Research Progress on Roof Extraction and Photovoltaic Resource Assessment Methods Based on UAV Image Segmentation

With the transformation of the global energy structure and the proposal of the goal of “carbon peak” and “carbon neutrality”, solar photovoltaic power generation, especially distributed photovoltaic system, has gradually become an important direction of new energy development with its advantages of small installed capacity and high energy utilization efficiency. However, in China, there is still a lack of scientific and automated methods for rooftop photovoltaic resource assessment. This paper reviews the progress of methods of rooftop extraction and photovoltaic resource assessment. First, it revisits the application of traditional methods in building extraction and analyzes their limitations. Then, it highlights the use of deep learning methods, including convolutional neural networks (CNNs) and Transformers, in rooftop contour extraction, as well as hybrid network structures that combine the both. Finally, the paper discusses the current challenges and future directions in this area. This paper provides theoretical and technical reference for high-efficient evaluation and utilization of rooftop photovoltaic resource.