Electric Tower Extraction in Overhead Line Based on Multi-Data Fusion

Lidar has been applied widely in intelligent applications of overhead line inspection. However, automatic extraction of electric towers from overhead lines remains a tricky problem considering both the accuracy and the efficiency. To address this, we propose a method for extracting electric tower point clouds based on the fusion of images and point clouds. The mature 2D object detection technique is introduced to conveniently locate the tower in an image, and then the tower point cloud can be extracted by the proposed spatiotemporal alignment of point clouds and images, through a coarse tower extraction and fine extraction. Experiments show that the extracted electric tower point clouds are complete and accurate, and the extraction speed is superior to traditional methods.