Research On Accuracy Compensation Of Weigh-in-Motion System Based On Video Analysis

Highly accurate weigh-in-motion systems(WIMs)are important for truck overload detection,weight-based tolling,and structural health monitoring of roads and bridges.WIMs identify the dynamic force of each axle of the vehicle on the road surface in order to complete the weighing of the whole vehicle.However,WIMs in practice due to the complex road traffic environment,although the system is formally put into use before the accurate parameter calibration,but there are still some vehicles can not be accurately weighed phenomenon.Since different drivers have different driving behaviors,the change of vehicle driving status when different types of vehicles pass through the weigh-in-motion area is the main reason that affects the weighing accuracy.Therefore,this paper analyzes many influencing factors that affect the weighing error of WIMs in order to extract the information of vehicle label and axle type as well as driving state information of vehicles in the weigh-in-motion area.And the experimental research is carried out to analyze and compensate the impact of different driving status on the weighing accuracy of vehicles,which is mainly as follows:(1)The framework of accuracy compensation method for WIMs based on video analysis.Firstly,the basic situation of WIMs is introduced,then the sources affecting the weighing error of WIMs are thoroughly analyzed,secondly,the framework of WIMs accuracy compensation based on video analysis is constructed,and finally,each part of the overall idea of WIMs accuracy compensation is briefly described.(2)Establishing the marker and axle type detection model of vehicles in the weigh-inmotion region.Since the number of features in the axle type detection region of the weigh-inmotion region vehicles is much larger than the features in the vehicle marker region,resulting in the difficulty of accurate detection of small targets.In view of the fact that existing target detection algorithms cannot meet the detection requirements of such scenarios with large target spans,the detection performance of the model in this application scenario is tested by experimental validation,drawing on many advantages of existing target detection algorithms,such as multi-scale convolutional feature fusion,anchor-based mechanism and regression classification.(3)Weigh-in-motion region vehicle’s driving state information extraction method.Aiming at the existing vehicle detection and tracking model with poor stability,influenced by light,frame by frame detection omission and frequent change of vehicle ID number when multiple targets are tracked,etc.A detection-based target tracking model adapted to fixed scenes is proposed to improve the stability of target detection while ensuring its real-time performance.The fixed interval trajectory is smoothed using the RTS(Rauch-Tung-Striebel)method on the Kalman filter-based tracking algorithm.Finally,the stability of the model for vehicle tracking in this application scenario is tested by experimental verification.(4)Fusing weigh-in-motion data to analyze the impact of vehicle driving status change on weight accuracy and compensate the experimental validation.The video of the weighing process of different types of vehicles passing through the weigh-in-motion area and the corresponding weighing results are collected,and the vehicle label and axle type information as well as the driving state information of the corresponding vehicles are extracted.Then,we analyze and compensate the impact on the weighing accuracy according to the different driving status changes of vehicles,in order to verify the effectiveness and reliability of the accuracy compensation method of WIMs based on video analysis proposed in this paper.Establishing weigh-in-motion accuracy compensation system based on research and data support to realize video analysis and accuracy compensation of vehicle weighing process.