à part 
> rest 
ordi- 
.. Aft- 
d the 
The 
ar to 
urse, 
ution 
100 
nd- 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
nr. of points 
Nr. type of zone 1 zone 2 zone 3 all zones with height 
of points rms nr. of rms — nr. of rms. :nr. of rms nr. of > < 
GCPs diff. points diff. points diff. points diff. points Survey height 
99 all points 9.8 4030 8.3 5524 10.4 3017 9.3 12571 7995 | 4576 
points with 7.4 3754 7.4 5386 8.5 2835 p 11975 7479 | 4496 
diff. « 20m 
40 all points 11.1 4030 9.2 5524 10.7 3017 10.2 12571 7715 | 4856 
points with 8.0 3678 8.2 5357 8.6 2816 8.2 11851 7135 | 4716 
diff. « 20m 
30 all points 11.4 4030 9.1 5524 10.7 3017 10.3 12571 7247 | 5324 
points with 8.4 3642 8.1 5363 8.7 2821 8.3 11826 6689 | 5137 
diff. < 20m 
20 all points 12.8 4030 10.4 5524 11.7 3017 11.5 12571 7458 | 5113 
points with 8.8 3545 8.9 5233 9.2 2760 9.0 11538 6611 | 4927 
diff. « 20m 
10 all points 15.0 4030 12.1 5524 13.0 3017 13.3 12571 6237 | 6334 
points with 9.9 3306 9.9 5030 9.6 2639 99 10975 5456 | 5519 
diff. « 20m 
4 all points 15.2 4030 13.8 5524 14.2 3017 14.4 12571 8358 | 4213 
points with 9.8 3360 9.6 4634 9.9 2529 9:8 10523 6525 | 3998 
diff. < 20m 
Table 4. rms differences of computed DEM and Survey DEM heights; rms values are given for zones 1,3 of twofold and zone 2 of threefold 
stereoscopic coverage for all points and for points with absolute value of the height difference less than 20m 
  
  
* forested areas, trees and built-up areas: 79 % 
* other areas (fields,meadows): 12 % 
® image matching errors: 9 % 
The percentage of matching errors is higher than in the 
whole set of conjugate points where it was found to be 
about 4% by manual check. Thus, we believe that the 
accuracy of the computed DEM will be better than 8m. 
Table 4 and Figure 10 in graphical representation for 
zone 2 show that by varying the number of ground 
control points a saturation effect can be noticed, here 
too for numbers beyond 30. 
144 
124 
104 
rms differences 
  
-i- diff « 20m 
  
-€- all points 
  
  
  
  
0 T T T T 
  
T T I T 
T 
0 10 .20..30 40 350 60 70 80 90 100 
number of GCPs 
Figure 10. rms-differences to Survey DEM for zone 2 and 
various numbers of ground control points 
75 
5. Conclusions 
Software for the derivation of digital elevation models 
from 3-line scanner imagery has been tested success- 
fully on airborne data of the MEOSS camera. The the- 
oretical standard deviations are in the range of 1-2 pix- 
els and the heights compare well with the Survey DEM. 
The test was performed under the heavy demands of 
high resolution of the airborne imagery and rough atti- 
tude behaviour of the aircraft. Thus, one can be opti- 
mistic for the satellite data of MEOSS and MOMS mis- 
sions. 
Automation of the start of matching in the image pyra- 
mid is one of the next goals.