The Design And Implementation Of Bus Passenger Flow Statistics System Based On Mobile Edge Computing

With the continuous advancement of urbanization,road traffic congestion continues to increase,and the role of public transportation becomes more and more significant.Bus is the most commonly used public transportation option for residents.Effective statistics of bus passenger flow can not only facilitate residents to reasonably plan their travel modes,but also help bus companies to flexibly arrange transportation capacity.The existing passenger flow statistics methods is mainly divided into two kinds: the traditional image processing method and the convolutional neural network method.The traditional image processing method usually runs on the embedded platform with good real-time performance but low accuracy.Although convolutional neural network method has high accuracy,it needs large-scale matrix calculation,so it is difficult to run on the embedded platform.Therefore,it usually uploads video to the central cloud through the core network for processing,which not only takes up a lot of network ba ndwidth,but also has high delay and poor real-time performance.Mobile edge computing,as one of key technologies of 5G,has provided new ideas for passenger flow statistics system due to its proximity,low delay,soft and hard resources and other characteristics.Combined with the characteristics of mobile edge computing,this paper designs and implements a bus passenger flow statistics system based on mobile edge computing.By deploying the passenger flow statistics application on the mobile edge computing server at the edge of the network,the monitoring video at the upper and lower bus doors is received and processed,and then the passenger flow statistics task is completed.This thesis first introduces the application background,related principles and technologies,then gives the overall design and module design of the passenger flow statistics system based on mobile edge computing,and then focuses on the three modules of target detection,tracking and counting deployed in the edge server.A passenger detector based on yolov3 was designed and implemented by taking human head as detection target.Then,on the basis of high precision detector,a multi-target tracker based on kalman filter and Hungarian matching algorithm is designed and implemented.Finally,the passenger trajectory is analyzed,and a new method based on dual alignment is proposed.Based on the above modules,this thesis implements the passenger flow statistics system and tests the system accuracy and processing speed.Experimental results show that the system designed in this thesis is 92% accurate,processing speed is higher than 10 FPS,better meet the needs of bus passenger flow statistics.