Use of edge detection filter for monitoring urban change

Observations concerning the level and distribution of building damage after a destructive earthquake are of primary importance in the immediate aftermath for planning response efforts, and later on for better understanding the effect of shaking on buildings. After recent earthquakes (the 2003 Boumerdes and Bam events) and the 2004 Indian ocean tsunami, very high-resolution (VHR) satellite imagery (QuickBird and Ikonos II) proved to be a valuable source of spatial information. For damage detection, textural change indexes have proved to be a useful means of describing and detecting changes between a temporal sequence of ‘before’ and ‘after’ images (Adams, 2003). However, for conventional spectrally-based approaches (e.g. band difference/ratio, principal component analysis) there are several factors that can hinder the direct extraction of useful information from an image, such as illumination differences and variable geometry angles (elevation and azimuth of sun and satellite). This paper explores an object-oriented approach to the detection of buildings damage caused by the Bam earthquake, which minimizes these sources of error. It also introduces a novel “edge detection” approach to characterizing visible changes accompanying building collapse.