Target Recognition of Live Working Robots for Distribution Networks using LIDAR Point Cloud

Abstract

In order to satisfy the requirements of the operation tasks of the live working robots for distribution networks, an approach to electrical equipment recognition using the LiDAR point cloud is proposed. After the point cloud is collected, the preprocessing operation is performed to eliminate interference points and retain the point cloud of the operation area. The electric pole and insulating cross arms are preliminarily detected based on the geometric features, and then the geometric constraints between the electric pole and cross arms are used to iteratively optimize the recognition results by the least square method. Using the geometric information provided by the obtained structure composed of the electrical pole and cross arms, the point cloud of power lines is further processed to recognize them. Experiments show that the proposed approach can segment and recongnize the electric poles, cross arms and power lines from the LiDAR point cloud of the operation area quickly, accurately and robustly, providing sufficient information for the robot to perform various subsequent operation tasks.