Research On Public Traffic Passenger Flow Statistics System Based On Surveillance Video

With the development of urbanization and the increase of traffic congestion,bus passenger flow detection plays an important role in the realization of reasonable traffic planning and optimization of bus operation routes.The popularity of video surveillance technology in public transportation systems has led to the development of bus passenger flow statistics based on surveillance video in the field of computer vision.For the complex scenes in the surveillance video of bus passengers getting on and off,this paper proposes the detection of passengers getting on and off based on the deep neural network method,and then designs a data association matching multi-target tracking algorithm based on the detection results of each frame of surveillance video.Aiming at the reduction of the predicted tracking effect caused by the uncertainty of the trajectory of passengers getting on and off,and the tracking trajectory drift,etc.,combined with the detection results,an independent tracking mechanism is adopted for the problem target,thereby improving the accuracy and robustness of the overall algorithm.This article first decided to use the passenger’s head area with obvious features and easy extraction as the detection target based on the surveillance video scene and the observation of the passengers getting on and off.Combined with the characteristics of the target area,the research on the structure and parameters of the Single Shot Multi Box Detector(SSD)network of the deep convolutional neural network method simplifies the network of the SSD target detection algorithm while ensuring high precision,and improves the speed of detecting the target.In target tracking,a data association method based on the Kuhn-Munkras(KM)algorithm is proposed to determine the tracking target by comparing the Intersection over Union(IOU)of the detection target of each frame of the video to improve the effect of multi-target tracking;At the same time,the Kalman filter is used to solve the problems of detecting target loss,tracking trajectory drift,etc.,to ensure the accuracy of target tracking.Finally,a counting method based on the length of the target motion trajectory is used to complete the statistics of passenger flow on and off the vehicle.The experimental results show that the bus passenger flow statistics system developed in this paper uses video data collected from the video monitoring area for bus passengers to get on and off to verify the algorithm.The accuracy of the algorithm in this paper reaches 89.8%,and the running rate is about 15.6 per second.The frame has good robustness and realizes the counting of passenger flow.