International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 Inte; 
  
     
  
  
  
Afte 
3.1 DSM editing It's worth to notice that while DTM represents the real com 
morphology of the terrain, the satellite derived DSM represents poss 
On the whole, and at first look, the DSM appears to be the surfaces of features in the scene, as the bare soil in non- histc 
satisfactory over the areas where the correlation has been vegetated areas, the tree canopies in wooden areas and the distr 
successful. The use of a large number of tie points helps in building obstructions in urban environment. In figure 6 a series ther 
getting better oriented epipolar images and improves the of transects are reported in order to show discrepancies between 
possibility of a good image matching and, hence, allows to the cartographic DEM and the DSM under evaluation. Plots in 
obtain a good DSM with few failure areas. Notwithstanding, a the following figure 7 shows the differences along the transects : 
little lack of data where the matching has failed, particularly in between the terrain models. >. 
the urban area corresponding to buildings with high reflectivity = 
and bare soil facing the sun, is present. fra ai) Oo 
Moreover, a border effect is also affecting the DSM. Elevation 9995000 1 | G 
in some cases are considerably out of range. For these reasons it dox 3 
has been necessary the editing of data in order to cut-off the 4933000- & 
outliers and fill the area where the matching fails. 
A noise removal filter and a smoothing have been applied 2931000 
3.2 DSM accuracy assessment 
4929000 
Once edited the DSM, its accuracy has been evaluated by Fi 
comparing elevation with an available and accurate DTM 4927000 con 
derived from the digitalisation of 1:5000 maps with a gridding 
space equal to 10 meters. For the comparison, both models must 49250001 
have the same gridding space. For this reason, the extracted 
DSM has been interpolated with a kriging algorithm and | 
resampled to the lower spacing. Differences in elevations 457200 
between the new and the available terrain model can be then The € 
evaluated at each node. 49210001 — LT discu: 
4678000 1680000 1682000 1684000 4686000 1688000 1690000 1692000 1694000 softw. 
; ; : : mode 
Figure 6. Transects for comparison between terrain models ration 
shown on the DSM Ther 
1er 
certai 
evenly 
The F 
withoi 
coeffi 
with r 
WGSt 
compt 
final r 
the lat 
It has 
wider 
| carriec 
[ 
| 4l | 
| Result: 
| terms 
p North 
e indepe 
| orthoir 
—— TEE the aui 
12000 4000 6000 8000 10000 12000 14000 
known 
orthore 
J : ; ; - 2 4 : ; n eT ; The ov 
Figure 7. Comparison between the elevations (m) of the Digital Elevation Model obtained from vectorialization of maps checke 
(black solid lines) and elevations as provided by the satellite Digital Elevation Model (red lines). Numbers on the left-above have to han e 
be related to the previous Figure 6. In the x-axis the progressive distances (m) along the transects M D 
scale-d 
of crit; 
overlap 
1052