PSFE-Net: Semantic Segmentation Network for Airborne LiDAR Transmission Corridor Scenes Inspection

In transmission corridor scenes, it is difficult for networks to learn the local features of the pylon point clouds because of their high overlap and complex structure. Therefore, a semantic segmentation network called PSFE-Net is proposed in this paper: According to the symmetry of the pylon structure, the symmetry features of the pylon point clouds are extracted by the pylon symmetry feature extraction module in this paper, which effectively solves the problem that the local features of the pylon point clouds are difficult to learn. The approach is employed in an experiment for semantic segmentation of point clouds in suburban transmission corridor scenes, and good results are obtained. The mIoU of point cloud semantic segmentation reached 89.9%.