关键词: 锚框聚类, 实时隐患检测, 无人机避障, 多传感器融合

Abstract: In the process of electric power distribution network inspection, the traditional UAV inspection mode often has a lag, which makes it difficult to discover and deal with the hidden tree obstacles in the inspection channel in a timely manner, leading to the impact on the operational safety of distribution lines. To solve this problem, this study proposes a real-time hidden danger detection technology that integrates LiDAR and PTZ camera, realizing real-time detection and analysis of tree obstacles in the inspection channel by carrying LiDAR and high-precision camera on the UAV and integrating the YOLO model of genetic K-mean anchor frame clustering. The system can automatically identify and locate tree obstacles, and accurately photograph the hidden parts through the gimbal camera to generate a real-time report on the hidden problems. Through testing in complex environments, the results show that the system is able to significantly improve the timeliness and accuracy of hidden trouble detection and significantly shorten the time from discovery to action. It is concluded that the proposed real-time hidden danger detection technology significantly improves the efficiency and safety of distribution network inspection, provides strong support for the intelligent management of electric power distribution networks, and has a wide range of application prospects.

Key words: anchor box clustering, real-time hazard detection, unmanned aerial vehicle obstacle avoidance, multi-sensor fusion