Multimedia system for interactive visual recognition of computer-generated objectsFig 2- A mo<strong>de</strong>l built from the unknown objectFig. 3 The mo<strong>de</strong>l of similitu<strong>de</strong> from thedatabaseFig. 4 The mo<strong>de</strong>l forward and afterthe <strong>de</strong>creaseThe visible mo<strong>de</strong>ls of the images look like linear simple drawings, in which the distinctpoints can be obtained and <strong>de</strong>tected through analysis. The visible mo<strong>de</strong>l integrates the curves ofthe Italian mathematician Guido Grandi (from the XVIII cent.). The operator classifies alwaysthe object by clicking the correspon<strong>din</strong>g image.The images from the database that have already been classified can be ad<strong>de</strong>d to thereference set, with assigned labels from the operator. They are subsequently used to improve thebuil<strong>din</strong>g of the mo<strong>de</strong>l required for the class<strong>ific</strong>ation through approximation.3. Mobile architecturesAlthough on the first portable system, visual interactive systems were implemented on afixed workstation (200 Mhz, 64 MB RAM, serial ports and Compact Flash), the camera mo<strong>de</strong>lwas compact Sharp which permitted the direct capture of images. Images from another room canbe also loa<strong>de</strong>d through the PDA port; Sony was chosen because it offers ultimate performance,Linux support and utilities to interact with Java Personal, supplementary to the position of the80
Multimedia system for interactive visual recognition of computer-generated objectsroom. Java Personal has a greater flexibility than MIDP (Mobile Interface Device Profile) carriesis launched on PDAs and mobile phones.The mechanism of recognition is completely <strong>de</strong>fined in abstract class, credit titles lacking someinterfaces for data management. Because of the generic class, interaction handling of images, theGUI (Graphical User Interface) prompts the implementation of a few methods. IVS are based onthe ability of human perception and grouping of similar regions, perceived features, outline ofdata and recognized sign<strong>ific</strong>ant differences. It uses the capacity of the computer memorizedcouples of images, quantifying the features and calculating the distance space of abstract featuresfor some forms and colours. The architecture was specially <strong>de</strong>veloped for the recognition in thefield of insular objects, where the available time for the class<strong>ific</strong>ation of each image iscomparable with the one for the acquisition of the image. A similar system was <strong>de</strong>scribed in thepresent, where, installed in a room with WiFi IEEE 802. 11b enabled PDA Toshiba-e800. Weredivi<strong>de</strong>d among PDAs and a laptop covering a zone of 100 m, around the host. The PDAtransmits on the radio interface each image obtained recently to host. The host calculates thevisible initial mo<strong>de</strong>l, classifies the or<strong>de</strong>r candidates using his stored reference images, and theyreturn the parameters of the mo<strong>de</strong>l. And the numbers in<strong>de</strong>x of the candidates from the PDA. ThePDA then displays the candidates of the tree from the database of the images of reference.If the user adjusts the mo<strong>de</strong>l (using his random registration), parameters of dodged-upmo<strong>de</strong>l of the by-path transmitted to the host, a new mo<strong>de</strong>l and a class<strong>ific</strong>ation is a calculated andcommunicated to PDA. The file is kept on the PDA. With this system we did experiments inwith flower images, from their natural landscapes and repeated some experiments from thedatabase with other subjects. From research results that for recognition on the age with across 20times quicker than the utilization of <strong>de</strong>sktop because of the adjustment. Is recognized that it waseasy at the thin bor<strong>de</strong>r because some flowers of reference can be easy distinguished on littledisplay of PDA. The calculus in net enforced with no sign<strong>ific</strong>ant <strong>de</strong>lay with the exception of thearrival of each new image. The most important drawback of portable <strong>de</strong>vices is that they can beeasy operate to the daylight, because the display is near invisible, still more many, the colour<strong>de</strong>pth and the resolution differs nice and long ago from a place to other.4. ConclusionThe experiments <strong>de</strong>monstrate that an iterative class<strong>ific</strong>ation in problems of multipleclasses is much more than the automatic current class<strong>ific</strong>ations and much more than the humanclass<strong>ific</strong>ation without algorithms. Of tacked sights current, the technology of memory and CPUis proper for the systems of tacked sights, portable computers, which cases can operatein<strong>de</strong>pen<strong>de</strong>ntly or in way customer host radio. The vast availability, you launch we apply thetraining, industrial check and medical diagnosis of visible symptoms.Bibliography:1. Enescu Florentina Magda , PhD Thesis, 20082. Ipate, F.E., “ Dezvoltarea aplicaţiilor <strong>de</strong> baze <strong>de</strong> date în Oracle 8 şi Forms 6 “, Editura BICALL, 20003. Michael Hernan<strong>de</strong>z, “Proiectarea bazelor <strong>de</strong> date”, Editura Teora, 20034. Jie Zou, Abhishek Gattani, George Nagz, Rensselaer Polztechnic Institute, Troy, NY5. J. Zou, Computer Assisted Visual Interactive Recognition: CAVIAR, Rensselaer PolztechnicInstitute PhD Thesis, May 200481