International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
3.2 Methodology 
The main steps followed in the proposed damage assessment 
method are illustrated in Figure 3. First, the post-event aerial 
photograph of the area was pre-processed using histogram 
equalization technique to provide better discrimination between 
the buildings and their shadows. Next, the buildings were 
selected one-by-one using the vector building boundary 
information. Then, for each building, the shadow-producing 
edges were determined. To do that, a simple algorithm was 
developed. The illumination angle was available from a previous 
study conducted by San (2002) as 135? from the x-axis. A buffer 
zone was generated along the shadow edges of the buildings. 
This was followed by the execution of the watershed 
segmentation algorithm. For each building, a binary-colored 
output representing the shadow and non-shadow areas was 
generated. Finally, the accuracy assessment was carried out by 
comparing the analyzed buildings with the reference data. 
  
  
Post-event Vector 
Aerial Building 
Photograph Boundaries 
  
  
  
  
  
  
  
Shadow Edge 
Detection «— 
Building 
Selection 
  
  
  
  
  
  
Buffer Zone 
Generation 
  
  
  
  
Watershed 
Segmentation 
  
  
  
  
Building 
Condition 
Assessment 
  
  
  
Figure 3. Damage detection using watershed segmentation 
3.2.1 Building Selection and Shadow Edge Detection: 
To select the vector building polygons, each polygon was 
assigned a unique identification code. In addition, the edges of 
cach polygon were also given numerical codes. The edges and 
the corresponding (x, y) coordinates of a building (#175) are 
illustrated in figure 4. The labeling of the edges was necessary to 
identify the shadow casting edges of a building being assessed 
and to relate these edges with the corresponding shadows. 
  
  
Edge 
x Y Number 
230 419 1 
239 398 1 
239 424 4 
230 419 4 
249 402 3 
2/39 424 f 
240 398 2 
249 402 2 
(a) (b) 
  
  
  
  
Figure 4. (a) The edges of building 4175 and (b) The format of 
vector data 
After selecting a building, a minimum-bounding rectangle was 
generated using the vector information by finding the minimum 
and maximum x-y coordinates (totally four points). Then, these 
points were connected to each other and the minimum-bounding 
rectangle was constructed. A buffer bound was then generated 
via expanding the minimum-bounding rectangle from its edges 
about six pixels. The bound was created in order to take into 
account the shadow regions produced by the buildings. Next, the 
shadow producing edges of the selected building were detected 
using a simple algorithm. The algorithm works as follows. The 
corner points are found from the vector information. This is 
simply finding the points that share the same end point on 
adjacent edges. For example, since both edge 4 and edge 1 share 
the same end point, (x=230,y=419), this end point is selected as 
a corner point (Figure 4b). Then, the Euclidean distances (di, d», 
d; d), shown in figure Sa, are computed between the corner 
points of the building and the corner of the minimum-bounding 
rectangle in the illumination direction. The computed distances 
are then sorted. If there is only one maximum distance, the edges 
that contain the same corner point are selected as the shadow 
edges. If on the other hand, there are two maximum distances 
then, the edge that contains those corner points is the shadow 
edge. These two cases can be illustrated with an example. If d; > 
d, » d; » d;, then the shadow edges are determined as edge 1 and 
edge 2 (figure 5b). This is because the corner point connecting 
these edges possesses the farthest distance (d,). If the ranking is 
d, = d, > d, > d,, then edge 1 is selected as the shadow edge 
since it is the only edge containing the farthest distances d, and 
du. 
  
di [- 
Shadow E = Edge 2 
producing 3 d 
edges 
         
Edge 
A Ca A Anleof 
es illumination 
  
  
  
(a) (b) 
  
Figure 5. (a) The Euclidean distances and the angle of 
illumination. (b) The shadow producing edges of building # 175 
644 
Inter 
  
3.2.2 
A t 
shad 
divic 
builc 
was 
the | 
purp 
shad 
of tl 
signi 
Bu 
( 
buil 
  
Figur 
edge: 
3.2.3 
The 
of fl 
sour 
need 
regic 
segn 
and 
builc 
respe 
rand 
orier 
The 
binai 
regic 
to sl 
area: 
in bl 
M: 
fa 
ar 
  
Figur 
trans 
trans