Extraction of Water and Riverine Sand using Deep Learning on Multispectral Remote Sensing Images

The study of land-use-land cover (LULC) has become a necessity with the advancement of the urbanization process. Increased erosion of soil, increased silting, and sedimentation of the rivers are key effects that require study and analysis. Deep learning has a significant impact on classification tasks, particularly in the field of remote sensing image analysis. The proposed framework classifies LULC classes by employing the characteristics of deep learning. In this work, we compared the proposed method with the traditional machine learning methods in extracting water and riverine sand from multispectral remote sensing images. Further, we analyse the impact of Stochastic Gradient Descent (SGD) and Adam optimizers. The Adam optimizer implemented in this work gives higher accuracy than other combinations.