Safety Evaluation of Unconventional Signalized Intersection Based on Traffic Conflict Extreme Model

To assess the safety of unconventional signalized intersections, a cross sectional analysis based on traffic conflict extreme model is proposed. The traffic conflicts and traffic flow data were extracted from 96 h video data, collected from three signalized intersections in Nanjing, using computer vision techniques. Further, a data level-processing level-prior level hierarchical Bayesian peak over threshold (POT) model is proposed. The model parameters were estimated using the Markov Chain Monte Carlo Simulation approach. The safety effect was calculated based on the predicted traffic crashes using the odds ratio technique. The results show that the safety of the unconventional signalized intersection with outside left lane is 21.8% higher than that of the conventional signalized intersection. The average left-turning radius, average hourly left-turning traffic volume, and heavy vehicle ratio of the left-turning volume have significant impacts on traffic crash risk. The larger the average left turn radius, the lower the traffic crash risk; the greater the average hourly left turn traffic volume and heavy vehicle ratio of left-turning volume, the higher the traffic crash risk. In conclusion, the hierarchical Bayesian POT conflict extreme model has a high utilization rate of the traffic conflict data, and can effectively describe the instability and heterogeneity of the traffic conflict extremes. Overall, it has a broad application prospect. © 2022, Editorial Department of China Journal of Highway and Transport. All right reserved.