A Method for Detecting Hidden Faults in Power Lines by Combining Visible and Thermal Infrared Images

Abstract. Hidden faults in power lines pose significant safety hazards, severely threatening the safety of human life and property, necessitating rigorous inspection. Current methods for detecting hidden faults face two primary challenges: firstly, the inability to accurately locate power line components in thermal infrared images, and secondly, the reliance on empirical thresholds for fault determination, which results in low detection accuracy. To address these issues, this paper proposes a method for detecting hidden faults in power lines by combining visible and thermal infrared images. The method initially leverages the high discernibility of visible images to precisely identify the locations of power line components using a deep learning-based object detection model. Subsequently, a matching algorithm based on regional features is employed to register visible and thermal infrared images, thereby obtaining the precise locations of components in thermal infrared images. Finally, thermal infrared images are used to measure the temperature of components. By comparing these measured temperatures with the surface temperatures of components under normal operating conditions, hidden faults are effectively identified. Experiments were conducted on transmission lines in Jiashan County, Jiaxing City, Zhejiang Province. Out of 1526 power line components, 14 faults were detected, with manual verification confirming 13 as genuine faults. Compared to manual inspection results, the proposed method exhibited no missed detections and only one false positive, achieving a detection accuracy of 0.93.