Và KV Twp Ka DR a a a AO PN 
rates may be dropped (linear model). Six weighted 
constraints keep the orbital positions of the two sensors 
within statistical limits from the expected nominal orbits 
or the ephemeris data. Twelve absolute constraints are 
enforced in order to keep projection centres moving 
strictly along appropriate elliptical orbital segments. 
Thus, linear and quadratic rates of change in X, Y, Z, 
which are eventually used as given values, are calculated. 
The minimum number of full control points is 4 for the 
linear model of attitude changes and 6 for the quadratic. 
3. TEST DATA 
A stereo SPOT panchromatic level 1A model over W. 
Switzerland was acquired (Figure 1). The inclination of 
the sensor’s optical axis was 23.4° R and 19.2° L 
respectively, leading to a B/H ratio of ca. 0.8. The 
acquisition dates were 20.7.1988 and 27.8.1988 with 
significant radiometric differences between the two 
images, particularly in agricultural areas. The elevation 
range was 350 - 3000 m. The following preprocessing 
was applied to the original digital images: 
e reduction of periodic and chess pattern noise 
e Wallis filtering for contrast enhancement 
e Wallis filtering for edge enhancement (optional) 
136 points of good to very poor definition covering the 
whole image format and with a height range of 350 - 
N 
2100 m were used as control and check points. They were 
measured with an accuracy of approx. 5 m in 1:25,000 
topographic maps. Their pixel coordinates were 
measured in one image manually, and in the second (a) 
manually, and (b) with least squares template matching 
using the first image as reference. The following versions 
were computed with Kratky's model: 
(a) linear versus quadratic rates of change for the rota- 
tion angles 
(b) 6, 10, and 30 control points 
(c) pixel coordinates of second image determined 
manually versus through matching 
The results (without exclusion of any point) are listed in 
Table 1. 
The linear and quadratic models give similar results. This 
was the case in previous investigations too. The quadratic 
model gives slightly better results in Y, but worse results 
in Z with 6 control points. There is a big improvement in 
Z from 6 to 10 control points but no improvement from 
10 to 30 points. There is an large improvement in Z when 
the pixel coordinates in the second image are measured 
by matching. For the 10 control point version, the manual 
measurements lead to a by 85% worse RMS in Z as 
compared to the matching measurements! Version linear, 
10 control points, matching measurements was selected 
for the further work. 
NEUCHÂTEL 
jum 
eu, FRANCE 
GENEVE | 
   
   
  
     
     
  
FRIBOURG 
A 
20 km 
E 
À control point 
  
Figure 1 SPOT stereo model with 10 control points. The test region for 
orthophoto generation (map sheet 1225) is also marked. 
359