Research on the long tail distribution of pedestrian flow at signalized intersections

Abstract

Pedestrian crossing at signal-controlled intersections is a hot and challenging problem in slow traffic research. On the one hand, the signal-controlled intersection is a typical mixed traffic environment where pedestrians crossing the street as well as vehicles, will constantly conflict and threaten road traffic safety. On the other hand, compared with other parts, the tail of pedestrian flow occupies a large amount of intersections’ space and time resources, which greatly hinders the passage of right-turning vehicles. Therefore, it is significant to study the characteristics of pedestrian crossing, especially the characteristics of the tail of pedestrian flow crossing at the intersections. It can not only enrich relevant research and provide reference for optimizing the timing of intersection signals, but also improve the pedestrian crossing environment and alleviating traffic congestion. In this paper, the video data of the intersection of Yanta North Road and Jianxi Street, the intersection of Xiaozhai West Road and Hanguang Road, as well as the intersection of Keji Second Road and Taibai South Road were collected by UAV. The track data of pedestrians and vehicles were extracted by OpenCV, YoloV3 and DeepSort. The pedestrian quantity-time curve was plotted for single one way and multiple two-way street crossings. It is found that there is a good fitting effect between the curve and one-dimensional Gaussian function. Based on this, the actual speed and average waiting time of right-turning vehicles passing the crosswalk are compared respectively under the long tail distribution of different significance degrees. The result shows the long tail distribution of pedestrian crossing can obviously reduce the speed of right-turning vehicles and prolong the average waiting time.