'HOTOGR AMMETRY 
D PHOTOCLINOMETRY 
  
^HOTOGR AMIMETRY 
2D PHOTOCL INOMETRY 
  
PHOTOGRAMMETRY 
2D PHOTOCL INOMETRY 
  
ige lines 
er, the use of patches 
in form of spurious 
he patch edges. So, 
cale topography, the 
y from two facts: 
across the considered 
| in the image data is 
s to build the terrain 
albedo features are 
ige texture). But for 
ruled out completely 
spreading of fresh 
ith varying albedo, 
ie illumination, such 
e of contrast. Under 
1s to be fulfilled. 
ous source of error. 
verify other model 
hotoclinometry, we 
ifferent illumination. 
> derived topography, 
1 the up-and-down 
1996 
  
  
Fig.4a. Region III :terrain models derived from 
photogrammetry (top) and 2D photoclinometry (bottom). 
sun direction. Indeed, we found evidence for albedo features in 
the shaded reliefs, e.g. at the rim of the large crater in Fig.3a. 
Second, the surface model used to start the iterations strongly 
effects the speed of convergence. Generally, small features 
converge before large ones, and the more the starting model is 
in accord with the real topography the faster convergence is 
reached. The iterations are stopped if the height residuals 
decrease below a threshold value. So, if the starting topography 
does not match the real topography on a large scale convergence 
is reached for small-scale features but not for large ones, 
allthough the residuals are small. Since the global shape model 
used as the starting surface model differs significantly from the 
large scale photogrammetric solution convergence was probably 
not reached for the large scale topography. This would explain 
why the large scale craters are much deeper in the 
photogrammetric models. Strong arguments for this 
explaination were obtained in a recent analysis, in which we 
used a smooth photogrammetric terrain as the starting model for 
photoclinometry. It turned out that the large scale topography 
of the starting model remained unchanged, i.e. large craters 
were now as deep as in photogrammetry, but the small scale 
features that could not be resolved by photogrammetry were 
added. This method looks very promissing and will be 
discussed in more detail in a forthcoming paper. 
     
  
12x1000 
11 
E 
5 10 an 
2 7 PHOTOGRAMMETRY 
— 20 PHOTOCL INOMETRY 
9 
line I 
line II 
14x 1000 
13 
m 12 
& 
= 11 
"m 10 - PHOTOGRAMMETRY 
AZ 
    
—— 2D PHOTOCLINOMETRY 
  
I T T T T 1 T T T T I 
0 50 100 150 200 
sample [pixels] 
1 
250 
Fig.4b. Region III: height profiles along image lines 
6. CONCLUSIONS 
There are significant differences in the terrain models of Ida 
derived from photogrammetry and from two-dimensional 
photoclinometry. These differences are due to the limitations of 
both methods and can be classified in terms of spatial scale. 
The photogrammetrically derived terrain models are expected to 
reliably show surface features with scale lengths larger than the 
patch size. Smaller-scale topography is not adequately 
resolved. Especially, surface features with very high spatial 
frequencies such as scarps, may result in topography gaps or 
even blunders. In contrast, the two-dimensional 
photoclinometric analysis can resolve the small-scale surface 
features with a reliability determined by model parameters such 
as albedo and photometric function parameters. The large-scale 
topography is determined by the large-scale properties of the 
starting surface model, that is, with this respect it is reliable 
only as the starting model is. 
Photogrammetry and two-dimensional photoclinometry in 
combination may give us terrain models much improved over 
those as of the two models alone. 
249 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
  
  
  
  
  
   
  
   
  
  
  
  
   
  
   
  
  
  
   
    
  
   
   
   
  
   
      
    
  
   
   
   
    
     
   
   
   
  
    
    
   
   
    
ETE TRS 
ant y 
rm TEE 
  
IRGEND M REI