Research On Vehicle Trajectory Reconstruction Method Based On Discrete Monitoring Data

In the process of intelligent transportation system becoming more and more perfect,the application value of vehicle trajectory information is constantly discovered.Vehicle trajectory contains important traffic laws,which can be used to achieve traffic monitoring,location prediction,vehicle tracking,route recommendation and other applications.Therefore,the research focus of intelligent transportation is to extract and analyze vehicle trajectory data.With the development of technology,vehicle location information can be easily obtained by means of monitoring video,GPS positioning,etc.However,due to the influence of many factors such as natural environment and equipment erection,vehicle location information is often discrete,and the vehicle trajectory reconstructed from discrete monitoring data is often missing,which seriously affects the analysis and mining of trajectory,reducing the actual effect of the application.Therefore,effective methods are needed to reconstruct vehicle trajectory from discrete monitoring data,and fill in the missing trajectory to improve the quality of trajectory data.In order to solve these problems,discrete monitoring data selects traffic monitoring video data and intersection license plate detection and recognition data.This paper constructs a vehicle re-identification image library based on the discrete monitoring video.On this basis,a vehicle re-identification method based on multi-hypergraph features fusion combined with temporal-spatial correlation is proposed,which is used to reconstruct vehicle trajectory from discrete video points.To solve the problem of missing vehicle trajectories,a filling method of missing vehicle trajectories based on Floyd route finding algorithm is proposed,which improves the effectiveness of the discrete traffic data.The main research work of this paper includes the following points:(1)Based on discrete traffic monitoring video data,a vehicle re-identification data set containing traffic parameters is constructed.In order to effectively use the traffic monitoring video data,this paper uses vehicle detection and tracking algorithm to identify the vehicles in the video,then uses traffic flow parameter extraction method to obtain the temporal-spatial information related to each vehicle,and stores the vehicle images at multiple monitoring points together with the traffic parameter information.A vehicle may appear in multiple monitoring images.Therefore,the images of the same vehicle in different positions are divided into a group of image sequences by matching the re-identification program and manual screening.The image library of cross-camera vehicle re-identification is constructed,which not only improves the available value of video data,but also provides a data basis for the research of vehicle re-identification methods.(2)Based on the constructed vehicle re-identification image library,a vehicle re-identification method based on multi-hypergraph features fusion method is proposed.In this paper,the constructed vehicle re-identification image database is used as an experimental data set,SIFT,HSV and ORB are used to extract different features of vehicle images,each feature is represented by hypergraph,multiple features are fused and analyzed by means of multi-hypergraph features fusion,image similarity vectors are obtained by hypergraph learning,temporal-spatial correlation analysis is carried out on vehicles in combination with traffic parameters in the data set,image similarity is rearranged in combination with temporal-spatial distance,finally image sequences of query images in other positions are found,and cross-camera vehicle re-identification is realized.The comparative test proves the advancement of this method in vehicle re-identification.(3)Based on the vehicle position data of the two regional road networks,the filling of missing vehicle tracks based on the Floyd algorithm is completed.The original data used in this paper are the vehicle license plate,location,time,longitude and latitude coordinates of intersections and other data collected by monitoring equipment.However,due to the reason of equipment installation density and accuracy of data collection,there are various data missing in the original data.In order to complete the missing information in the vehicle trajectory,data preprocessing operations such as redundant data deletion,classification and sorting and intersection information supplement are carried out.Aiming at the sorted trajectory data,a method for filling the missing trajectory of the vehicle based on Floyd route finding algorithm is proposed,which realizes the filling of the missing trajectory in the vehicle driving process and improves the integrity and usable value of the data.