Texture-based seismic damage assessment on radar data: a preliminary comparison between COSMO/SkyMed and TerraSAR-X datasets

This study focuses on remote sensing technology as a disaster monitoring tool. It emphasizes on Synthetic-aperture radar (SAR) applications to extract geo-information relevant to damage assessment on the block level from single post disaster imagery. The procedure undertaken was previously developed by our group, based on discovered correlations between texture measures on radar images and the extent of seismic damage in any given urban block. Ground truthing was based on a “Damaged Area Ratio” (DAR) damage indicator, computed as the area ratio of the damaged buildings to the block area. The damaged buildings were detected using data from high-resolution airborne sensors, thus only high levels of damage, mainly with ceiling partial or complete collapse, were considered due to the limitation of the space borne technology in detecting slight to moderate damages as well as the sandwich damages. The urban areas in the studied cities were allocated into a number of blocks, where DAR was calculated for each block. After that, damage categorization was applied using thresholds on the DAR values of the selected blocks. This work continues the investigation on the linear correlation between the textural features and the calculated damage indicator DAR. For that purpose, data acquisitions were analysed from two different SAR satellite sensors, TerraSAR-X and COSMO/Sky-Med. As test cases, damages from two earthquakes were analysed with different geometric resolutions: L’Aquila 2009 using High Resolution Spotlight images and Haiti 2010 using Strip Map images. The data were analysed with similar techniques for the sake of an objective comparison on the variations on the linear correlations. The funding and support of the Italian Department of Civil Protection through the “Progetto Esecutivo 2012-13”, as well as the support from the German Aerospace Agency through the LAN 1240 project are gratefully acknowledged.