Abnormal Driving Behavior And Road Traffic Safety Risk Identification Method Based On Trajectory Data

The problem of urban road traffic safety is becoming increasingly serious,and traffic accidents are the traffic problem with the greatest loss of life and property and the widest social impact.Research has shown that drivers’ abnormal driving behavior is the main cause of traffic accidents.With the continuous growth of the number of drivers and the rapid development of urban development planning and road infrastructure construction,more and more types of driving habits have been stimulated among different drivers.However,driving behavior is difficult to analyze and characterize,and cannot be applied to the field of traffic safety.With the rapid development of network communication and data collection technology,vehicles operated by transportation enterprises have generated massive trajectory data,providing a new source of data for studying driving behavior recognition and road traffic safety risks.This paper proposed a method for identified abnormal driving behaviors of drivers on different segments of urban roads by mined GPS(Global Positioning System)trajectory data generated by operating vehicles,and constructed a road safety risk indicator system based on abnormal driving behavior data.Based on this identification method,a road safety risk assessment model was established to study the impact of trajectory data,driver abnormal driving behavior,and road safety risks.The main research contents are as follows:(1)A method of GPS trajectory data processing and road segment matching was proposed.Due to the inability of GPS original trajectory data to meet the accuracy requirements of research targets,a method of data preprocessing based on data characteristics was proposed.After that,in order to meet the road segment matching problem of trajectory data,the road network was segmented and labeled used the road indefinite length method,and the HMM model was used to corrected the small range offset of trajectory points and matched the road segments.GPS trajectory data with road segment numbers were obtained to provided data input for the abnormal driving behavior recognition model.(2)Abnormal driving behavior recognition model was constructed.The relationship between traffic safety and various abnormal driving behaviors was described.Based on the different behavioral characteristics of driving behaviors,the identification methods of various driving behaviors were pointed out,and corresponding identification feature indicators and threshold determination methods were summarized.After that,the process and algorithm for identifying abnormal driving behaviors based on threshold values were designed,and a model for identifying abnormal driving behaviors based on trajectory data was established,provided a high latitude,operable,and real-time method for identifying driver abnormal driving behavior data for urban traffic safety management,and provided a data source for urban traffic safety risk identification.(3)The extraction of driving feature parameters and clustering of driver driving styles were completed.In order to solve the respective limitations of two methods for identifying abnormal driving behavior based on threshold value and unsupervised learning,the two methods were organically combined.Firstly,based on factor analysis of driving behavior characteristic parameters,the comprehensive reconstruction of driving characteristics was realized.Afterwards,K-means clustering was used to classify the driving style of drivers,and the classification results can serve as a basis for driver supervision.(4)A road safety risk scoring model was established and its effectiveness was verified.The recognition threshold for abnormal driving behavior was determined,and the quantification of abnormal driving behavior data on road segments was completed.An evaluation index system was established based on abnormal driving behavior data,and the weight of the index was determined based on AHP-EW integrated weighting method.The fuzzy comprehensive evaluation method was used to complete urban road safety risk assessment,and the effectiveness of the model was verified by an example.The proposed model can achieve the quantification of urban road safety risk based on abnormal driving behavior data,providing theoretical and data support for traffic management departments to accurately locate and scientifically supervise areas with potential safety hazards in the road network.