Abstract: In order to improve the path planning efficiency of unmanned aerial vehicle (UAV) when patrolling transmission lines, a new method is proposed to enable UAV to independently carry out intelligent path planning to avoid surrounding obstacles during patrolling. By collecting the point cloud data of transmission lines, the best point cloud segmentation method is selected to perform real-time semantic segmentation of the collected point cloud data, and the point cloud data with semantic attributes are merged into voxels. Euclidean symbolic distance fields (ESDFs) are used to reconstruct the distance and angle between the obstacle surface and surrounding obstacles in real time, and the two-dimensional image recognition technology is used to accurately take target photos and identify insulators in the transmission channel to adjust the frame angle and focus of the UAV. The experimental comparison between the proposed method and some of the most advanced methods shows that the position error of the UAV automatic patrol method is less than 10 cm. Compared with manual inspection, the efficiency of the proposed method can be improved by 57.98%～62.88%, which can be applied to large-scale inspection of transmission lines.