Research On Power Line Autonomous Inspection Based On Heterogeneous Point Cloud Association

At present,the country’s power grid construction is in a period of rapid development.The use of traditional manual inspection methods to inspect high-voltage towers and transmission lines is no longer suitable for the development of the times.Human resources also exist in the use of pilots to control drones for inspection In the case of waste,using intelligent UAV inspection methods instead of traditional methods can not only reduce operation and maintenance costs,but also greatly improve inspection efficiency.It has far-reaching research significance and broad application prospects.Autonomous navigation of drones is the foundation of intelligent inspection.This paper focuses on the scientific research of autonomous inspection of power poles and towers by UAVs,and proposes an autonomous navigation method using laser maps and vision sensors.Based on the premise of a known laser map,the problem of how to generate a patrol inspection path is studied.The problem of acquiring point cloud data using a depth camera and calibrating the pose of the aircraft by means of point cloud registration is discussed.This paper first proposes a path planning method based on the point cloud structure of the electrical tower.Using hierarchical search,the method of dividing the bounding box is used to find the position of the insulator on the electrical tower and the position of the ground wire.Observation position and attitude information of the drone.After that,because the coordinates of the laser point cloud map deviated from the actual geographical environment,relying on the laser point cloud map for path planning alone,it was difficult for the drone to completely capture the location to be inspected.It was decided to use a depth camera to obtain the RGBD point cloud.Point cloud registration calibrates the pose of the drone.Researched how to use the depth camera to obtain point cloud data,and used the European clustering segmentation method to extract the required point cloud data,and then used the voxel filtering method to reduce the amount of point cloud data and reduce the algorithm program Calculate pressure.Finally,a heterogeneous point cloud registration method from coarse to fine is proposed.The FPFH features were extracted from the point cloud data and laser point cloud previously collected by the depth camera,and the feature matching was performed on the two point clouds using the improved RANSAC algorithm to complete the rough registration of the point clouds.Then,the kd-treebased ICP algorithm was used.Perform fine registration on the point cloud data that has completed the coarse registration,and finally obtain the transformation relationship between the point cloud data in the two coordinate systems,and play a role in correcting the pose.Experiments prove that the autonomous navigation method proposed by this paper is more efficient and practical.