Design And Implementation Of Passenger Flow Monitoring And Early Warning System Based On Deep Learning

In recent years,the rapid development of urban rail transit,increasing passenger flow.With the continuous growth of passenger flow,the scale of line network and the rapid growth of passenger flow bring great challenges to the safety of rail transit operation.The rail transit passenger flow monitoring and early warning on artificial control,unable to real-time access to the station hall,and other important platform monitoring area of structured information of the passenger flow,the overall technical means a single,monitoring precision is not enough,real-time performance is poorer,lack of intelligent monitoring and quantitative means,high precision of passenger information cannot be effectively to high speed access to accurate real-time traffic information,Moreover,it is difficult to support accurate and reliable passenger flow prediction and early warning and reasonable and effective decision-making.The shortcomings of passenger flow monitoring restrict the development of rail transit informatization.Passenger flow monitoring and early warning system based on deep learning is a system integrating real-time passenger flow monitoring,statistics and analysis.The system mainly includes five modules,one is real-time passenger flow information display,the second is passenger flow statistics and analysis,the third is pedestrian track tracking display,the fourth is abnormal behavior recognition,the fifth is offline video detection.The real-time passenger flow information display module realizes the extraction of basic information such as passenger flow,passenger flow density and pedestrian speed,and provides the display function of different dimensions and granularity of passenger flow monitoring information.The passenger flow statistics and analysis module carries on the statistics and analysis to the passenger flow information,realizes the visual expression of the data,and presents it in the form of charts;The pedestrian track tracking display module realizes the track tracking of pedestrians and displays the track in real time.The abnormal behavior identification module is based on pedestrian track tracking,which can identify the behaviors of pedestrians intruding into dangerous areas or work areas.The system not only provides data support for passenger flow control of rail transit stations,but also provides help for safe operation of stations.The passenger flow monitoring and early warning system is suitable for the current automation and information management mode,which provides convenience to the rail transit control and safety staff,and can effectively solve the problems of real-time data collection and data processing lag.In terms of algorithm,this system uses deep learning algorithm to process video images,and targets are detected by means of YOLOV3 algorithm.The DEEPSORT algorithm was used to extract the trajectory of the target.Behaviour analysis method based on motion trajectory is used to distinguish abnormal behaviors,and illegal rules are set according to different abnormal types.When the rules are triggered,an alarm will be displayed.In the development of the system,B/S technology system is adopted.Web development framework chooses SSM framework;The proxy server uses NGINX to achieve load balancing;RTMP streaming media server is used for video collection and real-time transmission to the client,My SQL is used for database,and Echarts plug-in is used for data graphic display.This article completely in accordance with the software engineering technology,from the background research,key technology research,demand analysis,the general design of the system,the detailed design of each module,as well as the system test narrative.At present,the basic functions of the system have been developed,and the system has passed the test and entered the trial operation stage.During the trial operation stage,the system is deployed in XXX(Beijing)Network Technology Research Institute,and will be subsequently deployed in a subway station of a subway company in the south.During the trial run,all functions were normal,and the problems of real-time data collection and data lag were solved.