Full text: Proceedings of the Symposium "From Analytical to Digital" (Part 2)

2. PROGRAMMING SYSTEM 
Programming language is standard FORTRAN 77, the program system is 
being developed and implemented on PRIME minicomputers 550 and 750. 
However, an optimized version has also been produced to enable parts 
of the program with a high computational load to run on the CYBER 205 
vector computer. 
2.1 Epipolar geometry 
After orientation of the images by image correlation or conventional 
methods, then the position of points in the stereo partner, 
corresponding to a selected point, can only lie on a well defined line 
- the epipolar line. Because of this, the search for corresponding 
points is greatly accelerated. Further, the possibility of gross 
errors, due to correlation between similar but not identical points, 
is reduced, since the search area is restricted to this line. «This 
considerably increases the reliability and has the advantage that more 
points can be used. For instance, edges with components normal to the 
epipolar line can be used. If the search were not then restricted to 
this line, false correlations (gross errors) would be obtained at 
other points along the edge. 
To make full use of the epipolar geometry and simplify the procedure, 
both photographs are rectified over the stereoscopic area to be 
evaluated. This makes all epipolar lines parallel. If this direction 
is taken as the x-direction, then all corresponding points have the 
same y-value /10/. 
2.2 Grid generation 
The grid points used for the DTM are generated by a process of point 
densification. The technique starts with a very coarse grid, which 
gets finer until the required point density is reached. This is done 
in a hierarchy of 3 stages. For the coarse grids a low-pass filter is 
used to create an appropriate coarse image. At this level, accuracy 
of point identification is low. However, as the grid gets finer the 
photo resolution increases and the accuracy improves. This technique 
differs from the "progressive sampling" method used by Makarovic /8/, 
which densifies the points only in certain areas. At each grid level 
the search process described below can be used. Care must be taken to 
avoid gross errors at an early stage, since this will obviously effect 
the following stages. To assist in this, a visual: control is 
supported by the program. 
2.3 Search process 
In finding suitable points 4 steps in the search process have been 
identified. At the end of the search process it would be possible to 
interpolate points again in a regular grid. However, this technique 
is not used here and is not further discussed. 
= 565 = 
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.