Curvelet-based Change Detection on SAR Images for Natural Disaster Mapping

Abstract

This paper focuses on the use of SAR data in the context of natural disasters. A Curvelet-based change detection algorithm is presented that automatically extracts changes in the radar back-scattering from two TerraSAR-X acquisitions – pre-disaster and post-disaster - of the same area. After a logarithmic scaling of the geocoded amplitude images the Curvelet-transform is applied. The differ-entiation is then done in the Curvelet-coefficient domain where each coefficient represents the strength of a linear structure apparent in the original image. In order to reduce noise the resulting coefficient differences are weighted by a special function that suppresses minor, noise-like structures. The resulting enhanced coefficients are transformed back to the image domain and brought to the original scaling, so that the values in the difference image describe the increase and the decrease with respect to the amplitude value in the initial image. This approach is applied on three sample data sets: flood, forest fire, and earthquake. For all scenarios including natural landscapes and urban environ-ments as well areas with changes in the radar amplitude are clearly delineated. The interpretation of the changes detected in the radar images needs additional knowledge, e.g. pre-disaster maps. The combination of both could possibly deliver a robust and reliable database for the coordination of res-cue teams after large-scale natural disasters.