3. TESTING THE ECONOMY OF PSA 
A comparison of different methods of data acquisition of DTM's 
underlines the efficiency of Progressive Sampling using PSA. A 1000 
by 1000 m square area of very rough terrain was measured using con- 
ventional methods - grids, profiles and contourlines- and using 
Progressive Sampling. The terrain was covered with images of the 
scale 1:14000, which were taken from a flying height of approxi- 
mately 2180 m with a distance between the perspective centres of 
about 1140 m. The height variations in the area were up to 160 m. 
A subset of all the results is now presented. It includes 
1, IE profiling with distance registration and scan width of 
m, 
2. a regular grid with 40 m gridwidth, 
3. a regular grid with 10 m gridwidth, 
4. the capturing of contourlines at an height interval of 5 m and 
5. Progressive Sampling starting with a 40 m rough triangle network 
in different approaches: 
- without prior information on the terrain, 
- with a lot of prior information (4446 points) and 
- with an optimal subset (1894 points) of morphologic information 
describing only the main terrain characteristics. 
The results are compared with respect to time and number of points 
for data capture and accuracy of the derived DTM. 
Morphologic information consisting of breaklines as they are ridges, 
scarps etc. and structure-lines were measured before the data cap- 
turing loop. Roads, rivers and railways were stored on different 
files as were dead areas, which can be included in the contour 
interpolation program CIP. In total 6655 points (nearly 4900 points 
on terrain lines and about 1600 points as dead area descriptors) 
were measured in 2 hours and 9 minutes as prior information. For the 
DTM- generation with CIP 4904 points of them are included. 
The following table presents a summary of results for that compari- 
son. The accuracy of the DTM is derived from 23 points which were 
distributed in the whole area and were measured with better accuracy. 
-]5 -