In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
335 
Figure 2. Results of the radargrammetric stereo processing for 
the summer acquisitions (A) and autumn acquisitions (B) 
Furthermore, the acquisitions acquired with an incidence angle 
of more than 56° were not used for DSM calculation due to 
strong ambiguities in the images. 
The visual inspection of the test results achieved with different 
disparity angle settings showed the following: 
• DSMs calculated with stereo pairs with a small 
disparity angle (9°) show strong artifacts and noise. 
• The artifacts and noise reduces with an increasing 
disparity angle. 
• The results achieved with a disparity angle of -20° 
showed visually the best results. 
• Areas with invalids caused by layover and shadow 
artifacts (no matching possible) increase with a larger 
disparity angle. 
Scenario 
LE 90 
[m] 
No 
Points 
Mean 
Ascending summer (20°), raw 
5.8 
634 
-2.4 
Descending summer (20°), raw 
6.4 
634 
-2.0 
Descending October (20°), raw 
6.7 
473 
-1.9 
Ascending October (9°), raw 
16.2 
458 
-4.2 
Ascending October (11°), raw 
8.9 
453 
-2.9 
Ascending October (20°), raw 
6.5 
452 
-2.6 
Asc / Desc Oct. (20°), raw 
5.1 
616 
-2.2 
Asc / Desc Oct. (20°) edited 
4.7 
616 
-2.2 
Asc / Desc Oct. (20°) edited (cal.) 
3.1 
616 
-0.2 
Table 1. Statistical analysis for test scenarios using different 
orbit directions and disparity angles (raw = unedited DSM, cal. 
= calibrated) 
In a next step, the statistical analysis was performed. The 
distribution of the DGPS points in combination with the 
evaluation result for the DSM calculated for the October scenes 
(disparity angle -20°, combination of ascending and descending 
orbit) is depicted in Figure 3. Table 1 shows the statistical 
analysis for some representative test results. 
In contrast to the visual impression, the statistical analysis show 
better results for the DSM calculated with the summer 
acquisitions compared to the DSM of the October acquisitions. 
A reason might be that the reference points were acquired in 
parallel to the summer acquisitions. 
However, the statistics confirm that the quality improves with a 
larger disparity angle. The accuracy could further be improved 
by combining of ascending and descending orbit direction, 
which also reduces the number of invalid pixels. Moreover, the 
editing (smoothing, outlier removal) improves the visual 
impression and the statistical result. In a last step, the DSM was 
calibrated, improving the statistical results even more. 
Difference 
Meter 
» -19,464--9,753 
» -9,752--4,309 
° -4,308 - -2,574 
•> -2.573--0.951 
° -0,950- 1,307 
» 1.308-4,650 
• 4.651-9,310 
• 9,311-17.630 
Figure 3. Evaluation result for the DSM (based on October 
scenes, disparity angle -20°, combination of ascending and 
descending orbit) 
2.3.3 Orthorectification: Based on the edited DSM from the 
October acquisitions, an orthorectification of the TerraSAR-X 
images was performed with Infoterra’s in-house 
orthorectification processor. For a part of the test site street 
vector data was available. The overlay of the street vectors and 
the orthorectified images showed a very good match of 1 to 2 
pixels (2.5 meter spacing). 
In the frame of the research project, the DSM and the 
orthorectified TerraSAR-X images were provided to RapidEye 
AG, where both data sets were used for co-registration and 
orthorectification of RapidEye data. The achieved geo-location