Satellite Image Segmentation using Modified U-Net Convolutional Networks

Abstract

The object detection in satellite imagery is a primary and elaborate one receiving lot of interest in latest years and performs an essential function for wide range of applications. After the massive fulfillment of Deep learning techniques in computer imaginative and prescient discipline, they're presently being studied in the context of satellite imagery for unique functions like object identification, object tracking, object classification, semantic segmentation of aerial/satellite images. Although diverse assessment research associated with object detection from satellite/aerial imagery are carried out, this observation provides an assessment of the latest development in the discipline of object detection from satellite imagery with the use of deep learning. This paper elaborates the detection of roads, buildings, solar panels and vehicles using Modified U-Net Convolutional networks and achieves more accuracy compared to the previous ones.