Deep Learning Technique For Effective Segmentation Of Water Bodies Using Remote Sensing Images

Segmenting bodies of water in remote sensing images is a difficult job. In this work, the results of the implementation of a deep learning-based method for separating bodies of water in satellite imagery are presented. The suggested method makes use of a U-Net in addition to a pair of Deep Convolution Neural Networks (DCNNs). To simplify matters, two DCNNs for feature extraction and feature selection are used. Images of bodies of water are segmented using the U-Net. The suggested method is tested using two datasets. The findings demonstrate that the Unet method can successfully segment water bodies with an accuracy of over 82%. The U-NET approach has the potential to be used for a wide range of applications related to water bodies segmentation. the U-Net architecture has been proven to be very effective for the semantic segmentation of water bodies. It can capture fine-grained details and incorporates multi-scale context information. Furthermore, it has a fully convolutional network (FCN) structure which allows for pixel-level labeling. In comparison, DCNNs are limited in their ability to capture fine-grained information and are not as well suited for pixel-level labeling.