RAPID DISASTER ANALYSIS BASED ON SAR TECHNIQUES
Keywords: Disaster Analysis, SAR, Change Detection, Curvelet Filtering, Morphological Approach
Abstract. Due to all-day and all-weather capability spaceborne SAR is a valuable means for rapid mapping during and after disaster. In this paper, three change detection techniques based on SAR data are discussed: (1) initial coarse change detection, (2) flooded area detection, and (3) linear-feature change detection. The 2011 Tohoku Earthquake and Tsunami is used as case study, where earthquake and tsunami events provide a complex case for this study. In (1), pre- and post-event TerraSAR-X images are coregistered accurately to produce a false-color image. Such image provides a quick and rough overview of potential changes, which is useful for initial decision making and identifies areas worthwhile to be analysed further in more depth. In (2), the post-event TerraSAR-X image is used to extract the flooded area by morphological approaches. In (3), we are interested in detecting changes of linear shape as indicator for modified man-made objects. Morphological approaches, e.g. thresholding, simply extract pixel-based changes in the difference image. However, in this manner many irrelevant changes are highlighted, too (e.g., farming activity, speckle). In this study, Curvelet filtering is applied in the difference image not only to suppress false alarms but also to enhance the change signals of linear-feature form (e.g. buildings) in settlements. Afterwards, thresholding is conducted to extract linear-shaped changed areas. These three techniques mentioned above are designed to be simple and applicable in timely disaster analysis. They are all validated by comparing with the change map produced by Center for Satellite Based Crisis Information, DLR.