Research On Real-time Traffic State Identification And Accident Risk Early Warning Model Based On Traffic Flow

Reducing the occurrence of traffic accidents is the main task faced by the expressway operating units for a long period.At present,only the prevention of traffic accidents cannot meet the realistic and future development requirements of freeway.The paper Introduced early warning ideas into traffic safety management,build a real-time and efficient traffic accident risk warning model,reduce freeway traffic accidents,and provide safe and efficient experience for freeway travelers.Based on the influence of traffic flow on traffic safety,the paper fully exploits the potential regularity and characteristics between freeway traffic accidents and traffic flow features.Based on traffic flow,a real-time accident risk warning model will be constructed,which provide safety management measures for freeway traffic management.Firstly,comparing study of traffic flow parameters with traffic safety correlation of different landform highways.Drawing a conclusion that the natural traffic volume,traffic equivalent,passenger-to-goods ratio,percentage of follow-up,congestion,and time occupancy are related to two types of high-speed traffic safety.The range of correlation coefficient with plain freeway is [0.8,0.9],the range of correlation coefficient with mountain highway is [0.7,0.8].The traffic flow parameters are used to lay the foundation for the study of the regularity between traffic flow and traffic safety.Then,the traffic flow data is subjected to data pre-processing such as abnormal data elimination,data conversion,Z-score normalization.The original 17 variables in the traffic detector are reduced to 6 variables.6 traffic flows are taken from the perspective of the security domain.The parameter combination is mapped to the state space,and BP neural network is designed to classify the traffic safety domain boundary to achieve the safe state and non-safe state classification,and the model validity evaluation index is proposed.The experimental results show that the data pre-processing greatly improves training effect of the model.The BP neural network model in the security domain boundary estimation is better than the support vector machine(SVM)and the logistics regression model.The correct rate of traffic state recognition is 83.4%,and the accident correct rate is 66.7%,the false positive rate was 6.6%.Finally,the operating safety of vehicles is analyzed under different traffic flow conditions.The principal component comprehensive evaluation method is used to evaluate the mixed traffic flow state.The probability density estimation method is used to establish the risk warning level standard,and the actual traffic flow detection data is used to verify the validity of the early warning model and the rationality of the grading standards.Through the research of early warning model of traffic accident risk on freeway,this paper can transform the highway traffic safety treatment problem into active warning and improve the traffic safety level.The safety management measures are provided for the expressway operation department,research of the paper has certain theoretical significance and social significance.