International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
2.6 HRG/HRS DEM 
H. Raggam is the only investigator who produced DEM from 
HRG images. He computed a stereopair with the HRG image 
and the HRS1 image. The tests we've done during the in 
flight commission phase show better results when using 
HRS2 image mainly because of its better FTM (Rudowski, 
2003) 
Nevertheless we can see several interesting things in those 
results : 
- there is a little altimetric bias with reference but up today 
we didn't receive H. Raggam's report. The noise of those 
DEM might be the main cause. 
- The standard deviation is better on area 4 because there is 
less invalid area in this DEM, lower B/H ratios are often 
better solutions to produce DEM in high relief areas. 
- The standard deviation is a little bit higher for the two 
other areas, but if we look at the DEM we can see that 
most of the time it can be a real noise (buildings or 
vegetation). 
   
figure 10: example DEM profiles 
black = reference, pink = HRS DEM, blue = HRG/HRS DEM 
2.7 Comparative results: 
All the differences have been done for each DEM but to 
compare them we decided to compute statistics on the 
rigorous overlap. We have then 3 areas on sites 2 (150 — 
500m) which is of moderate relief, site 4 (60 — 1200m) of 
significant relief and site 6 (0 - 200m) which is a flat area. 
(Fia 11, 12,13). 
   
2 
rence 
Fig 11: refe 
  
p) 
Fig13 : Reference on site 6 (dark Zone not in the overlap) 
The next results reflect only the 3 overlapping areas 
(intersection between areas 2, 4 and 6 from each provider) : 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Area 2 Area 4 Area 6 
W. Kornus Min -8 -20 -6 
Spacing:10m Max 15 33 27 
Mean 0,8 1.2 10,9 
St. dev. 3,8 6,6 7,1 
P. Reinartz Min 1 -10 4 
Spacing:15m Max 23 37 32 
Mean 9,2 10,2 19,5 
St. dev. 3,8 7,3 6,5 
H. Raggam Min -11 -126 -9 
Spacing:10m | Max 16 105 24 
Mean 1 1.3 8,5 
St. dev. 4,5 21,6 7,8 
  
  
Table 14: Results on overlapping areas. 
On those areas, we can mention several points: 
- P. Reinartz DEM does have an altitude bias with 
reference, nearly 10 m for areas 2 and 4 and 20m for area 
6. 
- . W. Kornus and H. Raggam DEM don’t have any bias in 
those areas but have one (nearly 10m) in area6. The bias 
of altitude is higher for each provider in area6 but also 
the standard deviation; this can be explained by the fact 
that this area is mainly a city area, the space DEM 
describe more or less the top of the buildings and the 
reference is a real ground reference. 
- Area 4 is a quite hilly area and the standard deviation for 
W. Kornus and P. Reinartz is only around 7m. 
- If standard deviation is so high for H. Raggam in area4, 
it’s because we can clearly see in this DEM some 
correlation defects which have not been taken off the 
statistics, not being declared as invalid areas, probably for 
lack of time. 
F 
  
  
  
   
Inte 
2.8 
The 
pro 
exat 
Rei 
rep«