A framework for obtaining spatiotemporal data of vehicle wheel loads based on simplified 3D reconstruction and Weigh-in-Motion data

Loads play a fundamental role in structural analysis, and spatiotemporal data of vehicle load is promising for refined analysis of bridge structures. However, due to the complexity of real-world conditions, accurately and robustly locating wheel loads remains a focus for current vision-based vehicle load monitoring (VLM) systems. Based on geometric relationships among vehicle components, this study proposes an innovative simplified 3D reconstruction method for positioning vehicle wheel loads. Key components rather than the entire vehicle's bounding box are used for 3D reconstruction, enabling more precise key-point localization. Furthermore, by leveraging Weigh-in-Motion (WIM) data, license plate detection, and geometric relationships, the reconstruction process is significantly simplified. This eliminates the need for camera intrinsic parameters, and reduces reliance on wheel detection. Finally, the proposed method is validated through a field test on a representative long-span bridge and achieves remarkable comprehensive performance. The mean error (ME) and mean absolute error (MAE) of vehicle center positioning are 0.034 m and 0.091 m, respectively. Besides, the ME and MAE for lateral width estimation are 2.9% and 4.0%, respectively. Furthermore, the simplification prevented data loss for 11.2% of trajectory points in the case study, while maintaining a real-time throughput of 28 FPS on an RTX 4060 Laptop. Additionally, the deployment can be achieved at tripod‑level height, making it convenient for practical engineering applications. The proposed framework offers a comprehensive solution for accurate and robust vehicle wheel load monitoring on bridge decks, providing a data foundation for bridge assessment and maintenance based on actual traffic flow.