Automatic detection of earthquake-damaged buildings using DEMs created from pre- and post-earthquake stereo aerial photographs

This study presents detection of the collapsed buildings caused by, earthquake,, using digital elevation models (DEMs) created from pre- and post -earthquake stereo aerial photographs. It is based on the idea that changes in the height, of collapsed buildings can be detected. The study was conducted using aerial: photographs taken before (1994) and after (1999) the lzmit earthquake in. Turkey. DEMs were created automatically from both pre- and post-earthquake. aerial photographs at 5m spatial resolution. The DEMs created from the two epochs were then differenced and the difference DEM obtained was analyzed building by building to detect those buildings whose height difference between pre- and post-earthquake exceeded a specified threshold level. The producer's accuracy for detecting the collapsed buildings was found to be 84'. The results show that the difference DEM of the pre- and post-earthquake aerial images can reveal successfully the collapsed buildings caused by the earthquake. It was found that determining an optimum threshold level between change and no change pixels of the difference DEM is important.