Istanbul 2004 
ir images for 
consists of 
x technique for 
atching points 
ge space, and 
> interpolated 
erated DTMs, 
both versions 
ate whether an 
the obtained 
rent strategies 
t distinguishes 
meters that can 
¢ default value 
n this test. An 
io refine the 
1 visual quality 
| are visualised 
yellow, green) 
1e more or less 
the unreliable 
led description 
Supresoft Inc., 
pcess still have 
nesh size using 
et. 
) 3.1 compared 
uns the results 
wo 3.1 vs. the 
fin. — Max. 
.2m — 18.4m 
. 55 — 6.7m 
„Sm = 5.5m 
.9m - 11.9m 
0.8m — 8.3m 
2m — 4.3m 
mpared to the 
Ain. — Max. 
.9m — 16.8m 
6m — 7.7m 
1m — 4.2m 
3m — 18.6m 
0.7m — 7.4m 
2m — 4.3m 
measurements 
r than the ones 
ranges become 
automatically 
homogeneously 
between 0.65m 
difference: the 
num-maximum 
ompared to the 
unders. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
In model 116 117, like in the DTM generated with Image 
Station, the most blunders occur on the steep slopes, while the 
height error distribution in the flat areas are very similar for 
both systems. 
Unlike the smoothing effect occuring in the Image Station 
datasets, both Virtuozo software versions produce very rough 
surfaces (figure 6). A loss of geomorphological information as 
appearing in the Image Station data (figure 5) is not obvious. 
The results of both tested versions of Virtuozo achieved similar 
results, except for the models 116 117 and 212 213. In both 
cases, version 3.3 achieved a smaller standard deviation and 
minimum-maximum range. Generally, the values for AZ show 
that Virtuozo 3.3 tends to measure slightly higher than Virtuozo 
3.1 but seems to be less affected by blunders especially for the 
two problematic models 116 117 and 212 213. 
Figure 6: Manually derived DTM (left) and DTM generated by 
Virtuozo 3.3 (right) 
3.4 Comparison of the systems 
Figures 5 and 6 visualise the general difference between the 
acquired DTMs on Image Station and Virtuozo: On the one 
hand, a strong smoothing effect in the Image Station data and 
on the other hand the rough surface produced by Virtuozo 
3.1/3.3. 
The accuracies resulting from the comparison of the acquired 
DTMs (tables 1,2 and 3) show that in average the Virtuozo 
System has mean height errors close to Om with standard 
deviations mostly between about 0.7 and Im while the data 
produced on Image Station differs with a larger variance 
between about 0.7m and l.6m in the models covering flat 
terrain. Both systems show problems in model 116 117, where 
Virtuozo exceeds standard deviations of 2.5m and Image 
Station obtains 3.1m. Furthermore, a strong smoothing effect 
causes a loss of geomorphological information especially in 
rough terrain (compare figure 5) The differences between 
manually measured and automatically derived data are caused 
by the steep slopes in this model, as shown in figure 7. 
Concerning the flat areas, both systems achieved throughout 
good results wihtout big blunders or gaps. Contour lines from a 
subset of model 116. 117, the model for which the worst results 
in terms of standard deviations were obtained due to the slopes, 
are comparatively illustrated (figure 7). 
  
  
  
  
   
  
  
  
  
   
     
Figure 7: Im contour lines: Wild S9, Image Station and 
Virtuozo 3.3 from top to bottom (Model 116 117). 
Compared to the contour lines derived from the manually 
measured DTM, the contour lines based on the Virtuozo 
measurements are more detailed regarding geomorphological 
information while the contour lines based on Image Station 
measurements are too smooth. 
A comparison of the contour lines for the model with the best 
result (223 224), derived from Image Station and analytical 
plotter, with the difference grid shows that the errors of Image 
Station data result from the smoothing applied during DTM