RU-Net: Solar Panel Detection From Remote Sensing Image

With increasing impact of global climate change, huge efforts are needed to reduce greenhouse gas emissions. The rooftop solar panels installation is one of the mechanism. In this paper, we focus on distribution and deployment degree of rooftop solar panels, and identify locations and total surface area of solar panels within a given geographic area in tackling the climate change. A comprehensive database of the location of solar panels on rooftops is important to assist analysts and policymakers in defining strategies for further expansion of solar energy. The deep learning method was used for the detection of solar panel location and their surface using the aerial imagery. While focusing on light weight image segmentation and low-resolution images, we proposed a two-branch solar panel detection framework consisting of classifier and segmentation branch, which was trained using the public data set of remote sensing images. This work provided an efficient and scalable method to detect solar panels, achieving an area under the curve (AUC) of 0.97 for classification and intersection over union (IOU) score of 0.84 for segmentation performance.