Battle Damage Assessment for Building based on Multi-feature

Assessment of battle damages is of great significant for tactical planning and provides the role of auxiliary decision-making for achieving accurate strike operations. In this study, we propose a robust method to detect damaged buildings using satellite images and building footprints before and after the event. This method uses multi-feature to detect the change of buildings, and has strong robustness to the change of light and local small deformation. By using the context model to describe the neighborhood of buildings, the detection accuracy is improved. The building footprint is used to accurately assess the changes in building level and form the neighborhood environment. The experimental results show that the proposed algorithm can effectively improve detection accuracy and obviously better than single feature damage detection.