Research On Identification And Tracking Of Moving Vehicles On Bridges

Bridge structure plays an important role in the development of national economy.With the rapid increase of the number of vehicles in our country,the vehicle load on bridges is increasing.Therefore,under the long-term vehicle load,the damage and fatigue degree of the bridge structure increase,which seriously affects the normal service life of the bridge structure.Therefore,it was very important to develop a detection system which can detect the load of moving vehicles on the bridge and track it in real time.In the first chapter,it mainly introduced the research status of traditional and new weighing system,and the application of computer vision technology in intelligent monitoring and recognition.And combined with the current research pain points,proposed the main research content of this paperIn the second chapter,this paper proposes a vehicle load identification method based on strain signal,which combines time-frequency analysis and CNN network to solve the problem of mobile vehicle load identification.Firstly,the strain signal in the middle of the bridge span was transformed into a two-dimensional time-frequency signal with more signal characteristics by time-frequency analysis method,and the time-frequency signal was transformed into a numerical matrix with fixed size of 64 ×64 by bilinear interpolation;secondly,the CNN network was used for regression training of the numerical matrix to establish the mapping relationship between the numerical matrix and CNN network.Finally,based on the coupling theory of vehicle and bridge,the strain signal of two axle and half vehicle in the bridge span under various speed and axle load was simulated by MATLAB and ANSYS.The feasibility of the method was verified by using CNN network and CWT transformation.At the same time,this paper also simulated the load identification accuracy of CNN network under different pavement roughness,and verified the effectiveness of the method in complex pavement.In the third chapter,the traditional compressed sensing target tracking algorithm has tracking drift,which leaded to the problem of inaccurate target tracking results.This paper maked the following improvements: firstly,the region with the highest probability of target existence was identified by the traditional compressed sensing target tracking algorithm;secondly,the coordinates of the target region were modified by Kalman filter;finally,the classifier was updated in the modified target region to prepare for the next tracking calculation.The feasibility of this method was verified by laboratory and field experiments.In the fourth chapter,the tracking frame scale of the traditional compressed sensing target tracking algorithm was fixed,which can not adapt to the visual size and attitude transformation of vehicles on the bridge from far to near.Therefore,this paper proposed a scale adaptive target tracking algorithm.Firstly,the traditional compressed sensing target tracking algorithm was used to identify the area with the highest probability of target existence;secondly,the target frame adaptive adjustment method proposed in this paper was used to automatically adjust the size of the target frame to cover most of the target area;finally,on the basis of the target tracking frame with modified scale,the classifier and boundary threshold of the tracking algorithm were updated.At the same time,the feasibility of this method was verified by laboratory test and field test.