Detection of water body using very high-resolution UAV SAR and sentinel-2 images

Abstract

The extent of water body has far-reaching effects on agriculture, flood control, and ecological studies. Synthetic Aperture Radar (SAR) imaging technique can be operated in all weather, day and night circumstances. Due to the numerous advantages of SAR imaging technique over other conventional image acquisition practices, it has been used for the detection of the waterbody. Subsets of quad-pol, georeferenced (L-band) SAR imagery of UAV platform (provided by JPL, NASA) of Mondah, Gabon region and optical imagery by Sentinel-2 of the same region is used for the extraction of the water body. After preprocessing of UAV SAR image, Yamaguchi Decomposition was carried out and volume scattering image array (T33) has been used for the extraction of the waterbody. T33 array element of the coherency matrix represents volume back-scattering responses from the area of acquisition. Since the surface of the water body (either smooth or rough water surface) shows negligible volume back-scattering, water bodies can be easily delineated using thresholding and then applying the SVM classification method. The area covered by water reflects most of the radiations falling in the Green color frequency range and strongly absorbs Near-Infrared part of the electromagnetic spectrum. Taking advantage of this unique behavior of water surface while interacting with the electromagnetic spectrum, Normalized Difference Water Index (NDWI) is used for the extraction of waterbody from Sentinel-2 optical image. Finally, the SVM classified outcomes for extracted water area from both the images were compared. The harmonizing information from the, (UAV SAR and Sentinel-2 multi-spectral) images have been used for the quick and precise recognition of waterbody.