International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
scattered failing area exist that can be interpolated well with 
close pixels. 
  
  
  
  
  
Number of check points (CPs) 97 
Mean of errors (absolute values) s 26.08 m 
Mean of errors 17.98 m 
Maximum error (absolute value) 108.00 m 
Standard deviation of errors 37:31 m 
  
dz>im n=92 94.8% 
dz>2m n=81 83.5% 
dz>3m n=75 77.3% 
dz>4m n=69 71.1% 
dz>5m n=60 61.9% 
dz>6m n=58 59.8% 
dz>7m n=52 53.6% 
dz»8m n=50 51.5% 
dz»9m n=44 45.4% 
dz>10m n=43 44.3% 
Frequency of errors 
  
  
  
  
Table 5. The statistics of DEM test with editing 
But comparing table 4 and table 5 doesn’t show this 
improvement because in the case of the whole extracted DEM 
the failing areas are very wide and numerous. Extraction of true 
heights for these areas from neighboring pixels is very hard and 
then some false heights are produced in these areas. Existence 
of these heights in edited DEM and thereby in selected points is 
the reason of large standard deviation in the whole edited DEM. 
In general above experiments show that for generated DEM, the 
RMSE of errors within 95% confidence level was found nearly 
equal to 4.89 meters. Naturally this value for RMSE is produced 
in areas that image matching is executed well and forest areas 
have a little extent. In images with large forest areas where the 
image matching fails to find the corresponding pixels in large 
areas, interpolation is not a good way for DEM editing and 
RMSE is not a satisfied result. 
5. CONCLUSION 
Creating Digital Elevation Model by digitizing contour lines 
from topographic maps or through stereoscopic semi automated 
methods from aerial photographs are proven methods. However, 
DEM generation from satellite stereo images is still not a 
common practice. The DEM generated from satellite stereo 
pairs have some significant advantages over other sources, viz: 
1. World wide availability of satellite data without any 
restriction (often available as archived data) as against restricted 
and non availability of topographical maps and aerial 
photographs 
2. Large area coverage per scene 
3. Moderately high resolution 
4. Faster processing through sophisticated software and little 
manual effort 
5. Low processing cost 
Comparing some other studies reveal that of the different image 
pairs processed, this SPOT HRS images yield the better results. 
Kim and Kang produced a DEM from SPOT panchromatic 
stereo images with RMSE equal to 12.4 to 14.4 m (Kim, 2001). 
Also with KOMPSAT images with 6.6 m resolution the DEM 
extracted was produced with RMSE equal to 8-13 m (Kim, 
2001). In the same area using SPOT images with 10 m 
392 
resolution one DEM with 32 m RMSE was extracted (Kim, 
2001). Comparing the ability of DEM extraction from some 
images showed that using ASTER, Radarsat, IRS-1c and SPOT 
images with 15, 12.5, 6 and 10 m resolution RMSE equal to 30, 
65, 35 and 32 m were produced (Siva Subramanian , 2003). 
There are some reasons for these results. The results of SPOT-5 
are better because the temporal difference between the stereo 
pair is only a few seconds, whereas some pairs acquired through 
other satellites require a larger time cycle. Results also show 
that matching is good in case of the images with this condition 
due to short temporal differences. On the other hand the proper 
b/h ratio yields more consistent and better results over various 
terrains. The other reason is better resolution of these images in 
comparing with the others. 
This study shows that SPOT-5 HRS stereo images have a good 
potential for DEM generation. 
References 
Kim, S., Kang, S., 2001. Automatic genera'tion of a SPOT 
DEM: Towards coastal disaster monitoring. Korean Journal of 
Remote Sensing, 17 (2), pp. 121-129. 
Reinartz, P., Lehner, M., Müller, R., Rentsch, M., Schroeder, 
M., 2003. First Results on Accuracy Analysis for DEM and 
Orthoimages Derived from SPOT HRS Stereo Data over 
Bavaria. www ipi.uni-hannover.de/html/publikationen/ 
2003/workshop/baudoin.pdf. 
Siva Subramanian, K.S., Singh, A., Sudhakar, M., 2003. 
Evaluation of Digital Elevation Models Created from Different 
SatelliteImages. 
www.gisdevelopment.net/technology/rs/mi03079.htm. 
Toutin, T., 1995. Generating DEM from Stereo-Images with a 
Photogrammetric Approach: Examples with VIR and SAR data, 
EARSeL Advances in Remote Sensing. 4(2), pp. 110-117. 
Marra, M.. Maurice, K., Ghiglia, D., Frick, H., 2001. 
Automated DEM Extraction Using RADARSAT ScanSAR 
Stereo Data. 
www.gisdevelopment.net/technology/rs/mi03079pf.htm. 
Kim, T., Im, Y., 2001. Automated DEM extraction from the 
KOMPSAT-1 EOC images, 
www.crisp.nus.edu.sg/-acrs2001/pdf/291Kim.pdf.