01.12.2012 Views

Automatic generation of elevation data over Danish landscape

Automatic generation of elevation data over Danish landscape

Automatic generation of elevation data over Danish landscape

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.

2 Methods <strong>of</strong> determining <strong>elevation</strong> <strong>data</strong><br />

� grid and unstructured raw <strong>elevation</strong> <strong>data</strong> from different sources/producers<br />

A mathematical model must therefore be established which can handle the problem <strong>of</strong> historical, present<br />

and future <strong>elevation</strong> <strong>data</strong>, captured from different sources/producers, with different accuracies and structures<br />

(grid or raw <strong>elevation</strong> <strong>data</strong>). A solution to this problem, could be the ‘multi grid method’. The ‘multi<br />

grid method’ is a highly effective iterative solution strategy which can solve large areas with thinly determined<br />

points with the help <strong>of</strong> an adjustment system. The great advantage with the ‘multi grid method’ is<br />

that the adjustment system is independent <strong>of</strong> the numbers <strong>of</strong> unknowns.<br />

The starting point could be: two clouds <strong>of</strong> structured or unstructured points, where the <strong>elevation</strong> <strong>data</strong> is<br />

independent and has different accuracies. Should one <strong>of</strong> the clouds <strong>of</strong> points consist <strong>of</strong> grid points, this<br />

grid will be looked at, as a non-structural cloud <strong>of</strong> points.<br />

This is a classic adjustment problem, the solution being, an adjustment is done by means <strong>of</strong> the least<br />

squares measurement principle. (For the readers, who are familiar with least squares measurements,<br />

skip the next 1½ pages). The principle <strong>of</strong> least squares measurement is, that the random values (residuals)<br />

(vi) have to be minimum, which means:<br />

v<br />

m<br />

2 2 2<br />

2<br />

2<br />

� 1 v � 2 v � � � � �<br />

3 v � m �v i<br />

i�1<br />

� min<br />

When the observations are independent and have different weights the principle <strong>of</strong> least squares measurement<br />

demands that the weight function minimises like:<br />

� �<br />

m<br />

c � c v<br />

2<br />

� c v<br />

2<br />

� ��<br />

��<br />

��<br />

� � c<br />

2<br />

� c v<br />

2<br />

� � min<br />

2 2 2 3 3 m i i<br />

i � 1<br />

This function can be noted as matrix form:<br />

�<br />

�<br />

�vvv���v� 1 2 3 m<br />

�<br />

�<br />

�<br />

� �<br />

�<br />

�<br />

�<br />

�<br />

c1<br />

c<br />

2<br />

c<br />

3<br />

�<br />

c<br />

� �<br />

� �<br />

� �<br />

� � �<br />

� �<br />

� �<br />

� �<br />

� �<br />

m<br />

� 1<br />

�<br />

2�<br />

�<br />

3<br />

�<br />

� �<br />

�<br />

m�<br />

v<br />

v<br />

v<br />

This matrix form can be written more simply as:<br />

T<br />

� � v �C�<br />

v<br />

The equations for the observations can be written in matrix format as:<br />

l � A � x<br />

- v<br />

The above equation can be rewritten as:<br />

v � A � x<br />

- l<br />

T<br />

and together with equations � � v �C�<br />

v the equation will be:<br />

� = (A�x - l) T �C � (A�x - l)<br />

= (x T �A T �C - l T �C) � (A�x - l)<br />

= (x T �A T - l T )�C�(A�x-l)<br />

= x T �A T �C�A�x - l T �C�A�x - x T �A T �C�l + b T �C�l<br />

v<br />

37

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!