Urban damage assessment using multimodal QuickBird images and ancillary data: the Bam and the Boumerdes earthquakes

Remote sensing has proved its usefulness for the crisis mitigation through situation report and damage assessment. Visual analysis of satellite images is conducted by analysts, however automatic or decision aid method are desired. We propose a semi-automatic damage assessment method based on a pair of very high spatial resolution (VHR) images and some ancillary data. It is applied to two disaster cases, for which the QuickBird images acquisition conditions differ. For each case, the two images also have very different viewing and illumination angles. Hence their comparison requires a preliminary registration; an automatic method adapted to VHR images is described. Then several change features are extracted from the buildings, and their relevance to assess damage on buildings is evaluated. Some textural features allow a damage assessment, but correlation coefficients are more efficient. Finally, a step toward the full automation of the method is done, skipping the supervision step of the classification process. We show the robustness of the global approach for both disaster cases with average performances closed to 75 % when 4 damage classes are discriminated, up to 90 % for a intact/damaged detection.