Simple image processing techniques for near-real time inundation monitoring using TerraSAR-X imagery

1 ABSTRACT X-band SAR data are well suited for deriving inundated surface areas. The TerraSAR-X sensor with its different scanning modi allows for the derivation of spatial and temporal high resolved inundation masks. The paper presents preliminary studies on the application of different image processing techniques for extracting so called watermasks in the framework of the German-Vietnamese WISDOM project focussing the Mekong Delta region in Vietnam. The product is required as intermediate to (i) calibrate inundation model scenarios, (ii) estimate the extent of affected areas, and (iii) analyze the scope of prior crisis. Two simple image processing approaches have been tested and evaluated according to accuracy and processing performance of product generation. The first method is based on the assumption that water surfaces are forward scattering the radar signal resulting in low backscatter signals to the sensor. It uses multiple grey level thresholds and image morphological operations. The second approach presumes that inundated areas appear homogenous as compared to their surrounding objects because of more simple surface characteristics. The Local Homogeneity is calculated as combination of the gradientand the standard deviation filtered TerraSAR-X image. Homogenous areas are separated from the heterogonous ones by thresholding the resulting image. Subsequent image morphology and object based classification allows for eliminating false objects. The analysis was based on two TerraSAR-X Stripmap images in the plain of reeds area in the Mekong Delta which is characterised by seasonal flooding and dry periods. The two images have been acquired over the dry season in 2007 and the rain season in 2008. With the algorithms applied multitemporal inundation masks will be created during WISDOM project runtime as major input for subsequent inundation modelling. Along with this water level and -depths are planned to be calculated using the watermasks together with high resolution digital elevation models.