)04 
the 
rm 
nay 
noi 
| is 
our 
hin 
the 
our 
our 
ige, 
ult, 
t of 
her 
ick 
Wn 
e Is 
tain 
rust 
dge 
rust 
are 
" in 
ours 
inear 
; the 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
nearest contours. This approach is selected because it is rather 
robust, comparing to the option of calculating these heights 
from DTM (calculated bicubic surface patches). Extracted line 
segments are connected, smoothed, and included into TIN. 
New surface normals and surface patches are also calculated 
and this is final DTM based on contours. From Figures 7 and 8 
it is obvious that these elements contribute significantly to the 
morfological quality of the DTM. 
  
Figure 7. Shaded terrain without automatically extracted 
geomorfological elements 
    
s 
Figure 8. Shaded 
terrain with 
geomorfological elements included in DTM 
automatically extracted 
The final objective is to obtain geomorfological elements that 
would enable DTM generalization and also data reduction. 
Namely, it should be possible to build new DTM by 
interpolating semi-regular (similar to progressive and 
composite sampling) grid of heights from final contour DTM 
and to supplement it with extracted terrain forms, without 
significant loss of DTM quality. It is to expect large data 
reduction. Results show that the quality of extracted terrain 
forms is quite sufficient for contour DTM improvement, but it 
has to be enhanced if one wants to get sound basis for DTM 
data reduction and generalization. Better connection of 
extracted line segments into longer strings is desirable. Also, 
classification of these forms regarding their significance for 
DTM quality is required. 
655 
7. CURRENT PROJECT STATUS 
All the map layers with contours are vectorized and corrected 
on map sheet edges. Almost all map layers with spot heights 
are also digitized. Digitization of layers with hydrography is in 
progress. It is still open question how to integrate these data 
within DTM in the best way. 
Verification of contour data is in progress. This also includes 
some initial DTM quality assessment. Available data for this 
are provided by orthophoto projects. These are based on aerial 
photogrammetry and mostly for 1:5000 mapping scale. As it 
can be seen form Figure 9, there is a large area (Belgrade is 
included) covered with these data. It is to expect that there are 
going to be more data available from unfinished (currently in 
progress) and future orthophoto projects and from other sources 
as well for contour DTM quality assessment. 
  
Figure 9. Available 1:5000 DTM for quality assessment of 
1:25000 DTM (Serbia without Montenegro) 
Few initial tests have shown that the accuracy of acquired 
contour data DTM is in the expected range. For flat terrains 
RMSE is about 1-2m, while there is significant loss of 
accuracy for hilly terrains where RMSE is ranging from 5 to 
10m. This decrease in accuracy is mostly related to errors in 
photogrammetric measurements for wooded areas. However, 
further, more extensive testing is required in order to verify 
these preliminary estimates. It is to expect realization of these 
test in the next few months. 
8. CONCLUSIONS 
Even though it could be argued that building country wide 
DTM using existing maps is obsolete method, it is still 
justified in cases of limited resources. Efforts spent in the past 
on capturing these data are enough reason by itself. Having in 
mind meticulous procedures used for making these maps it is 
natural to expect that DTM made by using such data should