Point Cloud-Based Modelling of Footbridge Crossing Operating Railway Lines

In the renovation and maintenance of existing railway bridges, point cloud-based modeling techniques are essential for constructing digital twins of physical entities. However, the strict safety requirements imposed on construction adjacent to operating railway lines often restrict on-site laser scanning, leading to incomplete raw data. To address this challenge, a semantic segmentation method based on local directional features is proposed, which enables the identification of bridge components under such constrained conditions without relying on cross-sectional information. Following a workflow of data enhancement and semantic classification, noise points in the raw point cloud can be effectively removed, while semantic labels are assigned to bridge members according to their principal orientations. A practical case study on the footbridge at Yiwu Railway Station demonstrates the feasibility and robustness of the proposed approach. The resulting digital model provides a reliable foundation for subsequent structural performance evaluations, offering support for both demolition planning and operational safety assessment in complex railway construction scenarios.