International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
  
explanations for that (this might come from a transcription 
error in the reading of the DEM). This DEM seems to be 
quite noisy but if we look at the profile we can see as well a 
little planimetric delocalization. Those problems could 
probably be explained by modelisation errors (Kaczynski, 
2004). 
   
ig 31: Above: Profile in R. Kaczynski DEM (purple line) 
Below: green = R. Kaczynski DEM; black = ref N_50 
370< reference height<470 height exaggeration = 4 
3.4.5 A. Suchkov DEM analysis shows, from statistics, that 
A. Suchkov DEM is really very close to the reference. It is a 
bit noisier than the reference but the sampling is not the 
same. The difference image (Fig 32) is really light, that 
means close to zero and the profile is also comparable to the 
reference. 
  
  
    
A RIT i RT EEE ES Mea DE. £x 
Fig 32. Difference : A. Suchkov DEM — refN 50 
  
Fig 33: Above: Profile in A. Suchkov DEM (purple line) 
  
Below: pink = A; Suchkov DEM; black = ref N_50 
370< reference height<470 height exaggeration = 4 
The North area is a flat area, the reference sampling is lower 
than the produced DEM but with a good accuracy (roughly 
2m). There are some really good results, particularly the Dem 
produced by A. Suchkov which has absolutely no bias with 
the reference and a standard deviation lower than 4 meters. 
3.5 Results on South areas with 25 m sampling 
3.5.1 Two contiguous areas have been studied ( Fig 34) : 
-  S_25 1 (620 < height < 1340) 1,3 km x 10 km (North) 
- S25 2 (610 « height « 1680) 8,7 km x 10 km (South) 
EET Y EE m. o*  — 7 ii 
WET CF 
  
    
       
  
  
    
  
    
  
   
ory s 
- Fig34: References S 25 1 (upper) and S 25 2 (lower) 
     
    
  
  
St. No 
min max |mean| Dev. match. 
  
  
P. Reinartz -82 58 101] 18,7 | 0% 
  
K.Jacobsen] -51 38 12,2 | 12,9 17,1% 
  
S_25 1 JR. Kaczynskif -22 43 [3135| 10.9 |1,2% 
  
D. Poli 1 -12 24 6,7 | 6,4 | 0% 
  
D. Poli 2 -9 20 $81 87 109 
  
_P. Reinartz | -19] 58 8,5 | 24,4 | 0% 
  
K. Jacobsen] -100 44 8,8 | 15,1 19,9% 
  
S 25 2]R. Kaczynski]. -158 168 |210,3| 26,6 | 594 
  
D. Poli | „26 31 3.0 7,9 |. 0% 
  
  
  
D. Poli 2 „20 28 4,4 | 6,9 | 0% 
  
  
  
  
  
  
  
  
Table 35: Accuracy and matching quality on S_25 areas 
Min and max values are relative values for R. Kaczynski 
results. 
The "no match." Column gives the percentage of declared no 
correlation areas. P. Reinartz and D. Poli DEMs are complete 
DEM, that means that even if there were bad matching areas, 
their DEM production proçess filled the holes, the filling 
processes are always g "hetter than nothing" solution, it can 
      
    
      
     
   
   
   
   
  
    
     
   
    
   
   
  
   
   
     
    
    
       
   
     
   
      
     
     
    
     
  
Intei 
be p 
stati: 
don' 
can’ 
Tho: 
beca 
sign 
decl: 
any 
Rein 
prob 
Ever 
in R 
corri 
the I 
Exat 
  
This 
and 
proc 
are | 
matc 
à S 
The . 
the s 
and t