Recognition of Live Equipment Models in Substations Based on 3D Point Cloud Data

During the expansion and modification of substations, scenarios arise where construction equipment operates near live bodies, requiring a secure distance between the equipment and the live bodies. Semantic segmentation of point clouds of live equipment is a pivotal step in the detection and identification of live equipment, and the calculation of safety distances is predicated on the outcomes of this segmentation. This study leverages point cloud data with three-dimensional spatial information, acquired via a novel laser scanning measurement system, as the foundational data set. A PointNet++ architecture is meticulously designed to extract features from the point cloud data, thereby facilitating semantic segmentation of the point cloud data pertaining to live equipment within substations. The research findings indicate that the proposed methodology achieves a segmentation accuracy of $88.5 \%$ for the experimental point cloud data set, with an overall segmentation precision of $84.5 \%$. This approach has successfully accomplished reliable semantic segmentation of extensive multi-scale point cloud data in the context of live equipment model recognition, thereby fulfilling the analytical requirements for the identification of live equipment in substations.