查看文章

Application of different water identification indices and their modified form with a threshold is a common practice in surface water identification from multispectral images. Implementation of the statistical features of water present in such images to improve the accuracy of existing approaches is a novel application. A dynamic threshold selection is more suitable for the detection of sediment–water. In consideration of the facts, the present study proposed a hybrid approach for automatic surface water detection. Fuzzy c-means, NDWI, and a statistical feature: gradient are used to classify and therefore identify surface water. The study area, the river basin of Sundarban, is chosen due to its nature of water bodies such as wide rivers, narrow water streams, and sediment–water. The algorithm works with minimum human interaction. The method is validated by applying on Sentinal-2 and WorldView-2 images having a …