Abstract: To solve the problem of low efficiency and poor accuracy of traditional classification methods when processing LiDAR point cloud data for visual inspection of large-scale transmission lines, an intelligent solution is proposed to automatically separate power lines, towers, ground and vegetation in transmission line corridors. Firstly, power line points are extracted accurately according to curvature and neighborhood features. Secondly, ground points and non surface points are separated by cloth filtering method.Tower points and vegetation points are identified from non ground points based on cylindrical model. Finally, in this study, three sections of point cloud data of transmission line corridor with different density were selected for the experiment. The results showed that the method had good performance in the identification of power line, tower, ground and vegetation, and the overall classification accuracy was more than 90%. The user accuracy and mapping accuracy of power line gradually increased with the increase of cloud point density. The experiments prove that the method is applicable to many types of LiDAR point cloud data, and its theoretical method has a reference value for solving the automatic classification problem of visual inspection of transmission line corridors, which provides a scientific basis for fully improving the efficiency and accuracy of transmission line UAV inspection.