Extract Power-line Key Points From LiDAR Data

The power-line point cloud obtained by LiDAR can represent 3D geometric information of power lines, which has important applications in power line inspection. However, the rapid growth of point cloud data has brought challenges to data transmission, storage, and processing. An alternative approach is to preserve the structural information of relevant objects from point cloud data. This structure information is also useful in some important tasks, such as obstacle troubleshooting and sag measurement. In this paper, we aim to extract power-line key points from LiDAR data to represent the power-line structure. Unlike most existing methods that use 3D line fitting, we propose a data-driven approach that exploits point features in the 2D plane and 3D space. Our method first projects 3D points onto a 2D plane, which can largely reflect the distribution of points. It then uses the local features of points to extract power-line key points. We demonstrate the effectiveness and efficiency of our method using an original data set consisting of 50 power-line point clouds. The results show that our method is intuitive, easy to understand, and suitable for practical applications.