Istanbul 2004 
  
'ct, especially 
.1) can lead to 
cted from the 
he roads and 
etect roads by 
2.2) assuming 
ver to avoid 
rid resolution 
1 times and 
the grid is set 
lirections and 
segmentation 
An=6 degrees 
> considerably 
a to boundary 
classified and 
  
   
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
   
Figure 2.3- Segmentation 
  
Figure 2.4- Classification (roads) 
2.3 Final DTM 
Using the points derived from the last step, we can determine 
the first approximation of the DTM, iteratively comparing the 
DTM with the original DSM. If the height difference is less 
than 0.3 meter the original DSM points are selected and added 
to the DTM, whereupon the DTM is recalculated. This iterative 
processing ends when the number of points added during the 
present iteration is less than 100 points. Although some areas 
(see area A figure 2.4) have no data, by the end of the 
processing all DTM points have been added (figure 2.5). In 
some cases, where there are no roads, we can manually add 
seeds points to overcome the problem. The number of added 
points gradually decreases in each iteration. 
Total number of points in the DSM is : 466,652. 
Number of points built the first approximate DTM is: 104,772. 
Number of points was added is as follow: 
First iteration: 50,807 points. 
Second iteration: 25,759 points. 
Third iteration: 8,981 points. 
Total number of points added after 10 iterations : 163456. 
The final result of the DTM is presented y figure 2.6. 
  
Figure 2.5- All points data used to calculate the final DTM 
        
Figure 2.6- The final DTM 
3. Generating the DTM using the Robust method 
with Orthogonal Polynomials. 
3.1 The robust method outline 
The essence of the robust method is computing the height of the 
measured points by means of an interpolation function on the 
basis of neighbouring points and comparing the resulting height 
to the measured height. Points describing buildings will be 
characterized by a large positive difference vs. surface points 
characterized either by a large negative or a small positive 
difference. For computing a new height of a certain point, use is 
made of a  bi-dimensional interpolation function, the 
coefficients of which will be extracted by a correlation process 
in which all the points within a certain radios around the point 
under consideration will take part. For a function of higher 
order, a singularity in the coefficient matrix is possible; the 
resulting solution (if resulting at all) is unstable and influenced 
by the measurement errors. 
Using orthogonal polynomials will permit the usage of 
interpolation functions of the polynomial kind, with no 
restriction on the degree of the polynomial, this make it easier 
to remove building or other objects from the DTM. 
The process is iterative, where in each iteration a smaller AH 
(difference between original points and the polynomial) is set. 
The same processing is made in two directions and the final 
DTM is a combination between the two.