Research On Traffic Flow Detection Technology At Urban Traffic Intersections

With the continuous improvement of people's living standard,more and more people choose to buy to get the convenience of travel,but the massive increase of cars has brought huge pressure to the road traffic.Researchers have realized that it is increasingly urgent to build intelligent transportation systems(ITS)to solve the problem of urban road traffic congestion.The premise of intelligent traffic system to realize its function is to obtain real-time and detailed road traffic information,such as current traffic flow,congestion,driving speed and vehicle distance,so as to scientifically control urban roads.Therefore,vehicle detection and tracking research has been an important research direction for researchers.By analyzing the current vehicle detection and tracking technologies,it is found that there are still problems of missing and wrong detection,especially for vehicles with small targets in remote scenes with low detection accuracy and high missed detection rate.In view of the above technical problems,the research work and main innovations of this paper are as follows:1.The basic principles of vehicle detection and traffic flow statistics are studied,and the implementation principles and existing advantages and disadvantages of different algorithms are analyzed andintroduced in detail.Finally,it determines which detection algorithm is adopted in this paper.Based on the detection,the tracking and traffic flow statistics tasks are realized.2.This paper analyzes the reasons for the poor detection effect of small target vehicles in complex traffic sections,proposes an improved real-time detection algorithm of YOLOv3 vehicles combined with FPN to improve the overall detection effect of small target vehicles,and designs two levels of difficulty experiments.The experimental results show that the detection algorithm proposed in this paper improves the detection accuracy from 92%to about 95%,and the recall rate is also improved.3.The process principle of kalman filter was introduced in detail,and an intelligent vehicle flow statistics platform interface based on PyQT was designed and built for tracking and counting experiments,which effectively unified detection,tracking and vehicle flow statistics.The experimental results show that the accuracy rate can be improved to 92.7%by using this intelligent platform for vehicle statistics,and the missed detection rate can be reduced by about 3%,which is closer to the real number of vehicles and has good engineering application value.