Tracking And Modeling Pedestrian Movements At Intersections Based On UAV-Video

With the rapidly increases in the number of “shared bicycles”,in high-density urban traffic areas,due to the lack of corresponding bicycle lanes,the driving space and safety cannot be guaranteed,more and more cyclists are beginning to occupy pedestrian ways.This deepens the characteristics of mixed-traffic in China and brings new challenges to urban traffic management.The conflicts among pedestrians and cyclists are becoming severe,especially at the complex intersections,which led the pedestrians to decelerate or change direction for evasiveness,increasing the travel delays and threatening pedestrian safety.Considering that the existing pedestrian simulation model is mostly based on the foreign urban transportation system in which the bicycle lane has been set,neglecting the conflict between pedestrians,bicycles and motor vehicles at the intersection,and it is not suitable for the motion simulation of the domestic pedestrian flow.Therefore,how to analyze the complex behavior patterns of pedestrians under the mixed-traffic flow,combined with the actual crossing behavior of pedestrian,establish an improved pedestrian model suitable for domestic intersections,which is not only necessary for the traffic planning,management and design,but also important for the establish of the domestic urban traffic simulation system.To this end,this paper proposes an unmanned aerial vehicle(UAV)-based method for tracking and simulating pedestrian movements at intersections.Based on the social force theory,the conflict model is firstly developed to deals with the new problem of large amount of interference between bicycles and pedestrians on the crosswalk.And combine with the pedestrian crossing behavior at intersections,the positive impact of pedestrian green flash signal phase on pedestrian movement is transformed into the signal attractive force and a pedestrian simulation model which suitable for the mixed traffic environment in China is established.In addition,the high-resolution videos of intersection acquired by a UAV are innovatively used to recognize and position moving targets with a parallel feature pyramid network.A real pedestrian trajectory is extracted,and the maximum likelihood estimation method is performed to calibrate the improved social force model.Finally,in order to verify the validity of the proposed model,the prediction results of the traditional social force model are introduced for comparison and the real pedestrian trajectory data is also used to verify the simulation effect of the new model.The study selects an intersection in Shenzhen as the experimental area and the results show that the improved pedestrian model can more accurately and reasonably describe the pedestrian random behavior under complex conditions.It is of great significance for the study of domestic pedestrian behavior and provides new idea for urban traffic managers to designing and optimizing the traffic organization at the intersection.