Abstract
Bright curvilinear features in Synthetic Aperture Radar (SAR) images arising from the geometry of urban structures have been successfully used for estimating urban earthquake damage, using single pre- and post-event high resolution amplitude SAR images. In this paper, further automation of the process of selecting candidate curvilinear features for change detection is proposed, based on a model selection using priors derived from idealised building geometry. The technique is demonstrated using COSMO-SkyMed data covering the 2010 Port-au-Prince earthquake.