Extracting the flood extent from satellite SAR image with the support of topographic data

Abstract

A satellite SAR image is broadly used in monitoring and evaluating flood disasters. It is difficult to automatically extract the flood extent from RADARSAT SAR images because of the confusion of the gray value between the flood and shade in imagery gray. Therefore, extracting the flood extent from satellite SAR image with the support of digital topographic data is explored here. It included several steps as follows. Firstly, RADARSAT SAR imagery was filtered by enhanced frost filter with window size of 7 pixels by 7 pixels, and geo-registered to the topographic map. Secondly, flood extent was primarily extracted from RADARSAT SAR imagery. Thirdly, DEM was created from the digital topographic data by using GIS software. Fourthly, simulated SAR imagery was created from DEM. Finally, the simulated SAR imagery was registered to RADARSAT SAR imagery, and the shade from the simulated SAR image was used to mask the mislabeled flood extent from RADARSAT SAR due to its shadow influence. It is shown that the flood extent can be accurately, semiautomatically extracted from RADARSAT SAR by the method. It also is shown that the method is not only suitable for RADARSAT SAR imagery, but also suitable for other satellite SAR imagery with the different imaging mode, especially for SAR image covering mountainous area.