Research On Passengers Monitoring Technologies In Subway Station Based On Computer Vision

As an important part of rail transit,subway carries the important task of relieving the pressure of urban passenger transport.In recent years,the subway line in China has been expanding and the service capacity is developing.Subway has become an irreplaceable part of people’s life.However,in recent years,the safety accidents in subway stations happen continually,subway station internal personnel circulation is large,real-time monitoring of passengers is necessary.When staff spot suspicious people on surveillance video,they need to follow them up.The accidents of passengers falling down on escalators or stairs occur constantly.Whether due to human’s carelessness or escalator goes wrong,if they cannot be handled in time,it is likely to cause secondary injury to passengers.It can effectively reduce the harm caused by the fall by finding the passenger falling down in time and taking emergency measures.As an important node of urban rail transit line,subway station has a great pressure on passenger flow.The congestion often occurs in subway stations.At this time,not only the transportation service capacity is reduced,but also exist the risk of stampede due to congestion.If the passenger flow of subway station can be detected in real time and emergency plan can be made in advance,the safety accidents caused by congestion can be avoided effectively.The circumstances in subway station is complicated and has a huge circulation.People often lose objects in the station.The abandoned objects will not only affect the traffic capacity,but also exist threatening terrorist attacks.At this time,it is particularly important to find the abandoned objects in time and arrange staff to dell with them.In China,the layout of subway station monitoring system is impeccable,but the intelligent detection level needs to be improved.The usual means of investigation is to let the staff watch the monitoring all the time.If there is a dangerous event,it will be reported immediately.This method not only is a waste of manpower,but also has great possibility of missing inspection.In recent years,with the rapid development of computer vision,it has achieved remarkable achievements in object detection,object tracking,image identification,and it is of great value in many fields.Therefore,it is necessary to introduce the computer vision algorithm into the monitoring system of the subway station,which can not only make the function of real-time and precise monitoring of the safety status and service status of the subway station come true,but also economize a lot of financial and resources,and improve the overall operation level of the subway station.So as to guarantee the safety of passengers and the ordinary working condition of the station,this paper proposes a subway station passenger monitoring technology based on computer vision with the help of monitoring video.Aiming at four problems:Passenger tracking,fall detection,passenger flow detection,and abandoned object detection,four algorithms is proposed:(1)In allusion to the issue of passenger tracking,an object tracking algorithm based on KCF and DSST is proposed.The new object tracking algorithm combines the tracking filter of KCF with the scale filter of DSST to make the tracking characteristics of high tracking speed and adapting to the object variational scale come true.The algorithm is tested in some scenes of subway station,and good tracking results are obtained.(2)In allusion to the issue of passenger fall detection,a detection algorithm based on human joints information is proposed.The algorithm uses Mask R-CNN to segment the human body region,uses Open Pose algorithm to identify human joints,and realizes the purpose of judging whether the passenger falls or not by analyzing the joints information.The self-made fall video data set is used to verify the algorithm,which can achieve an effect of fast and accurate recognition of falls.(3)In allusion to the issue of passenger flow detection in subway station,a passenger flow detection method based on human head detection using YOLOv3 is proposed.In this method,the human head detection model is obtained by using the data set labeled with the head trained by YOLOv3 algorithm,so as to detect the human head and achieve the purpose of passenger flow detection.This method is used to detect several videos in subway station,and good detection results are obtained.(4)In allusion to the issue of abandoned object detection,an algorithm based on double Gaussian mixture background modeling is proposed.The algorithm detects the static foreground in the video image through the Gaussian mixture background model with different learning rates,and realizes the function of checking the abandoned objects.Through the relevant video verification,we can achieve the purpose of abandoned detection.This paper studies the key technologies of passenger monitoring in metro station based on computer vision,and proposes four detection algorithms for key scenarios,which can make the purpose of real-time and accurate detection come true.There is an important implication to make intelligent safety monitoring come true and ensure the safe operation of urban rail transit.