The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
139 
(a) TerraSAR-X image 
(b) Digital Map 
Figure 5. The example of the validation point used in this study. 
Table 11 shows the pixel location accuracy. The accuracy for X 
is worse than 10 m, while that for Y about 5 m. It is noted that 
the errors include the errors of extracting intersections from 
SAR images due to the difficult interpretation. However, the 
fact that the geometric errors of range (X) direction are larger 
for both sites implies that the quality of the used DEM affects 
the pixel location accuracy of EEC products. 
The DEM elevation errors were estimated using the simulation 
results of table 4. About 10 m of pixel location accuracy 
corresponds to 6 m of elevation errors in case of 32 degree of 
the incidence angle. This value includes the effects of the low- 
spatial resolution of referenced DEM as well as the actual 
elevation error. 
Figure 6 shows the comparison between the pixel location 
accuracy for X and the elevation at Kagawa site. The pixel 
location accuracy is partly related with the elevation (r = 0.13, 
without the point with 350 m of elevation). Further analysis is 
required to better understand the exact dependencies on the 
pixel location accuracy in mountainous areas. 
Site 
Toy a 
Kagawa 
Number of GCPs 
20 
30 
dX (m) 
-1.0± 12.9 
-11.0± 12.8 
dY (m) 
-1.1 ±4.7 
2.4±4.6 
dXY (m) 
9.8±8.7 
12.6± 12.3 
Table 11. Pixel location accuracy evaluated by digital map 
2500. 
Figure 6. Comparison between the geometric accuracy and the 
elevation. 
4. CONCLUSIONS 
TerraS AR-X satellites of high-spatial resolution images are 
being investigated recently for preparing topographic maps. 
One of the critical points for SAR system is geometric accuracy 
of the products as it is a side-looking system and the geometric 
accuracy of the ortho-products is strongly influenced by the 
accuracy of the DEM used for production. 
We focused on the applicability of sophisticated high-spatial 
resolution TerraS AR-X data for our objectives, and the study 
evaluated the geometric accuracy of TerraS AR-X by the 
following two methods, using comer reflector (CR) and by 
comparison with a digital map. 
The geometric accuracy of SSC products was firstly estimated 
from comparing the positions of the CR on TerraS AR-X image 
and in-situ data. It showed that the accuracy is better than 1 m. 
Then the accuracy of the orthorectified EEC products was 
evaluated. It revealed that the geometric accuracy was better 
than about 5 m in the flat area. Secondly, the accuracy was 
evaluated using a digital map at a scale of 2500 in order to 
evaluate the accuracy in mountainous areas. The measured 
accuracy degraded to more than 10 m due to the poor DEM 
quality used for the orthorectification. 
Further analysis is necessary for more accurate evaluation 
because the validation points were not enough. However, the 
fact that these results show the same tendency revealing the 
validity of the two methods as well as the reliability of the 
estimated accuracy.