International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004 
  
- dark impervious surface, like asphalt and 
darkened concrete. 
* Soil and/or dry vegetation. 
* Shadow (N. Haala, V. Walter, 1999) 
The selection of the samples was followed by the 
estimation of each class's samples by studying 
characteristics such as vectors of mean values 
intensions for each class (Fig. 9), histograms of each 
sample and the feature spaces of the samples. 
Figure 12: The synthetic image 
3.2.2. Estimation of the classification: Urban 
2500 regions, due to the heterogeneous nature they show 
concerning urban characteristics, are really hard to 
classify. Moreover, this problem is increased as the 
3800 spatial resolution of the images increases, because 
of the upgrade of the details and the spectrum range 
1000 SZ of the pixels. 
500 
  
2000 
  
[n this particular application the problem in the 
  
  
  
  
4 separating classes like streets and buildings. In their 
majority rooftops are consisted of different materials 
like metal, concrete, tile and others. 
1 2 bands 3 
Figure 9. Vectors of mean values intensions in every 
layer The materials that consist the rooftops have similar 
spectral behavior with the materials that consist the 
roads, like asphalt and concrete. The result of this 
fact is that the boundaries between roads and 
buildings are not distinct in the classified ima 
(Fig. 13, 14) 
  
HEALTHY GREEN VEGETATION 
ROOFTOP/METAL /TILE 
ASHPALT/ DARKENED CONCRETE 
CONCRETE/ WEATHERED ASHPALT 
SHADOW 
SOIL 
TREES/SHRUB VEGETATION 
  
ge 
t 
c 
  
  
  
  
  
  
  
  
Table 10. Classified image legend 
The next step was the classification of the synthetic 
| image (Fig. 11). 
  
Figure 13. Part of the synthetic | image 
  
Figure 11 : Classified image of the Municipality of 
Thessaloniki 
  
Figure 14. Classified image 
218 
  
0 classification was detected in the difficulty of