Research On Tunnel Traffic Safety Risk Evaluation Based On The Data Of Radar-Video Integrated Sensors

Road traffic safety is an essential component of urban transportation system development.With the deepening of safety research,safety assessment methods have also changed:from statistical analysis and evaluation based on accident data to traffic risk identification based on driving behavior data.Driving behavior is the result of the interaction between humans,vehicles and the environment.Traffic risk identification based on driving behavior data can effectively perceive the running state of vehicles and play a positive role in reducing the probability of accidents.However,due to the accuracy of data collection and the efficiency of data transmission,existing research is difficult to support the microscopic characterization of bad driving behavior.A new round of data acquisition technology innovation makes it easier to obtain high-precision and lowlatency vehicle trajectory data.The trajectory data contains rich driving behavior information,which provides strong support for the research of road traffic safety risk assessment driven by driving behavior data.Based on the data mining of the radar-video integrated sensor,this thesis proposes the identification principle and algorithm flow of bad driving behavior by selecting indicators that can reflect the characteristics of bad driving behavior and establishes a road traffic safety risk assessment model based on driving behavior data.The specific research contents are as follows:(1)Analyze the characteristics and advantages of the radar-video integrated sensor data and give the data processing flow.Compare the characteristics of traditional bad driving behavior identification data and integrated data,summarize their advantages and limitations,and clarify the advantages of integrated data in accurate identification of bad driving behavior and road traffic safety risk assessment.Based on the feature fields,give the analysis and processing flow of single device data extraction,multi-device data association and abnormal data.(2)Propose the identification methods of seven bad driving behaviors including speeding,rapid acceleration,unstable speed,rapid deceleration,abnormal low-speed,and bad car-following.Utilizing the integrated data,select the behavior characterization indexes,calibrate the threshold of the indexes,give the recognition principle and algorithm flow of each bad driving behavior,and analyze the spatial and temporal distribution of each bad driving behavior.(3)Establish a road traffic safety risk assessment model driven by driving behavior data and propose a method for dividing risk levels and grading thresholds.Introduce the theory of the entropy method and establish the evaluation index system,which has the first-level indicator as the traffic safety entropy and the second-level indicator as the bad driving behavior rate.Based on the entropy method,optimize the extreme value processing of secondary indicators and the calculation method of indicator weight.Based on the K-means algorithm,select the cluster number with the largest silhouette coefficient as the risk level number,and determine the optimal grading threshold with the highest model recognition accuracy as the goal.(4)Empirical research.Bad driving behavior rates have different spatial distributions.Considering this difference,the left lane of Jiaozhou Bay Tunnel in Qingdao is selected to carry out a verification study.Based on the bad driving behavior recognition algorithm,the bad driving behavior rate data of each road unit is calculated,and the objective weights of each index are given by using the improved entropy method.Then the traffic safety entropy value of each unit is obtained.The feasibility of the model is verified by comparing the distribution of accident data and traffic safety entropy data.Combined with the K-means clustering effect,the traffic operation risk is divided into two levels:high risk and low risk.The hierarchical threshold is 0.0507,and the grading accuracy is 92%.