International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
  
> Post-earthquake 
Spot imagery 
Topographic 
3 maps : 
Spot imagery : 
Geometric correction 
p; Relative radiometric A 
| Exclusion areas n] 
I. Band substitution 51 
| res to IHS transformation | 
b Image differencing " 
v 
|. Integration with vector ans J@— [Vector data ] 
| Comparison x ground rn poamage 
  
    
     
  
  
*il 
EE 
  
  
  
  
  
  
Figure 2. Damage assessment using Spot imagery 
  
  
    
  
   
Vector data 
Visual interpretation 
Training data 
Threshold decision 
E Video imagery | 
li Frame grabbing 2 
| Image enhancement | 
| Colour and edge feature 
| Texture feature layers | 
  
i34 
  
  
    
  
€ Ec 
  
  
  
  
  
Ir Multi level slicing | 
|. Actual damage id 
p: Comparison with actual damage 4 
  
  
  
  
Figure 3. Damage assessment using video imagery 
change detection to avoid spurious results (Jensen, 1996), 
relative radiometric correction was applied to normalize pixel 
values of the multitemporal data sets. Vegetation areas and 
water bodies were excluded from the analysis. Band 
substitution, which does not change radiometric qualities of any 
of the data (Jensen, 1996), was applied to merge Spot data set. 
The panchromatic band was substituted directly for Band 2. 
Intensity values of the merged data set were used for the change 
detection. Image differencing (Singh, 1989), which is one of the 
simple change detection algorithms, was applied in the 
methodology. To assess the damaged areas at the regional and 
local level, the result was thresholded and aggregated into 
parcel level. 
The analysis of video imagery was carried out in two steps: (1) 
visual interpretation and (ii) digital image processing. Visual 
interpretation was carried out at the parcel level. In digital 
image processing, frame grabbing was the first step to convert 
analogue video frames to digital ones. Astrostack software was 
used to improve the quality of the frames 
(http://www.innostack.com). Colour indices (hue), edge feature 
layers (edge) and local variance (saturation variance and edge 
variance), which is one of the texture parameters, were used to 
detect damaged areas. Threshold values derived from training 
data set of damaged areas were used for multi level 
thresholding of feature layers. At the end of the analysis, the 
result was compared with the actual damage information 
observed in the video imagery. The overall process of aerial 
video imagery analysis for damage assessment was shown in 
Figure 3. 
3. DATA ANALYSIS 
3.1 Analysis of User Information Requirements 
According to the results of the analysis, information 
requirements differ depending on government hierarchy and 
activities of the agencies. Although at the national level, there is 
a need for overall information about the damage, at the local 
level detailed information becomes most important for the user. 
Moreover, each organization requires different types of 
information in terms of scale, detail and characteristic based on 
their activities. For search and rescue operations, only 
information on collapsed buildings, as well as their inhabitants 
and use are required. On the other hand, for emergency aid 
activities, the number of people who survived the disaster is 
important, as they need to know food, accommodation and 
medication requirements. Rapid data gathering following the 
disaster is required, as after 72 hour, the chance for exposed 
and/or injured people to survive approaches zero. 
The analysis has shown that remotely sensed data without 
integration with baseline data are not enough by themselves to 
fulfil the information requirements of the user. Baseline data 
showing the pre-disaster situation, such as population, road 
network, land use, ownership information, are critical for an 
optimal use of the potential of remotely sensed data. Moreover, 
for an effective use and flow of information derived from 
remote sensing technology, there is a need for organizational 
improvements. In the time of emergency, the main challenge is 
dissemination of different types of information, which requires 
an information network between emergency agencies, as 
information is only valuable if it reaches to the right 
organization at the right time. 
3.2 Analysis of Spot Imagery 
The results of the Spot imagery analysis were evaluated at 
regional and local levels (Figure 4). At the regional level, 
damage assessment using Spot imagery showed both significant 
overestimation and underestimation of damaged areas. Due to 
smoke coming from fire at the Tupras Oil Refinery, there was 
underestimation in the western part of the area. Overestimation 
in the northern part of the image shows the need for 
orthorectification for hilly areas. Moreover, differences in the 
incidence angle, the pixel-by-pixel registration requirement, 
which can cause spurious changes in the change detection 
analysis (Jensen, 1996), clouds and shadows were other 
obstacles for change detection. Differentiating between 
damaged areas and change values due to external factors is a 
688 
In 
ch 
th 
as 
da 
Hk 
art 
  
  
  
Acc 
evel 
(tot: 
bu il 
Figu 
nort 
ovel 
In c 
exte 
info 
abot 
Des] 
over 
area: