Abstract:

While surface waters play a crucial role on terrestrial life, this resource is under threats due to both human and natural activities. Extracting water bodies is an important task for the management of water resource. Traditional remote sensing based methods have difficulties in the water body extraction because of the coarse spatial resolution and spectral reflectance heterogeneity of satellite images. This paper presents a new approach for detecting water bodies from water index images based on the modified fuzzy c-means (MFCM). The approach was designed to improve the water extraction accuracy. It was applied to Landsat 8 OLI images captured over the Bach Dang estuary situated in Hai Phong and Quang Ninh. Results showed the MFCM method improved the accuracy of water extraction in comparison to the density slicing method in term of kappa coefficients, with 0.87 and 0.84, respectively