Change detection using multiscale segmentation and Kullback-Leibler divergence: Application on road damage extraction

This paper addresses the problem of change detection from very high resolution remotely sensed images and its application on road damage extraction in case of major disaster. The proposed methodology is based on the multiscale image segmentation using the Haar wavelet in order to define the appropriate unit of analysis for the comparison step. The Kullback-Leibler divergence is then applied as a similarity measurement to identify changed regions. This strategy is adapted to solve the road damage extraction problem by applying the Dempster-Shafer theory (DST). The images acquired during the earthquake that hits Port-au-Prince (Haiti) on 12 January 2010 are used in the experimentations and the obtained results demonstrate the accuracy and the efficiency of the described method.