Research On Traffic State Perception Technology Based On Video Structured

In recent years,with the rapid development of my country’s economy and increasingly convenient transportation,my country’s car ownership has increased exponentially,and the risk of pedestrian travel has gradually increased.In our country’s urban road traffic monitoring system,there are a large number of actual urban road intersection monitoring videos.Through the surveillance video of the intersection section,you can understand road traffic and road condition information.But it is unable to judge the traffic state of the road in real time,so that abnormal events such as congestion evolve into more complex events,which eventually leads to a sharp increase in traffic pressure,thereby affecting normal traffic.Therefore,this article aims at the increasingly serious situation of my country’s urban road traffic monitoring pressure combined with the actual situation of urban road traffic intersection monitoring.Research a kind of traffic state perception technology to effectively collect and process traffic information and calculate structured traffic flow data to perceive the traffic state.This paper aims at the problems of small pedestrian and vehicle targets in the scene of congested traffic at complex intersections,and serious occlusion between vehicles and vehicles,pedestrians and vehicles,and pedestrians and pedestrians.Design and construct road traffic vehicle and pedestrian data sets using the abundant road surveillance video in the current video surveillance system for road traffic intersections in my country.Design a pedestrian-vehicle mixed detection algorithm,measure structured traffic flow data and analyze congestion.In this paper,F-CSPnet is designed to be added to the feature extraction network on the basis of the YOLO V4 target detection algorithm,and modify the loss function and NMS algorithm.Realize pedestrian and vehicle detection and tracking and calculate road structured data combined with video structured methods to perceive traffic status.This technology improves the pedestrian and vehicle mixed inspection algorithm according to the actual application of the road intersection scene,so as to improve the detection accuracy and operating efficiency of the algorithm.Combine and integrate technologies such as human and vehicle target detection,target tracking,and structured traffic data calculation,and perceive the status of road traffic and traffic through video structured means.This paper adopts the human-vehicle hybrid detection algorithm to build a road traffic perception system,detects and analyzes the operating status of people and vehicles in the video surveillance area and regional traffic status,detects and recognizes the road congestion status,calculates structured traffic flow parameter data,and realizes Perception of the traffic state of the detected road section.The system performs realtime statistics on the number of vehicles and pedestrians on the road monitoring site,and stores the vehicle structure parameters in the database.The average detection accuracy of pedestrian and vehicle targets has increased from 91.86% to 94.12%.