Vehicle Load Point Monitoring Data

This is the relevant data in "Monocular Homography Estimation and Positioning Method for the Spatial-Temporal Distribution of Vehicle Loads Identification".

Real-time measurement of the traffic load spatial-temporal distribution is crucial for bridge health monitoring and operation maintenance. Despite the significant progress made by computer vision-based measurement methods, their accuracy and automation still require improvement. This paper proposes a novel method that eliminates the need for on-site calibration. Firstly, the YOLO-v5 model is employed to detect vehicles in surveillance videos. Subsequently, this paper proposes a prediction model for the vehicle equivalent concentrated load that combines a pre-trained convolutional neural network (CNN) coding model and a BP neural network. The prediction model's error is constrained within 3%. Finally, this paper presents a homography matrix calculation algorithm based on the geometric priori information of lane lines, which enables the transformation of image coordinates into actual coordinates without on-site marking. The effectiveness of the proposed algorithm has been evaluated through a field test and benchmarked against traditional methods. The result shows that the proposed method outperforms traditional approaches, manifesting in a notable reduction in the need for manual calibration and a substantial improvement in the accuracy of the model.