Research On The Method Of Single Side Passenger Flow Detection Of Subway Channel Based On The Improved DaSiam-RPN Algorithm

The normal operation of the subway station is of great significance to the operation safety of the whole urban subway network.Accurately grasping the passenger flow operation status of the station is the fundamental basis to realize the scientific management and control of passenger flow.In view of the urgent needs of the accuracy and real-time of the passenger flow detection of the subway channel,this paper carries out the research on the detection method of the single side passenger flow of the subway channel based on video,In order to reduce the working pressure of subway field personnel.This paper first introduces the subway station operation security in the macro environment,and analyzes from the following four aspects: the demand analysis of subway station safety,the demand analysis of passenger flow information detection in subway station,the demand analysis of passenger flow detection in one side of two-way channel in subway station,and the passenger flow detection method at the present stage of subway station,The importance and necessity of subway passenger flow information detection are further clarified.Secondly,by comparing the one-stage method YOLO series and the two-stage method R-CNN Series in the target detection algorithm,according to the experimental results,the YOLOv3 algorithm with higher detection accuracy and faster detection speed is selected;Then the DaSiam-RPN algorithm and deep sort algorithm with better tracking effect are compared through experiments,and the DaSiam-RPN algorithm with better tracking accuracy is selected according to the experimental results;Finally,summarize the experimental results of the three and four chapters,propose an improved DaSiam-RPN algorithm based on adaptive threshold,use the field data of the subway,and compare the algorithm in this paper with the DaSiam-RPN algorithm.The results show that the algorithm in this paper can achieve good detection effect.The incremental learning model is introduced into DaSiam-RPN algorithm based on adaptive threshold.By increasing the adaptive threshold,the threshold setting is more reasonable,and the reasonable distractors can be further screened out;At the same time,through repeated iterations of experiments,the proportional relationship between the weight factor and the interference weight is determined in the subway channel scenario,so as to improve the tracking accuracy of subway station channel passenger flow;This algorithm can achieve real-time and accurate detection and tracking of the passenger flow on one side of the subway channel.When there is an abnormal situation,the subway staff can make the first response,so as to improve the operation safety of the subway.