Research On The Method Of Expressway Toll Path Recognition Based On The Networking Toll Data

With the adjustment of the national highway charging method,from the shortest circuit charging to the actual driving path charging,the scientific identification of the charging path has become the basis for the accurate operation of the high-speed network charging system.The existing toll path identification relies on the vehicle detector distributed in the road network to identify the vehicle driving path,but the problems such as detector equipment failure,communication interruption or unstable working conditions lead to abnormal charging phenomenon in actual use.To solve this problem,it is very important to establish a set of toll path recognition method independent of the detector system.The online charging data stores the information of vehicle access and time.This research focuses on the processing of toll data and the extraction of effective information for the construction of toll path recognition system.The basic ideas are as follows: denoising and fusion processing of toll data,and extracting traffic time information from it,and building a classification model to treat the paid vehicles according to the characteristics of traffic time distribution Path identification.In this paper,based on the analysis of toll data and the characteristics of high-speed road network,combined with big data processing technology and machine learning algorithm,the following research work is completed:(1)This paper introduces the format and characteristics of high-speed toll data,explains the feasibility of extracting traffic time data from toll data,explains and defines the related concepts of multi-path selection,and explains the differences and connections among reasonable path,effective path and toll path.(2)This paper analyzes the causes of abnormal toll data,and analyzes the necessity of traffic time data fusion between ETC(automatic toll)and MTC(manual toll).On this basis,the idea of data preprocessing is put forward: firstly,the data with missing fields or logic errors is eliminated and the effective data is screened out from the remaining data,then combining the characteristics of etc and MTC data,a dynamic method is proposed for data fusion.(3)The double-layer mapping method is proposed.In view of the fact that most toll stations in high-speed road network are located in a long road section,so they will not interfere with the road finding process,the data of high-speed road network are stored in two-layer data structure respectively,which can fully store the information of road network and maximize the number of compression nodes to save computing resources.(4)An ant colony algorithm considering the constraint of direction angle is proposed.Because the drivers do not fully grasp the road network information,they often choose the experience path rather than the strict shortest path.According to the pheromone concentration,the ant colony algorithm may converge to multiple paths,which can describe the above characteristics of road finding better.On this basis,the path set is constrained by the trajectory direction angle to accelerate the convergence of the algorithm,which can be used to solve the multi reasonable path problem in high-speed road network.(5)The EM algorithm is used to estimate the parameters of the mixed Gaussian distribution of travel time between OD pairs with multi-path.The BIC criterion is used to select the model,and the effective path set for toll path recognition is selected from the reasonable path set.Then,the hidden parameters obtained by EM algorithm are used as the probability of each path to be selected,and the sub Gaussian distribution parameters are used to calculate the probability of specific travel time in each effective path,so as to establish a Bayesian classifier to identify the toll path of vehicles.(6)The data of highway toll collection and road network in H Province are used for example verification.The toll data in 2018 is preprocessed to build a high-speed road network model.Take H east railway station to T railway station as an example to calculate the reasonable path set,extract the traffic time data of class 1 trucks between the two stations for processing and calculation,and build a Bayesian classifier.Because there is no travel path information in the original travel time data,it can't be used to verify the performance of the classifier,so we use the Monte Carlo simulation algorithm to build 50000 simulation data with path labels as the input of the test set,and the experimental recognition rate reaches 92.86%.Finally,the remaining data that can not be accurately identified are analyzed,and the vehicle inspector scheme is used as an auxiliary means to ensure the system efficiency and improve the overall recognition accuracy.Improve the scientificity and practicability of the method.