A Contemporary Model for Segmentation of Satellite Images Using Neural Networks Through the U-Net Model

Analysis and identification of vegetation-flora, regions, roads, and indication of water resources present is very much crucial for the continuing progression and augmentation of remote region is significant around the world. Segmentation of Satellite images has never been an easy task as it is not only categorized upon which is presented in the Satellite images but also analyzed on each and every per-pixel level. Here our main aim is to get a map representation from a satellite image. We have demonstrated and established an efficient and robust satellite image classifier whose task is to accurately detection and extraction of precise data in satellite images which are basically of the category of land, water, buildings, vegetation, and roads. The crucial part of the presented work is U -net model which is based on CNN model, the proposed model is more accurate, robust and efficient for entire prediction system for varying satellite images. The segmentation is done using a few U-Net Convolutional Neural Networks. U -net is considered one of the F-CNN (Fully Convolutional Neural Networks). The more sophisticated form of CNN, known as F-CNN, concentrates mostly on Segmentation problems while CNN just employs it for Classification issues. F -CNN helps us to reduce the number of network parameters to a large extent due to using no connected layers.