International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
4.1 DDTM 
The approach that was chosen for the generation of the dense 
DTM uses refined interpolation techniques applied to 3D digital 
map. The planimetric and altimetric information that is 
contained in them can in fact adequately describe the territory 
(height points, contour lines), the built entities (sections, 
vertices) and the buildings (centroids of known heights). The 
DDTM that were used to generate orthoprojections of satellite 
images were generated using the GeneDDTM software 
implemented in Visual Fortran at DIGET at the Politecnico di 
Torino (Dequal et al, 2002). 
  
  
   
  
  
  
Figure 3 — Part of the DDTM used for the generation of the true 
orthophoto (on the right) and the relative digital map at 1:2000 
scale (on the left). 
4.2 Application field 
The fields in which precision orthoprojection of high resolution 
satellite images could be more profitable than the usual 
procedure were investigated on the basis of the operative 
characteristics of the sensors. From this point of view, a test 
was first carried out of the conditions within which the 
displacements of the objects that can be put down to the 
presence of great discontinuities (buildings and infrastructures) 
results not to be negligible compared to the expected mapping 
scale to which the orthophoto should refer. 
Displacements due to the presence of 3 different classes of 
buildings defined according to their height: h;=10m, h,=30m, 
h:=100m were considered. Then the negligibility limits were 
reported that were considered to be equal to the tolerance of the 
orthophoto (0.5 mm at the map scale) for the following scales: 
1.2000, 1:5000 and 1:10000. These tolerances resulted to be 1 
m, 2.5 m and 5 m, respectively. 
In the graph of Figure 4 it is shown how with a variation of the 
mean view angle of the scene (y) the entity of the displacement 
is increased due to the buildings. The three curves that are 
reported refer to the three different classes of buildings that 
were considered. The variability of the mean view angle was 
limited to an interval of (0-50) gon, considering the operative 
characteristics of the satellites, where the orientation capability 
of the sensor with respects to the nadir position never exceeds 
this value. 
An analysis of the graph allows us to see how, for a building 
height equal to 30 m, taken as an example (central curve), it is 
easy to establish the following mean view angle values so that 
the displacements are kept within the permitted tolerance for 
the orthophotos at different scales: 
e lower than ~ 2.5 gon for an orthophoto in a 1:2000 scale; 
e lower than ~ 6 gon for an orthophoto in a 1:5000 scale; 
e lower than ~ 11 gon for an orthophoto in a 1:10000 scale. 
Mean Off-Nadir Angle (gon) 
0.001 5 10 15 20 25 30 35 40 45 50 
   
  
  
   
   
     
   
| 100 r-— —ÓÓ— 
idu hz10m 
000900? —e— h=30m 
th for 1 apenensseases 5 i hz100m 
| 090992 c ris |. mmm ol 1:0000 
EC” 775 [1:1000 [0e tol] 1:5000 
[x € | . : 
|. 8 SE 5000 sn toll 1:10000 
| 1 1:2000 | 
i i 
| | 
| A meer ere me == 
Figure 4 — Entity of the displacements (due to the presence of 
buildings) in function of the mean view angle. 
It has been shown that orthoprojection of non-nadir satellite 
images requires in most cases a more accurate approach and 
that altimetric height data of the buildings cannot be neglected, 
above all in urban areas. 
4.3 Tests 
In order to verify the efficiency and real incidence on an 
application case when adopting a “precision” approach for the 
generation of orthophotos, a non-nadir EROS A1 image (GSD 
= 1.9 m) was processed first using a traditional DEM (50 m 
steps) and then a DDEM previously generated from 3D digital 
map using the previously described software. The scene refers 
to the city of Cuneo (Piedmont). 
It can be seen in figure 5 how the viaduct is reported in its 
correct position thanks to the altimetric information that is 
derived from tic DDTM, but the radiometric values that 
identify it are also repeated for the portions of scenes that were 
hidden from the sensor and which are "uncovered" after 
orthoprojection. 
This lack of information, in a “rigorous” approach, is resolved 
by removing the radiometric value from images acquired from 
other points of view and whose orientations are known. These 
can be images from the same sensor or from another sensor 
(multi-sensor aprroach). 
If no other data are available, it is possible to resolve this lack 
of information using a masking of the hidden arcas with a 
background value that is easy to identify, as shown in figure 6 
and figure 7 (test on a QuickBird image). In this way, the 
interpretation of orthoprojected satellite images is easier for 
unskilled users that are not deceived from erroneous data. 
  
  
  
  
Figure 5 — Details that show the overlapping in correspondence 
to the viaduct using the DDEM (on the right) and the 
duplication of radiometric tones over the hidden area. 
   
  
  
    
  
   
   
   
   
  
  
  
    
  
  
  
  
   
  
  
  
   
   
   
  
   
  
   
   
    
   
    
   
   
   
   
   
   
   
   
  
  
  
  
  
  
  
  
  
  
   
  
  
  
  
   
   
   
   
Internatic 
  
Figure 6 
Duplicati 
masking 
  
Figure 7 
Duplicati 
of the san 
The here 
to the us 
satellite i 
buildings 
analysis 
confirme: 
view ang 
102. The 
the builds 
not negli 
DDEM c 
made. Th 
way rem 
model. 
interpreta 
especially 
Referenc 
Bello, M 
integratec