A High-Precision Laser Point Cloud Segmentation Algorithm For Substations Based On Improved Pointnet++

Accurate semantic segmentation of 3D point cloud maps is crucial for precise navigation and positioning of substation automation inspection robots during patrol missions. However, existing substation point cloud segmentation algorithms exhibit high complexity and demonstrate slow processing speeds for large-scale substation point clouds, failing to meet operational requirements. To address this issue, this paper proposes an improved PointNet++ point cloud semantic segmentation network incorporating Transformer modules and constructs a dedicated substation point cloud dataset. Experimental validation confirms that the proposed enhanced PointNet++ network achieves 90.1% overall accuracy, providing an effective solution for developing intelligent inspection robots for substation equipment.