Identification And Evaluation Of High Safety Risk Intersections Based On Abnormal Driving Behavior

In recent years,with the gradual deepening of research in the field of traffic safety,traffic accident data and accident cause analysis have become key analysis objects.The traffic safety research system based on statistical methods or traffic conflict theory is relatively perfect.Compared with the accident data,the driving trajectory data has large volume and strong real-time,which can make up for the delayed evaluation of traffic accident data.Abnormal driving behaviors can be identified from the track data,which may lead to traffic safety hazards.Studying the distribution of abnormal driving points in the road network can evaluate the traffic safety problems,especially for the intersection location with poor safety,find the causes of traffic hazards,and establish a perfect intersection safety evaluation method.In this paper,an improved identification method of rapid deceleration driving behavior is established,and the abnormal driving behavior data(rapid deceleration data)is identified from a large number of driving trajectory data.Map matching the abnormal data points,and screen the hot spots of abnormal driving behavior through nuclear density analysis.Finally,analyze the driving influencing factors,evaluate the traffic safety characteristics of the intersection,verify and improve the problems by using the existing data,and determine a complete set of intersection traffic safety evaluation model,which can better assist the comprehensive evaluation research of the intersection.This paper has certain significance in the field of evaluating the traffic safety of urban intersections.It can provide support for the design and transformation of urban intersections by analyzing the safety of traffic design and traffic facilities.The main content of the paper can be divided into the following three parts:Firstly,the driving trajectory data is preprocessed,and the three most common abnormal driving behavior data are analyzed by using indicators such as occurrence frequency,data volume and severity,that is,rapid deceleration,rapid acceleration and sharp turning behavior data.Finally,the rapid deceleration driving behavior that can most affect driving safety is selected for research.Aiming at the problem of commonly used fixed threshold screening methods,this paper establishes a set of improved rapid deceleration behavior recognition method including a variety of scenes and parameters,which is verified by actual data,so as to screen abnormal driving data from driving trajectory data.Secondly,the filtered abnormal driving behavior data is used for vector map location matching,and the nuclear density level of each area of the road network is analyzed.The nuclear density level is classified into low density,medium density,high density,high density and very high density.Finally,the hot spot area of abnormal driving is determined,and the core location is mostly near the key intersection,that is,the location of the intersection with high safety risk is determined.Finally,the traffic risk factors of intersections are studied,and the root causes of abnormal driving behavior in intersection areas are analyzed,that is,the safety impact caused by intersection traffic design,order and environment.Sort out the problems existing in the existing statistical analysis and traffic conflict methods,and use the abnormal driving behavior data as a supplement to fully evaluate the internal and external traffic safety of the intersection.Propose the intersection driving and conflict risk score IDCs and cluster the safety level.Finally,the influencing factors of intersection safety are listed,the correlation degree is determined by grey correlation analysis,the weight of various safety factors is determined by SOM neural network,a complete intersection safety grade evaluation model is constructed,and the intersection grade classified by kernel density of trajectory data is used for verification.