stanbul 2004 
  
  
DETECTION OF THE EARTHQUAKE DAMAGED BUILDINGS FROM POST-EVENT 
AERIAL PHOTOGRAPHS USING PERCEPTUAL GROUPING 
M. A. Guler *, M. Turker ® 
* Middle East Technical University, Computer Center, 06531, Ankara, Turkey — aguler@metu.edu.tr 
? Middle East Technical University, Graduate School of Natural and Applied Sciences, Geodetic and Geographic 
Information Technologies, 06531, Ankara, Turkey — mturker(@ 
metu.edu.tr 
  
Commission III, WG 111/4 
KEY WORDS: Remote Sensing, Earthquakes, Detection, Segmentation, Aerial, Edge, Urban, Building. 
ABSTRACT: 
The collapsed buildings due to Izmit, Turkey earthquake that occurred on 17 August 1999 were detected from post-event aerial 
photographs using the shadow analysis and the perceptual grouping procedure. The selected area of study is a part of the city of 
Golcuk, which is one of the urban areas most strongly hit by the earthquake. The area contains a total of 282 buildings, of which 79 
are collapsed and 203 are un-collapsed. First, the Canny edge detector was applied to detect the edges between the cast shadows and 
the surroundings. Second, the output edge image was converted into vector line segments through a raster-to-vector conversion 
process. These line segments were then grouped together using a two-level hierarchical perceptual grouping procedure. The 
boundaries of the buildings were available and stored in a GIS as vector polygons. Therefore, after the perceptual grouping 
procedure, the damage conditions of the buildings were assessed on a building-by-building basis by measuring the aggreement 
between the detected line segments and the vector building boundaries. The results obtained is satisfactory. The overall accuracy was 
found to be 72.6%. Of the total 79 collapsed buildings, 63 were detected correctly by the proposed approach, giving 79.7% 
producer’s accuracy. : 
1. INTRODUCTION 
On 17 August 1999, a strong earthquake (magnitude 7.4) 
occurred in north-west of Turkey. This devastating seismic 
activity proved to be one of the most deadly earthquakes to 
strike Turkey in recorded history. According to Turkish 
Government figures, about 17000 people lost their lives. 
There were more than 48000 people injured and thousands of 
buildings were totally or partially damaged. The region is in 
the first degree earthquake zone. Approximately, 20% of the 
total population live in this region. In addition, most of the 
industrial complexes and power plants are also established in 
this region. 
The identification of damaged buildings after such a 
destructive event is a vital issue to get information about the 
extent and the location of the hard hit areas. Of course, the 
assessment of the damaged buildings can be carried out 
accurately through a field survey. However, this would 
require a lot of resources and time. Therefore, a rapid 
assessment of the damaged buildings is required for 
dispatching resque teams and emergency services to hard-hit 
areas. The post event aerial and space images have become 
important data sources for the identification of the damaged 
areas. Using pre- and post-event SPOT HRV images, Turker 
and San (2003) detected the Izmit earthquake induced 
changes. The change areas were detected by subtracting the 
near-infrared channel of the merged pre-event image from 
that of the post-event image. The overall accuracy for the 
change areas was found to be 83%. In a recent study, Turker 
and Cetinkaya (in press), detected the collapsed buildings 
caused by the 1999 Izmit, Turkey earthquake using digital 
elevation models (DEMSs) created from the aerial photographs 
taken before (1994) and after (1999) the earthquake. The 
DEMs created from two epochs were differenced and the 
difference DEM was analyzed on a building-by-building 
basis for detecting the collapsed buildings. The producer’s 
accuracy for collapsed buildings was computed as 84%. 
Further, Turker and San (in press) utilized the cast shadows 
to detect the collapsed buildings due to 'Izmit, Turkey 
earthquake. The available vector building boundaries were 
used to match the shadow casting edges of the buildings with 
their corresponding shadows and to perform analysis in a 
building specific manner. Of the 80 collapsed buildings, 74 
were detected correctly, providing 92.50% producer's 
accuracy. 
This study presents a different approach for detecting the 
collapsed buildings due to earthquake. The proposed 
approach is based on perceptual grouping and utilizes the 
relationship between the buildings and the cast shadows. It is 
assumed that if a building is fully damaged and collapsed due 
to earthquake, it will not have corresponding shadows. The 
digital processing of the shadow producing edges of the 
buildings can therefore provide very useful cues for detecting 
the collapsed buildings. The building boundaries are 
available and stored in a GIS as vector polygons. The 
agreement is measured between the shadow casting edges of 
the buildings and the corresponding vector boundaries 
through perceptual grouping. The decision about the 
condition of a building assessed is taken based on the degree 
of the agreement between the two data sets.