Research On Passenger Flow Statistics System Based On Machine Vision

With the wide application of video monitoring system in the current society,people have more and more research in the field of intelligent monitoring.Among them,passenger flow statistics technology has a good development prospect in video monitoring.Passenger flow statistics can provide more intuitive and effective decision-making basis for public places,but it is difficult to achieve complex passenger flow statistics by artificial counting,infrared sensing and other technologies,so machines are used.Vision technology can achieve better passenger flow statistics,so it is of great significance to study the passenger flow statistics based on machine vision.In this paper,the system of passenger flow statistics is designed and implemented under the background of monocular camera oblique downward shooting,which can complete the real-time detection,tracking and counting of passenger flow.Firstly,the video sequence image is preprocessed to reduce the irrelevant information in the video image,enhance the useful information and simplify the image data,and improve the image processing speed.In the aspect of human body recognition,a moving object detection method based on the combination of background difference method of Gaussian mixture background modeling and three frame method is adopted to improve the accuracy of moving object detection.Considering the problem of partial occlusion of human head and the shortcomings of Hough transform detection circle,an improved Hough transform circle detection method is proposed.This method can detect the arc quickly and improve the accuracy of Hough transform circle detection.It improves the real-time performance of pedestrian detection.This paper analyzes the common methods of moving target tracking,and proposes a tracking method based on Kalman filter to match the head contour feature frames.Kalman filter algorithm is used to predict the next frame motion of the current target.After the optimal prediction results are matched with the pedestrian target of the next frame,the target tracking chain is updated in real time,and the in and out marker lines are set to achieve the goal of customers.Traffic statistics.On the basis of the improvement of the above passenger flow statistics algorithm,this paper designs a complete passenger flow statistics system,and finally makes an experimental analysis of the statistical system.The experimental results show that the passenger flow statistics system can detect pedestrians and track them in real time,improve the accuracy and robustness of passenger flow statistics,and lay a good foundation for further research.