Figure 4: Part of a SPOT XS scene, ground resolution 20m 
x 20m, 1200 x 1300 pixels, composite of the three bands. 
  
$7 3 Yol 
A (Gf : = 
© Ar A Ne S. 
  
  
  
  
Figure 5: Lines extracted from band 1 of the SPOT XS data 
(Fig. 4) using a window size of 15 x 15 pixels. 
92 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
  
  
  
Figure 6: Lines extracted from the SPOT XS data (Fig. 4) 
using a window size of 5 x 5 pixels. 
  
  
  
  
  
Figure 7: Result of selecting water pixels from all line pixels 
of Fig. 5 and road pixels from all line pixels of Fig. 6 by 
multispectral classification. The result has been vectorized 
using the sub-pixel positions of the lines and the methods 
mentioned in Section 3. The bold line represents the river. 
All other lines are roads. 
   
   
  
  
  
  
  
  
  
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
    
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
  
  
  
  
  
   
   
   
   
  
  
  
   
  
  
   
   
    
almost con 
roads insid 
The use of 
ing all kinc 
recognize r 
necessary t 
ing maps o 
it with the 
5 
The result 
allow to a 
images. T 
segmentat 
the width 
the image 
resolution 
and TM fe 
to a lot of 
means. C 
enables us 
other is tl 
which sho 
among a | 
by the exa 
Further we 
more knov 
autobahn 
railways a 
(Grün anc 
detection 
ment of tl 
important 
objects wi 
System. 
In this pa 
ages only. 
extraction 
1994a, 19 
pyramid v 
sults in su 
achieved « 
Busch, A 
hafter Ok 
dels, H. 
Berlin, pp 
Busch, A 
without u 
Eder, K. 
and Remc 
tional So: 
97. 
Canny, J. 
tion. IEE 
Intelligen: 
Deriche, 
IEEE Tra 
ligence 1: