anbul 2004 
etween 
) 
atr, c 
;. unknown 
> two shift 
tion of the 
ve as an 
on the next 
he MOLA 
application 
       
  
interior 
ientation 
  
  
  
described. 
presented 
ery of the 
have been 
s mission. 
he exterior 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
orientation and the calibration data of the interior orientation as 
well as the MOLA DTM are used as input for the DIM. 
The CCD arrays of the HRSC consist of 5176 active pixels 
each, which yields a swath width of about 65 km on the surface 
of Mars. The strips can have a length of up to 300.000 lines, 
spanning about 4.000 km on the surface. Due to a limited 
bandwidth between Mars and Earth only the nadir channel is 
able to operate at full resolution. Generally the resolution of the 
two stereo channels has to be reduced by a factor of 2 and the 
remaining channels by a factor of 4. To obtain an equivalent 
scale the nadir channel has to be resampled to the resolution of 
the stereo channels for the matching. Depending on the covered 
region on Mars the imagery shows areas with high texture and 
areas with hardly any texture (Figure 4). 
  
Figure 4: Left: Part of orbit 68 with good texture. Right: Part of 
orbit 68 with very low texture 
The matching was carried out on the original level-2 data 
(Roatsch, 2000) which means no geometric corrections have 
been applied to the imagery prior to matching. Since the 
minimal flying altitude of the spacecraft is about 300 km above 
the surface the spacecraft isn't exposed to turbulences which 
aerial cameras here on earth experience. Thus the spacecraft 
moves with hardly any disturbances along its trajectory 
resulting in imagery without jitter. 
In orbit 22 the two photometry channels were also available at 
the same resolution as the stereo channels and were used for the 
matching resulting together with the nadir channel in five 
overlapping strips. 
32 Results 
In a first evaluation the distribution of the tie points in the 
image strip is evaluated. In Figure 5 it is noticeable that the tie 
points are evenly distributed over the whole image strip. Most 
of them are 3-fold points. There are some areas visible which 
don't contain any point. This doesn't pose a problem for the 
determination of the absolute orientation because the used 
approach (Ebner et al., 2004) shifts the trajectory on the whole 
by applying six biases, one for every element of the exterior 
orientation. Even in orbit 68, with large empty areas due to low 
texture (Figure 4), we were able to determine the exterior 
orientation of the camera. Problems raise at determining a DTM 
in those areas using image matching methods. 
849 
A 
+ 
4 
+ 
ke to 
; 
+ 
1 
(ode 
Ath 
3 
NS 
SE 
x t 
++ 
"E 
  
  
ET 
MET +3 
y sh. +, "m 
our ++ 
et irm. 
sp 
     
      
     
t1 
E 
Gi 
+ T 
£* 
it M 
amat 
t + a 
A 
+ 
9 
zt 
tH JA 
T 
= 
a 
qot 4 
t 
I" 
  
Figure 5: Distribution of the tie points in orbit 18. 
+ 2-fold, 
agé v ss, 
3-fold point, north is up