Research Progress on Distributed Photovoltaic Roof Extraction Methods Based on UAV LiDAR Point Clouds

With the growing global demand for renewable energy, distributed photovoltaic (PV) systems have become a significant direction for new energy development. Building roofs are essential carriers for distributed PV systems, and the extraction of roof is closely linked to the evaluation of solar energy potential. This paper reviews the research progress of building roof extraction methods based on UAV LiDAR point clouds. Firstly, the application of traditional methods in point cloud segmentation of building roof was reviewed, such as morphological algorithms, active contour models, region growing algorithms, random sample consensus algorithms (RANSAC), and clustering-based algorithms. These methods, while effective, face certain limitations when processing large-scale point cloud data. In contrast, PointNet and its variants, which can directly process point clouds, have emerged as key deep learning methods. Recently, Transformer-based models have shown advanced performance in point cloud classification and segmentation tasks. Finally, the challenges and future trends of UAV LiDAR-based roof extraction are discussed, providing technical support for evaluating the potential of urban distributed PV systems and formulating distributed PV power generation plans.