Pedestrian Warning Algorithm Research Based On Traffic Intersection Video

In China,with the increase in the number of cars,more and more developed public traffic,people travel more quickly and easily.But,there are also more and more traffic accidents.According to the statistics of the relevant departments,the accident mortality rate of road traffic is the highest among all types of traffic accidents.In these traffic accidents,a large part is due to pedestrians do not comply with traffic rules and motor vehicle drivers did not pay attention to the intersection's condition.Through the actual scene of the study,timely alerting the distant vehicle when the pedestrian crossing at the intersection of roads will be good way to avoid such traffic accidents.On the other hand,with the rapid development of machine vision technology and image processing,through the analysis of traffic junction video analysis,you can determine whether there are pedestrian targets into the monitoring camera,which in accordance with the results of the timely warning to the vehicles,to avoid the occurrence of traffic accidents.This paper mainly studies some key problems involved in the pedestrian warning algorithm based on road intersection video.The main contents are as follows:(1)The detection of moving targets and the identification of pedestrian targets.In the pedestrian target detection part,this paper compares several common foreground detection algorithms,and obtains the foreground detection algorithm suitable for the road intersection through theoretical research and experimental comparison at the road intersection.In the identification of pedestrian target,this paper introduces the algorithm of feature extraction and classification of the moving targets.With regard to feature extraction,this paper mainly discusses several common principles and extraction of the image features.Through experimental comparison,this paper verified the validity of the histogram of Oriented Gradient(HOG)feature extraction.A multi-class classification method based on Support Vector Machine(SVM)is proposed.The experiment proves that this method has high classification accuracy in the road intersection scene.(2)Pedestrian target tracking and movement direction judgment.In the road intersection,there are multiple pedestrian targets at the same time.Thus,it's a multi-target tracking problem in this scenario.In this paper,the performance difference between the multi-target tracking algorithm based on Kalman filter and the multi-target tracking algorithm based on Joint Probabilistic Data Association(JPDA)is compared.Experiments show that the multi-target tracking algorithm based on JPDA is superior to the tracking performance in road intersection scene.(3)Parallel processing of algorithm to improve real-time performance of pedestrian early warning algorithm in the road intersection scene.Timely warning of pedestrian targets requires high real-time performance to reflect its value in avoiding traffic accidents.On the other hand,in order to ensure the accuracy of the algorithm,this paper adopts the foreground detection algorithm,pedestrian target classification and pedestrian target tracking algorithm with large computational complexity.This paper proposes the parallelization of the above three-part algorithm,through the way of thread scheduling to ensure high real-time performance.Experiments show that the method is effective and feasible,and the computation time is obviously reduced.