Urban damage estimation using statistical processing of satellite images

Remotely sensed satellite imagery of an earthquake-affected region can significantly assist in estimating the severity of infrastructure damage. Modern high-resolution satellite systems have been launched to provide users optical imagery with submeter accuracy, which enable the possibility of sensing damage for individual structures by means of pre- and postevent imagery. Recognizing these advancements, herein, we focus our study on the region of Barn, Iran, which was devastated by a moment magnitude M-w = 6.6 earthquake on December 26, 2003, causing approximately 43,200 lives lost. The recognition of urban structures within the Bam region is obtained by performing morphological filtering and intensity thresholding, which is further optimized through a statistical procedure. By overlaying the recognized structures with the pre- and postevent images, three object-based change detection methods are presented. The performance of change indices resulting from the three change detection methods is evaluated by means of a histogram-based classification approach. Damage estimation results are presented using easily interpretable maps, wherein individual structures are rendered with colors representing the severity of damage.