Point Cloud Extraction of Tower and Conductor in Overhead Transmission Line Based on PointCNN Improved

The point cloud deep learning algorithm PointCNN was improved using a clustering algorithm, to replace the way random sampling of point clouds that the original algorithm used in hierarchical convolution. The improved algorithm was applied to the task of automatic point cloud identification for key objects in Transmission channels, such as tower and conductor. The results showed that the overall recognition accuracy of the original algorithm was 89.17% and the overall recognition accuracy of the improved algorithm was 91.83 ^% , the improved deep learning algorithm was proven to get better results in the recognition for tower and conductor point clouds.