Traffic Conflict Prediction Model for Bottleneck Section of Expressway Construction Area Based on Video Recognition

Abstract

The highway enters the peak period of maintenance and repair, and the safety level of the bottleneck section of the expressway construction area is low at present, and the current status of related safety research is lacking. This study overcomes the problem of difficult data collection, and utilizes the data collected by the UAV video to fully exploit the advantages of rich data and relatively low cost. Based on the improved conflict identification TDTC model, vehicle video recognition and tracking, analysis of the relationship between traffic conflict and observable factors such as traffic volume, average speed, and car ratio, and comparing the applicable conditions of various models. It is found that the traffic conflict data set is relatively discrete and the variance and mean are very different, which is consistent with the negative binomial distribution. The negative binomial prediction model of traffic conflict is established and verified, and the number of traffic conflicts in the typical bottleneck section of the expressway construction area is confirmed. There is a significant correlation between traffic flow factors. The number of traffic conflicts increases first and then decreases with the increase of traffic volume, and reaches a peak when the traffic volume is 850pcu/h. As the average speed increases, the proportion of large vehicles increases, the number of traffic conflicts also keeps increasing. Increase. The above results show that the traffic volume and the proportion of large vehicles in the construction area can be reduced through traffic organization, and speed limit measures can be taken to improve the safety level of the construction area.