A WIM-Integrated Method for Bridge Traffic Load Distribution Identification

This study proposes a full-bridge traffic load monitoring method integrating weigh-in-motion (WIM) systems with computer vision to address the need for dynamic traffic load perception in bridge health monitoring. A collaborative analysis framework is established, consisting of vehicle detection, trajectory tracking, cross-camera trajectory fusion, and WIM matching, overcoming the limitations of spatial and temporal resolution in traditional and singular monitoring and enabling real-time visualization of traffic load distribution. First, a vehicle detection model with multi-scale feature enhancement is constructed based on deep convolutional neural networks, incorporating a channel-spatial dual attention mechanism to optimize target representation under complex backgrounds. Then, accurate vehicle trajectory tracking is achieved through multi-object tracking within each camera view. Furthermore, a cross-camera matching strategy is proposed, jointly modelling appearance and dynamic features to fuse vehicle trajectories across different perspectives. Finally, WIM data are precisely matched with visual detection results based on timestamps and lane information, enabling accurate vehicle load acquisition. Experimental results demonstrate that the proposed method effectively improves detection accuracy, trajectory continuity, and data matching precision, providing reliable support for bridge structural safety assessment and intelligent traffic management.