GEN m 
  
  
  
  
5.0 -. Flight altitude h 4.0 km 
- Strip length 18.6 km 
Strip width 3.2 km 
Update point interval DOR 200.0 m 
DEM-X-interval XIV 40.0 m 
DEM-Y-interval YIV 640.0 m 
E Focal length c 100.0 mm 
4.0 — Image error + 5.0 um 
2 
ZGEN 
ne | IN XGEN 
3.04 
ne \ 
N \ \ 
\ \ 
\ \ 
X LJ 
2.0 \ A 
h 4.0 km \ \ 
18.6 km a 
3.2 km . 
interval 0.2 km \ YGEN \ 
> 100.0 mm \ 
Yo + 5.0 um 1.04 À N S 
x = . "— NÀ = 
— se, - — eu 
M———M — === EA on 
1 T | i 1 
2.5 5.0 7.8 10.0 12.5 
Sensor angle — o gon 
Y 
\ Figure 8. Over-all model accuracy in dependence of the sensor angle 
\ / 
Pul 
e of the strip In Fig. 9 the mutual dependency and influence of the intervals DOR be- 
T pet tween the update points Ps the intervals XIV of the DEM points P in 
tipeoordinete X flight direction (X-direction) and the sensor angle a are represented. 
A B This figure reveals the following results: 
a. In each case angular arrangement of the sensor lines A and C 
improves the accuracy of the strip model 
b. Smaller XIV-intervals between the DEM points and greater DOR-inter- 
vals between the update points P. increase the accuracy, provided 
that the image errors GC“ are asdumed constant. But really image 
errors cannot be constan? at increasing DOR intervals, because of 
increasing interpolation errors (see Fig. 6). 
As the DOR-interval determines the bandwidth of the normal equation 
system and this width influences the computation time with power of 
two, in practice a compromise between the required accuracy and the 
computation time has to be found. 
Of greatest interest is the error propagation along the strip. In 
Fig. 10 the mean square height error ZGEN of strips are represented 
3 in dependence on the length of the strips which are absolutely oriented 
by four control points in the corners of the strips. 
=. 353. -