Spatio‐Temporal Alignment of On‐Board Sensors for Railway Systems

To address the challenges of obstacle detection in railway environments, this paper proposes a highly robust perception method based on light detection and ranging (LiDAR) and camera fusion. The core of this method is an innovative spatio‐temporal alignment framework designed to resolve the critical issue of spatio‐temporal misalignment in conventional point cloud and image data processing. For temporal synchronisation, the framework employs the precision time protocol (PTP) to achieve microsecond‐level time synchronisation between the LiDAR and camera, thereby improving cross‐modal temporal consistency at their acquisition timestamps. In the spatial domain, the method eliminates the dependency on specific calibration targets by introducing an adaptive and targetless calibration strategy based on natural scene features. This strategy constructs a multi‐channel scoring function to maximise the mutual information (MI) between point cloud and image features, and it leverages a particle swarm optimisation (PSO) algorithm to efficiently solve for the sensors' six‐degree‐of‐freedom (6‐DoF) extrinsic parameters. Experimental results demonstrate that the proposed method achieves high‐precision online calibration in railway scenarios. By providing an integrated solution to the spatio‐temporal misalignment problem, this study effectively improves the performance of multi‐sensor fusion, significantly enhancing the reliability and accuracy of perception for railway perimeter security.