Research On Urban Rail Passenger Flow Statistics Technology Based On Computer Vision

In recent years,the rapid development of urban rail transit has not only dredged the urban road traffic and shortened the commuting time of citizens,but also sharply increased the passenger flow of subway.When emergencies come,how to guarantee the safety of citizens in the underground network has become a major problem in subway operation.The real-time and accurate acquisition of passenger flow information can provide strong decision data support for improving the efficiency of operation and organization of subway operation management department and enhancing its ability to cope with sudden large passenger flow.In this paper,computer vision technology is applied to real-time passenger flow monitoring of urban rail transit.By further optimizing the current target detection algorithm with better performance,and designing the corresponding target tracking and passenger flow counting algorithm,accurate and efficient real-time statistics of subway passenger flow in the monitoring scene are realized.The main contents of this paper include:(1)Firstly,the basic principle of several vision-based passenger detection methods and the feasibility of their application in urban rail monitoring scenarios are explored.After a comparative study and test on the hogs +SVM detection method based on motion characteristics and machine learning,it was determined to use the single-shot Multibox Detector(SSD)algorithm based on deep learning as the detection algorithm for passenger targets.(2)Secondly,the framework and principle of SSD algorithm are studied,and the training,testing,learning contents and methods of SSD network are discussed.Aiming at the problem of insensitivity of small targets in traditional SSD algorithm,the detection algorithm is improved by means of densenet-based(dense connection network)basic network and structural adjustment prediction network,so as to achieve better classification and prediction of passenger targets.(3)Then,the tracking algorithm and passenger flow counting strategy corresponding to the detection algorithm are studied.Namely,the tracking class in OpenCV is used as the support of the algorithm,and the detection result-based KCF algorithm is used to track the passengers in video image in real time.The corresponding trajectory analysis and passenger counting algorithms are designed to realize accurate calculation and statistics of passenger flow and provide algorithm support for subsequent experimental verification.(4)Finally,the modified SSD passenger detection algorithm designed in this paper is tested by using video data intercepted by guangzhou metro hd monitoring system.The results show that the improved detection algorithm solves the problems existing in the traditional SSD algorithm.Taking the recognition rate and the rate of omission and error detection as indicators,the paper evaluates the results of the target tracking,track analysis and passenger counting algorithm designed in this paper.The results show that the passenger flow statistics algorithm designed in this paper based on computer vision can meet the needs of subway monitoring scene.