A Three-Dimensional Fault Localization Method for Distributed Photovoltaic Systems Based on Multi-Source Image Mapping

In practical photovoltaic (PV) operation and maintenance, precise localization of fault positions is essential for subsequent repair tasks. However, the challenge in PV fault localization lies in establishing accurate spatial correspondence between 2D images and 3D point clouds, which hinders the implementation of 3D fault positioning. To address this issue, this study proposes a 3D fault localization method integrating infrared images, visible-light images, and 3D point cloud mapping. It consists of three key stages: First, it utilizes the YOLOv10 object detection model to localize surface defects in infrared images and extract 2D fault coordinates; Subsequently, a coordinate mapping relationship is established through three pairs of matched feature points between PV images and 3D point clouds; Finally, with the coordinates of the faulty region in the 2D image, the 3D coordinates of the faulty region in the point cloud are solved by the optimised coordinate mapping algorithm. Experimental validation conducted on a self-constructed dataset demonstrates the effectiveness of the proposed method.