International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 
  
  
18.0 - 
16.0 
14.0 
12:0 
10.0 
8.0 
6.0 
4.0 
2.0 
0.0 
   
Le 3 
d 
em. Bat 9.2 
- 
44.8 
DSM accuracy [m] (1c) 
   
HRS/HRG HRS 
Region Growing Image 
Matching (10m grid) 
5.2 
   
  
    
    
HRS/HRG 
ISAE (45m grid) 
    
   
   
  
  
mountainous 
moderate/flat 
     
     
HRS 
      
Terrain Type 
Figure 9: Summarized standard deviations of height differences between the produced DSMs and reference DTM 
Figure 9 gives a summary of the obtained standard deviations 
of the height differences between the produced DSMs and 
the reference DTM. As can be seen, they depend on the 
terrain type, on the number of employed viewing directions 
(cameras) and on the used method, which leads to the 
following 3 simple statements: 1. The DSM accuracy in 
mountainous terrain is lower than in moderate and flat 
terrain, which is obvious due to the higher probability of 
occlusions and due to the higher impact of horizontal errors. 
In addition, homogenous image patterns, e.g. in forest areas, 
which also obstruct the matching process, produce gaps in 
the point cloud and later in the DSM. Here is the biggest 
potential of accuracy improvement by manual interaction. 2. 
Three viewing directions (HRS/HRG) are better than two 
(HRS only). Although the nadir viewing camera HRG does 
not geometrically contribute to a better height accuracy, its 
presence, however, supports the accuracy and reliability of 
the matching process, especially in mountainous regions, 
where it also helps to bridge occlusions. Nevertheless, HRG 
imagery, if available at all, does not cover the whole HRS 
scene and therefore it is not always possible to use a three 
viewing (HRS/HRG) approach. 3. The DSM generated with 
region growing image matching are more accurate than the 
ISAE-DSM, which probably is at least partly due to the 
different grid spacing. The actual reasons have not been 
analysed in this study. 
The study demonstrates, that DSM production using SPOT-5 
data is possible with an absolute accuracy of better than 5 m 
(lo). In mountainous areas the accuracy is worse due to 
occlusions obstructing the automated mass point generation 
process, especially if no nadir viewing HRG imagery is 
available. The presented results still include all errors of the 
automatic matching process and also the difference between 
the produced surface model and reference terrain model. 
Therefore it is expected, that the accuracy values still can be 
considerably improved by manual editing and appropriate 
filtering, filling the gaps in the automatically generated point 
cloud and excluding blunders and points on top of vegetation 
or artificial objects. 
ACKNOWLEDGEMENT 
We would like to express our gratitude to Assumpció 
Térmens, who implemented the functional model into our 
bundle adjustment program and to Cristina Ruiz, who 
measured the control points. Our sincere thanks also go to the 
Institute of Photogrammetry and Geoinformation at the 
University of Hanover (Prof. C. Heipke) for leaving us the 
region growing matching software and to Rupert Miiller of 
the Remote sensing Technology Institute of the German 
Aerospace Center DLR, who provided us with a software 
tool to read the SPOTS ancillary data. 
REFERENCES 
Alamüs R., Kresse W., Langner M., 2000: "Accuracy 
potential of point measurements in MOMS-images using a 
rigorous model and a rational function". International 
Archives of Photogrammetry and Remote Sensing, Vol. 33, 
B4, pp. 515-517, Amsterdam, The Netherlands. 
Colomina, I, Navarro, J., Térmens, A.,1992: “GeoTeX: a 
general point determination system”, International 
Archives of Photogrammetry and Remote Sensing, Vol. 29, 
Comm. III, pp. 656--664, Washington D.C, USA. 
Ebner H., Kornus W., Ohlhof T., 1992: „A simulation study 
on point determination for the MOMS-02/D2 space project 
using an extended functional model", International 
Archives of Photogrammetry and Remote Sensing, Vol. 29, 
B4, pp. 458-464, Washington D.C, USA. 
Fratter C., Moulin M., Ruiz H., Charvet P., Zobler D., 
2001:“The SPOT-5 Mission”, 52" International 
Astronautical Congress, Toulouse, France. 
Heipke c Kornus W., 1991: “Nonsemantic 
photogrammetric processing of digital imagery — the 
example of SPOT stereo scenes” in: Ebner, Fritsch, Heipke 
(Eds.): Digital Photogrammetric Systems, ISBN 3-87907- 
234-5, Wichmann Verlag Karlsruhe, Germany, pp. 86-102. 
Krzystek P., 1991: “Fully Automatic Measurement of Digital 
Elevation Models”, Proceedings of the 43th 
Photogrammetric Week, Stuttgart, pp. 203-214. 
Otto G., Chau T., 1989: “Region growing algorithm for 
matching of terrain images”, Image and vision computing 
(7) 2, pp. 83-94. 
SPOT Image, 2002: “SPOT Satellite Geometry Handbook, S- 
NT-73 12-SI, Edition 1, Revision 0", 15. 01. 2002 
SPOT Magazine No. 31, 2000: “The secrets of SPOT-5 
Supermode”, Sept. 2000. 
Vol XXXV, Part Bl. Istanbul 2004 
  
  
        
    
   
     
    
    
    
   
      
      
   
    
    
  
   
    
    
    
  
  
  
     
  
   
    
  
      
  
   
   
    
    
   
  
   
KEY 
ABS" 
As pa 
Melbc 
Geom. 
planin 
The li 
the up 
identi 
and) h 
For M 
in Cla 
possil 
an ac 
impro 
in Sap 
This r 
Centre 
the SI 
Montt 
precis 
elevat 
house 
comm 
gener: 
ERD/ 
1.1 SI 
the s 
launcl 
identi 
instru 
panch 
veget: 
globa 
stereo 
terrai 
HRS 
inclin 
track. 
opera 
pixel 
samp 
every 
to inc