Extraction of Solar Photovoltaic Panels Based on High-Resolution Remote Sensing Images

Accurately and efficiently determining the installation positions, distribution, and total area of solar photovoltaic panels over a large area is important for investments and applications in photovoltaics. High-resolution optical satellite remote sensing imagery enables rapid and accurate extraction of ground-level objects. This provides the data foundation for automated extraction of photovoltaic panels based on deep learning. This paper utilizes high-resolution remote sensing imagery of solar photovoltaic panels. It employs the DeepLabv3+ semantic segmentation algorithm with the global convolutional network (GCN) to achieve a larger receptive field and enhance the network's global feature extraction capabilities. This improves the segmentation accuracy of photovoltaic panels. Experimental results demonstrate that the improved DeepLabv3+ model can achieve high-precision automated extraction of solar photovoltaic panels in complex remote sensing backgrounds.