Trajectory Reconstruction And Delay Extraction At Signalized Intersections

At present,the main contradiction of China's traffic is that the increasing vehicle ownership does not match the road network system.However,the road network coverage cannot be expanded in a short time,so it is necessary to combine existing road resources to improve the use efficiency of the road network by using various optimization methods.The vehicle trajectory data contains rich traffic flow information,and analyzing these traffic flow information can help improve the efficiency of the road network.A large number of vehicle trajectory data can be obtained by traffic information acquisition equipment.However,in practice,the number of vehicle trajectory data is incomplete due to vehicle occlusion,equipment failure or human factors,which will have a great impact on traffic flow analysis and deep data mining.On the basis of reviewing the relevant research at home and abroad,combined with VISSIM simulation experimental data and field video data,this paper uses a variety of methods to reconstruct vehicle trajectories in different scenarios,and extracts vehicle delays by using the reconstructed vehicle trajectories.The specific work is as follows:(1)According to the relevant literature and the actual survey data,the operation characteristics of vehicles passing through the signalized intersection are analyzed in depth.This paper divides vehicle passing through signalized intersection into four stages,which are the constant speed stage,the deceleration stage,the parking stage and the start acceleration stage.There are uniform speed stage,deceleration stage and start-up acceleration stage in trajectory reconstruction stage.(2)In the case of missing part of the data,that is,a complete vehicle trajectory with one or more blank data,the vehicle trajectory is reconstructed by segments.In this scenario,three different reconstruction methods,kinematics method,least squares method and Hermite interpolation method,are applied to segment vehicle trajectory reconstruction using VISSIM simulation experimental data and field video data,and then the three reconstruction results of the same data source are compared and analyzed to show the advantages and disadvantages of the three reconstruction methods.The results show that the Hermite interpolation method is more suitable and effective.(3)Vehicle trajectory reconstruction in surveillance video scene.Video surveillance scene refers to the video data acquired through the intelligent monitoring system of highway vehicles,commonly known as Jack video.There will be some restrictions on the shooting angle,and subsequent vehicles will be occluded by the front car or not in the monitoring area.Therefore,this paper assumes that the stop time and position of the first four vehicles and the start time and position of the first four vehicles in each cycle can be acquired through the video card,and the vehicle trajectory of the follow-up vehicles can be reconstructed using the known data of the first four vehicles.This paper presents a method to determine the arrival time of vehicles by using the probability of vehicle arrival distribution.When the traffic flow is small,the arrival time of vehicles is determined according to the principle of Poisson distribution,and the parking position is determined according to the average headway distance at the intersection.Then,according to the characteristics of vehicle deceleration,the vehicle deceleration process is described by using the hyperbolic sine function,and the results are obtained.The time and position of vehicle starting deceleration and the track of vehicle deceleration;then the average starting time interval of vehicle is calculated according to the average headway of HCM2000 intersection,and then the starting acceleration time of subsequent vehicles is obtained.Finally,according to the characteristics of vehicle starting acceleration,the starting acceleration process is described by a linear reduced variable acceleration motion model to determine the vehicle returning to normal.The time point and position of the driving speed,as well as the vehicle starting and accelerating trajectory.Then the reconstructed results are compared with the original vehicle trajectory.(4)Using the reconstructed vehicle trajectory in the surveillance video scene,this paper extracts the vehicle delay.According to the model,the time point of vehicle returning to normal running minus the time of vehicle starting deceleration and then minus the time of vehicle passing through the intersection at normal speed is the delay caused by vehicle at the intersection.Finally,the delay of vehicle trajectory extraction is compared with simulation delay and manual video extraction delay.The results show that the absolute error of average periodic delay is less than 5 seconds and the relative error is less than 20%.