Automatic extraction of buildings damaged by tsunami following 2011 East Japan Earthquake using aerial images

In this paper, we present a method to extract damaged buildings by using aerial images observed pre- and post-tsunami. Stereo aerial images are used to generate a digital surface model (DSM) of the area. Regions corresponding to buildings are identified in the pre-tsunami image using a segmentation algorithm. Damaged regions are extracted by considering the height difference within a region between the pre- and post-disaster images. Validation using manually interpreted damage assessment as a reference showed that the proposed method successfully extracted damaged buildings with 63.9% accuracy when no GCPs were available. We conclude that our method can automatically generate maps of buildings damaged not only by a tsunami but also by other disasters.