Detecting damaged buildings using a texture feature contribution index from post-earthquake remote sensing images

The rapid and accurate detection of damaged buildings after an earthquake is critical for emergency response. As roof textures of damaged buildings are changed, damaged buildings may be accurately detected by the texture heterogeneity among buildings. However, existing researches have scarcely detected damaged buildings quantitatively by the texture heterogeneity among buildings from post-earthquake images. Therefore, this paper proposes an approach for the detection of earthquake-damaged buildings using texture feature contribution from post-earthquake images. In the proposed approach, a texture feature contribution index (TFCI) is defined by the information gain for selection of the optimal texture features, and a building texture damage index (TFdam) is defined for detection of damaged buildings. The proposed approach primarily comprises the following three steps: 1) building image objects are obtained using a pre-earthquake vector map to segment post-earthquake images; 2) the optimal texture features are selected by the TFCI to establish the texture feature vector; 3) damaged buildings are detected using the TFdam. The town of Leigu in Sichuan Province was selected as the study area to validate the performance of the proposed approach. The accuracy of 93.45% and kappa coefficient of 0.869 were achieved. Furthermore, the results indicate that the proposed approach can accurately detect earthquake-damaged buildings.