very big point<br />

cloud is needed. The point cloud has to be accurate with high resolution. To make such a<br />

point cloud from images any software needs a lot of time for computation. Also, computation<br />

of big amount of points in the data from scanning takes a lot of time. This is not because of<br />

low PC, but because of software limitations. Some scanners are able to make more dense<br />

point cloud with better accuracy than others. With higher resolution <strong>and</strong> better accuracy<br />

scanners become smaller, <strong>and</strong> also distance range is decreasing. Though, for small object<br />

smaller distance is required. <strong>Photogrammetry</strong> may produce also very good results, if only the<br />

material is non-reflective/shiny, or glass.<br />

In the project photogrammetric software has been used, such as PhotoModeler Scanner <strong>and</strong><br />

Topcon ImageMaster, where both are able to create 3D surface from photographs, taken by<br />

non-metric camera. Both softwares provide camera calibration. The results at the end were<br />

quite good, but the sculpture is quite shiny <strong>and</strong> reflects the light, <strong>and</strong> this situation is not<br />

required for image matching <strong>and</strong> creating 3D points. Therefore, the model results with a lot of<br />

noise. Also, the time needed for DSM creating was very long <strong>and</strong> this part still needs some<br />

improvement. Free software from Autodesk – Project Photofly 2.0 was also tested. The<br />

application creates mesh <strong>and</strong> may be used by non-experienced users. Nevertheless, the results<br />

are not so accurate or the resolution is lower, it still may be used in most common situations.<br />

From laser scanning, Konica Minolta VI-910 was used for data acquisition, but also very<br />

interesting h<strong>and</strong>held small scanner from Artec was used. Both are created for small range<br />

distances; nevertheless Artec seems to be better for smaller objects. It has better accuracy <strong>and</strong><br />

higher resolution. In both cases, softwares recommended by the companies were used for<br />

modelling. Geomagic software was also used for post processing afterwards. Polygon Editing<br />

Tool from Konica Minolta seemed not working properly with new Windows 7 x64 bits, but<br />

Artec was even able to use most of the processors cores for computations.<br />

The best model was achieved by applying small structured scanner from Artec with 0,5mm of<br />

resolution <strong>and</strong> accuracy up to 0,1mm. The model created using Artec software works good,<br />

but in the author‟s opinion Geomagic software works better, where smoothing may be done in<br />

the respect to curves.<br />

vi


TABLE OF CONTENTS<br />

1 INTRODUCTION .............................................................................................................. 1<br />

1.1 Project background ...................................................................................................... 1<br />

1.2 Cathedral <strong>and</strong> sculpture background information ....................................................... 1<br />

1.3 Three-dimensional printers .......................................................................................... 4<br />

1.4 Structure of the report .................................................................................................. 5<br />

2 THEORETICAL BACKGROUND .................................................................................... 6<br />

2.1 Review of similar projects ........................................................................................... 6<br />

2.2 <strong>Photogrammetry</strong> .......................................................................................................... 7<br />

2.2.1 Aerial <strong>Photogrammetry</strong> ........................................................................................ 7<br />

2.2.2 Close – range photogrammetry ............................................................................ 8<br />

2.2.3 Focusing ............................................................................................................... 9<br />

2.2.4 Image scale <strong>and</strong> accuracy .................................................................................. 10<br />

2.2.5 Coordinate systems ............................................................................................ 11<br />

2.2.6 Sensor principle .................................................................................................. 11<br />

2.2.7 Camera calibration - Interior orientation .......................................................... 11<br />

2.2.8 Exterior orientation ............................................................................................ 15<br />

2.2.9 Collinearity equations ........................................................................................ 16<br />

2.2.10 Direct Linear Transformation (DLT) ................................................................. 17<br />

2.2.11 Absolute orientation ........................................................................................... 17<br />

2.2.12 Bundle triangulation .......................................................................................... 17<br />

2.2.13 Matching types of digital images ....................................................................... 20<br />

2.3 Three-dimensional digitizers; Three-dimensional scanning ...................................... 21<br />

2.3.1 Time-of-light (TOF) measurements .................................................................... 22<br />

2.3.2 Phase-shifting measurement techniques ............................................................ 22<br />

2.3.3 Triangulation-based measurements ................................................................... 23<br />

2.4 Three-dimensional digital modelling <strong>and</strong> visualisation ............................................. 25<br />

2.4.1 Solid triangulated surfaces ................................................................................. 26<br />

2.4.2 CAD models ........................................................................................................ 27<br />

2.4.3 Visualisation ....................................................................................................... 27<br />

3 INSTRUMENTS AND SOFTWARE .............................................................................. 28<br />

3.1 Computer - workstation ............................................................................................. 28<br />

3.2 <strong>Photogrammetry</strong> ........................................................................................................ 28<br />

3.2.1 Camera: Canon EOS-1Ds (Body) ...................................................................... 28<br />

3.2.2 Lens: Canon 28mm f/2.8 .................................................................................... 29<br />

3.2.3 Topcon ImageMaster Software .......................................................................... 29<br />

3.2.4 PhotoModeler Scanner Software ....................................................................... 30<br />

3.2.5 Autodesk Project Photofly - Photo Scene Editor 2.0 Software .......................... 30<br />

3.3 Three-dimensional digitizing ..................................................................................... 31<br />

3.3.1 Konica Minolta VI-910 laser scanner ................................................................ 31<br />

3.3.2 Artec MHT three-dimensional scanner .............................................................. 32<br />

3.4 Geomagic Software; post-processing, modelling <strong>and</strong> visualisation .......................... 32<br />

4 EXPERIMENT ................................................................................................................. 34<br />

4.1 <strong>Photogrammetry</strong> ........................................................................................................ 34<br />

4.1.1 Image acquisition ............................................................................................... 34<br />

4.1.2 Topcon ImageMaster Software .......................................................................... 37<br />

4.1.3 PhotoModeler Scanner Software ....................................................................... 39<br />

4.1.4 Autodesk Project Photofly Photo Scene Editor 2.0 Software ............................. 42<br />

vii


4.2 Three-dimensional scanning ...................................................................................... 44<br />

4.2.1 Konica Minolta VI-910 laser scanner ................................................................ 44<br />

4.2.2 Artec MHT three-dimensional scanner .............................................................. 46<br />

4.3 Geomagic Software ................................................................................................... 47<br />

5 RESULTS, COMPARISSON AND ANALYSIS ............................................................ 49<br />

5.1 <strong>Photogrammetry</strong> ........................................................................................................ 49<br />

5.1.1 Camera calibration (PhotoModeler Scanner <strong>and</strong> Topcon ImageMaster) ......... 49<br />

5.1.2 Photogrammetric project concerning modelling of the sculpture (PhotoModeler<br />

Scanner <strong>and</strong> Topcon ImageMaster) .................................................................................. 51<br />

5.1.3 Autodesk Project Photofly Photo Scene Editor 2.0 Software ............................. 55<br />

5.2 Three-dimensional scanning ...................................................................................... 56<br />

5.3 Geomagic Software, modelling, post-processing software ....................................... 58<br />

6 DISCUSSION AND CONCLUSION ............................................................................... 61<br />

REFERENCES ......................................................................................................................... 67<br />

APPENDIXES ......................................................................................................................... 72<br />

viii


LIST OF FIGURES<br />

Figure 1.1 Nidaros Cathedral nowadays – west, front side ........................................................ 2<br />

Figure 1.2 Actual sculpture of Archbishop in Nidaros Cathedral .............................................. 4<br />

Figure 2.1 Photogrammetric principle of 3D measurement; example on convergent image<br />

configuration (Luhmann et al., 2006) ......................................................................................... 8<br />

Figure 2.2 Focusing <strong>and</strong> depth of field (Luhmann et al., 2006) ............................................... 10<br />

Figure 2.3 Perspective centre <strong>and</strong> principle distance (Luhmann et al., 2006) ......................... 12<br />

Figure 2.4 Effect of radial – symmetric distortion; example (Luhmann et al., 2006) .............. 13<br />

Figure 2.5 Effect of radial – asymmetric <strong>and</strong> tangential distortion; example (Luhmann et al.,<br />

2006) ......................................................................................................................................... 14<br />

Figure 2.6 Effect of affinity <strong>and</strong> shear deviations; example (Luhmann et al., 2006)............... 14<br />

Figure 2.7 Interior orientation (Luhmann et al., 2006) ............................................................ 15<br />

Figure 2.8 Exterior orientation for terrestrial photogrammetry (Luhmann et al., 2006) .......... 16<br />

Figure 2.9 Methods <strong>and</strong> data flow for orientation <strong>and</strong> point determination (based on<br />

(Luhmann et al., 2006)) ............................................................................................................ 20<br />

Figure 2.10 Epipolar geometry ................................................................................................. 21<br />

Figure 2.11 Basic triangulation principle applied in scanning<br />

(http://www.impactstudiostv.com/Studios) .............................................................................. 23<br />

Figure 2.12 System setup with one bar pattern <strong>and</strong> one camera; example of structured-light<br />

scanning (http://www.rob.cs.tu-bs.de/en/research/projects/shape/) ......................................... 24<br />

Figure 3.1 Camera Canon EOS-1Ds (Canon, 2011) ................................................................ 28<br />

Figure 3.2 Lens Canon 28mm f/2.8 (Canon, 2011) ................................................................. 29<br />

Figure 3.3 Konica Minolta VI-910 laser scanner (Minolta <strong>and</strong> GmbH, 2011) ........................ 31<br />

Figure 3.4 Artec MHT scanner (Artec, 2011) .......................................................................... 32<br />

Figure 4.1 Photograph from image acquisition ........................................................................ 36<br />

Figure 4.2 Results from Topcon ImageMaster Camera Calibration ........................................ 37<br />

Figure 4.3 Textured mesh from one photo pair in Topcon ImageMaster ................................ 39<br />

Figure 4.4 Residuals on control points in camera calibration in PhotoModeler Scanner ........ 40<br />

Figure 4.5 Coloured point cloud of small part of the sculpture in PhotoModeler Scanner ..... 42<br />

Figure 4.6 Data flow using Autodesk Project Photofly software (Autodesk, 2011) ................ 43<br />

Figure 4.7 Final Maximum mesh in Autodesk Photo Scene Editor 2.0 Project Photo Fly ...... 44<br />

Figure 4.8 Merged point cloud in Polygon Editing Tool from Konica Minolta ...................... 45<br />

Figure 4.9 Merged <strong>and</strong> smoothed point cloud in Artec Scanner software ............................... 47<br />

Figure 4.10 Registration of point clouds in Geomagic; model of deviations .......................... 48<br />

Figure 5.1 Results of camera calibration from Topcon ImageMaster ...................................... 50<br />

Figure 5.2 Textured model from PhotoModeler Scanner; sculpture was sprayed with dust ... 54<br />

Figure 5.3 Shaded model from PhotoModeler Scanner, representing “hair”, 1mm resolution 54<br />

Figure 5.4 Textured model from Autodesk Project PhotoFly 2.0 ............................................ 55<br />

Figure 5.5 Model from Autodesk Project PhotoFly; converted <strong>and</strong> improved in Geomagic .. 56<br />

Figure 5.6 Model from Artec MHT scanner <strong>and</strong> Software ...................................................... 58<br />

Figure 5.7 Model from Artec MHT scanner, but data processed in Geomagic, which is<br />

perceived as the best one. ......................................................................................................... 60<br />

Figure 6.1 Final model from Geomagic, using Artec MHT data ............................................. 64<br />

Figure 6.2 Visual comparison of model – more detailed part of the sculpture – “hair”; top left<br />

Geomagic - “Artec”, top right Geomagic - “Autodesk”, bottom left Geomagic – “Konica<br />

Minolta”, bottom right “PhotoModeler Scanner” .................................................................... 65<br />

ix


Figure 6.3 Visual comparison of model – more sharp part of the sculpture – “collar”; top left<br />

Geomagic - “Konica Minolta”, top right Geomagic - “Autodesk”, bottom Geomagic – “Artec”<br />

.................................................................................................................................................. 66<br />

LIST OF TABLES<br />

Table 2.1 Simplified workflow for modelling: from point cloud to final model ..................... 26<br />

Table 3.1 Specification of Canon lens 28mm f/2.8 (Canon, 2011) .......................................... 29<br />

Table 5.1 Comparison on results of camera calibration using photogrammetric softwares .... 50<br />

Table 5.2 Comparison of photogrammetric software: ImageMaster, PhotModeler Scanner ... 53<br />

Table 5.3 Comparison on scanning systems: Artec, Konica Minolta, processing in Geomagic<br />

.................................................................................................................................................. 59<br />

Table 5.4 Comparison on results of alignment scans in Geomagic software .......................... 60<br />

LIST OF APPENDIXES<br />

A. 1 Specification of Canon Camera EOS-1Ds ....................................................................... 72<br />

A. 2 Report - results of camera calibration using PhotoModeler Scanner software ................ 76<br />

A. 3 Comparison on photogrammetric software used in the project including system<br />

requirements, import/export files, main features camera calibration <strong>and</strong> modelling<br />

manipulation tools .................................................................................................................... 78<br />

A. 4 Comparison of 3D digitizers: main features, specification: used methods, light, range,<br />

resolution, accuracy, etc. .......................................................................................................... 81<br />

A. 5 Comparison of software from scanning systems; main features, system requirements,<br />

model manipulation tools. ........................................................................................................ 83<br />

A. 6 Geomagic Qualify: main features, import/export files, system requirements <strong>and</strong> main<br />

manipulation tools used in the project ...................................................................................... 86<br />

x


1 INTRODUCTION<br />

1.1 Project background<br />

INTRODUCTION<br />

Sculptures are exposed to different damages, like atmospheric <strong>and</strong> so on. For many years,<br />

they have been getting to wear down <strong>and</strong> it is possible to see the signs of erosion. Something<br />

has to be done to preserve cultural heritage. Because of these problems many replicas are<br />

being done <strong>and</strong> they are replacing original forms inside the museums. Some of them are too<br />

fragile to use old methods <strong>and</strong> make plaster casts. There is a risk that touching the sculpture<br />

will destroy it completely.<br />

Nowadays, there are geodetic techniques which allow making three-dimensional digital<br />

models without touching the objects. It may be photogrammetric or laser scanning methods.<br />

The accuracy <strong>and</strong> resolution has been also improved over the time <strong>and</strong> it is possible to get<br />

0,2mm resolution with 0,05mm accuracy (Artec S scanner). This kind of model may be<br />

displayed on a screen <strong>and</strong> virtual museum may be created. Digital models can be also printed,<br />

using special three-dimensional printers which allow recreating the 3D shape of the object.<br />

These methods seem to be the best, because of no risk of destroying the original sculpture.<br />

Moreover, these printers are able to make a copy with good accuracy as well, usually with<br />

0,1mm layer thickness, where 3D dots size is around 0,05mm in diameter.<br />

The project is being done on The Restoration Workshop of Nidaros Cathedral request. They<br />

are trying to reconstruct all the sculptures <strong>and</strong> shapes of the Cathedral. Reconstruction process<br />

takes a long time <strong>and</strong> it is very tedious. Many characters have been reconstructed in plaster<br />

form or sculptures in stone. Around 4000 casts has been made, for preservation copy as a<br />

proof. In 1983 the Archbishop Palace was burnt. A lot of work <strong>and</strong> priceless documentation<br />

was lost in the flames.<br />

In November 2003, Vinn Design company from Oslo (Bache, 2003), have done some<br />

scanning in Nidaros. They made model of Prinsipp <strong>and</strong> Kong Sverre <strong>and</strong> made a real replica<br />

afterwards. The results were quite good, although some parts were not exact reflection of the<br />

real sculpture. Some small details were missing <strong>and</strong> sharp edges were being smoothed. The<br />

Cathedral needs more detailed <strong>and</strong> accurate models. As some of the sculptures are very small,<br />

or some details are very small, a proper method for acquiring the 3D data <strong>and</strong> modelling<br />

should be applied. To get better replica the higher resolution <strong>and</strong> more accurate model is<br />

needed.<br />

In the project investigation on methods of making detailed models has been done. Methods<br />

hereby mean existing photogrammetric software in use of non-metric commercial cameras<br />

<strong>and</strong> close range laser scanning. The software has been tested <strong>and</strong> the results are compared.<br />

The object of investigation is a sculpture – a head of Archbishop from the Cathedral, probably<br />

of Pål Bårdsson. It is a relatively small object, about 25x25x30 cm <strong>and</strong> there are a lot of small<br />

details <strong>and</strong> sharp edges, some small details are even smaller than 1mm.<br />

1.2 Cathedral <strong>and</strong> sculpture background information<br />

Nidaros Cathedal (Nidarosdomen/Nidaros Domkirke) is a church located in city centre in<br />

Trondheim. Figure 1.1 shows the front side of the Cathedral. It was established in 1152 <strong>and</strong> at<br />

that time it was Norwegian archdiocese until its abolition in 1537. After the reformation, it<br />

1


INTRODUCTION<br />

had been the cathedral of the Lutheran bishops of Trondheim (or Nidaros) in the Diocese of<br />

Nidaros. Romanesque <strong>and</strong> Gothic style are represented by architectural style of the cathedral.<br />

The style was also based on English cathedral in Canterbury. This is the largest chapel in all<br />

Sc<strong>and</strong>inavian countries, where a king of Norway Olav II was buried. He interposed Christian<br />

religion in his country <strong>and</strong> for a very long time, his grave was a place of pilgrimages. Because<br />

of this phenomena, 40 years after his funeral the construction of the cathedral began.<br />

Nowadays, it is not possible to see the silver coffin of Olav, as it was moved to Denmark in<br />

1537 <strong>and</strong> remelted into coins.<br />

Building work on the cathedral started in 1070 <strong>and</strong> it was finished about 1300. Nevertheless,<br />

the cathedral had been damaged several times, mostly by fire, firstly in 1327 <strong>and</strong> again in<br />

1531. The nave west of the transept was destroyed <strong>and</strong> it was rebuilt long time after, in early<br />

1900s. Moreover, in 1708 it burned down completely except for the stone walls. After then,<br />

the cathedral had been through major rebuilding <strong>and</strong> restoration, which started in 1869,<br />

initially led by architect Heinrich Ernst Schirmes, <strong>and</strong> nearly completed by Christian Christie.<br />

In 2001 it was officially completed. However, maintenance of the cathedral is an on-going<br />

process. More information can be found on webpages: http://www.nidarosdomen.no/nb-NO/<br />

(Nidaros) <strong>and</strong> http://snl.no/Nidarosdomen (Leksikon et al., 2009).<br />

Figure 1.1 Nidaros Cathedral nowadays – west, front side<br />

2


INTRODUCTION<br />

Generally, the sculptures in the cathedral represents roles in society or the Bible, like Jesus,<br />

Maria, kings, queens, etc. rather than portraits of real person. However, the sculpture used in<br />

the project (see Figure 1.2) is a medieval sculpture of Archbishop Pål Bårdsson. Gerhard<br />

Fisher, who wrote the jubilee book for the restoration of the cathedral (1965) presumes that it<br />

was Bårdsson, archbishop in the period 1333 – 1346. Fisher‟s guess is based on the<br />

assumption of building period a part of the church where the sculpture is located. This place is<br />

on the south side, in the Octagon – the oldest part of the cathedral. Having in mind that the<br />

building history is insecure when it comes to date, the portrait must be considered as a pure<br />

guess. Unfortunately, there is no information about who made the sculpture. The name of the<br />

artist is a relatively modern phenomenon.<br />

Pål Bårdsson died in 01.02.1346 in Nidaros cathedral, but the place of his birth is unknown.<br />

He was elected for archbishop in 1333. When it comes to his education, it is mentioned in the<br />

history that he was a cannon in Bergen from 1309. Later, he became magister in 1313. Pål<br />

represented also bishop in Bergen in law <strong>and</strong> diplomatic fields. Around 1320 he was mostly<br />

spending his time <strong>and</strong> do business abroad. He got also his PhD from the University in Orel<strong>and</strong><br />

in both canonical <strong>and</strong> Rome law. What is more, he had a degree of a professor, which very<br />

few Norwegians had at that time. From 1327 he was also a chancellor of the king. It may be<br />

concluded that he was very wise <strong>and</strong> good man. More information can be found on webpages:<br />

http://www.katolsk.no/organisasjon/mn/erkebiskoper/15 (Organisasjon <strong>and</strong> Martinsen, 1996)<br />

<strong>and</strong> http://snl.no/.nbl_biografi/Pål_Bårdsson/utdypning (Leksikon <strong>and</strong> Dybdahl).<br />

3


INTRODUCTION<br />

Figure 1.2 Actual sculpture of Archbishop in Nidaros Cathedral<br />

1.3 Three-dimensional printers<br />

Since 2003 there has been a large growth in the sale of 3D printers, mostly because the cost of<br />

them has declined. The technology is being applicable in many different fields. Nowadays,<br />

there is large number of competing technologies <strong>and</strong> companies that produce such 3D<br />

printers. The main differences are in the way layers are built to create parts. Printers may use<br />

different tools for cutting the proper 3D surface, but also different materials. Some of them<br />

create bigger objects, whereas some are created for smaller <strong>and</strong> more detailed ones.<br />

Resolution, the most important issue in printing a sculpture with small details, is given in<br />

layer thickness, X-Y resolution in dpi. Usually layer thickness is around 0,1mm, while in X-Y<br />

resolution is comparable to that of laser printers. The particles (3D dots) are around 0,05mm<br />

to 0,1mm in diameter.<br />

4


1.4 Structure of the report<br />

INTRODUCTION<br />

In the beginning the report contains theoretical background about photogrammetry <strong>and</strong><br />

scanning method. Information about camera calibration <strong>and</strong> computed features are given.<br />

Also, there is a description of some basic – principle equations which may be met in<br />

photogrammetry. When it comes to scanning, there is a briefing about laser scanning <strong>and</strong><br />

structured – light scanners <strong>and</strong> the principle that they use – triangulation.<br />

Next section describes instruments used such as camera with lens, scanners <strong>and</strong> computer.<br />

There is also basic information about software: photogrammetric <strong>and</strong> used by canners for<br />

acquisition <strong>and</strong> also for post-processing.<br />

Later on, there are the results, which are divided into <strong>Photogrammetry</strong> <strong>and</strong> Scanning. There is<br />

also at the same time comparison of those methods. Some discussion may be found there as<br />

well.<br />

The last section is discussion <strong>and</strong> conclusions. It contains discussion on what the results<br />

showed <strong>and</strong> which method was the best for the project purpose. There are also figures<br />

comparing models from different methods, which are focused on more detailed areas of the<br />

sculpture.<br />

5


2 THEORETICAL BACKGROUND<br />

2.1 Review of similar projects<br />

THEORETICAL BACKGROUND<br />

Over the time, there were done similar projects concerning documentation of cultural<br />

heritage. Some of them were more precise <strong>and</strong> some were more generalized. For instance in<br />

paper 3D Color Imaging For Cultural Heritage Artefacts Hess <strong>and</strong> Robson (Hess <strong>and</strong><br />

Robson, 2010) are showing how to use 3D laser scanners <strong>and</strong> close range photogrammetry for<br />

this purpose, they also made some comparison. In the paper An Optical Three-dimensional<br />

Measuring Techniques for a Detailed Non-contact Data Acquisition…Floth <strong>and</strong> Breuer (Floth<br />

<strong>and</strong> Breuer, 2010) are showing mostly laser scanning techniques for data acquisition <strong>and</strong><br />

process chain. However, in the What is the Future of Metric Heritage Documentation <strong>and</strong> Its<br />

Skills? (Blake, 2010) Blake is focused on future software that could be used for cultural<br />

heritage preservation. He describes new methods <strong>and</strong> shows that online web – based software<br />

may be future.<br />

In some projects, referring to paper works, the same software has been used. For example<br />

speaking of ImageMaster – photogrammetry software, was used by (Wojtas, 2010) in Offthes-helf<br />

Close-range <strong>Photogrammetry</strong> Software for Cultural Heritage Documentation at<br />

Stonehenge. She also used PhotoModeler Scanner for this purpose <strong>and</strong> she also made camera<br />

calibration. Nevertheless, it must be mentioned, that she used PI-3000 which is previous<br />

version of Topcon ImageMaster. She tested this software on Stonehenge rock. Also, (Kochi et<br />

al.) in PC-based 3D Image Measuring Station With Digital Camera…, used PI – 3000<br />

software. Their case of study was historical ruins of Parthenon but also Hdriaphooroi –<br />

statues. What is more, (Chmielewski <strong>and</strong> Szulwic) in Niemetryczne zdjęcia cyfrowe w<br />

fotogrametrii bliskiego zasięgu w systemie Topcon PI-2000, Chmielewski <strong>and</strong> Szulwic also<br />

used the same software, however their test object was architectural building, but Kazanaa<br />

(Kazanaa, 2010) , used the software on the old fountains abbey stone (English heritage).<br />

These methods were applied for non-metric commercial cameras <strong>and</strong> make process automatic<br />

<strong>and</strong> easy to use. For the modelling of the telamon in the archaeological museum ImageMaster<br />

was also applied in connection to laser scanning data – using TOF ZScan system of Menic<br />

Software - (Brutto <strong>and</strong> Spera, 2011). In the same combination as in the project ImageMaster<br />

<strong>and</strong> the same scanner Konica Minolta is used by Barnett (Barnett et al.) while Recording<br />

Prehistoric Rock Art. PiCalib software for camera calibration from Topcon was also used in<br />

Master Theses supervised by Ph.D. Regina Tokarczyk at University of Science <strong>and</strong><br />

Technology (<strong>AGH</strong>, Kraków) (Pawlak <strong>and</strong> Pulchny, 2009), but also in (Jędrzejek <strong>and</strong> Łącka).<br />

Some experiments with PhotoModeler Scanner have made Alyilmaza with others (Alyilmaza<br />

et al., 2010) during Measurement of Petroglyphs (Rock of Arts…), where the objects of<br />

interest were drawings on the rock like Qobustan petroglyphs. From laser scanning,<br />

mentioned before, Konica Minolta was also used for a purpose by Barba (Barba et al., 2011).<br />

They used this scanner with Tele lens for documentation of artefact from the Roman period,<br />

found in Spanish city Merida. For modelling, visualisation <strong>and</strong> texture they used Geomagic<br />

<strong>and</strong> Photoshop.<br />

Geomagic, software that is able to h<strong>and</strong>le point clouds from laser scanning <strong>and</strong> make models<br />

was applied by (Akca et al.) in Recording <strong>and</strong> modeling of cultural heritage objects with<br />

coded structured light projection systems <strong>and</strong> other similar papers, done by the same persons,<br />

where objects: Weary Herakles, the Khmer Head, Herakles Farnese - sculptures , Lady<br />

6


THEORETICAL BACKGROUND<br />

praying painting were used for documentation. They mentioned that they had used optoTOP-<br />

HE <strong>and</strong> optoTOP-SE, Breuckmann GmbH, systems for 3D digitalization, acquisition, <strong>and</strong><br />

recording. Moreover, they compared two similar applications: Geomagic <strong>and</strong> Polyworks,<br />

which both seems to be good. Though, (Tucci et al., 2011) in Effective 3D Digitization of<br />

Archeological Artifacts… also used Geomagic for documentation of small heritage objects.<br />

They used triangulated based laser scanner NextEngine‟s 3D Scanner HD, which is generally<br />

used for surveying small <strong>and</strong> medium objects. Bigger artifacts were scanned by Z-Scan by<br />

Menci Software photogrammetric system. Texture was captured simultaneously.<br />

For small objects h<strong>and</strong>held laser scanner seems to be very convenient <strong>and</strong> easy to use.<br />

Bonfanti with others (Bonfanti et al., 2010) used such a scanner – H<strong>and</strong>yScan 3D model HZ,<br />

which is triangulated based LiDAR system. For modelling <strong>and</strong> visualisation RiscanPro <strong>and</strong><br />

3D Reshaper were applied in their project.<br />

There is much more photogrammetric software available on the market, or laser scanners,<br />

which use triangulated based system for small objects surveying. Clorama software from<br />

3Dixpplorer is one of the photogrammetric software which is able to reconstruct 3D shapes<br />

from pictures. Pictran or iWitness may be used in example for camera calibration, but they are<br />

not able to create a mesh.<br />

There are also some open source softwares, available even via internet, which are able to<br />

create 3D models from pictures. It is believed these types of software are the future. They are<br />

low cost <strong>and</strong> they do not require very specific photographs. The user does not have to know<br />

photogrammetric principles <strong>and</strong> all the calculations are made in one step. The results of<br />

course are worse than using other software <strong>and</strong> there is very little info about the accuracy.<br />

However, in many cases this might be enough. Deseilligny <strong>and</strong> Clery (Deseilligny <strong>and</strong> Clery,<br />

2011) tested one of these software – Apero <strong>and</strong> Micmac. Also, Balestrinia <strong>and</strong> Guerra<br />

(Balestrinia <strong>and</strong> Guerra, 2011) used Arc3D for 3D reconstruction. On the market there are<br />

also visible Autodesk Project PhotoFly <strong>and</strong> Photosynth which are also open source licensed<br />

<strong>and</strong> produce point cloud or mesh, or both<br />

2.2 <strong>Photogrammetry</strong><br />

<strong>Photogrammetry</strong> is the science of making measurements from photographs. By using<br />

photographs it is possible to determine geometric properties of objects by specific<br />

measurements <strong>and</strong> computations. It is possible to create three dimensional models as well.<br />

By proper calculation, camera station for each photograph may be described. From<br />

intersection of light rays from each photo position of 3D point may be obtained. For this<br />

purpose usually bundle adjustment with least-squares optimization is applied.<br />

<strong>Photogrammetry</strong> can be dated to the mid-nineteenth century <strong>and</strong> words may be explained as:<br />

Photo – light<br />

Gram – drawing<br />

Metry – measurement<br />

2.2.1 Aerial <strong>Photogrammetry</strong><br />

The camera is mounted in an aircraft <strong>and</strong> is usually pointed vertically towards the ground.<br />

Multiple overlapping photos of the ground are taken as aircraft flies along a flight plan. These<br />

7


THEORETICAL BACKGROUND<br />

photos are processed in a stereo mode – nowadays usually with 3D glasses connected to the<br />

workstation – to see two photos at once in stereo view. These photos are usually used for<br />

making Digital Elevation Model (DEM) <strong>and</strong> orthophotomaps.<br />

2.2.2 Close – range photogrammetry<br />

The camera is located close to the object, h<strong>and</strong>held or on a tripod. Usually this type of<br />

photogrammetry work is non – topographic. It is used usually to model buildings, engineering<br />

structures, vehicles, cultural heritage objects, etc.<br />

To be able to describe an object in 3 dimensions there is a need to be at least two pictures of<br />

the object taken in a proper way. At the beginning of photogrammetry pictures were analogue<br />

<strong>and</strong> specific instruments were used to match pictures <strong>and</strong> use them for map making for<br />

instance. Nowadays everything is computerized <strong>and</strong> there are plenty of cheap, commercial<br />

digital cameras which may be used for this purpose. There are also available metric cameras,<br />

very accurate, with certificates of calibration, made for geodetic measurements. Their<br />

specification is that they have fixed settings like: principle distance, aperture, etc.<br />

Knowing camera parameters <strong>and</strong> location of the cameras, which is computed during camera<br />

orientation (interior, exterior, relative <strong>and</strong> absolute orientation – or in one block during bundle<br />

adjustment), which is mainly to coordinate photos in the space, it is possible to compute 3D<br />

point from triangle – rays intersection (using at least two photos). Figure 2.1 shows this<br />

principle. As these parameters are known, rays may be recreated. After the identification of<br />

common points on each image it is possible to gain the position of this point in 3D space,<br />

which is the intersection of the recreated rays.<br />

Figure 2.1 Photogrammetric principle of 3D measurement; example on convergent image<br />

configuration (Luhmann et al., 2006)<br />

Firstly, the object of interest should be examined, if it is big/small <strong>and</strong>, what kind of features<br />

it contains, etc. In respect to shape of the object, the proper camera with lens (wider angle or<br />

not) should be chosen. If applying lens with wider angle the image scale may be bigger <strong>and</strong><br />

more details may be captured. But, it may also lead to bigger error (distortion etc.) <strong>and</strong> more<br />

photographs needed to cover the entire object.<br />

8


THEORETICAL BACKGROUND<br />

For computation of orientation, control points should be measured. In some cases they should<br />

be marked, for better accuracy or to measure them with Total Station to get Global<br />

coordinates. An another situation, marking is not needed at all, because natural features<br />

(natural points, lighter or darker spots for instance) may be used for orientation – then the<br />

computations are in local coordinate system.<br />

No matter what types of control point are used, these have to be measured. Some softwares<br />

are able to mark points in the image <strong>and</strong> reference the same points between the images<br />

automatically (if special coded points used), some are able to measure semi-automatically <strong>and</strong><br />

in others all measurements have to be done by the user. Usually, if automation is available the<br />

software has option for sub-pixel measurements by applying computer vision methods. Every<br />

software has requirements, when it comes to amount of points which are necessary for<br />

computation. It depends on applied algorithms <strong>and</strong> calculation methods. Unfortunately it is<br />

hard to say what exact type the specific software uses, the companies do not want to share this<br />

information. Nevertheless, the main principle is that computations are based on bundle<br />

adjustment <strong>and</strong> least-square normalisation.<br />

Next, the software computes all the necessary parameters to get information about orientation<br />

<strong>and</strong> link the photos between them. The same moment there is accuracy assessment, <strong>and</strong> at this<br />

moment the user should have a look on the results, if they are sufficient. Sometimes the<br />

corresponding points may be wrongly referenced or measured at a wrong place. This<br />

assessment helps to find those errors. It is also possible to remove lens errors from photograph<br />

<strong>and</strong> transform them from central to orthogonal projection.<br />

From that point, point cloud – 3D may be generated. In the past, it was done manually, but<br />

todays methods give an opportunity to generate 3D points automatically, using advanced<br />

computer vision. Image matching with sub-pixel accuracy give good prospects.<br />

Unfortunately, in some cases there might occur miss-match <strong>and</strong> points have to be checked by<br />

the user. For generation of Digital Surface Model (DSM) the area on the image should be<br />

specified. It helps software to limit time for image-matching, while not all the photograph<br />

contains the object of interest. From that point, cleaning the data is being started. Modelling<br />

<strong>and</strong> visualisation tools are described further in the report.<br />

2.2.3 Focusing<br />

For accurate measurements, the object of interest has to be sharp. Then, the object has to be in<br />

a depth of field area, where the object will be sharp. The object is sharp when the diameter of<br />

the circle of confusion u’ does not exceed a diameter of ca. 1pix for digital sensors. Hence,<br />

not only the object point P focused in distance a is sharply defined, but all points between P1<br />

<strong>and</strong> P2 appear sharp. Figure 2.2 illustrates this situation. Depth of field is dependent on f –<br />

number (aperture), distance to the object <strong>and</strong> focal length. What is more, the depth of field<br />

increases by stopping down, enlarging distance to object <strong>and</strong> reducing focal length.<br />

9


Where:<br />

T – depth of field<br />

R – the furthest sharply distance<br />

S – the nearest sharply distance<br />

u’ - circle of confusion<br />

THEORETICAL BACKGROUND<br />

Figure 2.2 Focusing <strong>and</strong> depth of field (Luhmann et al., 2006)<br />

2.2.4 Image scale <strong>and</strong> accuracy<br />

Scale of the image m is connected with distance to the object, principle distance <strong>and</strong> sensor<br />

size (image format).<br />

Where:<br />

h – object distance<br />

c – principle distance<br />

X – distance in object space<br />

x’- distance in image space<br />

In order to achieve sufficient accuracy <strong>and</strong> delectability of fine detail at the scene, the image<br />

scale M has to be defined appropriately with respect to the available imaging system <strong>and</strong> the<br />

exterior environmental conditions.<br />

10<br />

(2.1)<br />

Accuracy is connected to the image scale, because if the bigger scale, the more details are<br />

visible in the photo. What is more, configuration of the image may have influence on the<br />

accuracy, especially when it comes to Z direction. It is connected with height-to-base ratio<br />

(h/b) <strong>and</strong> if the cameras were situated parallel or convergent to each other. Also, generation of<br />

3D point cloud – resolution is dependent on image scale


2.2.5 Coordinate systems<br />

THEORETICAL BACKGROUND<br />

From image to model is a long way. Just after recording, photo is in image coordinate system<br />

(defined by camera); while in the end the model is in global coordinate system, or at least in<br />

object local system. Several coordinate systems may be specified:<br />

� Image coordinate system<br />

It is two dimensional image-based reference system of rectangular Cartesian coordinates x’,<br />

y’. Usually the origin of the image of frame coordinates is located at the image centre. In<br />

digital images, origin is located in left upper corner <strong>and</strong> values of pixels are specified by rows<br />

<strong>and</strong> columns, <strong>and</strong> thus mainly the coordinate system has to be transformed to the centre of<br />

projection on the image<br />

� Camera coordinate system<br />

It is similar to the image coordinate system, it is only moved by z’ axis – principle distance c<br />

– the origin on this system is the spatial position of the perspective centre<br />

� Object coordinate system<br />

It is also known as world coordinate system, or global, Cartesian XYZ coordinate system. The<br />

system is used for all the photographs <strong>and</strong> object. It is defined by reference points on the<br />

object, if the points are measured in global coordinate system, or if not - it may be defined as<br />

local coordinate system of the object, where all the photographs are referenced to the one<br />

(chosen by the used).<br />

� 3D instrument coordinate system<br />

Coordinate system is oriented by one of used 3D measuring machine. This is not directly<br />

related to any superior system or particular object.<br />

2.2.6 Sensor principle<br />

“Sensors consist of a large number of light-sensitive detector element, that are arranged on<br />

semi-conductor modules either line-wise or matrix-wise (line or matrix sensor). Each detector<br />

element (sensor element) generates an electric charge that is proportional to the amount of<br />

incident illumination falling on it. The sensor is arranged such that the charge at each<br />

individual element can be read out, processed <strong>and</strong> digitized.” (Luhmann et al., 2006)<br />

There are sensors based in CCD (Charge Coupled Device) or CMOS technology – they differ<br />

on used material <strong>and</strong> transportation of the electro-magnetic information. CMOS sensors<br />

(complementary metal oxide semi-conductor) seem to be better in many ways. They have less<br />

power consumption, lower manufacturing costs, high image frequencies, high dynamic range<br />

<strong>and</strong> low image noise.<br />

2.2.7 Camera calibration - Interior orientation<br />

Camera calibration is a process to get specification of the camera, parameters of interior<br />

orientation. Some cameras have a certificate, where camera calibration has been done in the<br />

lab. But, commercial cameras do not have it in the beginning. In these cameras there is a<br />

possibility to change lens, aperture, focal length, which changes camera features.<br />

In addition to make accurate 3D measurements – good (accurate) camera calibration is<br />

needed. At this moment there are softwares on the market which provide calibration for<br />

commercial cameras. Some of them may be used for processing at the same time.<br />

11


THEORETICAL BACKGROUND<br />

To make camera calibration, pictures of special calibration chart should be taken with fixed<br />

settings: fixed focal length, focus, principle distance, aperture; <strong>and</strong> it is more secure to at least<br />

fix the lens with a tape to prevent eventual moving. They should be taken also on the same<br />

day with the same conditions, e.g. temperature <strong>and</strong> humidity. The test field may differ. Most<br />

of software companies developed their own calibration pattern; however it is usually possible<br />

to create a new one. Using pattern that the company recommends to use – calibration<br />

computation is automatic, otherwise mainly all the points have to be measured manually or at<br />

least semi-automatically. Every software has his own recommendations how to take pictures<br />

(from which angles etc.) <strong>and</strong> how many. It is also possible to make self – calibration on the<br />

field, but the results may not be as accurate as using test field. Without camera calibration a<br />

slightly different equations are used <strong>and</strong> it is very hard to find the error when it occurs.<br />

No matter what kind of software is used the same features, listed below are computed. The<br />

algorithms may be slightly different but the principle is the same – to get the most accurate<br />

results – with Least Square Method (LSM).<br />

Perspective centre <strong>and</strong> distortion<br />

Mathematically, the perspective centre is defined by the point of central perspective that is the<br />

point through which all straight lines from all image rays pass. Perspective centre depends on<br />

chosen aperture. This point may be defined both external <strong>and</strong> internal of objective. In the ideal<br />

case these points are equal. Similarly in ideal case, principle distance c is equal to the image<br />

distance a’, still there is always some error. What is more, lens distortions, which will always<br />

occur, also change interior geometry. Geometric basics of perspective centre <strong>and</strong> distance<br />

shows Figure 2.2.The shift of the point on the image may be computed:<br />

Where:<br />

τ – angle of incidence<br />

τ’- exit angle<br />

r’ – image radius of an image point<br />

∆r‟- radial shift due to distortion<br />

Figure 2.3 Perspective centre <strong>and</strong> principle distance (Luhmann et al., 2006)<br />

12<br />

(2.2)


THEORETICAL BACKGROUND<br />

Computed features of camera calibration; interior orientation are:<br />

� Principle Point (image coordinates)<br />

Nadir of the perspective centre with image coordinates (x’0, y’0= for st<strong>and</strong>ard cameras<br />

approximately equal to the centre of the image)<br />

� Principle distance<br />

It is normal distance to the perspective centre from the image plane; approximately equal to<br />

the focal length of the lens when focused at infinity: c≈f’<br />

� Parameters of functions describing imaging errors, distortions, etc. (∆r’):<br />

o Radial distortion<br />

It is the major imaging error, the most effective, in most camera systems. It is connected to<br />

the lens <strong>and</strong> its error. It is related with refraction at each individual component lens within the<br />

objective. It may also change while changing focusing distance <strong>and</strong> the object distance. Figure<br />

2.4 illustrates the effect of radial imaging error.<br />

Distortion curve is usually modelled with a polynomial series with parameters K1 to Kn<br />

((Luhmann et al., 2006) - based on Brown, 1971):<br />

Figure 2.4 Effect of radial – symmetric distortion; example (Luhmann et al., 2006)<br />

13<br />

(2.3)<br />

Parameter K2 is often correlated with K1. Parameter K3 can normally be determined to<br />

significant value only in special cases, f.e. for fish eye lens.<br />

o Tangential (asymmetric or decentric) distortion<br />

Radial – asymmetric distortion, called tangential or decentric. It is mainly caused by decentric<br />

<strong>and</strong> misalignment of individual lens elements within the objective<br />

Figure 2.5 illustrates the effect of decentric distortion.<br />

By following function this distortion may be compensated ((Luhmann et al., 2006) - based on<br />

Brown 1971):<br />

( )<br />

( )<br />

(2.4)


THEORETICAL BACKGROUND<br />

Figure 2.5 Effect of radial – asymmetric <strong>and</strong> tangential distortion; example (Luhmann et al.,<br />

2006)<br />

o Affinity <strong>and</strong> shear of the image coordinate system<br />

These components represent error in due to orthogonallity <strong>and</strong> scale image coordinate system<br />

to coordinate axes.<br />

Figure 2.6 represents an example of the effect of this error.<br />

The following function can be used to provide an appropriate correction:<br />

Figure 2.6 Effect of affinity <strong>and</strong> shear deviations; example (Luhmann et al., 2006)<br />

Parameters of tangential asymmetric, distortion B1, B2, <strong>and</strong> affinity <strong>and</strong> shear C1, C2 often<br />

exhibit high correlation with other values of interior orientation.<br />

o In some cases, there are some other additional parameters describing imaging<br />

error, however they have less influence <strong>and</strong> are rather very small.<br />

14<br />

(2.5)


THEORETICAL BACKGROUND<br />

Total correction<br />

The individual terms used to model the imaging errors of most typical photogrammetric<br />

imaging systems can be summarised as follows:<br />

Interior orientation<br />

Knowing these parameters, the (error-free) imaging vector x’ can be defined with respect to<br />

the perspective centre (hence, the principle point).<br />

Figure 2.7 illustrates geometric principle of interior orientation.<br />

[<br />

] [<br />

Where:<br />

x’p, y’p – measured coordinates of image point P’<br />

x’0, y’0 – coordinates of the principle point H’<br />

∆x’, ∆y’ – axis – related correction values for imaging errors<br />

2.2.8 Exterior orientation<br />

Figure 2.7 Interior orientation (Luhmann et al., 2006)<br />

This step is to orient camera in the space, it consists six parameters which describe this<br />

position in camera coordinate system. As the project is in close range photogrammetry, not<br />

aerial, only the example on close range situation is shown.<br />

Six parameters are: three coordinates X0, Y0, Z0 of the principle point in image coordinate<br />

system <strong>and</strong> three rotation angles α, κ, ω– around three axes x, y, z; what is shown in Figure<br />

2.8.<br />

15<br />

]<br />

(2.6)<br />

(2.7)


THEORETICAL BACKGROUND<br />

Figure 2.8 Exterior orientation for terrestrial photogrammetry (Luhmann et al., 2006)<br />

Matrix R of rotation for exterior orientation is a set of combination of sinus <strong>and</strong> cosinus<br />

functions of rotation angles mentioned before. Matrix X0 describes position of the camera.<br />

2.2.9 Collinearity equations<br />

These equations are the core in the photogrammetry. From these equations it is possible to get<br />

object coordinates from image coordinates (exterior orientation). These equations are also<br />

base for other computations, like Bundle Adjustment or Direct Linear Transformation.<br />

Collinearity equations:<br />

( ) ( ) ( )<br />

( ) ( ) ( )<br />

( ) ( ) ( )<br />

( ) ( ) ( )<br />

Where:<br />

x’, y’, z’ – image coordinates of corresponding points (measured on the picture)<br />

z’ – is usually equal principle distance z’=-c<br />

x’0, y’0 – coordinates of the principle point H’ in image coordinate system<br />

∆x’, ∆y’ – correction terms of the image coordinates<br />

X, Y, Z – object coordinates of corresponding points (measured on the picture)<br />

X0, Y0, Z0 – object coordinates of principle point H<br />

These equations describe the transformation of object coordinates (X, Y, Z) into corresponding<br />

image coordinates (x’, y’) as functions of the interior parameters <strong>and</strong> exterior orientation<br />

parameters of one image.<br />

An alternative form may be given, if the object coordinate system is transformed by shift to<br />

the perspective centre <strong>and</strong> orientation parallel to the image coordinate system.<br />

These equations demonstrate clearly that each object point is projected into a unique image<br />

point, if it is not occluded by other object points. The formulas effectively describe image<br />

generation inside a camera by the geometry of a central projection.<br />

16<br />

(2.8)


THEORETICAL BACKGROUND<br />

The collinearity equations are fundamental equations of analytical photogrammetry. They are<br />

suitable for direct use of observations, also in an over-determined least-squares adjustment.<br />

They are used to set up the equation system for spatial intersection, space resection <strong>and</strong><br />

bundle triangulation.<br />

2.2.10 Direct Linear Transformation (DLT)<br />

These equations let determine the orientation of an image without need for approximate initial<br />

values. It is based on collinearity equations, extended by an affine transformation of the image<br />

coordinates. There is also no need for fixed coordinate system in the camera.<br />

DLT equations:<br />

Where:<br />

x, y – measured comparator or image coordinates<br />

X, Y, Z – 3D coordinates of the reference point<br />

L1 – L11 – coefficient, DLT parameters – from these the parameters interior <strong>and</strong> exterior<br />

orientation may be derived<br />

DLT needs minimum 6 points measured on the picture; but for better result, eventual error<br />

detection <strong>and</strong> accuracy assessment there is need to measure more than 6. What is more, using<br />

DLT equations is not easy to detect errors in reference points.<br />

2.2.11 Absolute orientation<br />

This orientation describes how the images are situated in the space in global coordinate<br />

system. It means the local coordinate system may be rotated, scaled <strong>and</strong> moved due to<br />

reference points. It is used mostly in aerial photogrammetry. Sometimes in terrestrial, though<br />

for small objects like sculpture it is not needed, because only the local coordinate system is<br />

used.<br />

2.2.12 Bundle triangulation<br />

17<br />

(2.9)<br />

Bundle block adjustment, multi–image triangulation, multi–image orientation is a method for<br />

the simultaneous numerical fit of unlimited number of spatially distributed images (bundles of<br />

rays). By use of tie points, single images are merged into a global model which the object<br />

surface can be reconstructed in three dimensions. The most important constraint is that all<br />

corresponding (humongous) images rays should intersect in the corresponding object point<br />

with minimum inconsistency (Least Square method). In an over – determined system of<br />

equations, an adjustment technique estimates 3D object coordinates, image orientation<br />

parameters <strong>and</strong> any additional model parameter, together with related statistical information<br />

about accuracy <strong>and</strong> reliability. It is one set of simultaneous calculations, triangulation is the<br />

most powerful <strong>and</strong> accurate method of image orientation <strong>and</strong> point determination in<br />

photogrammetry.


THEORETICAL BACKGROUND<br />

The bundle adjustment has a lot of advantages, because it may be applied with irregularly<br />

arranged <strong>and</strong> often unfavourable image configurations. It is more complex structure of normal<br />

system of equations. There is complex generation of approximate values for the unknowns<br />

<strong>and</strong> arbitrary oriented object coordinate systems. It is possible to combine adjustment of<br />

survey observations <strong>and</strong> conditions, where also several imaging systems can be calibrated<br />

simultaneously.<br />

Adjustment model - Gauss-Markov linear model - based on collinearity equations<br />

This principle is based on that the unknown parameters are estimated with maximum<br />

probability.<br />

Condition for the residuals results (L2 - normalization):<br />

∑ [ ]<br />

Mathematical model of the bundle block adjustment is based on the collinearity equations,<br />

which are mentioned before in equation (2.8) above.<br />

18<br />

(2.10)<br />

The structure of these equations allows the direct formulation of primary observed values<br />

(image coordinates) as functions of all unknowns‟ parameters in the photogrammetric<br />

imaging process. The collinearity equations, linearized at approximate values, can therefore<br />

be used directly as observation equations for least-square adjustment according to the Gauss-<br />

Markov model.<br />

It is principle of the image coordinates of homologous points which are used as observations.<br />

The following unknowns are iteratively determined as functions of these observations.<br />

� 3D object coordinates for each new point i (up, 3 unknowns each)<br />

� exterior orientation of each image (uI, 6 unknowns)<br />

� Interior orientation of each camera k (uc, 0 or ≥ 3 unknowns each)<br />

The bundle adjustment therefore represents an extended form of the space resection:<br />

Where:<br />

i – point index<br />

j – image index<br />

k – camera index<br />

(<br />

(<br />

St<strong>and</strong>ard form, the linearized model (functional model):<br />

With n observations <strong>and</strong> u unknowns, n>u.<br />

⏟̂<br />

⏟<br />

⏟<br />

̂⏟<br />

)<br />

)<br />

(2.11)<br />

(2.12)


Where:<br />

Solving the normal equations:<br />

Variance – covariance matrix:<br />

THEORETICAL BACKGROUND<br />

̂⏟<br />

⏟<br />

⏟<br />

⏟<br />

⏟<br />

⏟<br />

( ⏟<br />

19<br />

⏟<br />

⏟<br />

⏟<br />

⏟ )<br />

⏟<br />

⏟<br />

⏟<br />

⏟<br />

(2.13)<br />

(2.14)<br />

(2.15)<br />

The adjustment is solved iteratively. In this case the corrected approximate values in iteration<br />

k are used as new starting values for the linearized functional model of next iteration k+1,<br />

until the sum of added corrections for the unknowns is less than a given threshold.<br />

Precision <strong>and</strong> accuracy<br />

St<strong>and</strong>ard deviation a posteriori, the empirical st<strong>and</strong>ard deviation of unit weight is given by:<br />

With redundancy:<br />

̂ √<br />

St<strong>and</strong>ard deviation of a single unknown xj is given by:<br />

̂ ̂ √<br />

Where qjj are the elements of the principle diagonal of matrix Q<br />

(2.16)<br />

(2.17)<br />

RMS – root mean square. In many cases adjustment results are reported as RMS instead of<br />

above defined st<strong>and</strong>ard deviation. The RMS value is the square root of the mean squared<br />

difference between n given nominal values Xnom <strong>and</strong> corresponding adjusted observations Xobs<br />

√ ∑( )<br />

(2.18)


THEORETICAL BACKGROUND<br />

RMSE – root mean square error – RMS error of adjusted observations with respect to mean<br />

of adjusted observations<br />

√ ∑( ̅)<br />

20<br />

(2.19)<br />

To orient two images or more, tie points have to be defined. These points may be natural<br />

points which are visible on both pictures. However, they may be marked before. These points<br />

may be defined in a specific shape – coded points, where every software may have different<br />

pattern. In some cases the points (or at least few of them) should have known coordinates<br />

(XYZ), for orientation to global coordinate system, but for close range <strong>and</strong> in the project it is<br />

not necessary. Points should be arranged around the object <strong>and</strong>/or could be on the object.<br />

Picture should cover as best (as biggest area as possible) the object with marked points around<br />

it. Referenced points should not be located in a common plane, this situation may give errors.<br />

Depending on methods (algorithm, transformation etc.), or step in orientation different<br />

amount of control points is needed. In Figure 2.9 is specification of point determination on<br />

each field.<br />

Figure 2.9 Methods <strong>and</strong> data flow for orientation <strong>and</strong> point determination (based on<br />

(Luhmann et al., 2006))<br />

2.2.13 Matching types of digital images<br />

Matching is a process in computer vision that matches pictures to each other. These are<br />

methods which are used to identify <strong>and</strong> uniquely match identical objects features.<br />

Feature-based matching<br />

This is usually the first step <strong>and</strong> it is mainly identifying as many corresponding features as<br />

possible in all images. Specifying area of interest or any other information can be used to limit<br />

the search space <strong>and</strong> minimalize mismatches. This matching is done by extracting distinct


THEORETICAL BACKGROUND<br />

features from images, <strong>and</strong> then identifies those features that correspond to each other by<br />

comparing features attributes <strong>and</strong> location. Types of features may be: edge elements, line<br />

segments, curve segments, circles, ellipses, regions, where cost function may be used.<br />

Correspondence analysis based on epipolar geometry – see Figure 2.10. In the figure the<br />

situation is simplified, it depicts two cameras looking at point X. Virtual image plane is placed<br />

in front of the focal point of each camera to produce unrotated image. OL <strong>and</strong> OR represent the<br />

focal points of the two cameras. Points xL <strong>and</strong> xR are the projections of point X onto the image<br />

plane. When the projection of point xL is known, then the epipolar line eR-xR is known <strong>and</strong> the<br />

point X projects into the right image, on a point xR, which lie on this particular epipolar line.<br />

When the points xL <strong>and</strong> xR are known, their projection lines are also known. If the two image<br />

points correspond to the same 3D point X the projection lines must intersect precisely at X.<br />

Furthermore, the epipolar lines are parallels to the line OL-OR between the focal points.<br />

Figure 2.10 Epipolar geometry<br />

Area-based multi-image matching<br />

It is more precise <strong>and</strong> accurate matching than the one mention above. Correlation <strong>and</strong> least -<br />

squares method are used in this method. In addition, geometric information, epipolar<br />

geometry can be used to improve accuracy <strong>and</strong> reliability. This type of matching compares<br />

the grayscale values of patches of two or more images, trying to find features based on<br />

similarity in those grayscale value patterns.<br />

Object-based matching, Symbolic<br />

This matching use as well geometric but also radiometric object model, where the<br />

corresponding image regions are determined iteratively. The algorithm uses cost function <strong>and</strong><br />

description of symbols. In this matching, the relation between sample <strong>and</strong> correct choice<br />

stimulus is established arbitrarily by the experimenter.<br />

2.3 Three-dimensional digitizers; Three-dimensional scanning<br />

Three-dimensional digitizer is a device that analyses a real world objects or environment to<br />

collect data on its shape <strong>and</strong> if possible e.g. colour. The collected data are usually 3D point<br />

clouds which can be used to construct digital, three dimensional models. This may be used for<br />

several purposes, like industrial design, reverse engineering, etc. <strong>and</strong> of course documentation<br />

of cultural artifacts. Every scanner (or set of scanners inside one company) has its own<br />

software for data acquisition <strong>and</strong> processing. The software for processing, when the point<br />

clouds are captured, has been also developed separately by different companies.<br />

21


THEORETICAL BACKGROUND<br />

In the beginning the purpose of the model, or the object, or both has to be investigated <strong>and</strong> in<br />

respect to this, proper scanner should be used. In some cases it is needed to put additional<br />

targets, for registration of scans, which are taken from different positions. Some companies<br />

required their own coded targets. Some scanners do not need any special targets on the object<br />

for referencing the scans. Aligning is done by approximate specification of corresponding<br />

features which are easy to define <strong>and</strong> algorithms compute needed parameters, by applying<br />

high precision matching computations.<br />

Some scanners are also available to capture a texture – to record photos, <strong>and</strong> put the real<br />

colour on the 3D points. Model has real appearance afterwards. During the data acquisition,<br />

everything depends on used scanner. Usually, for the close-range the object should be visible<br />

<strong>and</strong> separated from background if possible. If the background is far enough, it is not recorded,<br />

because of range limitations of the scanner. For scanning it is also important, that the object is<br />

not made from reflective material/ shiny/ or glass. Some scanners may cope with these<br />

materials. Mostly, when the scanner uses laser or light projection, the light may be reflected<br />

(or go through the object – glass) <strong>and</strong> then the model may result in noise, badly described 3D<br />

position of the points.<br />

After data acquisition it is time for selection of the data. At this stage, the background <strong>and</strong> all<br />

the necessary surfaces should be removed. These cosmetics help software with later alignment<br />

<strong>and</strong> consequence with better results. Next, registration of the point clouds should be done. As<br />

mentioned before, for this purpose corresponding points should be measured/ specified on the<br />

scans. Here, again, software specifies how many points are necessary. Later, the fine<br />

registration is being computed, very accurate computations, where all the scans are being<br />

align simultaneously. Next, the scans may be merged into one regular point cloud. After this<br />

modelling / edition of the 3D surface may be done <strong>and</strong> this is described further.<br />

2.3.1 Time-of-light (TOF) measurements<br />

This type of measurements is preferred at longer ranges. A pulse TOF detects the time a laser<br />

beam is reflected back to the receiver detector. The accuracy depends on how precisely we<br />

can measure the travel distance of the light; which is 3,3 picoseconds is the time taken for<br />

light to travel 1mm. In these scanners we may say that accuracy is relatively constant for the<br />

whole volume of measurement, but increasing the distance the accuracy will be decreasing.<br />

Also, with increasing distance dense resolution of the point cloud is decreasing. These types<br />

of scanners are able to gain about 5000 - 10000 points per second. Distance range is from<br />

200m +. Of course, these numbers are approximate – every laser scanner – every company<br />

may have different specifics. These types of scanners are also used in aerial scanning –<br />

LiDAR (Laser Detection <strong>and</strong> Ranging).<br />

2.3.2 Phase-shifting measurement techniques<br />

Time-of-light method may also be realized by amplitude modulation, using phase difference.<br />

This technique is usually more precise than normal TOF; but there is a higher possibility to<br />

get big – gross error. It is because the distance is computed by the difference in a phase of<br />

sent beam <strong>and</strong> reflected. It is easy to find the difference but it may be not obvious to get the<br />

amount how many times the phase was repeated. These types of scanners are able to gain<br />

even 100000 points per second. Distance range is from up to 80m <strong>and</strong> accuracy about 1-2mm.<br />

Of course, these numbers are approximate – every laser scanner – every company may have<br />

different specifics.<br />

22


2.3.3 Triangulation-based measurements<br />

THEORETICAL BACKGROUND<br />

Laser scanners may be based on different principles <strong>and</strong> use slightly different techniques,<br />

motors, mirrors, laser beams, structured-light patterns <strong>and</strong> so on. The triangulation principle is<br />

very old; it was demonstrated by Greeks for navigation <strong>and</strong> by astronomers more than 2000<br />

years ago.<br />

The sensors recording the data, images, at first, of course, were analogue, but during the time<br />

they were improving <strong>and</strong> now, information is transformed into digital form.<br />

When previous methods may be used for long distances this one may be used only for short<br />

range. Triangles are basis for many measurements techniques, in laser scanning that principle<br />

was developed in the early 80‟s. The triangle is as follows: laser source, which project a beam<br />

of light on an object; camera – sensor, which collect the position of a beam; <strong>and</strong> object of<br />

interest. Figure 2.11 explains triangulation principle. Knowing position <strong>and</strong> angles between<br />

laser source <strong>and</strong> camera it is easy to compute 3D point on the object. This method is much<br />

more precise than the others. This type is mostly used for acquiring high resolution <strong>and</strong><br />

detailed data when it comes to small objects. For cultural heritage <strong>and</strong> preservation these<br />

scanners are common to use. Laser scanning is good non-contact way to produce 3D complex<br />

models of sculptures <strong>and</strong> other object for documentation. This method may be used for other<br />

purpose as well. Distance range, quality <strong>and</strong> time rate may differ very much in these scanners.<br />

There are many different types upon this principle <strong>and</strong> a lot of companies make such<br />

scanners.<br />

Material which may be scanned with current technology include clay, was, rubber, plastic,<br />

bone, wood, ferrous metal, nonferrous metal, glass, stone, <strong>and</strong> ceramics, but not every type is<br />

capable to scan those surfaces.<br />

Figure 2.11 Basic triangulation principle applied in scanning<br />

(http://www.impactstudiostv.com/Studios)<br />

23


THEORETICAL BACKGROUND<br />

Single point scanners<br />

The scanner uses only one single ray beam projection <strong>and</strong> Camera sensor receives only one<br />

point at the time. The point is acquired on the intersection ray beam with an object that the ray<br />

meets. The Imaging lens may differ, for example there may be only one lens; or one lens with<br />

dual aperture mask( inserted next to the diaphragm of the lens creates two spots on CCD<br />

detector <strong>and</strong> a unique peak-position/ separation relationship); or two lenses on both sides of<br />

the project. These additional components give better accuracy.<br />

Slit scanners<br />

The principle is quite similar to single point scanners, though the projector generates laser<br />

plane instead of only one ray. This method gives more information at the time where whole<br />

profile may be received from one scan. The main inconvenience of these scanners is the<br />

compromise between the field of view <strong>and</strong> depth resolution. The other disadvantage is their<br />

relatively poor immunity to ambient light. Some improvements have been done <strong>and</strong> indoor<br />

acquisition is relatively good, but outdoors there might occur some problems.<br />

Pattern projection scanners; structured-light<br />

The projector generates a light pattern (no laser): moiré patterns, circular or line profiles<br />

(horizontal or vertical). “The most popular method use binary coded orphases shift fringe<br />

patterns. Grey-code binary images use multiple frames with increased resolution to encode a<br />

pixel on the CCD with its corresponding range. Subpixel resolution is obtained by detecting<br />

the edge transitions in the highest resolution image. Other methods use sinusoidal phase<br />

encoding the measure range.” (Blais, 2004) In practise, the more frames are used the better<br />

accuracy. These scanners are popular as well, because of low-cost projectors <strong>and</strong> full 3D<br />

volume can be acquired quickly in just few video frames. Disadvantage in comparison to strip<br />

scanners is smaller depth of view. Also, there might occur defocus (when the e.g. image is out<br />

of focus) “of the projected pattern due to fact that a larger projector lens is needed to collect<br />

as much light as possible from the light course.“ (Blais, 2004). Figure 2.12 below illustrates<br />

Pattern projection principle.<br />

Figure 2.12 System setup with one bar pattern <strong>and</strong> one camera; example of structured-light<br />

scanning (http://www.rob.cs.tu-bs.de/en/research/projects/shape/)<br />

24


THEORETICAL BACKGROUND<br />

H<strong>and</strong>-held scanners<br />

There are also scanners available on the market which are small, h<strong>and</strong>–held. They are usually<br />

used for very small objects <strong>and</strong> the most interesting thing is that they do not need marking any<br />

control points. Scanning is done by capturing frames <strong>and</strong> each frame is oriented (referenced)<br />

to the previous automatically (if the scanner is not moving too fast). Here for instance, in<br />

comparison to the scanners, one scan means several frames referenced to each other, every<br />

frame is a point cloud <strong>and</strong> set of this frames creates one big point cloud.<br />

2.4 Three-dimensional digital modelling <strong>and</strong> visualisation<br />

In computer graphics three-dimensional, digital modelling is the process of developing a<br />

mathematical representation of any surface or object via specialized software. The model may<br />

be represented in many different ways: 2D image, animation, short walk-through video <strong>and</strong> so<br />

on.<br />

In the project a 3D model is derived from point cloud (from scanning or photogrammetry) <strong>and</strong><br />

how to get this point cloud is described before. In modelling from these 3D points there is a<br />

need to filter the point cloud. All data are weighted with some kind of error <strong>and</strong> there is<br />

always some kind of noise that is unwelcome in modelling. Therefore, these bad defined<br />

points should be removed, smoothed, or by special filters the position of points may be<br />

changed by applying a specific method. Also, at the beginning the data contain a lot of points<br />

<strong>and</strong> they might be reduced as well, to improve work (it is meaningful especially in time<br />

needed for processing).<br />

Some objects are very complicated <strong>and</strong> it is very hard to cover the entire object with points.<br />

There, the software encloses a tool – “Filling holes”, where all these holes may be filled. The<br />

method may differ in every software.<br />

All these filters may be used on points; they might be used on triangulated mesh as well. It is<br />

easier to work <strong>and</strong> reduce noise <strong>and</strong> data before, using points. To get a surface from points,<br />

triangulation / wrapping / meshing tool should be employed. It is possible to use only points,<br />

but if some curves may be specified, they might be used as break lines. Not every software<br />

gives this opportunity <strong>and</strong> mostly is allows making Triangulated Irregular Network (TIN) or a<br />

skeleton (defined by lines, arcs, etc.). On created mesh it is usually possible to use similar<br />

filters as on points, what was mentioned before.<br />

Very important part for mesh is that the reparation tool should be run. It checks topology,<br />

errors of the surface, like if there are any intersections, etc.<br />

When the model of the surface is ready, model of the light <strong>and</strong> reflection may be created. It is<br />

more visual side. At this point, light <strong>and</strong> type source may be chosen. The user may set a<br />

position of a light, but also a material of the model <strong>and</strong> how much is it reflecting the light.<br />

It is also possible to put texture on the model. It may be default, only one colour, or from<br />

photographs. If photos are referenced to the model it is very easy to put the real texture on the<br />

model. It gives the model more realistic outlook.<br />

The last part is to make files, or hard copies to show the final model. It may be 2D image<br />

from one of the view, but it may be also 3D file. There is a set of different 3D files that may<br />

be exporter <strong>and</strong> viewed, one of them is 3D PDF, GIF or VRML. 3D PDF gives quite a lot of<br />

25


THEORETICAL BACKGROUND<br />

opportunities, because the model may be rotated, but also, the light source <strong>and</strong> material may<br />

be changed as well, <strong>and</strong> the file may be open in Adobe Reader. Other opportunity is creation<br />

of video file, key frames may be chosen by the user, <strong>and</strong> short video may be exported.<br />

Simplified workflow, with listed operations, in modelling <strong>and</strong> visualising from 3D point cloud<br />

to final product is shown in Table 2.1.<br />

Table 2.1 Simplified workflow for modelling: from point cloud to final model<br />

One regular 3D point cloud<br />

Filtering of point cloud Removing noise, occluded point<br />

Smoothing algorithms<br />

Some manual edition if needed<br />

Filling holes (if exists)<br />

Reducing number of points<br />

Triangulation Generation of TIN, 3D surface<br />

Filtering of TIN (if needed) Filling holes (if exists)<br />

Smoothing algorithms<br />

Removing spikes<br />

Reparation of surface (topology, intersections, etc.)<br />

Reducing number of triangles<br />

Model of light <strong>and</strong> reflection (visualisation)<br />

Light source<br />

Type of light<br />

Place of light<br />

Illumination<br />

Reflection<br />

Texture mapping Real texture from photographs<br />

Or unreal texture<br />

Some additional visual, display settings Depending on the software<br />

Export output 2D images<br />

3D files, for example:<br />

3D PDF, VRML, etc.<br />

Video files<br />

2.4.1 Solid triangulated surfaces<br />

Collection of 3D points, which create vertices that are connected with each other <strong>and</strong> creates<br />

triangles (TIN). This type usually defines the volume of the object they represent, like a rock<br />

etc. From this collection 3D models may be created usually in software using proper<br />

algorithms. This method is mainly used in connection to data derived from 3D scanning or<br />

photogrammetry, where point clouds can be gained. This method is used for shapes, forms,<br />

where exact shape of the object is hard to define, like rectangles, shapes which are easy to<br />

define may help to improve the model. Curves easy to define may be used as break lines <strong>and</strong><br />

have impact on created surface.<br />

26


2.4.2 CAD models<br />

THEORETICAL BACKGROUND<br />

These models may be also at first derived from 3D points, but the principle of this model is<br />

that the skeleton of the object creates model. Surface is represented as the boundary of the<br />

object, not its volume. These are easier to work than solid models. Shape of the object may be<br />

defined by straight lines, circles, ellipses etc. This method may be employed when the shape/<br />

boundary is easy to define, e.g. buildings.<br />

2.4.3 Visualisation<br />

After obtaining a 3D model, next step is to visualize the model. Every model may be<br />

displayed differently. The purpose of the model specifies the final look.<br />

� Texturing<br />

To gain a photo-realistic model, the texture from photographs should be used. These photos<br />

may be from scanning (if scanner acquires colour photographs) or from photogrammetry.<br />

There may be used different texture as well, which will depend on the operator who will<br />

choose the proper settings.<br />

� Rendering<br />

It is usually the last step. It is what the final scene will look like. It includes where the camera<br />

will be placed, where to put the light source <strong>and</strong> what kind of light – this will affect shadows,<br />

reflections, transparency <strong>and</strong> other special effects, like adding e.g. background features (grass<br />

etc.)<br />

The very last part is representation of the model – exporting of final method files like 2D<br />

images, animation, etc.<br />

27


3 INSTRUMENTS AND SOFTWARE<br />

3.1 Computer - workstation<br />

INSTRUMENTS AND SOFTWARE<br />

Computer used for the project purpose was very good <strong>and</strong> fast. The main specifications are:<br />

� Microsoft Windows 7, Service Pack 1, based on x64-bits system type,<br />

� Processor: Intel Xeon CPU X5650 2,67 GHZ, 2661 MHz, 6 cores,<br />

� 24 GB RAM physical memory,<br />

� Graphic card, display: NVIDIA Quadro FX 4800, 1,50 GB RAM, resolution:<br />

1920x1080<br />

� 1,82 TB hard disc drive<br />

From the description it is possible to see, that the PC features are good enough for used<br />

methods in the project. It is at least enough for running the software that the companies<br />

require to have.<br />

3.2 <strong>Photogrammetry</strong><br />

3.2.1 Camera: Canon EOS-1Ds (Body)<br />

The camera used in the project is professional digital SLR camera. CMOS 11 megapixels full<br />

35mm frame sensor with 8,4µm x 8,4µm pitch size. The camera covers almost every other<br />

type of photography from l<strong>and</strong>scape to portrait; photo journalism etc. The base of camera<br />

contains the large battery pack <strong>and</strong> allows for the integration of a vertical h<strong>and</strong> grip <strong>and</strong><br />

control system. EOS-1Ds sensor because of full frame, there is no crop in the field of view<br />

(without focal length multiplier). The sensor has slightly larger pixel pitch, than others<br />

cameras, because of its larger effective imaging area. This, in theory will lead to more<br />

sensitivity, dynamic range <strong>and</strong> lower noise. More specification may be found in Appendix A.<br />

1 <strong>and</strong> visual appearance of the camera in Figure 3.1.<br />

Figure 3.1 Camera Canon EOS-1Ds (Canon, 2011)<br />

28


3.2.2 Lens: Canon 28mm f/2.8<br />

INSTRUMENTS AND SOFTWARE<br />

The used lens is light in weight <strong>and</strong> quite small. Minimum Focus Distance is close – 305mm.<br />

Lens has a 5-blade aperture. Angle of horizontal view is 65 degrees <strong>and</strong> vertical view is 46<br />

degrees. Specification may be found in Table 3.1 <strong>and</strong> visual appearance of the lens in Figure<br />

3.2. Also, more information about the Canon camera <strong>and</strong> lens can be found on the webpage<br />

http://www.canon.com/ (Canon, 2011)<br />

Table 3.1 Specification of Canon lens 28mm f/2.8 (Canon, 2011)<br />

3.2.3 Topcon ImageMaster Software<br />

Figure 3.2 Lens Canon 28mm f/2.8 (Canon, 2011)<br />

The software is built on the PI-3000 photogrammetry software. The version used in the<br />

project was 2.0 <strong>and</strong> it trial version, available only for 30-days. The software allows accurate<br />

3D models from stereo photographic image taken with a st<strong>and</strong>ard digital camera.<br />

ImageMaster is able to create automated 3D surface models, Ortho Images <strong>and</strong> xyz point<br />

clouds – the latter at a fraction of the cost of laser scanner. It allows measurements, TIN,<br />

contour lines or cross section creation; calculations of the models, drawing plotting or 3D<br />

model display <strong>and</strong> editing <strong>and</strong> so on. Using coded targets or at least circular it is possible to<br />

measure their centre automatically with sub-pixel accuracy. Similar, the software is also<br />

available to recognize corners. The company also provides camera calibration – ImageMaster<br />

Camera Calibration Software with their own calibration chart. More information about the<br />

Software <strong>and</strong> other instruments that the company offer can be found on webpages:<br />

http://www.topcon-positioning.eu/ (Topcon, 2011), http://imagemaster3d.com/ (Topcon,<br />

2008), http://www.terrageomatics.com/ (TerraDat et al.), <strong>and</strong><br />

http://www.topconpositioning.com/ (Topcon, 2009)<br />

29


3.2.4 PhotoModeler Scanner Software<br />

INSTRUMENTS AND SOFTWARE<br />

The software provides the tools to create accurate, high quality 3D models <strong>and</strong> measurements<br />

from photographs. This process is called photo-based 3D scanning. Results are similar as<br />

from laser scanner. This scanning produces a dense point cloud from photographs of textured<br />

surfaces. The software includes also meshing, filtering <strong>and</strong> clean up tools. Matching may be<br />

done by using coded targets, or not coded or without any targets – then the SmartMatch<br />

function may be used by the software. It is also possible to measure circular target with subpixel<br />

precision, based on advanced computer vision techniques. This software provides also<br />

camera calibration with its own calibration chart. More information about the software can be<br />

found on the webpage: http://www.photomodeler.com/ (PhotoModeler, 2011)<br />

3.2.5 Autodesk Project Photofly - Photo Scene Editor 2.0 Software<br />

Autodesk‟s software, which is available for free to download from the Autodesk‟s website. It<br />

is open source software available for everyone. It allows anyone with a digital camera to<br />

create 3D models from photographs thanks to advanced computer vision technologies<br />

developed by Autodesk. Project Photofly automatically converts photographs into “Photo<br />

Scenes”, which includes photorealistic 3D model. Uploaded Images to a server are being<br />

processed <strong>and</strong> ready Photo Scene file is sent back once the computation is over. While<br />

waiting for computations there is no need to have open window – it is possible to ask for an email<br />

when it will be finished. When the Images are stitched properly, a draft mesh is<br />

computed automatically always as a result. However, we may refine the mesh, change quality<br />

for Mobile, St<strong>and</strong>ard <strong>and</strong> Maximum:<br />

� Mobile – fast, medium resolution mesh, suitable for viewing on mobile devices;<br />

� St<strong>and</strong>ard – high resolution textured mesh, best for visualization on the desktop, or<br />

� Maximum – very high density mesh, suitable for manipulating in external<br />

applications.<br />

The software also provides simple cleaning the 3D mesh, defining referenced points <strong>and</strong><br />

distances or defining coordinate system. It is also possible to make some measurements on a<br />

model, create lines <strong>and</strong> polylines.<br />

The software allows creating an animation <strong>and</strong> a movie <strong>and</strong> exporting to e.g. YouTube.<br />

Photo Scene Editor is very easy to use <strong>and</strong> anyone may use, because of free availability.<br />

There is also no need to have a very specific camera; it is commercial software for everyone.<br />

The Project Photofly 2.0 technology preview will expire on December 31,2012. More<br />

information about the software, <strong>and</strong> the software may be downloaded from the webpage:<br />

http://labs.autodesk.com/utilities/photo_scene_editor/ (Autodesk, 2011)<br />

Appendix A. 3 represents software comparison on many different fields <strong>and</strong> introduces to<br />

main features <strong>and</strong> used tools in the project.<br />

30


3.3 Three-dimensional digitizing<br />

3.3.1 Konica Minolta VI-910 laser scanner<br />

INSTRUMENTS AND SOFTWARE<br />

The laser scanner is based on triangulation, slit beam principle. Figure 3.3 represents this<br />

scanner. The laser beam is scanned using a high – precision galvanometric rotating mirror,<br />

<strong>and</strong> 640x480 individual points can be measured per scan. The VI910 is provided with three<br />

interchangeable lenses that can accommodate measurement objects of various sizes <strong>and</strong><br />

distances from the lens. The scanner incorporates the same automatic focus technology used<br />

in modern cameras. The CDD <strong>and</strong> RGB filter acquire rich, 24-bit full-colour images. Using<br />

these pictures it is able to get true-colour models. Complete fine scan (unsurpassed speed)<br />

takes 2,5 seconds <strong>and</strong> measures 307 200 points. 3D digitizer may be used without a host<br />

computer <strong>and</strong> recorded data are being saved onto Compact Flash memory card. Scanners<br />

integral LCD viewfinder can be used to set camera parameters <strong>and</strong> as view-finder to frame the<br />

shot or review the data. Thanks to these features it may be used wherever the subject is<br />

located. With dynamic range magnification mode very contrast objects are not a problem<br />

anymore. Fine mode gives precision of +- 0,008mm <strong>and</strong> accuracy of +- 0,10mm can be<br />

achieved on the Z – axis; X +-0,22mm <strong>and</strong> Y +- 0,16mm (with Fine mode <strong>and</strong> using Tele<br />

lens). The scanner is able to measure at distance 0,6 – 2,5m. However optimal 3D<br />

measurement range is 0,6-1,2m.<br />

Konica Minolta uses Editing Polygon Tool software (PET) for data acquisition <strong>and</strong><br />

processing. It controls scanner <strong>and</strong> it allows easily polygonize, edit, <strong>and</strong> convert the scanned<br />

data into any of several common formats. Multiple scans can be easily registered <strong>and</strong> merged<br />

into single model. Editing functions are available like: fill holes, filter irregular polygons <strong>and</strong><br />

noise, <strong>and</strong> perform smoothing.<br />

More information about the scanner system <strong>and</strong> software may be found on the webpage:<br />

http://www.konicaminolta.eu/ (Minolta <strong>and</strong> GmbH, 2011)<br />

Figure 3.3 Konica Minolta VI-910 laser scanner (Minolta <strong>and</strong> GmbH, 2011)<br />

31


INSTRUMENTS AND SOFTWARE<br />

3.3.2 Artec MHT three-dimensional scanner<br />

The scanner is based on light pattern projection, triangulation principle. This bulb flashes light<br />

pattern is projected onto the object <strong>and</strong> the camera record this pattern. Figure 3.4 represents<br />

MTH scanner. It is possible to get up to 15 frames per second, so moving around the object<br />

may be quite quickly with still good result. The scanner has also ability to capture the texture.<br />

3D resolution is 0,5mm with 3D point accuracy up to 0,1mm <strong>and</strong> 3D accuracy over distance<br />

up to 0,15% over 100cm. Working distance for this scanner is 0,4-1m <strong>and</strong> data acquisition<br />

speed up to 500000 points per second.<br />

Artec Scanner software is used for data acquisition <strong>and</strong> processing. It is possible to align<br />

scans <strong>and</strong> register into one coordinate system. After than Fusion all scans together is needed<br />

<strong>and</strong> optimizing the grid with fill holes <strong>and</strong> smooth tools may be applied. It is also possible to<br />

texture the model from photos taken by the scanner.<br />

Figure 3.4 Artec MHT scanner (Artec, 2011)<br />

Appendix A. 4 represents comparison of 3D digitizers <strong>and</strong> contains more details on their<br />

specification. There is also mentioned Artec S scanner, because this instrument has the best<br />

accuracy from all Artec scanners. There was a plan to test this one as well, but finally it was<br />

not able to borrow this one. Appendix A. 5 illustrates comparison on software that mentioned<br />

scanner use for data acquisition <strong>and</strong> processing.<br />

More interesting information can be found on webpages: http://www.artec3d.com/ (Artec,<br />

2011) <strong>and</strong> http://www.exact3dscanner.com/ (Metrology)<br />

3.4 Geomagic Software; post-processing, modelling <strong>and</strong> visualisation<br />

Geomagic is a leading provider of 3D software for creating digital models of physical objects.<br />

The company got numerous awards for technologies <strong>and</strong> owns six patents. There are three<br />

editions: Studio, Qualify <strong>and</strong> Wrap.:<br />

� Geomagic Studio is mainly used in reverse engineering, product design, rapid<br />

prototyping <strong>and</strong> analysis. It is created for transforming 3D scan data <strong>and</strong> polygon<br />

meshed into accurate, surfaced 3D digital models.<br />

32


INSTRUMENTS AND SOFTWARE<br />

� Geomagic Qualify is made for accurate, graphical comparisons between digital<br />

referenced models, <strong>and</strong> as-built parts for first-article inspection, production inspection<br />

<strong>and</strong> supplier quality management. Qualify is powerful report designer, <strong>and</strong> also the<br />

first inspection software that allows creating fully interactive 3D documents that can<br />

be viewed by Adobe Reader.<br />

� Geomagic Wrap enables the transformation of point cloud data to 3D polygon meshes<br />

for use in manufacturing, design <strong>and</strong> analysis.<br />

Mostly all of them are able to do/make the same things, there are just some differences that<br />

one is able to do something more than another <strong>and</strong> vice versa.<br />

Geomagic Qualify was used in the project for processing point clouds. It was tested in<br />

aligning <strong>and</strong> registration of point clouds into one coordinate system, removing noise, filters<br />

<strong>and</strong> mesh creation, also with some smoothing filters.<br />

Geomagic supports all 3D digitizers, cameras <strong>and</strong> scanners in XYZ/ASCII formats <strong>and</strong><br />

h<strong>and</strong>les ordered <strong>and</strong> unordered surface <strong>and</strong> volume data.<br />

Appendix A. 6 shows all main features of Geomagic Qualify, system requirements, main<br />

features <strong>and</strong> manipulation tools used in the project.<br />

More about the software capabilities may be also found on the webpage:<br />

http://www.geomagic.com/ (Geomagic, 2011)<br />

33


4 EXPERIMENT<br />

4.1 <strong>Photogrammetry</strong><br />

4.1.1 Image acquisition<br />

EXPERIMENT<br />

Image acquisition was taken few times, for tests. The sculpture is made of quite shiny material<br />

<strong>and</strong> therefore to get better results different settings were used. However, some of the settings<br />

were fixed for every acquisition.<br />

Fixed settings:<br />

� ISO = 100<br />

ISO system is in other words sensitivity to light of the sensor, also Film speed. Lower speed<br />

index requires more exposure to light to produce the same image density as a more sensitive<br />

film. In photography, the reduction of exposure corresponding to use of higher sensitivities<br />

generally leads to reduced image quality (like higher image noise, etc.) The higher film speed,<br />

the more grain on the image will occur.<br />

To achieve the best image quality the lowest film speed was used in the project.<br />

� Aperture = 22<br />

It is a hole or opening through which light travels. More specifically, the aperture is the<br />

opening that determines the cone angle of a bundle of rays that come to a focus in the image<br />

plane. The aperture determines how collimated <strong>and</strong> admitted rays are, which is of great<br />

importance at the image plane. If an aperture is narrow, then highly collimated rays are<br />

admitted, resulting in a sharp focus at the image plane. If an aperture is wide, then sharp focus<br />

is only for rays with a certain focal length. It also means, that sharp image will be around<br />

what the lens is focusing on, the rest will be blurred. Focusing principle was also described<br />

before in section 2.2.3. Moreover, the aperture determines how many of the incoming rays are<br />

actually admitted <strong>and</strong> this how much light reaches the image plane (the narrower the aperture<br />

-> the darker the image for a given exposure time).<br />

In order to get as sharp image as possible, the aperture was the narrower as possible. This<br />

setting changes highly lens distortion parameters <strong>and</strong> thus it have to be unchanged during the<br />

acquisition.<br />

� Focusing = ca. 0,5m<br />

The distance where the object is will be focused; the image will be sharp at this distance from<br />

the camera to the object. This distance was set in a way, to get sharp object – the sculpture<br />

<strong>and</strong> cover most of the image with the object. This setting also changes highly lens distortion<br />

parameters <strong>and</strong> it was fixed for the whole image recording<br />

� Format picture = 4064x2704pixels,<br />

This is the maximum resolution that is possible to choose, for the type of used camera. It<br />

depends on used sensor.<br />

� Image quality: JPG(the highest quality) + RAW (TIFF)<br />

The camera gives an opportunity to acquire in JPG, but also in RAW file. RAW file gives<br />

more opportunities with processing the image. It contains more information <strong>and</strong> uses no<br />

34


EXPERIMENT<br />

compression. RAW file contains minimally processed data from the image sensor <strong>and</strong> they<br />

have to be processed, or convert to print or edit them. Disadvantage is physical size of the<br />

image, which is a few times bigger than JPG.<br />

These were the most important settings, which stayed fixed <strong>and</strong> the same camera calibration<br />

was used for every picture. This was relevant especially for availability of comparison the<br />

results.<br />

Other camera settings which were slightly changing during the acquisition<br />

� Exposure time<br />

Shutter speed,, it is the effective length of time a camera‟s shutter is open. The total exposure<br />

is proportional to this exposure time, or duration of light reaching the image sensor.<br />

This setting was changing, because of different lighting conditions. The image acquisition<br />

was always inside the building <strong>and</strong> rather in dark places <strong>and</strong> this time was from ca. 1.5sec. to<br />

2.5sec. It is also connected with chosen ISO speed <strong>and</strong> aperture. Lower speed ISO <strong>and</strong> the<br />

narrower aperture needs more time for exposure in order to get better quality image, not too<br />

dark.<br />

� Light<br />

Acquisition took place in the building, always with lights. Firstly only lights in the room were<br />

used. Second – round light source – ring flash, that was fixed on the lens. Third – additional,<br />

separate two moving lights were implemented, in that case sheets were used to cover this light<br />

a bit due to too strong/ direct light<br />

� Additional<br />

Dust spray was used in addition to cover the sculpture <strong>and</strong> remove reflection – used only<br />

during one image recording, because the visual appearance was not very slightly for later<br />

texturing, etc.<br />

� White balance<br />

White balance is the global adjustment of the intensities of the colours (primary colours: R-<br />

red, G – green, B – blue). An important goal of this adjustment is to render these colours, as<br />

neutral colours, naturally. In camera it is possible to choose one of the auto programme;<br />

custom mode, or specify a number. The number here describes the colour temperature of the<br />

light in Kelvin.<br />

The setting was changed when light conditions were changed <strong>and</strong> custom mode was chosen<br />

from the menu. Camera is able to define white balance from a taken picture of white object<br />

(the camera is able to define the temperature from this photo), <strong>and</strong> this kind of picture was<br />

taken.<br />

� Background<br />

In general the background was white wall. In the beginning the background was usually the<br />

floor with the rest of the room, <strong>and</strong> the position of camera was changed. To improve results,<br />

the background was changed. The sculpture lay down on the pile of books <strong>and</strong> stood on the<br />

table. Because the additional lights which stood on the tripod were used, changing position of<br />

them was difficult. Only one wall was available. For complete image recording only the<br />

object of interest was moved – rotated. Also the camera always stood on the tripod in the<br />

same place – only height was changed.<br />

35


EXPERIMENT<br />

� Tripod<br />

For image recording tripod for the camera was always used, to reduce blur of applied long<br />

exposure time. Shutter time was also set for 0,5sec, to avoid camera movements <strong>and</strong> then blur<br />

while touching the camera.<br />

The best pictures were taken, with additional lights <strong>and</strong> without flash <strong>and</strong> these were chosen<br />

for processing. In PhotoModeler Scanner a few photos with dust were used as well, to<br />

compare the results. Figure 4.1 illustrates one of the photographs taken during image<br />

acquisition. It is possible to see that the photo is quite good; but it was not possible to avoid<br />

reflectiveness of shiny material <strong>and</strong> some pixels are “lost” (too bright, sometimes too dark).<br />

Additional control point around <strong>and</strong> on the sculpture were used for better bundle adjustment<br />

results. There was also scale bar to be able to scale the project. After all, a lot of images have<br />

been recorded <strong>and</strong> some of them may be found in folder: Canon\Photos including photographs<br />

for camera calibration.<br />

Figure 4.1 Photograph from image acquisition<br />

36


4.1.2 Topcon ImageMaster Software<br />

EXPERIMENT<br />

Camera calibration<br />

For camera calibration in Topcon the test field from the software was used. It was hanging on<br />

the wall, because of software requirements. Then five photographs were taken from every side<br />

(right, left, up, down <strong>and</strong> front). Afterwards camera parameters were computed automatically.<br />

Marking <strong>and</strong> referencing points is automatic. Figure 4.2 is a snapshot from calibration of the<br />

camera, which was used.<br />

The most important things which need to be remembered during the acquisition are: - first line<br />

of points should be as near as possible to the edge of the picture - <strong>and</strong> the chart should cover<br />

as much as possible the whole picture size. Also, very important for photogrammetry is to set<br />

focus length, principle distance <strong>and</strong> aperture <strong>and</strong> fix it for acquisition, because these<br />

parameters change the parameters of the calibration.<br />

New camera calibration was saved in a new project file in folder Topcon ImageMaster\Project<br />

files\Camera Calibration Canon 28mm<br />

Figure 4.2 Results from Topcon ImageMaster Camera Calibration<br />

<strong>Photogrammetry</strong> project<br />

During the work <strong>and</strong> processing the software was easy to use. The circular targets were easy<br />

to mark with subpixel accuracy – available in the software. Every point had to be clicked. The<br />

software was also able to measure corners with subpixel accuracy, but it was not used in the<br />

project. Next, the bundle block adjustment <strong>and</strong> all the computations worked fine. The<br />

software is able to use DLT equations instead of bundle adjustment, e.g. for an unknown<br />

camera. To coordinate system <strong>and</strong> scale, distances between three control points were used.<br />

These were measured with a scale bar. Afterwards, using stereo pairs the area for DSM<br />

creation was chosen. It is enough to draw the polygon on at least one photo pair. Photo pairs<br />

may be displayed in stereo view, when special polarizing glasses <strong>and</strong> 3D monitor are<br />

available. The polygon is being drawn simultaneously on the other photo pairs. To create<br />

DSM it is possible to choose only one photo pair or all of them <strong>and</strong> one mesh will be created.<br />

37


EXPERIMENT<br />

In the project both ways have been tested. After all, the best combination was to create DSM<br />

using separate pairs. The software may use cross-correlation or least-Square Matching as<br />

matching method. Computations take some time, <strong>and</strong> it needs more time for high resolution<br />

mesh. Also, exporting takes some time. Software does not give so many tools for mesh<br />

editing <strong>and</strong> improving. Some results have been exported <strong>and</strong> imported in Geomagic, It did not<br />

get any better results <strong>and</strong> no more investigations were done.<br />

Steps taken while using the software:<br />

� Import appropriate photos to the project. In the experiment 12 photos were used.<br />

� Orientation, bundle block adjustment. For different arrangement of photos, different<br />

amount of point is required. For this purpose at least 7 corresponding points was<br />

defined on every picture. If the coded or not coded circular points were used it is<br />

possible to find them automatically with subpixel accuracy. For computation, photopairs<br />

have to be defined by the user. The software use bundle adjustment but for some<br />

reasons it needs photo-pairs. Next stage is just to run the process <strong>and</strong> computations are<br />

being done automatically.<br />

� After that, photo-pairs may be displayed in stereo-view. Pictures are being calculated<br />

to remove distortion, change central projection to orthogonal to make possible<br />

displaying stereoscopic view, so it also respects orientation of the images.<br />

� Later on, it is recommended to specify area at least on one stereo pair for TIN<br />

creation– triangulated 3D point cloud.<br />

� Next, running TIN creation – there are some settings that may be changed. It is<br />

possible to create mesh only from one pair or use all of the pairs. It is also possible to<br />

create mesh with checked some filters (mean/ median) as well. Median filtering was<br />

mainly used in the project. For texturing only one photograph may be used, but also all<br />

of them, though one of them should be defined as basic one. When separate photopairs<br />

were used for TIN creation, one of them was set as basic.<br />

� The results may be displayed as 3D point cloud, shaded triangles or textured from<br />

pictures. Figure 4.3 illustrates view of the software with one textured mesh – using<br />

only one photo-pair.<br />

� Final model has not been reached. There were met some problems with TIN creation<br />

<strong>and</strong> merging the separate TINs. Exporting process was not the best solution, <strong>and</strong><br />

therefore, Geomagic was not used for any improvement. The main reason of stopping<br />

the research with that software was because a lot of noise was produced affecting with<br />

wrong surface, mainly because of mismatching in shining areas. The project file may<br />

be found in folder Topcon ImageMaster\Project files\ArchBishop<br />

38


EXPERIMENT<br />

Figure 4.3 Textured mesh from one photo pair in Topcon ImageMaster<br />

4.1.3 PhotoModeler Scanner Software<br />

Camera calibration<br />

For this calibration the software require 12 photos, from each side (left, right, up <strong>and</strong> down)<br />

<strong>and</strong> with rotation of a camera with +90º <strong>and</strong> -90º. In that case it is easier to put test field on<br />

the ground or on the table – depending on used focus, lens <strong>and</strong> object of interest. In the<br />

project the table was used. Test field for calibration was also used the one that software<br />

require. 3D frame was also applied, however it was hard to take good pictures <strong>and</strong> the<br />

software compute parameters without sufficient accuracy <strong>and</strong> therefore this method was<br />

skipped. Similarly to Topcon‟s software, the chart should cover as much as possible the whole<br />

picture size <strong>and</strong> the first line on points should be as near as possible to the edge of the picture.<br />

Afterwards the calibration was computed automatically. Marking <strong>and</strong> referencing points is<br />

automatic. New camera was saved in project file in folder PhotoModeler\Project files\Camera<br />

calibration. Figure 4.4 below is snapshot from the camera calibration project. In the picture<br />

residuals on each point is shown.<br />

39


EXPERIMENT<br />

Figure 4.4 Residuals on control points in camera calibration in PhotoModeler Scanner<br />

<strong>Photogrammetry</strong> project:<br />

To process the photographs there is a need to reference control points between photographs,<br />

to be able to compute orientation parameters. Because circulars points were used around the<br />

object, point marking could be automatic. At least six points between the pictures have to be<br />

referenced manually. Then project may be processed – initial process. After then automatic<br />

reference could be executed <strong>and</strong> all of the points can be used. This is a very helpful tool <strong>and</strong><br />

may improve results, because of using more points (this usually results with better<br />

adjustment). Also, when some points have not been marked for some reasons, they can be<br />

also marked with subpixel accuracy – when circular (only by defining rectangular area of<br />

point location on the photo). It is possible to process only few photographs in the project <strong>and</strong><br />

after that all of them. In the project though, every stipe of photographs were computed<br />

separately. After that, all photos were oriented together. For some reasons, the computations<br />

failed when all the photographs were taken into the account for the first “processing”.<br />

For DSM creation software may use option - “use the best photo pairs”, where residuals,<br />

height to base ratio <strong>and</strong> angle between photographs are being taken into consideration.<br />

Maximum project residual quality – the lower the better accuracy; Base to Height Ratio<br />

should be from 0,1 to 0,5; <strong>and</strong> angle should be less than 30º (photos should not be too<br />

convergent) – these are as recommendations from the software. Tool for selection of the best<br />

photo-pairs is very helpful, because photo-pairs do not have to be specified from the<br />

beginning like in Topcon‟s software, but the best photo-pairs may be chosen automatically.<br />

Then, using selected pairs <strong>and</strong> additional settings like resolution, the point cloud is being<br />

generated. Then, if the results are fine, point clouds may be processed – it means: reduce the<br />

noise <strong>and</strong> triangulate the surface. DSM creation with mesh processing also takes some time,<br />

especially when high resolution is needed. The results may be exported, what was done in the<br />

project, <strong>and</strong> afterwards the files have been imported in Geomagic. However, the software was<br />

not able to make big improvement, probably because of small amount of points, <strong>and</strong> the<br />

surface was already created. There was no more investigation considering improvement in<br />

Geomagic, because the test did not give significant conclusion.<br />

40


EXPERIMENT<br />

In the project there were 12 photos used – 6 pairs with few more pair, that the software<br />

founded good for DSM creation. Also, in comparison to the Topcon‟s software, all the<br />

available control points were used, because this software has an ability to mark <strong>and</strong> reference<br />

them automatically. Project files may be found in folder PhotoModeler\Project files<br />

Steps taken while using the software:<br />

� Import appropriate photos to the project.<br />

� Orientation, bundle block adjustment. For this purpose at least six corresponding<br />

points were measured at every picture. If the coded or not coded circular points are<br />

used it is possible to find them automatically with subpixel accuracy. It is also possible<br />

to reference the points automatically, however only when the pictures are already<br />

“initial” oriented. Next stage is just running process <strong>and</strong> computations are being done.<br />

� After that, the project was idealized – it means that used pictures are being<br />

transformed to ideal form. Idealizing tool uses calibrated camera parameters <strong>and</strong> the<br />

lens distortion is being removed from the images. Also, it re-maps (pixel by pixel)<br />

image by non-centred principal point <strong>and</strong> any non-square pixels. After then the project<br />

was to be reprocessed - recomputed, what is recommended in order to get better –<br />

more accurate results.<br />

� Later on, it is recommended to specify area at least on one picture to create DSM – 3D<br />

point cloud.<br />

� Next, running DSM creation – there are some settings that may be changed. It is also<br />

possible to create mesh in the same time using some filters as well, however mesh <strong>and</strong><br />

filters may be used afterwards. Setting, that may be changed for DSM creations are:<br />

o Sampling rate – 3D points are spaced approximately this distance apart on the<br />

surface; higher number results in more points; higher resolution<br />

o Distance from the main created surface – below or above it<br />

o Sub-pixel matching – recommend, for more accurate surface<br />

o Super-sampling factor, factor that also increase accuracy, but also time for<br />

computation, factor that is used for subpixel method<br />

o Matching region radius – radius that the algorithm is looking for pixels, to<br />

make matching between photo-pairs. Larger is slower with smoother results,<br />

but maybe useful if the texture features are large. Smaller is faster but may be<br />

noisier. Lower number reduces time but may produce more noise<br />

o Texture type – number that specifies if the matching algorithm should respect<br />

texture more or less, the lower number means that source images have a weak<br />

r<strong>and</strong>om non-repeating texture, where the biggest number indicates that images<br />

have a regular repeating texture. Also lower number produces a denser result,<br />

but sometimes with more noise.<br />

� The software uses stereo pairs for point cloud generation – it means “the best photo<br />

pairs” was chosen.<br />

� Based on points, TIN is being created, smoothing, fill holes filters may be used there<br />

as well. Every tool for point cloud modification has been used in the project with<br />

default settings, such as merging, noise reduction, smoothing <strong>and</strong> filling holes.<br />

� The results may be displayed as points, shaded triangles or textured from pictures.<br />

Figure 4.5 below represents snapshot from the software with mesh (mesh illustrates<br />

small area of the sculpture – “hair”). It is visible there are a lot of holes, even though<br />

photographs were taken around <strong>and</strong> cover that area, but there should be miss-match,<br />

because of different colour <strong>and</strong> reflection recorded on the images taken from different<br />

angle.<br />

41


EXPERIMENT<br />

Figure 4.5 Coloured point cloud of small part of the sculpture in PhotoModeler Scanner<br />

4.1.4 Autodesk Project Photofly Photo Scene Editor 2.0 Software<br />

The software is open source - available for free to download from Autodesk website. The<br />

software is able to create mesh from a set of the photographs. Interface <strong>and</strong> use of the<br />

software is very user friendly. There is no need to know basics of photogrammetry. These<br />

features make the software able to be used by everyone.<br />

Figure 4.6 is showing data flow in during the process. If in the sent photo scene, there are<br />

some bad stitched photographs, they may be stitched manually, by measuring at least four<br />

corresponding points on three photographs. In the project 154 photographs were used. In<br />

every photo at least 4 corresponding points were measured, for better alignment.<br />

Photogrammetric process is being done by sending the photos with point to the server <strong>and</strong><br />

ready photo scene is being sent to the user when it is finished. Unfortunately is hard to say<br />

what is the accuracy <strong>and</strong> resolution of the mesh. The mesh was also exported <strong>and</strong> imported in<br />

Geomagic. In the Autodesk‟s software it is possible to see some errors in the surface, <strong>and</strong><br />

therefore the model was improved in Geomagic. Nevertheless, the video file was exported as<br />

well from Autodesk‟s software <strong>and</strong> they may be found with project files in folder<br />

Autodesk<br />

42


EXPERIMENT<br />

Figure 4.6 Data flow using Autodesk Project Photofly software (Autodesk, 2011)<br />

Steps taken while using the software:<br />

� Creating a new Photo Scene by selecting appropriate photos<br />

� Next the photos are being uploaded to the system <strong>and</strong> processed (camera calibration,<br />

orientation, bundle adjustment – or something similar –the software does not say what<br />

kind of equations they use <strong>and</strong> how the 3D point cloud is being obtained)<br />

� Ready Photo Scene – oriented photographs <strong>and</strong> created mesh, is being send back to the<br />

software or to a given e-mail address, where the scene may be downloaded afterwards<br />

� The Photo Scene include automatically created Draft Mesh – low resolution mesh<br />

� Next, stitching of photos needs to be checked – some pictures may be not stitched<br />

correctly or be just not stitched. Stitching means here orientation or referencing<br />

photographs. To do the stitching at least (<strong>and</strong> only) four corresponding points (visible<br />

on at least 2 other pictures) has to be defined. It is also possible to unstitch the photos<br />

– if badly stitched, <strong>and</strong> measure the points. In the project to improve accuracy <strong>and</strong><br />

reduce misalignment on every photograph at least four points has been measured.<br />

Since, there is no information on given accuracy of the computations, to be sure, that<br />

the photos are stitched as best as they could, the corresponding points were measured.<br />

� The Photo Scene then is being recomputed – it is sent again to the server <strong>and</strong> ready<br />

model is being sent back, similarly as before<br />

� It is possible to define scale, by referenced distance<br />

� Also, some lines or points may be drawn on the model – it is also done in 3dimensional<br />

model.<br />

� If the model is satisfying <strong>and</strong> all the pictures are stitched correctly, the mesh may be<br />

recomputed to better accuracy <strong>and</strong> resolution. “Maximum” mesh setting is the best<br />

one. Mesh is also always textured from these photos that are used. Figure 4.7 is a<br />

snapshot from the software <strong>and</strong> project file. In the image, location of cameras (location<br />

of taken photographs with maintained sequence of recording) may be found <strong>and</strong><br />

created mesh itself as well.<br />

� After that the model may be exported to different files <strong>and</strong> it is possible to work on a<br />

model using different software.<br />

� The software also provide animation video creation <strong>and</strong> it allows to upload it to the<br />

YouTube<br />

43


EXPERIMENT<br />

Figure 4.7 Final Maximum mesh in Autodesk Photo Scene Editor 2.0 Project Photo Fly<br />

4.2 Three-dimensional scanning<br />

Data acquisition<br />

Artec scanner is h<strong>and</strong>held light <strong>and</strong> acquisition was done by moving scanner around the<br />

sculpture. In the same time view of scanned data was track, to have a view on what was<br />

scanned <strong>and</strong> what need to be scanned. Use of this scanner is very easy <strong>and</strong> starting/ stopping<br />

scan is done by pushing the button at scanner, or mouse click at software. Konica Minolta<br />

instead is quite heavy <strong>and</strong> bigger, <strong>and</strong> therefore it should st<strong>and</strong> on the tripod. In that case, the<br />

sculpture was moving; scanner was only moved to fix the view, mostly on vertical angle.<br />

Scanning is done by mouse click using PET software.<br />

4.2.1 Konica Minolta VI-910 laser scanner<br />

Laser scanner is quite heavy <strong>and</strong> for recording it stood on the tripod. For data acquisition<br />

TELE lens was used <strong>and</strong> 21 scans has been recorded. As it turned out TELE lens was not the<br />

best choice for this type of sculpture to cover the whole object in one scan – the object has to<br />

be farther from the digitizer, sometimes too far away, that the scanner could not work within<br />

his working distance. Therefore, MIDDLE lens was also used for improvement <strong>and</strong> 15 more<br />

scans were acquired. TELE lens seems to be better when it comes to accuracy, but when the<br />

object has to be far away the accuracy is not necessarily increasing. For the acquisition, the<br />

sculpture stood on the table or on the floor, to get the data from all of the sides. Scans were<br />

align <strong>and</strong> processed in PET software, however in many cases the software did not work,<br />

probably because this software is not meant to be use with Windows 7 64 bits operation<br />

system. Due to this problem, the scans were processed in Geomagic from raw data.<br />

Steps taken while using the scanner <strong>and</strong> software:<br />

� Plug every wire to connect laser scanner to the computer<br />

� Open Polygon Editing Tool Software<br />

� Turn on the scanner<br />

� Set some settings in the software if needed <strong>and</strong> find the scanner<br />

44


EXPERIMENT<br />

� Choose one scan; it is possible to set some setting there as well, but the most<br />

important is to set Fine Scan – to get the best results. Also filters are not required – it<br />

is possible to clean the scan afterwards. It is also possible to take a picture for<br />

texturing.<br />

� Move, rotate the object or the scanner or both to get the best view on the object<br />

� Take the scan – the scanner sets focus automatically.<br />

� Now save the scan <strong>and</strong> open it – clean, process <strong>and</strong> so on; or take another scan to<br />

cover the whole object of the interest. In the project 21 scans were acquired with<br />

TELE lens <strong>and</strong> 15 with MIDDLE lens<br />

� Next it is possible to register the scans into one coordinate system. In the beginning<br />

there is a need to register the scans with Initial Manual Registration tool – there one<br />

scan has to be defined as a base <strong>and</strong> second scan is beginning align into the first one<br />

<strong>and</strong> so on. It is possible to work only on one pair of the scans at the same time. To run<br />

the registration there is a need to define at least three corresponding points. Then the<br />

computational algorithm is being run. In the project, usually four or more points were<br />

used for better approximation.<br />

� After good initial registration it is time for Fine Registration – by “elements”. There<br />

also the base scan has to be chosen – all the scans will be registered into the chosen<br />

scan‟s coordinate system. Here all the scans may be chosen to register them all<br />

together.<br />

� After that the scans may be merged into one point cloud, cleaned, filtered, smoothed,<br />

fill holes <strong>and</strong> so on.<br />

� The Software automatically creates mesh from points, but it may be changed from<br />

view settings<br />

In the project steps until fine registration <strong>and</strong> merging were done, because of met the<br />

problems with the software. Scans were exported <strong>and</strong> imported into Geomagic <strong>and</strong> processed<br />

there afterwards. Also, the time was limited, final model in Geomagic was not achieved,<br />

although the results were quite promising. Figure 4.8 below illustrates the view of the PET<br />

software with merged scans from Konica Minolta scanner.<br />

Figure 4.8 Merged point cloud in Polygon Editing Tool from Konica Minolta<br />

45


EXPERIMENT<br />

4.2.2 Artec MHT three-dimensional scanner<br />

The 3D scanner is very easy to use. It is small <strong>and</strong> h<strong>and</strong>held. Scanning is being done by<br />

clicking the button which is on the scanner or using the software on PC. The object is being<br />

scanned with frames. By moving the scanner it is possible to cover the whole object even with<br />

a one single scan. If the digitizer is not being moved too fast, the frames are referenced<br />

automatically. In some cases, like in places hidden deep inside, when the object is<br />

complicated, there might be need to scan more times <strong>and</strong> then align the scans. In the project<br />

25 scans were acquired, but not all of them were used. For processing the scans Artec<br />

software was used, raw data were also exported <strong>and</strong> imported in Geomagic. Simple text file<br />

format was used for export to Geomagic.<br />

Global registration in Artec software when there are a lot of frames takes a lot of time <strong>and</strong><br />

therefore not all the scans were used. The scanner is able to capture very dens point cloud<br />

with high resolution <strong>and</strong> very good accuracy, but processing all the points is very arduous task<br />

for PC. Raw scan may be found in the folder Artec\Scans <strong>and</strong> final model in Artec\Project<br />

files<br />

Steps taken while using the scanner <strong>and</strong> software:<br />

� Plug every wire to connect the scanner to the computer<br />

� Open Artec Scanner software<br />

� Set some settings in the software if needed <strong>and</strong> find the scanner<br />

� Click Scan <strong>and</strong> start of the scan may be initiated from the software by the “Record” or<br />

by the button in the Scanner. It is also possible to take pictures for texturing<br />

� Scan the object – rotate slowly the scanner around the object of interest<br />

� It is possible to take one scan <strong>and</strong> then there is more frames in one scan; or take more<br />

scans with lower amount of frames in each scan. It general, the author‟s opinion is that<br />

more scans with lower amount of the frames should be taken. Of course, it depends on<br />

the object, situation <strong>and</strong> so on. Nevertheless, processing seems to be better when the<br />

scan contains less than 500-700 frames. Also, there should not be too many scans<br />

neither, some computations take too much time then.<br />

� After acquiring the data it is time for Registration – first Rough, than Fine – for every<br />

scan.<br />

� Next, aligning the scans into one coordinate system. One scan has to be chosen as a<br />

reference – all the scan will be aligned into this scan‟s coordinate system. Only one<br />

pair may be aligned in one time, it is possible to align two scans <strong>and</strong> then another<br />

without quitting running this tool. The software needs at least three corresponding<br />

points for aligning. In the project usually four or more points were measured for better<br />

initial approximation.<br />

� After that the scans has to be register more accurately <strong>and</strong> there is a time for Global<br />

Registration Tool to run – it usually lead to better results, more accurate. The<br />

algorithm is taking every frame into the account <strong>and</strong> recomputes again, also by<br />

analysing relations between the frames.<br />

� Having the scans in one coordinate system the software provides Fusion – merging all<br />

scans together – it usually leads to better results. The result is usually smoother model<br />

<strong>and</strong> possibility of reducing of points in mesh may be met. It is hard to say exactly what<br />

kind of tool Fusion is, <strong>and</strong> what it exactly does for sure. Smoother model looks better<br />

– visually, but it may be not necessarily better, because the accuracy may be lower <strong>and</strong><br />

some details may disappear.<br />

� Next, it is possible to filter the model, fill holes, clean <strong>and</strong> smooth it.<br />

46


EXPERIMENT<br />

� Artec software automatically creates mesh from points; it is possible to change it in<br />

the view settings. Screenshot from the software with smoothed model illustrates<br />

� Figure 4.9 below.<br />

Figure 4.9 Merged <strong>and</strong> smoothed point cloud in Artec Scanner software<br />

4.3 Geomagic Software<br />

The software is very powerful. It may open almost every type of format file from 3D<br />

scanning; it may be also point cloud or polygon or mesh from any other software. The<br />

software has many more tools when it comes to operations on the point cloud or mesh. Here,<br />

again processing huge amount of data like points or triangles needs a lot of time; however the<br />

results are being satisfactory. Scans from three-dimensional scanning were processed from the<br />

beginning using this software. The results from photogrammetry were also imported, but the<br />

only thing which was done is a bit of final mesh processing.<br />

Steps taken while using the software:<br />

� Import the scans<br />

� Aligning the scans (if not aligned). First by defining corresponding points – at least<br />

three – Manual initial registration. Only one scan may be aligned to another at the<br />

same time, however without exiting the tool <strong>and</strong> by choosing “next”, next scans may<br />

be aligned to already aligned ones.<br />

� Next, run Global Registration – accurate registration of all scans into one coordinate<br />

system;<br />

� Then “execute overlap” may be chosen <strong>and</strong> this option usually improves the results. It<br />

is a comm<strong>and</strong> after Global Registration <strong>and</strong> applies a slight bend the objects to move<br />

them into a more perfect registration.<br />

� Then the scans may be merged into one point cloud<br />

� Later on, the scans may be filtered, cleaned <strong>and</strong> so on. Cleaning the scans should be<br />

done before alignment <strong>and</strong> merging – it usually improves the results, by for instance<br />

removing the unnecessary points. Also some noise may be removed before merging;<br />

but this may be done afterwards. In the project both ways were tested <strong>and</strong> some noise<br />

47


EXPERIMENT<br />

was removed before merging but most of occluded points were removed afterwards,<br />

by selecting tool: “outliers” <strong>and</strong> “disconnected components”<br />

� Next, the mesh – TIN may be created. Running this tool it is also possible to use some<br />

filters. Those filters mostly smooth the surface <strong>and</strong> remove some noise. Smoothing is<br />

maybe not the best solution for the purpose of the project, but on the other h<strong>and</strong> model<br />

results with a lot of “bad” triangles which have to be repaired, <strong>and</strong> at the end, the<br />

resulting model is not the best neither.<br />

� The resulting model may be here filtered, cleaned, smoothed, the holes may be filled<br />

in <strong>and</strong> so on. In the project, most of the tools were used. First removing the noise with<br />

the least smoothing type <strong>and</strong> choosing “aggressive-curve” style to preserve curves in<br />

the model. Also, removing spikes smooths the model <strong>and</strong> make it nicer. For every type<br />

of model different amount for reducing should be used. If mesh has a lot of triangles<br />

the number for smoothing may be bigger. Some bad triangles however had to be<br />

removed manually. Holes were mostly filled with “Fill All” function, but the bigger<br />

ones were filled separately, <strong>and</strong> one by one.<br />

� Having a model it is also usually recommended to check the model with Mesh Doctor<br />

<strong>and</strong> repair the triangles. This tool was also used in the project, it is very important to<br />

receive right surface, the tool also may improve the results.<br />

� In the Geomagic Studio or Wrap it is possible to texture the model from pictures,<br />

though in the project, only Qualify was used <strong>and</strong> therefore only the colour of texture<br />

was being changed.<br />

Figure 4.10 below represents screenshot from the software after global registration <strong>and</strong> on the<br />

model, deviations on points are displayed.<br />

Figure 4.10 Registration of point clouds in Geomagic; model of deviations<br />

48


RESULTS, COMPARISSON AND ANALYSIS<br />

5 RESULTS, COMPARISSON AND ANALYSIS<br />

5.1 <strong>Photogrammetry</strong><br />

5.1.1 Camera calibration (PhotoModeler Scanner <strong>and</strong> Topcon ImageMaster)<br />

In order to get good results – accurate, from photogrammetric method, a good camera<br />

calibration is required. Because, commercial camera, non-metric was used in the project,<br />

calibration was done, to get all the necessary parameters of camera interior orientation <strong>and</strong><br />

distortions of the lens. In such cameras PhotoModeler scanner or ImageMaster could be used<br />

for this purpose. The companies give calibration field to have all measurements <strong>and</strong><br />

calculations automatic. It is very convenient <strong>and</strong> even people who do not have any<br />

photogrammetric background may do this. It is also possible to use different test field, but<br />

then some measurements have to be done manually <strong>and</strong> also taking appropriate photographs<br />

may be difficult to achieve required coverage.<br />

Some parameters are possible to get from EXIF information from the image file. EXIF –<br />

Exchangeable image format, is a st<strong>and</strong>ard that specifies the formats for images <strong>and</strong> ancillary<br />

tags used by digital camera. Having this on mind, the softwares mentioned before may treat<br />

some parameters as initial, <strong>and</strong> / or maybe fixed. For instance, ImageMaster may use Image<br />

resolution obtained fully automatic, but it may also be specified by the user by giving pixel<br />

size <strong>and</strong> image sensor size. From this point of view, it may be suspected that even if full<br />

automatic calibration is chosen, these parameters are treated as fixed <strong>and</strong> might be taken from<br />

EXIF information. It may be also suspected from calibration results, where no parameters<br />

considering image sensor size are shown.<br />

When it comes to accuracy, the results - numbers that the software gives us look quite good.<br />

In PhotoModeler last error from the last iteration is 0,847; Overall RMS is 0,810pix with<br />

maximum equal 0,463pix <strong>and</strong> maximum RMS 0,317pix, where the maximum vector is<br />

0,0404mm, what can be acclaimed as a good result of adjustment. ImageMaster gives<br />

st<strong>and</strong>ard deviation, which is equal 0,9842mm <strong>and</strong> maximum residual which is 2,6474mm.<br />

The results are also good, but maximum residual might be reconsidered for the very accurate<br />

measurements. It is quite hard to compare the results, since the softwares give slightly<br />

different deviations or different unit.<br />

When it comes to values of parameters, they are also different. Especially when it comes to<br />

Focal length, <strong>and</strong> difference is about 1,4mm, which is quite a lot for that feature. Also, when<br />

it comes to principal point, in X- direction difference is 0,9mm, but a bit smaller in Z –<br />

direction – 0,5mm. Though, the parameters of camera interior orientation are highly<br />

correlated, <strong>and</strong> when the software uses different image sensor size, for example the focal<br />

length may differ. Also, parameters of lens distortion are correlated <strong>and</strong> both softwares have<br />

slightly different results. Therefore, it is hard to say, which software gives more accurate<br />

parameters.<br />

When it comes to the user, it is easier <strong>and</strong> faster to use ImageMaster, because the software<br />

requires only five photographs for calibration, when PhotoModeler needs twelve. When it<br />

comes to the results <strong>and</strong> accuracy, there might by suspicion that PhotoModeler will result in<br />

better adjustment. Having more photographs taken from different angles may improve<br />

computations, may give better approximation <strong>and</strong> adjustment of the parameters, because more<br />

49


RESULTS, COMPARISSON AND ANALYSIS<br />

observations are given to the mathematical model. Table 5.1 below represents results of<br />

camera calibration from both softwares.<br />

From the user side, both calibration projects are easy, most of processing is being done<br />

automatically when test field from software applied. Figure 5.1 <strong>and</strong> Attachment A. 2<br />

represents in accordance results of camera calibration from Topcon‟s <strong>and</strong> PhotoModeler‟s<br />

software.<br />

Table 5.1 Comparison on results of camera calibration using photogrammetric softwares<br />

Application PhotoModeler Scanner Topcon ImageMaster<br />

Number of photographs used 12 5<br />

Interior orientation parameters:<br />

Focal lengths f 29,361075 mm 27,999864 mm<br />

Principle point Xp 18,020201 mm 17,081758 mm<br />

Principle point Yp 11,789349 mm 11,314753 mm<br />

Lens distortion parameters:<br />

Radial distortion K1 1,419 e-004 1,538941 e-004<br />

Radial distortion K2 -1,709 e-007 -2,057164 e-007<br />

Tangential (decentric) distortion P1 -6,230 e-006 -6,622480 e-006<br />

Tangential (decentric) distortion P2 -1,368 e-005 -0,8847591 e-005<br />

Format width Fw 35,761454 mm No info<br />

Format height Fh 23,799800 mm No info<br />

Pixel size Xr/Yr 8,4 µm 8,4 µm<br />

St<strong>and</strong>ard deviation 0,847 – last error 0,9842 mm<br />

Maximum residuals 0,463 pix (0,0039 mm) 2,6474 mm<br />

Figure 5.1 Results of camera calibration from Topcon ImageMaster<br />

50


RESULTS, COMPARISSON AND ANALYSIS<br />

5.1.2 Photogrammetric project concerning modelling of the sculpture (PhotoModeler<br />

Scanner <strong>and</strong> Topcon ImageMaster)<br />

When it comes to the terminal project, bundle block adjustment of photographs of the<br />

sculpture <strong>and</strong> also creating point clouds from photographs the situations is being changed. Of<br />

course, in both softwares: PhotoModeler Scanner of ImageMaster, camera calibration made<br />

before is being used for the project.<br />

PhotoModeler Scanner is able to mark all the circular points automatically, or coded points<br />

designed by the company. Nevertheless there was a slight problem with automatic referencing<br />

between them. At least six points in each photograph had to be referenced manually at first.<br />

The software needs approximate values to be able reference all the points automatically.<br />

Referencing <strong>and</strong> marking points does not have to be done in respect to photo pairs. Number of<br />

measured point is set automatically <strong>and</strong> link between the points (corresponding on the other<br />

photograph) specifies the correspondence. Points, if the automatic marking is used, are<br />

measured with sub-pixel accuracy. Also, when a point is not marked, it can be marked<br />

manually by the user, by specifying rectangular area where the circular point is. Afterwards,<br />

the project had to be “processed” – it is realising computations for orientation, bundle block<br />

adjustment. After that, having „initial‟ orientation, automatic referencing could be used <strong>and</strong> all<br />

marked points are being referenced.<br />

For some reasons, every stripe had to be processed separately. A stripe here means that few<br />

photos were taken from lower or higher position than others. The software could not manage<br />

to compute all of them together in the beginning. The result was error message <strong>and</strong> software<br />

was not able to compute anything at all, or very high residuals were seen on the other stripe.<br />

But, when every stripe was treated separately – results <strong>and</strong> accuracy was very good <strong>and</strong> after<br />

then, all photographs were processed <strong>and</strong> oriented together <strong>and</strong> final results of computations<br />

were satisfactory. For some reasons, software refuses to compute parameters of orientation on<br />

the beginning, but after “initial” orientation of every stripe, it worked completely fine, even<br />

though nothing was changed in the position of points or links between them. It may be<br />

because the software requires better approximation of initial parameters for computations, but<br />

any error message did not give any clue of that problem. In the project, 120 points were used<br />

in total for bundle adjustment. Having more points may give better accuracy, better<br />

adjustment – more accurate. It may be approved, because overall RMS error is 0,102 pixels,<br />

where the maximum residual is 0,544 pixels. Also, maximum RMS error is 0,255 pixels <strong>and</strong><br />

maximum vector length is 0,271mm, which can be perceived as accurate results.<br />

In ImageMaster photo pairs should be created from the beginning. Marking <strong>and</strong> referencing<br />

control points could be done in a single photo or in photo pair, at this stage it does not have<br />

any difference. It has to be mentioned that the same photos <strong>and</strong> the same amount were used<br />

using both software. Marking could be done together with Referencing, which is done by<br />

specifying a point with the same index number. Circular points or corners may be measured<br />

with sub-pixel accuracy as well, but in Topcon‟s software automatic measurement means only<br />

auto measure but by approximate click at the point. Because every point has to be clicked,<br />

only points on the sculpture <strong>and</strong> in the corners were used for bundle block adjustment, not all<br />

of them, from 7 to 19 tie points were used in one photograph. Computational part went very<br />

well, very small errors <strong>and</strong> residuals <strong>and</strong> all pictures were processed together without any<br />

problems. Maximum Residual, parallel is 0,21pixel 0,26mm in Image Coordinate System.<br />

51


RESULTS, COMPARISSON AND ANALYSIS<br />

Next stage was to idealize project <strong>and</strong> process again after then in PhotoModeler. Idealize<br />

project means to transform images <strong>and</strong> remove distortions caused by the lens also it respects<br />

position of principal point <strong>and</strong> non-squared pixels. This process is recommended to run <strong>and</strong><br />

the project should be “re-processed” in order to get better – more accurate results of<br />

orientation.<br />

In ImageMaster opening stereo pair is stereoscopic view initiate computations which<br />

transform images <strong>and</strong> display them stereoscopically (special equipment needed). It is a bit<br />

similar process that “idealizing” in PhotoModeler Scanner, but here images have to be also<br />

transformed in a way that stereoscopic view is possible to achieve.<br />

Next, the area for DSM creation has to be specified, at least in one photograph<br />

(PhotoModeler), or one photo pair in stereo mode in ImageMaster. It is needed especially to<br />

reduce time for computations, because then the algorithm uses the selected area for image<br />

matching <strong>and</strong> point generation.<br />

After then, the setting for DSM creation needs to be chosen. In ImageMaster there is not so<br />

many available options to choose, only filter (mean or median <strong>and</strong> eventually e.g. breaklines);<br />

how many specified photo-pairs should be used (from defined before), <strong>and</strong> which photographs<br />

should be used as a texture. In PhotoModeler instead every combination of photo pairs could<br />

be used, however not every pair will produce good mesh or any mesh at all. The software<br />

gives opportunity to select the “best pairs” , what is done automatically. Criteria here are:<br />

height-to-base ratio, angle between the photos <strong>and</strong> residuals – quality. There are also<br />

additional settings that can be changed during DSM creation. It may be a distance<br />

above/below the surface, sub-pixel sampling, matching area size <strong>and</strong> texture quality. In<br />

PhotoModeler at the same time the point cloud may be processed, merged into one surface,<br />

triangulate <strong>and</strong> smooth, but it may be done afterwards as well, which is rather recommended.<br />

It is better to choose lower resolution for point generation <strong>and</strong> check if the created point cloud<br />

represents good surface <strong>and</strong> then change the resolution to higher. It is also better to check the<br />

surface before modifying the points, filtering <strong>and</strong> triangulating. It may save time, because<br />

sometimes the chosen settings may give wrong surface <strong>and</strong> therefore there is no sense of<br />

waiting for a very long time for bad result. In Topcon software the result is already<br />

triangulated surface.<br />

In both software creation of dense high resolution DSM needs a lot of time. In both cases<br />

there were some parts where the algorithms created very good surface, but in some areas very<br />

bad. The sculpture is shiny <strong>and</strong> there were always some parts which reflect the light. There<br />

was no possibility to avoid this situation. What is more, the sculpture is very detailed <strong>and</strong> has<br />

places hidden deep inside which are covered by other parts of the object <strong>and</strong> therefore more<br />

photographs should be taken, to cover the entire object. If more photographs, the more time<br />

needed for computations <strong>and</strong> image matching <strong>and</strong> DSM creation. If the object would not be<br />

shiny photogrammetry method gives very good results, very little noise, but in this case it was<br />

very hard to get the whole model <strong>and</strong> model resulted with high amount of noise <strong>and</strong><br />

mismatch. Topcon‟s software was used only in Trial version <strong>and</strong> therefore, the final model<br />

was not achieved, because of time-limitations.<br />

Figure 4.3 in previous section illustrates textured model in ImageMaster using only one<br />

photo-pair. In PhotoModeler final model was not achieved neither, because long time was<br />

needed for point cloud generation <strong>and</strong> model resulting with big amount of noise. Figure 5.2<br />

<strong>and</strong> Figure 5.3 illustrates the result from PhotoModeler Scanner. The first one is a mesh using<br />

52


RESULTS, COMPARISSON AND ANALYSIS<br />

three photographs – four photo-pairs <strong>and</strong> second one is mesh using more photo-pairs but<br />

smaller area. The aim was to test the method if it will produce the good surface at all. These<br />

results are also from photographs, where dust-spray was used <strong>and</strong> there was less reflection. It<br />

is possible to see that in some places there was more noise <strong>and</strong> mismatching. Importing the<br />

results to the Geomagic did not improve the results so much <strong>and</strong> therefore the more results<br />

<strong>and</strong> omitted. Table 5.2 below compares the software – the main steps <strong>and</strong> features in respect<br />

to used object <strong>and</strong> author‟s opinion.<br />

Table 5.2 Comparison of photogrammetric software: ImageMaster, PhotModeler Scanner<br />

Step PhotModeler Scanner Topcon ImageMaster<br />

Camera calibration,<br />

image acquisition,<br />

processing<br />

Bundle block<br />

adjustment, control<br />

points marking,<br />

referencing,<br />

processing<br />

DSM creation,<br />

modification, view<br />

+ automatic marking, referencing <strong>and</strong><br />

computation<br />

- more photographs needed – more time for<br />

acquisition, but may produce more accurate<br />

results<br />

- hard to take picture with required coverage<br />

+ automatic marking all circular or coded<br />

points with sub pixel accuracy<br />

+ automatic referencing (if „initial‟ orientation<br />

firstly computed)<br />

- error during processing entire set of<br />

photographs, the software probably needs<br />

better approximation of values <strong>and</strong> refuses to<br />

process the entire project, or big error is<br />

coming out, sometimes few photographs has to<br />

be processed before processing entire set<br />

+ easy specification of area of interest<br />

+ selecting useful photo pairs<br />

- time needed for generation of point cloud <strong>and</strong><br />

triangulated surface<br />

+ easy modification of DSM <strong>and</strong> triangulation<br />

+ texture available from view settings<br />

- inconvenient view changing: rotation, moving<br />

<strong>and</strong> panning<br />

53<br />

+ automatic marking, referencing<br />

<strong>and</strong> computation<br />

+ only 5 photos needed – faster, but<br />

might produce less accurate results<br />

+ easy image acquisition with<br />

software requirements<br />

- no automatic marking control<br />

points<br />

+ automatic only by click by means<br />

of sub pixel accuracy on circular<br />

point of corner<br />

- no automatic referencing<br />

+ processing entire set of<br />

photographs<br />

- difficult specification of area of<br />

interest – (polygon is created on<br />

photo pair in stereo mode <strong>and</strong> is<br />

visible on every photo pair, when<br />

round object, difficult to specify<br />

proper polygon)<br />

- photo pairs specified from the<br />

beginning (good <strong>and</strong> bad, depends<br />

on purpose)<br />

- time needed for generation of<br />

triangulated surface (<strong>and</strong> only<br />

surface, without point cloud)<br />

- not much tools for DSM<br />

modification<br />

+ texture available from view<br />

settings<br />

+ easy view changing<br />

Exporting + easy <strong>and</strong> fast export, also with texture + easy export, also with texture<br />

- time needed for export


RESULTS, COMPARISSON AND ANALYSIS<br />

Figure 5.2 Textured model from PhotoModeler Scanner; sculpture was sprayed with dust<br />

Figure 5.3 Shaded model from PhotoModeler Scanner, representing “hair”, 1mm resolution<br />

54


RESULTS, COMPARISSON AND ANALYSIS<br />

5.1.3 Autodesk Project Photofly Photo Scene Editor 2.0 Software<br />

This software the author treats separately because of different features of the application. It is<br />

open source, very easy to use <strong>and</strong> does not require any background; the only knowledge<br />

which is needed is introduced by the Autodesk on the website. For commercial use it is very<br />

good software, produces quite good dense mesh without long arduous work in the office.<br />

Only few (at least 4 point corresponding) points are measured in the pictures if needed for<br />

stitching them <strong>and</strong> everything is computed by the server <strong>and</strong> ready surface is being sent back.<br />

The only thing that would be appreciated is the accuracy assessment <strong>and</strong> capability to choose<br />

higher resolution, or at least some information about it. Also, in some cases, some editing of<br />

the mesh could be improved. Probably for most of the users it is enough, for this project it<br />

was not. The mesh which was created looked very good, but there were few places where the<br />

surface was computed wrongly, it may be caused also by the reflection <strong>and</strong> the shape of the<br />

object. After all, having on mind that the software if freeware <strong>and</strong> everyone can use it, the<br />

entire impression is very high <strong>and</strong> for many cases this accuracy <strong>and</strong> resolution is enough.<br />

It is good tool, for modelling <strong>and</strong> also make CAD model – wire frame, because of capability<br />

of drawing lines <strong>and</strong> so on <strong>and</strong> export CAD file. This file includes only lines of points, not<br />

raw point cloud. There is also a very interesting tool available in the software – making video<br />

file <strong>and</strong> put it to the YouTube. The video file was also made during the project <strong>and</strong> may be<br />

found in folder Autodesk\Exports\Video. Figure 5.4 illustrates the model from Autodesk‟s<br />

software <strong>and</strong> Figure 5.5 illustrates the same model, but improved using Geomagic <strong>and</strong> without<br />

the texture.<br />

Figure 5.4 Textured model from Autodesk Project PhotoFly 2.0<br />

55


RESULTS, COMPARISSON AND ANALYSIS<br />

Figure 5.5 Model from Autodesk Project PhotoFly; converted <strong>and</strong> improved in Geomagic<br />

5.2 Three-dimensional scanning<br />

Scanners used in the project are very different. For instance Konica Minolta works with a bit<br />

longer distances. When it comes to light, Konica uses laser – (actually three scans, where<br />

output is actually one scan, which is average of these three scans). Laser – actually horizontal<br />

plane of laser is moving from up to down, by using rotating mirror. Artec instead uses flash<br />

bulb - no laser; a light pattern of horizontal lines. Artec, since it is h<strong>and</strong>held scanner, <strong>and</strong><br />

pattern is very small, scanning is done by acquiring single, small frames (up to 15 per<br />

second), which are being referenced online - through tracking system. One scan is a set of<br />

these frames. Key frames are automatically defined by the software <strong>and</strong> those build a<br />

skeleton. In respect to key frames all of frames in scan are registered into one complete scan.<br />

If some scans are badly referenced they might be moved to a new scan.<br />

In both softwares, that the digitizers use, aligning scans is easy <strong>and</strong> require three<br />

corresponding points to align two scans. Specific control points or spheres are not required.<br />

Measuring -marking corresponding points does not have to be very accurate neither. In both<br />

cases situation of initial alignment is very similar. In PET only two scans can be aligned at<br />

one time, in Artec also, but the aligned scan may be marked as align <strong>and</strong> added to the model<br />

of “aligned” scans <strong>and</strong> aligning next scans is becoming easier, because there is visible bigger<br />

area of object to align.<br />

56


RESULTS, COMPARISSON AND ANALYSIS<br />

Next, when initial aligning is done, fine registration may be computed. In PET it may be done<br />

on two scans, a few or all of the scans by applying “Fine registration”. Situation in Artec is<br />

similar, any amount of scans may be chosen for accurate registration. In Artec it is done by<br />

“global registration”.. There the algorithm recomputes every frame separately; it computes<br />

correlations between every frame. When there is a lot of scans <strong>and</strong> frames it may take a very<br />

long time.<br />

In PET registered scans are being merged <strong>and</strong> then model may be improved by applying some<br />

modifications like filtering, smoothing or filling holes. In Artec, the scans, all frames are<br />

being “fused”. There is visible improvement of model after the fusion – the model is<br />

smoother. After then, some cosmetics may be done in the software. No matter what kind of<br />

scanner of software, it is recommended to remove points from background, before alignment,<br />

because it may produce bad results, due to that algorithms use all points <strong>and</strong> background on<br />

every scan may differ.<br />

PET software is not fully compatible with 64 bit systems <strong>and</strong> Windows 7. Probably that was<br />

the reason of some problems that were met during using the application. The software became<br />

grey in some cases <strong>and</strong> mostly it just refused to work, by showing error messages. Because of<br />

this, model was done only until merging all the scans. The rest has been done using Geomagic<br />

software. Although the final model was not achieved (because of time limitations), it was<br />

looking quite promising.<br />

Artec is able to use full, or most of performance that computer gives. It still needs some time<br />

for computation, if a lot of data. However, the final model was made. One problem was met<br />

there, in some places, where the noise occurred. It was irremovable by any filters, <strong>and</strong><br />

therefore smooth brush has to be implemented. Moreover, the brush does not work good<br />

enough, because it soothed the area around <strong>and</strong> it might smooth some surfaces in unexpected<br />

way. At the end the model looks good, omitting few places. The model is very detailed.<br />

Figure 5.6 illustrates the final model from Artec Software.<br />

When it comes to number of scans which was used in the project, from Konica Minolta 20<br />

scans with TELE lens was used <strong>and</strong> 24 with MIDDLE. In PET all of them were merged. In<br />

Geomagic not every scan was used, because very few of them were not aligned properly.<br />

Using Artec scanner about 17 scans were taken. Each scan contains from 207 to 722 surfaces<br />

(frames) <strong>and</strong> about 1090 to 23014 polygons in one frame. In Artec software only few scans<br />

were used for modelling, because Global Registration for all frames would take too much<br />

time. In Geomagic instead all of them were taken into account, but few of them were skipped,<br />

because of miss-alignment with others.<br />

Accuracy in Artec software is very hard to define. The company gives only resolution which<br />

is up to 0,5mm <strong>and</strong> accuracy, which is 0,2mm. After global registration the Quality which<br />

shows the software was 0,2mm. It is assumed it is accuracy of alignment the scans (frames)<br />

(residuals between scans). Accuracy for Konica Minolta scans in PET software can be given<br />

in registration error. For tele lens average error was 0,527 with sigma 0,428. For middle lens<br />

average error was 0,348 <strong>and</strong> sigma 0,310. It is visible that alignment was slightly better with<br />

middle lens. It is probably connected with the distance to the sculpture. When tele lens was<br />

used, this distance has to be very close the limitation of the scanner. The object was quite far<br />

away; otherwise not entire surface was possible to be seen on the screen <strong>and</strong> scanned.<br />

The model in Artec resulted in 5 943 308 polygons, where in PET model contains 967 131<br />

polygons. Artec has about six times more polygons in mesh.<br />

57


RESULTS, COMPARISSON AND ANALYSIS<br />

Figure 5.6 Model from Artec MHT scanner <strong>and</strong> Software<br />

5.3 Geomagic Software, modelling, post-processing software<br />

In the end, from every softwares, the results – point clouds, surface or raw scans were<br />

imported to Geomagic. The software is able to h<strong>and</strong>le a lot of different file formats. It is<br />

possible to process raw scans from the beginning, align them <strong>and</strong> so one, what was done. The<br />

results of alignment look better. The software also needs a lot of time for some computations<br />

if there is a lot of data, heavy point clouds; however some of them seem to be faster. It is sure<br />

that from a technical point of view, it was better to use Geomagic than PET when it comes to<br />

Konica Minolta software.<br />

58


RESULTS, COMPARISSON AND ANALYSIS<br />

In the end, in author‟s assessment Geomagic produced the best model, the most detailed,<br />

which was the aim of the project. In connection to data from Artec <strong>and</strong> this software results<br />

seem to be the best. It may be compared with model made in Artec scanner as well. Konica<br />

Minolta produced also quite a good model, but fewer points make it more smoothed. Also,<br />

hidden places in the sculpture were very difficult to get in comparison to Artec scanner. With<br />

smaller <strong>and</strong> h<strong>and</strong>held scanner it is easier to get data in hidden areas, also working distance is<br />

important there. Table 5.3 below shows comparison of main features of the softwares from<br />

scanning <strong>and</strong> Geomagic: from data acquisition to final model; in author‟s assessment. Figure<br />

5.7 illustrates 2D image from Geomagic, where data from Artec scanner were used. There is<br />

also Table 5.4 where the results from scanning are compared, processing in Geomagic.<br />

Table 5.3 Comparison on scanning systems: Artec, Konica Minolta, processing in Geomagic<br />

Stage Konica Minolta, PET Artec, software Geomagic<br />

Data<br />

acquisition<br />

+ fast scan<br />

+ not so much noise<br />

- hard to get hidden areas<br />

- distance of working<br />

- heavy <strong>and</strong> big instrument<br />

- TELE lens for better accuracy<br />

can be used only with small<br />

things, due to working distance<br />

<strong>and</strong> size of the view<br />

Alignment + good alignment without special<br />

points<br />

Registration + final registration gives quite<br />

good result<br />

Merging - merging scans may give holes,<br />

even if the data were existing<br />

(probably because of<br />

Surface tools,<br />

editing<br />

misalignment)<br />

+ fast scan with a lot of<br />

data<br />

- losing tracking when<br />

rotated too much or too<br />

fast<br />

- sometimes too much<br />

data, which produce a lot<br />

of noise<br />

+ distance of working<br />

+ light <strong>and</strong> very easy to<br />

use<br />

+ easy to get all<br />

necessary data<br />

- in some cases<br />

misalignment of frames<br />

in one scan<br />

+ easy alignment of<br />

scans, without special<br />

points<br />

+ fine <strong>and</strong> global<br />

registration which gives<br />

good results<br />

+ fusion gives better<br />

model<br />

No info + easy use of filters,<br />

which mostly give good<br />

results <strong>and</strong> model is<br />

improved<br />

- some noise is difficult<br />

to remove, especially if<br />

close to the surface<br />

59<br />

Geomagic is not designed for<br />

any data acquisition, only<br />

processing<br />

+ easy, do not require special<br />

points<br />

+ good results<br />

+ good results, option usually<br />

improve results<br />

+ no reduction/ addition/<br />

edition of data while running<br />

merging tool, all scans/ data<br />

are just merged<br />

+ a lot of tools, filters for<br />

model improvement, which<br />

gives better model<br />

- in some cases difficult to<br />

remove noise without proper<br />

surface


RESULTS, COMPARISSON AND ANALYSIS<br />

Figure 5.7 Model from Artec MHT scanner, but data processed in Geomagic, which is<br />

perceived as the best one.<br />

Table 5.4 Comparison on results of alignment scans in Geomagic software<br />

Konica Minolta Polygon Editing Tool Artec 3D Scanner<br />

Alignment & global registration:<br />

Average distance 0,568 mm 0,240 mm<br />

St<strong>and</strong>ard deviation 0,304 mm 0,210 mm<br />

Maximum deviation 0,688 mm 0,614 mm<br />

Total scan used 19 (tele lens)<br />

23 (middle lens)<br />

17<br />

Number of triangles (final model(s)) 967131 polygons 5943308 polygons<br />

60


6 DISCUSSION AND CONCLUSION<br />

DISCUSSION AND CONCLUSION<br />

The best model was received using the most accurate scanner. Some companies advertise<br />

their projects as the best, but in this case Artec scanner turns out to be the best for the project<br />

purpose. Artec Company also has even a more accurate scanner, than it was used in the<br />

project. It is version S, which is capable to acquire 0,2mm resolution with 0,005mm accuracy,<br />

where distance of working is shorter. The only disadvantage is that it does not acquire texture.<br />

It is assumed that this scanner would give the best results, because of better accuracy <strong>and</strong><br />

resolution. Unfortunately, this scanner was not available for tests.<br />

Artec software is able to create a very good model from the scanned data. Some<br />

computations took a lot of time e.g. Global Registration <strong>and</strong> therefore only the necessary<br />

scans should be used for post-processing.<br />

Geomagic seems to be better software for processing, it allows to have bigger impact <strong>and</strong><br />

control what is actually being done with the data. The user may specify more accurate what<br />

kind of smoothing algorithms he/she needs <strong>and</strong> factor of smoothness she/he wants. It is<br />

possible to reduce the amount of points or triangles even while importing the files. Also, it is<br />

very easy to remove some points – in general. The tool, which was very appreciated <strong>and</strong> used<br />

a lot, was Select- outliers, because it very efficiently reduces noise. The “disconnected<br />

components” tool was also used for reducing unwanted points. It is very difficult to remove<br />

points - noise, which are close to the real surface <strong>and</strong> there the tool “select outliers” worked<br />

very well.<br />

“Noise reduction” tool used on points, gives also a possibility to select outliers <strong>and</strong> remove<br />

them. The algorithms move points to the average position <strong>and</strong> make the surface smoother,<br />

which was also a very good tool <strong>and</strong> used as well. A very smooth surface was not the aim of<br />

the project; however, in order to remove noise the tool was appreciated. While smoothing,<br />

some details may disappear, but when the aggressive-curve tool is applied it is not changing<br />

so much the curves. On the other h<strong>and</strong>, Noise reduction tool has to be used, in order to<br />

remove noise, which also has big impact on the surface <strong>and</strong> created mesh was not<br />

representing the real object. There should be some balance, because without smoothing,<br />

model contains a lot of noise <strong>and</strong> does not render the object in a good way. When the model is<br />

smoothed it looks more like a real object, noise is removed, but some small details may<br />

disappear.<br />

But, Geomagic software does not smooth the surface from the beginning, what Artec probably<br />

does during the Fusion. This especially makes Geomagic better software. Moreover, if more<br />

points used for modelling in Geomagic <strong>and</strong> more triangles are created while “Wrapping”<br />

(meshing, triangulation) the more detailed model will come out. Even if the Noise reduction<br />

tool or smoothing (low smoothing level) is being used, the details still remain visible on the<br />

model. Disadvantages here may be that the software needs more time for same computations<br />

<strong>and</strong> also the project file is getting bigger.<br />

Konica Minolta VI-910 turns to be quite a good laser scanner, however for this purpose it was<br />

very difficult to get the whole data, <strong>and</strong> the model contained a lot of holes. Also, in<br />

comparison to the Artec scanner it has slightly worse accuracy <strong>and</strong> lower resolution, which<br />

was the main determining feature. From Table 5.4 <strong>and</strong> section above, it is possible to see that<br />

Artec resulted with 0,210mm st<strong>and</strong>ard deviation <strong>and</strong> 0,614mm with maximum deviation<br />

61


DISCUSSION AND CONCLUSION<br />

between aligned scans, where Konica Minolta resulted adequately 0,304mm <strong>and</strong> 0,688mm.<br />

The values are not so much different, but bigger difference is in Average distance between<br />

scans which is in Artec 0,240mm <strong>and</strong> 0,568mm in Konica Minolta, which tells us that<br />

alignment here was worse. A comparison was made in Geomagic software. But, these<br />

numbers may not describe that the model will be better or worse or the model will have more<br />

or fewer details. It is some kind of assessment, but cannot be treated as the most important<br />

factor. St<strong>and</strong>ard deviation may be bigger, but visually the scans may be aligned better. Also, if<br />

the scans contain a lot of noise, even if the alignment of them looks very good from the<br />

numbers, the model may not be so accurate.<br />

The author has the impression here that Konica Minolta gives scans with less amount of<br />

noise, but the resolution seems to be worse than using Artec. On the other h<strong>and</strong>, if a lot of<br />

scans would be acquired the resolution will somehow increase while merging all the data <strong>and</strong><br />

the model may contain a lot of details. Still, during data acquisition using Konica Minolta, it<br />

was hard to get all the data – there is a lot of holes, but it may not be the worst problem <strong>and</strong><br />

those holes may be filled afterwards. Konica Minolta might also result in detailed model,<br />

rather slightly worse than from Artec. Yet, because of time limitations the final model was not<br />

achieved.<br />

For the purpose of the project, texture is not necessary. It would be nice to have one, in case<br />

you want to show people a digital model on the screen, to give more natural look. However,<br />

the model is needed to make a real replica, to use it for printing. In that case only 3D<br />

information is needed, <strong>and</strong> therefore texture may be skipped.<br />

<strong>Photogrammetry</strong> method works very well for modelling, when not high resolution is required,<br />

because waiting for point cloud takes a lot of time. Also, if the object is not glass, or any<br />

surface that is shiny <strong>and</strong> reflects the light, the method is proper. In the project the sculpture<br />

was a bit shiny <strong>and</strong> in some cases image matching did not worked well enough <strong>and</strong> software<br />

produced a lot of noise. Photogrammetric softwares still need some improvement when it<br />

comes to time of computation, especially when high resolution required. What is more, the<br />

software should be made in a way that would be able to make use of maximum performance<br />

that the computer gives. Or, at least the software should has an option, where it could be<br />

chosen or specified, what kind of PC is being used <strong>and</strong> how much of processor <strong>and</strong> memory<br />

could be used. Artec software is able to use all of the processor‟s cores, <strong>and</strong> also Geomagic is<br />

able to make use of most of the available performance. The rest applications that have been<br />

tested do not use all the power that PC may give.<br />

Autodesk software is also based on photogrammetry, but works differently. It is open source<br />

<strong>and</strong> it is very easy to use. Everyone is able to use this software, without mostly any<br />

background knowledge. The application produces quite good mesh, with texture <strong>and</strong> in many<br />

cases it is enough. Nevertheless, for the project the model was not too accurate <strong>and</strong> resolution<br />

was too low.<br />

Software for DSM creation or edition, manipulation point clouds still needs some<br />

improvements. Some works better than the others, however there should be options which<br />

could be changed according to computer performance. Everyone may use a different PC <strong>and</strong><br />

in the project the workstation was very good. During the computations the system worked<br />

slowly, or it was possible to discover that not every processor core was working. Not entire<br />

available memory was used either. Still computations took too much time, while the potential<br />

of PC was not used properly. Companies need to improve this field <strong>and</strong> give opportunity to<br />

62


DISCUSSION AND CONCLUSION<br />

change performance settings in respect to the used computer, also in respect to used operation<br />

system, especially x64bits, which gives more capability.<br />

In spite of all the problems which were met, the final model was done, the one that was giving<br />

the best results – from Artec MHT scanner. Artec <strong>and</strong> Geomagic software was used for this<br />

purpose, however in author‟s opinion Geomagic resulted in a better model. Figure 6.1<br />

represents this model. Also, 3D PDF file was also made in Geomagic <strong>and</strong> is possible to find<br />

that file in folder Geomagic\Artec\Exports\Files. The produced file is not as heavy as VRML<br />

file <strong>and</strong> the model may be viewed in 3D mode, rotate, etc. It is also possible to change the<br />

light <strong>and</strong> model display. Also, any other exported 2D images or 3D files <strong>and</strong> more 3D PDF<br />

files may be found on the DVD attached to the report.<br />

At the end of the report, in figure Figure 6.2 <strong>and</strong> Figure 6.3 visual comparison of some<br />

detailed area may be found. The first one is more concentrated on “hair” <strong>and</strong> a face of the<br />

Archbishop. It is easy to notice that the model from Artec gives more details, where Autodesk<br />

– significantly less. Also, probably if the model from Konica Minolta was finished it would<br />

probably look also quite good, but here a lot of small holes are visible. A similar situation<br />

may be found on the second figure, but there the view is more concentrated on more “sharp<br />

edge” of the collar.<br />

63


DISCUSSION AND CONCLUSION<br />

Figure 6.1 Final model from Geomagic, using Artec MHT data<br />

64


DISCUSSION AND CONCLUSION<br />

Figure 6.2 Visual comparison of model – more detailed part of the sculpture – “hair”; top left Geomagic - “Artec”, top right Geomagic -<br />

“Autodesk”, bottom left Geomagic – “Konica Minolta”, bottom right “PhotoModeler Scanner”<br />

65


DISCUSSION AND CONCLUSION<br />

Figure 6.3 Visual comparison of model – more sharp part of the sculpture – “collar”; top left Geomagic - “Konica Minolta”, top right Geomagic -<br />

“Autodesk”, bottom Geomagic – “Artec”<br />

66


REFERENCES<br />

REFERENCES<br />

Akca, D., Gruen, A., Alkis, Z., Demir, N., Breuckmann, B., Erduyan, I., Nadir, E., 2006. 3D MODELING OF<br />

THE WEARY HERAKLES STATUE WITH A CODED STRUCTURED LIGHT SYSTEM, in: Maas, H.-G.,<br />

Schneider, D. (Eds.), ISPRS Commission V Symposium 'Image Engineering <strong>and</strong> Vision Metrology'.<br />

International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Dresden, Germany, pp. 14-19.<br />

Akca, D., Grun, A., Breuckmann, B., Lahanier, C., 2007. HIGH DEFINITION 3D-SCANNING OF ARTS<br />

OBJECTS AND PAINTINGS, in: Gruen, A., Kahmen, H. (Eds.), Optical 3-D Measurement Techniques VIII,<br />

Zurich, Switzerl<strong>and</strong>, pp. 50-59.<br />

Akca, D., Remondino, F., Novák, D., Hanusch, T., Schrotter, G., Gruen, A., Recording <strong>and</strong> modeling of cultural<br />

heritage objects with coded structured light projection systems, pp. 1-7.<br />

(http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.67.5893&rep=rep1&type=pdf)<br />

Akca, D., Remondino, F., Novák, D., Hanusch, T., Schrotter, G., Gruen, A., 2007. Performance Evaluation of a<br />

coded structured light system for cultural heritage applications, in: Beraldin, J.-A., Remondino, F., Shortis, M.R.<br />

(Eds.), Videometrics IX, Proc. of SPIE-IS&T Electronic Imaging, San Jose (California), USA, pp. 1-12.<br />

Alyilmaza, C., Alyilmaza, S., Yakarb, M., 2010. MEASUREMENT OF PETROGLYHPS (ROCK OF ARTS)<br />

OF QOBUSTAN WITH CLOSE RANGE PHOTOGRAMMETRY, in: Mills, J.P., Barber, D.M., Miller, P.E.,<br />

Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image<br />

Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne,<br />

United Kingdom, pp. 29-32.<br />

Atkinson, A.D.J., Blasco, J.J.d.S., Bejarano, J.d.M., 2010. APPLIED 3D PHOTOGRAMMETRIC STUDIES<br />

FOR THE HISTORICAL HERITAGE OF EXTREMADURA (SPAIN), in: Mills, J.P., Barber, D.M., Miller,<br />

P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image<br />

Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne,<br />

United Kingdom, pp. 51-54.<br />

Atkinson, K.B., 1996. Close Range <strong>Photogrammetry</strong> <strong>and</strong> Machine Vision. Whittles Publishing., Caithness,<br />

Scotl<strong>and</strong>, UK.<br />

Bache, S.N., 2003. Scanning av skulptur på Nidarosdomen. Vinn Design, Oslo, Trondheim, pp. 1-4.<br />

Balestrinia, E.F.d., Guerra, F., 2011. NEW INSTRUMENTS FOR SURVEY: ON LINE SOFTWARES FOR 3D<br />

RECONSTRUCTION FROM IMAGES; ISPRS Commission V - WG 4; Poster Session, in: Remondino, F., El-<br />

Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction<br />

<strong>and</strong> Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>,<br />

Trento, Italy, pp. 1-8.<br />

Barazzetti, L., Remondino, F., Scaioni, M., 2010. AUTOMATION IN 3D RECONSTRUCTION: RESULTS<br />

ON DIFFERENT KINDS OF CLOSE-RANGE BLOCKS, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I.<br />

(Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement<br />

Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United<br />

Kingdom, pp. 55-61.<br />

Barba, S., Fiorillo, F., Coder, P.O., D‟Auria, S., Feo, E.D., 2011. AN APPLICATION FOR CULTURAL<br />

HERITAGE IN ERASMUS PLACEMENT. SURVEYS AND 3D CATALOGUING ARCHAEOLOGICAL<br />

FINDS IN MÉRIDA (SPAIN), in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS<br />

International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization of Complex Architectures".<br />

International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy, pp. 1-6.<br />

Barnett, T., Bryan, P., Ch<strong>and</strong>ler, J., Case study: Recording Prehistoric Rock Art; Pre-historic photogrammetry<br />

<strong>and</strong> laser scanning, pp. 1-2. (http://www.terrageomatics.com/pdfs/Rock%20Art%20-PI-3000.pdf)<br />

Blais, F., 2004. Review of 20 Years Of Range Sensor Development. Journal of Electronic Imaging 13(1), 231-<br />

240.<br />

67


REFERENCES<br />

Blake, W.H., 2010. WHAT IS THE FUTURE OF METRIC HERITAGE DOCUMENTATION AND ITS<br />

SKILLS?, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V<br />

Mid-Term Symposium 'Close Range Image Measurement Techniques'. International Society of <strong>Photogrammetry</strong><br />

<strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 98-102.<br />

Bonfanti, C., Chiabr<strong>and</strong>o, F., Spanò, A., 2010. HIGH ACCURACY IMAGES AND RANGE BASED<br />

ACQUIRING FOR ARTISTIC HANDWORKS 3D-MODELS., in: Mills, J.P., Barber, D.M., Miller, P.E.,<br />

Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image<br />

Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne,<br />

United Kingdom, pp. 109-114.<br />

Brutto, M.L., Spera, M.G., 2011. IMAGE-BASED AND RANGE-BASED 3D MODELLING OF<br />

ARCHAEOLOGICAL CULTURAL HERITAGE: THE TELAMON OF THE TEMPLE OF OLYMPIAN ZEUS<br />

IN AGRIGENTO (ITALY), in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International<br />

Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization of Complex Architectures". International<br />

Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, pp. 1-8.<br />

Chmielewski, K., Szulwic, J., Niemetryczne zdjęcia cyfrowe w fotogrametrii bliskiego zasięgu w systemie<br />

Topcon PI-2000, Sesja Jubileuszowa 60-lecia Katedry geodezji Politechniki Gdańskiej, pp. 1-8.<br />

(http://www.terrageomatics.com/pdfs/PI-3000%20Polish.pdf)<br />

Deseilligny, M.P., Clery, I., 2011. APERO, AN OPEN SOURCE BUNDLE ADJUSMENT SOFTWARE FOR<br />

AUTOMATIC CALIBRATION AND ORIENTATION OF SET OF IMAGES., in: Remondino, F., El-Hakim,<br />

S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong><br />

Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento,<br />

Italy, pp. 1-8.<br />

Earl, G., Beale, G., Martinez, K., Pagi, H., 2010. POLYNOMIAL TEXTURE MAPPING AND RELATED<br />

IMAGING TECHNOLOGIES OR THE RECORDING, ANALYSIS AND PRESENTATION OF<br />

ARCHAEOLOGICAL MATERIALS, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings<br />

of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'.<br />

International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 218-<br />

223.<br />

Egels, Y., Kasser, M., 2002. Digital <strong>Photogrammetry</strong>. Taylor & Francis Inc., London, UK, New York, USA;<br />

Canada.<br />

Fassi, F., Achille, C., Fregonese, L., Monti, C., 2010. MULTIPLE DATA SOURCE FOR SURVEY AND<br />

MODELLING OF VERY COMPLEX ARCHITECTURE, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I.<br />

(Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement<br />

Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United<br />

Kingdom, pp. 234-239.<br />

Floth, M., Breuer, M., 2010. AN OPTICAL THREE-DIMENSIONAL MEASURING TECHNIQUE FOR A<br />

DETAILED NON-CONTACT DATA ACQUISITION OF OBJECT SURFACES IN THE FIELDS OF<br />

CULTURAL HERITAGE, ARCHEOLOGY AND THE CARE AND PRESERVATION OF HISTORIC<br />

MONUMENTS, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS<br />

Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International Society of<br />

<strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 240-244.<br />

Fraser, C.S., Hanley, H.B., 2004. DEVELOPMENTS IN CLOSE-RANGE PHOTOGRAMMETRY FOR 3D<br />

MODELLING: THE iWitness EXAMPLE, in: Gruen, A., Murai, S., Fuse, T., Remondino, F. (Eds.),<br />

International Workshop on "Processing <strong>and</strong> Visualization using High-Resolution Images". INTERNATIONAL<br />

ARCHIVES OF PHOTOGRAMMETRY, REMOTE SENSING AND SPATIAL INFORMATION SCIENCES,<br />

Pitsanulok, Thail<strong>and</strong>, pp. 1-4.<br />

Fryer, J., Mitchell, H., Ch<strong>and</strong>ler, J., 2007. Applications of 3D Measurement from Images. Whittles Publishing,<br />

Caithness, Scotl<strong>and</strong>, UK.<br />

68


REFERENCES<br />

Gruen, A., Huang, S.T., 2001. Calibration <strong>and</strong> Orientation of Cameras in Computer Vision. Springer, Verlag,<br />

Berlin, Heidelberg, Germany.<br />

Grussenmeyer, P., L<strong>and</strong>es, T., Voegtle, T., Ringle, K., COMPARISON METHODS OF TERRESTRIAL<br />

LASER SCANNING, PHOTOGRAMMETRY AND TACHEOMETRY DATA FOR RECORDING OF<br />

CULTURAL HERITAGE BUILDINGS. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, pp.<br />

213-218. (http://www.isprs.org/proceedings/XXXVII/congress/5_pdf/38.pdf)<br />

Hanke, K., Grussenmeyer, P., 2002. ARCHITECTURAL PHOTOGRAMMETRY: Basic theory, Procedures,<br />

Tools. ISPRS Commission 5 tutorial, Corfu.<br />

Hess, M., Robson, S., 2010. 3D Colour Imaging For Cultural Heritage Artefacts, in: Mills, J.P., Barber, D.M.,<br />

Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range<br />

Image Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle<br />

upon Tyne, United Kingdom, pp. 288-292.<br />

Jędrzejek, M., Łącka, M., Wykorzystanie wzorców świetlnych i metod fotogrametrycznych do pozycjonowania<br />

ciała ludzkiego (PiCalib), WGGiIŚ. Akademia Górniczo-Hutnicza, Kraków. (http://galaxy.uci.agh.edu.pl/~zfiit/)<br />

Kazanaa, E., 2010. TOPCON‟S PI-3000 IN THE INVENTORY OF DATA RELATING TO SCULPTURAL<br />

AND PORTABLE ARTIFACTS, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of<br />

the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International<br />

Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 322-324.<br />

Kimpton, G.R., Horne, M., Heslop, D., 2010. TERRESTRIAL LASER SCANNING AND 3D IMAGING:<br />

HERITAGE CASE STUDY – THE BLACK GATE, NEWCASTLE UPON TYNE, in: Mills, J.P., Barber, D.M.,<br />

Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range<br />

Image Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle<br />

upon Tyne, United Kingdom, pp. 325-330.<br />

Kochi, N., Ito, T., Noma, T., Otani, H., Nishimura, S., Ito, J., PC-BASED 3D IMAGE MEASURING STATION<br />

WITH DIGITAL CAMERA AN EXAMPLE OF ITS ACTUAL APPLICATION ON A HISTORICAL RUIN.<br />

The International Archives of the <strong>Photogrammetry</strong>, <strong>Remote</strong> <strong>Sensing</strong> <strong>and</strong> Spatial Information Sciences;<br />

Commission V XXXIV, Part 5/W12, 195-199. (http://www.terrageomatics.com/pdfs/PI-<br />

3000%20ISPRS%20Paper%20-%20Historical%20Ruin.pdf)<br />

Kraus, K., 2007. <strong>Photogrammetry</strong>: Geometry from Images <strong>and</strong> Laser Scans, 2 ed. Walter de Gruyter GmbH &<br />

Co. KG., Berlin, Germany.<br />

Letellier, R., Schmid, W., LeBlanc, F., Eppich, R., Chabbi, A., Recording, Documentation, <strong>and</strong> Information<br />

Management for the Conservation of Heritage Places. The Getty Conversation Institute.<br />

(http://www.getty.edu/conservation/publications/pdf_publications/guiding_principles.pdf)<br />

Luhmann, T., Robson, S., Kyle, S., Harley, I., 2006. Close Range <strong>Photogrammetry</strong>: Principles, Techniques <strong>and</strong><br />

Aplications. Whittles Publishing, Caithness, Scotl<strong>and</strong>, UK.<br />

Marshall, G.F., 2004. H<strong>and</strong>book of Optical <strong>and</strong> Laser Scanning. Marcel Dekker, Inc, New York, USA; Basel,<br />

Switzerl<strong>and</strong>.<br />

McCurdy, P.G., Woodward, L.A., Davidson, J.I., Wilson, R.M., Ask, R.E., Manual of <strong>Photogrammetry</strong>. Pitman<br />

Publishing Corporation, New York, Chicago, USA.<br />

Nex, F., Rinaudo, F., 2010. PHOTOGRAMMETRIC AND LIDAR INTEGRATION FOR THE CULTURAL<br />

HERITAGE METRIC SURVEYS, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of<br />

the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International<br />

Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 490-495.<br />

Pawlak, A., Pulchny, K., 2009. Zastosowanie cyfrowych aparatów fotograficznych do fotogrametrycznego<br />

pomiaru wybranych części ciała ludzkiego / The application of digital cameras to photogrammetric measurement<br />

of chosen parts of human body., WGGiIŚ. Akademia Górniczo-Hutnicza, Kraków, p. 133.<br />

69


REFERENCES<br />

Prokos, A., Kalisperakis, I., Karras, G., 2011. AUTOMATIC POINT CLOUD GENERATION AND<br />

REGISTRATION WITH A STEREOVISION SLIT-SCANNER, in: Remondino, F., El-Hakim, S. (Eds.),<br />

Proceedings of the 4th ISPRS International Workshop 3D-ARCH; "3D Virtual Reconstruction <strong>and</strong> Visualization<br />

of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy pp. 1-6.<br />

Schewe, H., Holl, J., Gründig, L., LIMEZ - Photogrammetric Measurement of Railroad Clearance Obstacles. 1-<br />

7. (http://www.technet-gmbh.de/fileadmin/Photogrammetrie/Publikationen/Limez-english.pdf)<br />

Schewe, H., Ifert, F., 2000. SOFT TISSUE ANALYSIS AND CAST MEASUREMENT IN ORTHODONTICS<br />

USING DIGITAL PHOTOGRAMMETRY, in: Chikatsu, H., Heuvel, F.v.d. (Eds.), Invited <strong>and</strong> Presented Papers<br />

of the XIXth Congress; Technical Commission V. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong><br />

<strong>Sensing</strong>, Amsterdam, The Netherl<strong>and</strong>s, pp. 699-706.<br />

Stathopouloua, E.K., Valanisa, A., Lermab, J.L., Georgopoulosa, A., 2011. HIGH AND LOW RESOLUTION<br />

TEXTURED MODELS OF COMPLEX ARCHITECTURAL SURFACES, in: Remondino, F., El-Hakim, S.<br />

(Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong><br />

Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento,<br />

Italy, pp. 1-5.<br />

Tucci, G., Cini, D., Nobile, A., 2011. EFFECTIVE 3D DIGITIZATION OF ARCHAEOLOGICAL<br />

ARTIFACTS FOR INTERACTIVE VIRTUAL MUSEUM, in: Remondino, F., El-Hakim, S. (Eds.), Proceedings<br />

of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization of Complex<br />

Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy, pp. 1-8.<br />

Visnovcova, J., Gruen, A., Zhang, L., 2001. IMAGE-BASED OBJECT RECONSTRUCTION AND<br />

VISUALIZATION FOR INVENTORY OF CULTURAL HERITAGE, Proceedings of EVA 2001 Conference -<br />

Electronic Imaging <strong>and</strong> Visual Arts, Florence, Italy, pp. 118-122.<br />

Vosselman, G., Maas, H.-G., 2010. Airborne <strong>and</strong> Terrestrial Laser Scanning. Whittles Publishing, Caithness,<br />

Scotl<strong>and</strong>, UK.<br />

Wojtas, A.M., 2010. OFF-THE-SHELF CLOSE-RANGE PHOTOGRAMMETRIC SOFTWARE FOR<br />

CULTURAL HERITAGE DOCUMENTATION AT STONEHENGE, in: Mills, J.P., Barber, D.M., Miller, P.E.,<br />

Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image<br />

Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne,<br />

United Kingdom, pp. 603-607.<br />

Artec, G., 2011. Artec 3D Scanners<br />

http://www.artec3d.com/.<br />

Autodesk, 2011. Autodesk Labs Utilities - Exploring new approaches to design technology - Project Photofly 2.0<br />

Photo Scene editor<br />

http://labs.autodesk.com/utilities/photo_scene_editor/.<br />

Canon, 2011. Canon Global<br />

http://www.canon.com/.<br />

Geomagic, 2011. Geomagic reverse engineering <strong>and</strong> 3D inspection software - "The magic of making it simple"<br />

http://www.geomagic.com/.<br />

Leksikon, S.N., Brochmann, O., Tschudi-Madsen, S., Storsletten, O., 2009. Nidarosdomen<br />

http://snl.no/Nidarosdomen.<br />

Leksikon, S.N., Dybdahl, A., Pål Bårdsson – utdypning (NBL-artikkel)<br />

http://snl.no/.nbl_biografi/Pål_Bårdsson/utdypning.<br />

Metrology, E., Artec 3D Scanner from Exact Metrology - Effortless 3D Scanning<br />

http://www.exact3dscanner.com/.<br />

Minolta, K., GmbH, B.S.E., 2011. Konica Minolta Europe - The essentials of imaging<br />

70


http://www.konicaminolta.eu/.<br />

Nidaros, D., Nidarosdomen & Erkebispegården<br />

http://www.nidarosdomen.no/nb-NO/.<br />

REFERENCES<br />

Organisasjon, D.K.K., Martinsen, O., 1996. Archbishop Pål (1333-1346)<br />

http://www.katolsk.no/organisasjon/mn/erkebiskoper/15.<br />

PhotoModeler, 2011. PhotoModeler - Measuring & Modeling the Real World - accurate <strong>and</strong> affordable 3D<br />

photogrammetry measurement <strong>and</strong> scanning<br />

http://www.photomodeler.com/.<br />

Studios, I., 3D Laser Scanning Services, A Division of Impact Studios LLC<br />

http://www.impactstudiostv.com/.<br />

TerraDat, TerraGeomatics, Topcon, I.P., Topcon ImageMaster Photo - 3D photogrammetry survey software<br />

http://www.terrageomatics.com/.<br />

Topcon, 2009. Topcon Positioning System, Inc<br />

http://www.topconpositioning.com/.<br />

Topcon, 2011. Topcon Europe Positioning B. V.<br />

http://www.topcon-positioning.eu/.<br />

Topcon, I., 2008. Topcon ImageMaster - 3D Measurement & <strong>Photogrammetry</strong> software<br />

http://imagemaster3d.com/.<br />

71


APPENDIXES<br />

APPENDIXES<br />

A. 1 Specification of Canon Camera EOS-1Ds<br />

72


APPENDIXES<br />

73


APPENDIXES<br />

74


APPENDIXES<br />

75


APPENDIXES<br />

A. 2 Report - results of camera calibration using PhotoModeler Scanner software<br />

Status Report Tree<br />

Project Name: Camera calibration Canon 28mm.pmr<br />

Problems <strong>and</strong> Suggestions (0)<br />

Project Problems (0)<br />

Problems related to most recent processing (0)<br />

Information from most recent processing<br />

Last Processing Attempt: Tue Jun 07 12:50:00 2011<br />

PhotoModeler Version: 6.4.0.821 - final, full<br />

Status: successful<br />

Processing Options<br />

Orientation: off<br />

Global Optimization: on<br />

Calibration: on (full calibration)<br />

Constraints: off<br />

Total Error<br />

Number of Processing Iterations: 2<br />

Number of Processing Stages: 2<br />

First Error: 0.847<br />

Last Error: 0.847<br />

Precisions / St<strong>and</strong>ard Deviations<br />

Camera Calibration St<strong>and</strong>ard Deviations<br />

Camera1: Canon EOS-1DS [28.00 ok]<br />

Focal Length<br />

Value: 29.361075 mm<br />

Deviation: Focal: 9.7e-004 mm<br />

Xp - principle point x<br />

Value: 18.020201 mm<br />

Deviation: Xp: 0.001 mm<br />

Yp - principle point y<br />

Value: 11.789349 mm<br />

Deviation: Yp: 0.001 mm<br />

Fw - format width<br />

Value: 35.761454 mm<br />

Deviation: Fw: 6.7e-004 mm<br />

Fh - format height<br />

Value: 23.799800 mm<br />

K1 - radial distortion 1<br />

Value: 1.419e-004<br />

Deviation: K1: 1.4e-007<br />

K2 - radial distortion 2<br />

Value: -1.709e-007<br />

Deviation: K2: 3.3e-010<br />

K3 - radial distortion 3<br />

Value: 0.000e+000<br />

P1 - decentering distortion 1<br />

Value: -6.230e-006<br />

Deviation: P1: 4.1e-007<br />

P2 - decentering distortion 2<br />

Value: -1.368e-005<br />

Deviation: P2: 4.2e-007<br />

76


Quality<br />

APPENDIXES<br />

Photographs<br />

Total Number: 12<br />

Bad Photos: 0<br />

Weak Photos: 0<br />

OK Photos: 12<br />

Number Oriented: 12<br />

Number with inverse camera flags set: 0<br />

Cameras<br />

Camera1: Canon EOS-1DS [28.00 ok]<br />

Calibration: yes<br />

Number of photos using camera: 12<br />

Average Photo Point Coverage: 85%<br />

Photo Coverage<br />

Number of referenced points outside of the Camera's calibrated<br />

coverage: 0<br />

Point Marking Residuals<br />

Overall RMS: 0.108 pixels<br />

Maximum: 0.463 pixels<br />

Point 2 on Photo 11<br />

Minimum: 0.096 pixels<br />

Point 67 on Photo 9<br />

Maximum RMS: 0.317 pixels<br />

Point 140<br />

Minimum RMS: 0.054 pixels<br />

Point 67<br />

Point Tightness<br />

Maximum: 0.00056 m<br />

Point 140<br />

Minimum: 8.6e-005 m<br />

Point 94<br />

Point Precisions<br />

Overall RMS Vector Length: 3.73e-005 m<br />

Maximum Vector Length: 4.04e-005 m<br />

Point 141<br />

Minimum Vector Length: 3.63e-005 m<br />

Point 101<br />

Maximum X: 2e-005 m<br />

Maximum Y: 1.98e-005 m<br />

Maximum Z: 2.9e-005 m<br />

Minimum X: 1.63e-005 m<br />

Minimum Y: 1.64e-005 m<br />

Minimum Z: 2.75e-005 m<br />

77


APPENDIXES<br />

A. 3 Comparison on photogrammetric software used in the project including system requirements, import/export files, main features camera<br />

calibration <strong>and</strong> modelling manipulation tools<br />

Application PhotoModeler Scanner V 6. Topcon ImageMaster V 2.0 (Trial – 30 days) Autodesk Project<br />

Photofly 2.0<br />

System requirements<br />

Operating<br />

System<br />

Windows XP;<br />

Windows Vista;<br />

Windows 7;<br />

Windows 98, Me, NT, 2000 also, but there<br />

might be video <strong>and</strong> interface driver problems;<br />

64 bit version runs in compability mode;<br />

software is 32 bit <strong>and</strong> cannot take full advantage<br />

of the 64 bit operating system;<br />

Can be installed on a Intel – based Mac running<br />

Windows as well<br />

Main features<br />

Import Image:<br />

JPG, TIF, BMP, PCX, TGA, PNG, PCT, PSD,<br />

PPM, MAC, IFF, CAL, PCD, SGI, RGB, JP2,<br />

JPX, HDP, WDP<br />

3D file:<br />

DXF, OBJ, 3D Studio, Raw text file<br />

Export 3DS, 3DM, DXF, FBX, IGS, KMZ, MA, MS,<br />

OBJ, RAW, STL, WRL;<br />

Mesh – specific formats:<br />

BYU, FACET, IV, PLY, STL, TXT<br />

Animation:<br />

GIF, AVI<br />

Ortho – photo:<br />

JPG, TIF, BMP, PCX, TGA, PNG, PSD, IFF,<br />

CAL, SGI, RGB<br />

Table:<br />

TXT, CSV<br />

Main features Modelling tools (modelling every type of<br />

surface, shapes etc., DSM creation)<br />

Coded targets: automate the setup – initial<br />

marking <strong>and</strong> referencing<br />

Photograph h<strong>and</strong>ling: any number of<br />

Windows 2000;<br />

Windows XP;<br />

Windows VISTA;<br />

Windows 7<br />

CSV (scan points, GPS control, total stations, etc.,), DXF, TIN (e.g., site<br />

basemaps), IS scanning data, Laser scan point cloud data, Stereo JPEGS<br />

acquired with st<strong>and</strong>ard digital camera, GPT - 7000i imaging total station<br />

orientation data<br />

DXF, CSV, VRML, TIN, Ortho – Images, Point cloud data (with coloured<br />

RGB values)<br />

IS <strong>Remote</strong> Access Control<br />

Grid Scan, point measurement & image capture<br />

Stereo Imaging / <strong>Photogrammetry</strong><br />

Creation of Pass (tie) Points<br />

78<br />

Windows XP SP 3 or<br />

higher;<br />

Windows 7;<br />

32 – bit <strong>and</strong> 64 – bit<br />

systems<br />

Image formats<br />

DWG, OBJ, LAS,<br />

IPM, RZI<br />

Creating near accurate<br />

3D models from<br />

photographs using the<br />

web,


APPENDIXES<br />

Application PhotoModeler Scanner V 6. Topcon ImageMaster V 2.0 (Trial – 30 days) Autodesk Project<br />

Photofly 2.0<br />

photographs, add new photographs at any time, Bundle Adjustment<br />

Utilizes common point<br />

Camera support: images from digital, film or 3D measurements from stereo images (3D monitor supported)<br />

<strong>and</strong> shoot digital<br />

video cameras, Automatic Camera Calibration<br />

cameras,<br />

determines the position of the camera when the Tools<br />

image was taken, Use different cameras in the Creation <strong>and</strong> modification of points <strong>and</strong> polylines<br />

Harnesses the power<br />

same project<br />

Automatic generation of TIN (digital surface model at pre-defined mesh) of cloud computing to<br />

Camera Calibrator<br />

TIN / DSM editing <strong>and</strong> merging<br />

translate photos into<br />

Texture extraction <strong>and</strong> ortho – photos Creation of contours <strong>and</strong> cross sections<br />

detailed 3D models,<br />

Sub-Pixel Target Marking (based on advanced Texture mapping using digital photo input<br />

computer vision techniques)<br />

Creation of ortho – images<br />

3D models can be<br />

Edge Modelling: Model linear features even Distance measurement between two points<br />

manipulated by design<br />

when distinct end points cannot be matched Area calculations<br />

software (AutoCAD,<br />

across the photographs<br />

Constraints: Enter known relationships into the<br />

Volumetric calculations <strong>and</strong> differencing<br />

Inventro, 123D)<br />

project (e.g., parallel, perpendicular, collinear, Project Settings<br />

Sharing through<br />

etc.) to fine-tune processing<br />

Automated camera calibration for stereo image capture<br />

YouTube, iPad,<br />

Inverse Camera: Enter a few known control Coordinate transformations & translation<br />

iPhone, <strong>and</strong> iPod<br />

points or constraints <strong>and</strong> the focal length <strong>and</strong><br />

position of an unknown camera is automatically<br />

calculated for projects involving single or<br />

multiple photos<br />

Single Photograph Support: Use Inverse<br />

Camera with several known control points or<br />

object constraints to model from a single photo,<br />

This allows the use of photographs from<br />

historical archives, byst<strong>and</strong>ers, <strong>and</strong> third parties<br />

for producing 3D data<br />

Automated Tools for Surface Models:<br />

Automated marking, referencing, <strong>and</strong> surfacing<br />

aid in high-density surface model creation<br />

Bundle Adjustment<br />

Eos' Bundle Adjustment provides camera selfcalibration<br />

<strong>and</strong> full error propagation. The first<br />

stage of our Bundle provides automated relative<br />

<strong>and</strong> absolute camera orientation.<br />

Touch<br />

79


APPENDIXES<br />

Application PhotoModeler Scanner V 6. Topcon ImageMaster V 2.0 (Trial – 30 days) Autodesk Project<br />

Photofly 2.0<br />

Model edition/manipulation tools<br />

Cleaning the<br />

model<br />

Smoothing<br />

algorithms (point<br />

cloud/mesh)<br />

Reducing<br />

algorithms (point<br />

cloud/mesh)<br />

Clean outliers – remove points <strong>and</strong> small point<br />

clusters from the mesh that are separated from<br />

the main mesh <strong>and</strong> are probably noise<br />

Clear redundancy – remove duplicate points<br />

that are close to each other within the mesh<br />

Smooth points – adjust the points in the mesh to<br />

remove small bumps <strong>and</strong> imperfections<br />

Decimate triangles – reduce the number of<br />

triangles while retaining the overall shape<br />

Smooth triangles – adjust the triangles in the<br />

mesh to remove small bumps <strong>and</strong> imperfections<br />

Point decimation – reduce the number of points<br />

while retaining the shape of the point cloud<br />

Select – delete Select – delete<br />

TIN Filtering:<br />

� Median<br />

sorts the elevation around the designated point by ascending order,<br />

determines the median value, then replaces the elevation with the median<br />

value, With this filter, outlying, noise – line points can be removed,<br />

Another feature is that some uneven places can be retained, In order to<br />

increase the precision of each point it‟s necessary to narrow the<br />

measurement interval <strong>and</strong> measure many points,<br />

� Mean<br />

determines the average of the elevations around the designated point <strong>and</strong><br />

then replaces those elevations with that average value, It can be used to<br />

make an entire surface smooth, In order to increase the precision of each<br />

point, it is necessary to narrow the measurement interval <strong>and</strong> measure<br />

many points<br />

n/a<br />

Triangulation Triangulation – triangulate a point cloud to<br />

form a surface<br />

n/a<br />

Filling holes Fill Holes – fill enclosed areas that were not<br />

triangulated due to missing point data<br />

n/a<br />

Texture mapping n/a Texture mapping – function for mapping texture (an image) to a selected<br />

surface (TIN) on the model or registration screen or on the stereo screen<br />

Offset n/a Offset elevations – offsetting the elevations (Z coordinates) of selected<br />

TIN vertices<br />

Reversed n/a Reversed Triangles – reversing the front <strong>and</strong> back (normal vector<br />

triangles<br />

orientation of the triangle selected<br />

Clipping n/a Clipping – clipping area<br />

80<br />

All steps in one set –<br />

computed, made <strong>and</strong><br />

sent by server<br />

Mesh quality –<br />

Mobile, St<strong>and</strong>ard,<br />

Maximum


APPENDIXES<br />

A. 4 Comparison of 3D digitizers: main features, specification: used methods, light, range, resolution, accuracy, etc.<br />

Model Konica Minolta VI-910 Artec MHT 3D Scanner Artec S 3D Scanner<br />

Measurement method Triangulation light block method Triangulation light block Triangulation light<br />

method<br />

block method<br />

Autofocus Image surface AF (contrast method); active AF Yes Yes<br />

Light – receiving lenses (exchangeable) TELE: Focal distance f = 25 mm<br />

MIDDLE: Focal distance f = 14 mm<br />

WIDE: Focal distance f = 8 mm<br />

No info No info<br />

Ability to capture texture Yes Yes No<br />

Scan range (depth of field) / working<br />

distance<br />

0,6 to 2,5 m (2m for WIDE lens) 0,4 – 1 m 0,15 – 0,25 m<br />

Optimal 3D measurement range 0,6 to 1,2m Whole range Whole range<br />

Light source Laser Flash bulb (no laser) Flash bulb (no laser)<br />

Laser class Class 2 (IEC60825-1), Class (FDA) n/a n/a<br />

Laser scan method Galvanometer – driven rotating mirror Light pattern Light pattern<br />

X-direction input range (varies with 111 to 463 mm (TELE),<br />

148 to 371 mm 56 to 93 mm<br />

distance) / linear field of view<br />

198 to 823 mm (MIDDLE),<br />

359 to 1196 mm (WIDE)<br />

Y-direction input range (varies with 83 to 347 mm (TELE),<br />

214 to 536 mm 80 to 134 mm<br />

distance)<br />

148 to 618 mm (MIDDLE),<br />

269 to 897 mm (WIDE)<br />

Z-direction input range (varies with 40 to 500 mm (TELE),<br />

No info No info<br />

distance)<br />

70 to 800 mm (MIDDLE),<br />

110 to 750 mm (WIDE)<br />

Angular field of view, HxW No info 30x21° 30x21°<br />

3D resolution No info Up to 0,5 mm Up to 0,2 mm<br />

Accuracy / 3D point accuracy X: ±0,22 mm, Y: ±0,16 mm, Z: ±0,10 mm to the Z reference plane<br />

(Conditions: TELE/ FINE mode, Konica Minolta‟s st<strong>and</strong>ard)<br />

Up to 0,1 mm Up to 0,05 mm<br />

3D accuracy over distance No info Up to 0,15% over 100 cm Up to 0,15% over 100<br />

cm<br />

Exposure time n/a 0,0002 s 0,0002 s<br />

Video frame rate n/a Up to 15 fps Up to 15 fps<br />

Input time / Data acquisition speed 0,3 s (FAST mode) or<br />

1,5 s (FINE mode)<br />

Up to 500 000 points/s Up to 500 000 points/s<br />

Ambient lighting condition 500 lx or less No info No info<br />

Imaging element 3D data: 1/3 inch frame – transfer CCD (340 000 pixels)<br />

Colour data: 3D data is shared (colour separation by rotary filter)<br />

No info n/a<br />

81


APPENDIXES<br />

Model Konica Minolta VI-910 Artec MHT 3D Scanner Artec S 3D Scanner<br />

Number of output pixels / texture 3D data: 307 000 (FINE mode), 76 800 (FAST mode)<br />

1,3 mp<br />

n/a<br />

resolution; colours<br />

Colour data: 640 x 480 x 24 bits colour depth<br />

24 bpp<br />

n/a<br />

Data file size Total 3D data <strong>and</strong> colour data capacity:<br />

Depending on amount of Depending on amount<br />

1,6 MB (FAST mode) per data,<br />

3,6 MB (FINE mode) per data<br />

frames <strong>and</strong> texture<br />

of frames<br />

Dimensions (W x H x D) 213 x 413 x 271mm (8- 3/8 x 16 – ¼ x 10-11/16 in,) 260 x 180 x 187 mm 80 x 125 x 195 mm<br />

Weight Approximately 11 kg (25 lbs) 1,6 kg 1,6 kg<br />

Processing capacity No info 40 000 000 triangles / 1 GB 40 000 000 triangles / 1<br />

RAM<br />

GB RAM<br />

82


APPENDIXES<br />

A. 5 Comparison of software from scanning systems; main features, system requirements, model manipulation tools.<br />

Application Konica Minolta Polygon Editing Tool V2.20 Artec 3D Scanner<br />

System requirements:<br />

OS / Operating<br />

System<br />

Main features<br />

Readable formats /<br />

Scan data formats<br />

Output formats /<br />

Exporting<br />

Windows 2000 Professional SP4,<br />

XP Professional SP2 (x64 Edition not supported)<br />

83<br />

Windows XP 32,<br />

Vista 32 / 64,<br />

7 32 / 64<br />

CAM, WD, SCN, CDM, CDK<br />

General format: STL<br />

Artec formats<br />

Konica Minolta formats & STL, DXF, OBJ, ASCII points, VRML STL, PLY, WRL, OBJ, PTX, CSV, ASC, AOP<br />

Multi core processing No Yes<br />

Camera remote<br />

operation<br />

Main Functions /<br />

features<br />

Measurement,<br />

measurement reference distance setting,<br />

number of scans setting,<br />

laser power setting,<br />

high – quality setting,<br />

filter setting, etc.,<br />

Scanning,<br />

Data alignment,<br />

data merging,<br />

smoothing,<br />

data reduction,<br />

polygon check,<br />

hole filling with data interpolation,<br />

texture blending<br />

measuring tools<br />

Display modes:<br />

Wireframe,<br />

shading,<br />

texture mapping<br />

Light settings for texture<br />

Working distance settings<br />

Number of frames settings<br />

Scanning,<br />

multicapturing,<br />

scan alignment,<br />

global optimization,<br />

model fusion,<br />

smoothing,<br />

simplify,<br />

hole filling with data interpolation,<br />

repair tool,<br />

texture mapping,<br />

measuring tools<br />

Display modes:<br />

Points,<br />

Wireframe<br />

Shading<br />

Lighting, etc.<br />

Texture


APPENDIXES<br />

Application Konica Minolta Polygon Editing Tool V2.20 Artec 3D Scanner<br />

Model manipulation tools<br />

Aligning point 3 corresponding point need for manual initial registration;<br />

3 corresponding point need for manual initial<br />

cloud/mesh<br />

Any number of scans can be aligned with fine registration<br />

registration;<br />

Any number of scans can be aligned with fine<br />

registration <strong>and</strong> then global registration, which<br />

improves the results<br />

Merging point Scans seems to be slightly moved, again <strong>and</strong> some data seem to be lost – more holes “Fusion” tool, which seems to move a bit scans <strong>and</strong><br />

cloud/mesh<br />

smooth the model, sometimes resulting with more holes<br />

Cleaning the model Select - Delete Eraser brush – select <strong>and</strong> remove unwanted objects<br />

Singletons – removing occasionally encounter outliers –<br />

small areas or fragments untied to the model;<br />

� Filter by threshold<br />

all scene objects smaller than the specified size (number<br />

of polygons) determined by the threshold parameter will<br />

be removed;<br />

� Leave biggest objects<br />

Smoothing algorithm Smooth:<br />

leaves only the biggest objects in each frame<br />

Smooth – algorithm which allows smoothing out noise,<br />

(point cloud/mesh) � Element – makes selected elements smooth as well as to regularize surface outliers on the 3D model <strong>and</strong> makes it generally<br />

points density;<br />

smoother<br />

� Points – makes boundary of the points selected in the currently displayed<br />

element smooth as well as regularize surface points density<br />

Smooth brush – select <strong>and</strong> smooth small parts<br />

Reducing algorithm Subsample<br />

Simplify – optimization algorithm operation<br />

(point cloud/mesh) � Uniformly – Element/Points – reduces the data of the points of the selected<br />

element so that surface points density becomes uniform<br />

� Adaptively – Element/Points – reduces the data of the points of the selected<br />

element so that surface points density in simple-shaped areas is lower than in<br />

complicated-shaped areas<br />

Filling holes<br />

Modify – Element / Points – rebuilds the selected element / points t by deleting small<br />

polygons<br />

Fill Holes:<br />

Fill Holes – or<br />

� Manually (by manually selecting holes)<br />

Filling Holes using separate special tool Holes<br />

� Automatically (create polygon data automatically for the holes However,<br />

some complicated holes may not be filled<br />

Repair Check Polygons – checking for illegal polygons among those compromising the<br />

selected element<br />

Repair (mesh)<br />

84


APPENDIXES<br />

Application Konica Minolta Polygon Editing Tool V2.20 Artec 3D Scanner<br />

Adding points Subdivision – Element / Points– rebuilds the selected element / points by dividing<br />

large polygons<br />

n/a<br />

Triangulate Triangulate – Elements/ Polygon – dividing polygons of the selected element(s) / No info<br />

points into triangles,<br />

Triangulated point cloud is set from view menu<br />

Texture mapping Texture Blending – making the edges of colour image of registrated element smooth Texture mapping – applying texture onto the 3D model<br />

from photos<br />

85


APPENDIXES<br />

A. 6 Geomagic Qualify: main features, import/export files, system requirements <strong>and</strong> main manipulation tools used in the project<br />

Application Geomagic Qualify V 12<br />

System requirements:<br />

OS / Operating System Windows 7 (32 <strong>and</strong> 64 bit);<br />

Windows Vista SP1 (32 <strong>and</strong> 64 bit);<br />

Windows XP SP3 (32 <strong>and</strong> 64 bit)<br />

CPU / Processor Intel compatible 2 GHz / Dual or Quad core 2 GHz CPU‟s recommended<br />

Main features<br />

Readable formats / Scan<br />

data formats (input)<br />

Supports all 3D digitizers, cameras <strong>and</strong> scanners in XYZ / ASCII format:<br />

3PI, AC, ASC, BIN, SWL, BRE, BTX, CDK, CDM, RGV, RVM, VVD, COP, CWK, DBT, FLS, G3D, SURF – GOM, GPD, GTI, HYM,<br />

ICV, MET, MTN, NET, OPD, OPT, PIX, PMJ/ X, PTX, SAB2, SCN, MGP, SCN, STB, XYZ, XYZN, ZFS<br />

Output formats / Exporting Polygon Import / Export:<br />

3DS, DXF, IGS, LWO, NAS, OBJ, PLY, STL, VRML, WRP<br />

CAD Import / Export:<br />

IGES, STEP 203/214, VDA, PRO / ENGINEER (.PRT), PARASOLID (.X_T* <strong>and</strong> .X_B*), SAT<br />

Reports:<br />

3D PDF, HTML, Microsoft Word, PPT, Excel formats, CSV, Unicode data, SPC<br />

Available translators CAIA V4 & V5, NX, Solidworks<br />

Languages Chinese, English, French, German, Italian, Japanese, Russian, Czech, Brazilian Portuguese, Spanish<br />

Main Functions / features Scan data (Collect point <strong>and</strong> polygon data from all major 3D scanners, digitizers <strong>and</strong> hard-probing devices)<br />

Aligning scans (Weighted RPS alignment)<br />

Data merging,<br />

Smoothing,<br />

Data reduction,<br />

hole filling with data interpolation, defining exact shapes,<br />

Mesh check,<br />

Comparing (illustration of deviations, reports, etc.)<br />

Evaluate (Creation of cross-sections, detect <strong>and</strong> inspect geometrical features, calculate size, analyse fit, compare features, measuring<br />

distances, angles)<br />

Reporting,<br />

Automate:<br />

� Automate the inspection process from alignment to the generation of inspection reports<br />

� Record steps in the inspection process to inspect multiple parts automatically<br />

� Define macros for automating repetitive tasks<br />

Geomagic Blade Module (Optional)<br />

Texture mapping available in Geomagic Wrap or Studio<br />

86


APPENDIXES<br />

Application Geomagic Qualify V 12<br />

Display modes:<br />

Different colour (or texture)<br />

Reflectiveness<br />

Light sources<br />

Light type<br />

Transparency of the model<br />

And many more…<br />

Model manipulation tools<br />

Aligning point cloud/ mesh From full automatic to manual, needs at least 3 corresponding points for initial registration; global registration with any amount of scans;<br />

also “Overlap reduction” tool which improve the results <strong>and</strong> align the scans better<br />

Merging point cloud / mesh Merging all scan together; all points without any data lost or adding any points<br />

Cleaning the model Select – Delete points / triangles<br />

Crop Points – Deletes all points from an object except selected points<br />

Select:<br />

� Outliers - Select points that are at least a given distance from most others<br />

Smoothing algorithm (point<br />

cloud / mesh)<br />

Reducing algorithm (point<br />

cloud / mesh)<br />

� Disconnected Components – evaluates the proximity of points <strong>and</strong> select groups of points that are distant from other group<br />

Smooth:<br />

� Reduce Noise Points – Compensates for scanner error (noise) by moving points to statistically correct locations,<br />

� Reduce Noise (mesh) – compensates for noise such as scanner error by moving points to statistically correct locations, Noise can<br />

cause sharp edges to become dull or smooth curves to become rough<br />

� Remove Spikes – detects <strong>and</strong> flattens single – point spikes on a polygon mesh<br />

Sample:<br />

� Uniform – Reduces the number of points on flat surfaces uniformly, but reduces the number of points on curved surfaces to a<br />

specified density<br />

� Curvature – Reduces the number of points in flat regions but preserves points in high – curvature areas to maintain detail<br />

� Grid – Reduces the number of pints in an Unordered Point object by creating an evenly spaced set of points, regardless of curvature<br />

<strong>and</strong> original density<br />

� R<strong>and</strong>om – Removes a percentage of points r<strong>and</strong>omly from an Unordered Points object<br />

� Decimate – Reduces the number of triangles without compromising surface detail or colour<br />

Filling holes Fill Holes – Inserts ordered points into voids on the surface of an unordered point object<br />

Fill All – fills all selected holes on the active Polygon object<br />

Fill Single – fills the hole or holes that are chosen as click one at a time in the Graphics Area, according to the setting in the Fill Holes ribbon<br />

group; Ribbon Group: Specifies that the new mesh must match the curvature of the surrounding mesh:<br />

� Curvature;<br />

� Tangent – more tapering than Curvature;<br />

� Flat – new mesh is generally flat,<br />

� Complete – specifies that an entire opening will be filled;<br />

87


APPENDIXES<br />

Application Geomagic Qualify V 12<br />

� Partial – Specifies that a portion of a hole will be filled,<br />

� Bridge – Specifies that a bridge across a hole will be built, thus dividing it into separately fillable holes, This may be used to divide<br />

a complex hole into smaller ones that can be filled more correctly,<br />

Repair Normals – Comm<strong>and</strong>s that manipulate normal information on unordered point objects:<br />

� Repair – Manipulates, Flips <strong>and</strong>/or removes normal information on an unordered point object<br />

� Remove – Deactivates the shading of Point objects<br />

Scan lines – A set of comm<strong>and</strong>s that repair scan lines (which are created only by some scanning devices):<br />

� Interpolate – optimizes scanlines in preparation for Wrap /<br />

� Order – convents unordered scanline data to ordered data (applicable to Point objects captured with a line scanner)<br />

Mesh:<br />

� Mesh Doctor (Automatically repair imperfections in a polygon mesh), Looking for:<br />

o Non-Manifold Edges<br />

o Self-Intersections<br />

o Highly Creased Edges<br />

o Spikes<br />

o Small Components<br />

o Small Tunnels<br />

o Small Holes<br />

Also possible, to:<br />

o Clean<br />

o Defeature<br />

o Fill Holes<br />

o Make Manifold<br />

� Make Manifold - Set of comm<strong>and</strong>s that delete non-manifold triangles; Manifold triangle is one that is connected the others on all<br />

three sides, or on two sides if it lies on an edge)<br />

� Open - Deletes non-manifold triangles from an otherwise open manifold objects<br />

� Closed- from closed objects (volume enclosing) – On an open manifold objects, all triangles would be considered nonmanifold<br />

<strong>and</strong> the entire object would be deleted<br />

� Trim – Superimposes a plane on the object <strong>and</strong> removes all triangles on one side of that plane, or creates an artificial boundary line<br />

at the intersection:<br />

o Plane -Superimposes a plane on the object <strong>and</strong> removes all triangles on one side of that plane, or creates an artificial<br />

boundary line at the intersection<br />

o Sheet – cuts 3 dimensional chunk from a polygon object by slicing it with a 2 dimensional curve)<br />

� Repair Tool :<br />

o Repair Normals – repairs normal directions of polygons whose “front” <strong>and</strong> “back” sides are confused – a problem caused<br />

by wrapping a Noisy point object<br />

o Flip normals – reverses the normal direction of a Polygon mesh – flips the “front” <strong>and</strong> “back” sides of the entire mesh)<br />

88


APPENDIXES<br />

Application Geomagic Qualify V 12<br />

Adding points Add points – adds individual points to a plane in an unordered point object<br />

Offset Offset Points – Moves unordered points by a user – specified distance in their normal directions<br />

Offset – applies a uniform offset <strong>and</strong> thus makes an object larger or smaller<br />

Triangulate Wrap – Converts a Point object into an polygon object by converting the point cloud to a mesh – triangulation to TIN<br />

Texture mapping Texturing available in Geomagic Wrap or Studio, in Qualify texture is available only from default pallet<br />

Other Compensate Deviations – creates a polygon object that will serve as input to a rapid prototyping machine <strong>and</strong> which compensates for a<br />

consistent flaw in that machine<br />

And more…<br />

Akca, D., Remondino, F., Novák, D., Hanusch, T., Schrotter, G., Gruen, A., Recording <strong>and</strong> modeling of cultural heritage objects with coded structured light<br />

projection systems, pp. 1-7.<br />

Alyilmaza, C., Alyilmaza, S., Yakarb, M., 2010. MEASUREMENT OF PETROGLYHPS (ROCK OF ARTS) OF QOBUSTAN WITH CLOSE RANGE PHOTOGRAMMETRY,<br />

in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement<br />

Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 29-32.<br />

Artec, G., 2011. Artec 3D Scanners<br />

http://www.artec3d.com/.<br />

Autodesk, 2011. Autodesk Labs Utilities - Exploring new approaches to design technology - Project Photofly 2.0 Photo Scene editor<br />

http://labs.autodesk.com/utilities/photo_scene_editor/.<br />

Bache, S.N., 2003. Scanning av skulptur på Nidarosdomen. Vinn Design, Oslo, Trondheim, pp. 1-4.<br />

Balestrinia, E.F.d., Guerra, F., 2011. NEW INSTRUMENTS FOR SURVEY: ON LINE SOFTWARES FOR 3D RECONSTRUCTION FROM IMAGES; ISPRS Commission V -<br />

WG 4; Poster Session, in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong><br />

Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy, pp. 1-8.<br />

Barba, S., Fiorillo, F., Coder, P.O., D’Auria, S., Feo, E.D., 2011. AN APPLICATION FOR CULTURAL HERITAGE IN ERASMUS PLACEMENT. SURVEYS AND 3D<br />

CATALOGUING ARCHAEOLOGICAL FINDS IN MÉRIDA (SPAIN), in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-<br />

ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy,<br />

pp. 1-6.<br />

Barnett, T., Bryan, P., Ch<strong>and</strong>ler, J., Case study: Recording Prehistoric Rock Art; Pre-historic photogrammetry <strong>and</strong> laser scanning, Tertia Barnett<br />

Paul Bryan<br />

Jim Ch<strong>and</strong>ler, pp. 1-2.<br />

89


APPENDIXES<br />

Blais, F., 2004. Review of 20 Years Of Range Sensor Development. Journal of Electronic Imaging 13(1), 231-240.<br />

Blake, W.H., 2010. WHAT IS THE FUTURE OF METRIC HERITAGE DOCUMENTATION AND ITS SKILLS?, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.),<br />

Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong><br />

<strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 98-102.<br />

Bonfanti, C., Chiabr<strong>and</strong>o, F., Spanò, A., 2010. HIGH ACCURACY IMAGES AND RANGE BASED ACQUIRING FOR ARTISTIC HANDWORKS 3D-MODELS., in: Mills,<br />

J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement<br />

Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 109-114.<br />

Brutto, M.L., Spera, M.G., 2011. IMAGE-BASED AND RANGE-BASED 3D MODELLING OF ARCHAEOLOGICAL CULTURAL HERITAGE: THE TELAMON OF THE<br />

TEMPLE OF OLYMPIAN ZEUS IN AGRIGENTO (ITALY), in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH:<br />

"3D Virtual Reconstruction <strong>and</strong> Visualization of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, pp. 1-8.<br />

Canon, 2011. Canon Global<br />

http://www.canon.com/.<br />

Chmielewski, K., Szulwic, J., Niemetryczne zdjęcia cyfrowe w fotogrametrii bliskiego zasięgu w systemie Topcon PI-2000, Sesja Jubileuszowa 60-lecia Katedry<br />

geodezji Politechniki Gdaoskiej, pp. 1-8.<br />

Deseilligny, M.P., Clery, I., 2011. APERO, AN OPEN SOURCE BUNDLE ADJUSMENT SOFTWARE FOR AUTOMATIC CALIBRATION AND ORIENTATION OF SET OF<br />

IMAGES., in: Remondino, F., El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization<br />

of Complex Architectures". International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy, pp. 1-8.<br />

Floth, M., Breuer, M., 2010. AN OPTICAL THREE-DIMENSIONAL MEASURING TECHNIQUE FOR A DETAILED NON-CONTACT DATA ACQUISITION OF OBJECT<br />

SURFACES IN THE FIELDS OF CULTURAL HERITAGE, ARCHEOLOGY AND THE CARE AND PRESERVATION OF HISTORIC MONUMENTS, in: Mills, J.P., Barber,<br />

D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'.<br />

International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 240-244.<br />

Geomagic, 2011. Geomagic reverse engineering <strong>and</strong> 3D inspection software - "The magic of making it simple"<br />

http://www.geomagic.com/.<br />

Hess, M., Robson, S., 2010. 3D Colour Imaging For Cultural Heritage Artefacts, in: Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the<br />

ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>,<br />

Newcastle upon Tyne, United Kingdom, pp. 288-292.<br />

Jędrzejek, M., Łącka, M., Wykorzystanie wzorców świetlnych i metod fotogrametrycznych do pozycjonowania ciała ludzkiego (PiCalib), WGGiIŚ. Akademia<br />

Górniczo-Hutnicza, Kraków.<br />

Kazanaa, E., 2010. TOPCON’S PI-3000 IN THE INVENTORY OF DATA RELATING TO SCULPTURAL AND PORTABLE ARTIFACTS, in: Mills, J.P., Barber, D.M., Miller,<br />

P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement Techniques'. International Society of<br />

<strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 322-324.<br />

90


APPENDIXES<br />

Kochi, N., Ito, T., Noma, T., Otani, H., Nishimura, S., Ito, J., PC-BASED 3D IMAGE MEASURING STATION WITH DIGITAL CAMERA AN EXAMPLE OF ITS ACTUAL<br />

APPLICATION ON A HISTORICAL RUIN. The International Archives of the <strong>Photogrammetry</strong>, <strong>Remote</strong> <strong>Sensing</strong> <strong>and</strong> Spatial Information Sciences; Commission V<br />

XXXIV, Part 5/W12, 195-199.<br />

Leksikon, S.N., Brochmann, O., Tschudi-Madsen, S., Storsletten, O., 2009. Nidarosdomen<br />

http://snl.no/Nidarosdomen.<br />

Leksikon, S.N., Dybdahl, A., Pål Bårdsson – utdypning (NBL-artikkel)<br />

http://snl.no/.nbl_biografi/Pål_Bårdsson/utdypning.<br />

Luhmann, T., Robson, S., Kyle, S., Harley, I., 2006. Close Range <strong>Photogrammetry</strong>: Principles, Techniques <strong>and</strong> Aplications. Whittles Publishing, Caithness,<br />

Scotl<strong>and</strong>, UK.<br />

Metrology, E., Artec 3D Scanner from Exact Metrology - Effortless 3D Scanning<br />

http://www.exact3dscanner.com/.<br />

Minolta, K., GmbH, B.S.E., 2011. Konica Minolta Europe - The essentials of imaging<br />

http://www.konicaminolta.eu/.<br />

Nidaros, D., Nidarosdomen & Erkebispegården<br />

http://www.nidarosdomen.no/nb-NO/.<br />

Organisasjon, D.K.K., Martinsen, O., 1996. Archbishop Pål (1333-1346)<br />

http://www.katolsk.no/organisasjon/mn/erkebiskoper/15.<br />

Pawlak, A., Pulchny, K., 2009. Zastosowanie cyfrowych aparatów fotograficznych do fotogrametrycznego pomiaru wybranych części ciała ludzkiego / The<br />

application of digital cameras to photogrammetric measurement of chosen parts of human body., WGGiIŚ. Akademia Górniczo-Hutnicza, Kraków, p. 133.<br />

PhotoModeler, 2011. PhotoModeler - Measuring & Modeling the Real World - accurate <strong>and</strong> affordable 3D photogrammetry measurement <strong>and</strong> scanning<br />

http://www.photomodeler.com/.<br />

Studios, I., 3D Laser Scanning Services, A Division of Impact Studios LLC<br />

http://www.impactstudiostv.com/.<br />

TerraDat, TerraGeomatics, Topcon, I.P., Topcon ImageMaster Photo - 3D photogrammetry survey software<br />

http://www.terrageomatics.com/.<br />

Topcon, 2009. Topcon Positioning System, Inc<br />

91


http://www.topconpositioning.com/.<br />

Topcon, 2011. Topcon Europe Positioning B. V.<br />

APPENDIXES<br />

http://www.topcon-positioning.eu/.<br />

Topcon, I., 2008. Topcon ImageMaster - 3D Measurement & <strong>Photogrammetry</strong> software<br />

http://imagemaster3d.com/.<br />

Tucci, G., Cini, D., Nobile, A., 2011. EFFECTIVE 3D DIGITIZATION OF ARCHAEOLOGICAL ARTIFACTS FOR INTERACTIVE VIRTUAL MUSEUM, in: Remondino, F.,<br />

El-Hakim, S. (Eds.), Proceedings of the 4th ISPRS International Workshop 3D-ARCH: "3D Virtual Reconstruction <strong>and</strong> Visualization of Complex Architectures".<br />

International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Trento, Italy, pp. 1-8.<br />

Wojtas, A.M., 2010. OFF-THE-SHELF CLOSE-RANGE PHOTOGRAMMETRIC SOFTWARE FOR CULTURAL HERITAGE DOCUMENTATION AT STONEHENGE, in:<br />

Mills, J.P., Barber, D.M., Miller, P.E., Newton, I. (Eds.), Proceedings of the ISPRS Commission V Mid-Term Symposium 'Close Range Image Measurement<br />

Techniques'. International Society of <strong>Photogrammetry</strong> <strong>and</strong> <strong>Remote</strong> <strong>Sensing</strong>, Newcastle upon Tyne, United Kingdom, pp. 603-607.<br />

92

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!