The inspection of toll evasion has always been an important task in the operation and management of highways.Traditionally,the audit department mainly relies on video surveillance,manual technical analysis,and other people’s reports to crack down the evasion behaviors,which has the disadvantages of large workload and low auditing efficiency.With the perfection and maturity of the networking charges and increasingly sophisticated means of evasion,the expressway management departments urgently need to use new technologies to increase the level of evasion inspections.On the other hand,the high-speed road network operating system generates a large number of vehicle traffic flow data every day,such as inbound and outbound traffic records,cross-section or bayonet traffic records,and online charging split records.Although these data are fragmented from a single point of view,correlating these data can give a general idea of the behavior of vehicles in the highway network.When traffic evasion occurs on the vehicle,the flow data will inevitably show abnormal traffic behavior.Based on this,this paper combines the techniques of big data and data mining,applies intelligent computing and other methods,and proposes a research of vehicle evasion inspection system based on big data.In this paper,we first study the eleven types of traffic anomaly related to vehicle evasion,and give an algorithm based on traffic data to analyze the abnormal behavior of vehicles.Then,the vehicle credit degree is taken as the index of suspicious degree of vehicle evasion,and the multi-attribute utility model is used to calculate the vehicle credit degree.On this basis,the model was improved by using BP neural network.Finally,combined with actual projects,the vehicle evasion inspection system was developed and implemented.The main research work and innovations of the thesis include:1.Build a big data platform for evading inspections.Analyze and sort the data of toll collection and vehicle flow in each monitoring subsystem of the existing expressway,and design a data base set suitable for audit analysis requirements.Based on this,through the development of data acquisition programs,the automatic collection of data is accomplished.At the same time,a method based on the combination of partition and sub-table mechanism is proposed to realize the analysis and operation of data under big data.Information reconstruction technology is used to deal with the lack of information and information error in the charging system data by establishing the vehicle profile.2.Eleven types of abnormal behavior related to evasion fees were proposed.For each type of normal behavior,its mining data analysis algorithm based on traffic data is studied.Through the analysis of mass traffic data,a vehicle traffic behavior archive was formed.3.The vehicle credit index and vehicle evasion inspection model and algorithm are proposed.In this paper,the relationship between the number of abnormal behaviors and the vehicle credit degree is modeled,and the vehicle credit degree calculation method based on multi-attribute utility model is given.The analysis of the traffic data of a highway company in the past three years shows that the model has certain effectiveness.This paper also uses the BP neural network to improve the model.Experiments show that the BP model improves the accuracy of the audit from 33% to 67%,and has achieved satisfactory results.Finally,this paper also discusses a fuzzy evaluation model based on vehicle credit rating,which further improves the validity and correctness of credit calculation.4.Developed and implemented a vehicle evasion inspection application system based on big data.The system integrates functions such as vehicle traffic data collection,abnormal behavior mining analysis,and vehicle evasion fee auditing human-machine interface.It realizes real-time and automatic analysis of vehicle evasion costs and satisfies the requirements of practical applications.The system developed in this paper has been actually running on the line,and the system’s fugitive inspection can be very effective,which greatly improves the work efficiency of the audit department.The research results of this paper have a good reference for the application of big data technology. |