Construction of a Semantic Segmentation Network for the Overhead Catenary System Point Cloud Based on Multi-Scale Feature Fusion

Accurate semantic segmentation results of the overhead catenary system (OCS) are significant for OCS component extraction and geometric parameter detection. Actually, scenes complex, density point cloud data obtained through Light Detection Ranging (LiDAR) scanning is uneven due to character difference components. However, inconsistent points, it challenging complete better with existing deep learning methods. Therefore, this paper proposes a multi-scale feature fusion refinement structure neural network (PMFR-Net) cloud. The PMFR-Net includes prediction module module. innovations include double efficient channel attention (DECA) serial hybrid domain (SHDA) structure. (PCRM) used as network. DECA focuses on detail features; SHDA strengthens connection contextual information; PCRM further refines In addition, created released new dataset Based dataset, overall accuracy (OA), F1-score, mean intersection over union (MIoU) reached 95.77%, 93.24%, 87.62%, respectively. Compared four state-of-the-art (SOTA) methods, comparative experimental showed that achieved highest shortest training time. At same time, performance S3DIS public than other SOTA effectiveness DECA, structure, was verified in ablation experiment. show could be applied practical applications.