Research On The Passengers Flow Statistical Techniques Based On Machine Vision

Machine vision is an analog of biological vision by the using computers and related equipments,in recent years,the relevant technologies machine vision field have achieved rapid development,among them,passengers flow statistical techniques which based on machine vision have very important scientific value and extensive application value.In this article,the passengers flow statistical techniques which based on machine vision was studied,and a passengers flow statistical system was designed,this system can analyse and process real-time or non-real-time video information,then,get the passengers flow data,provide accurate passengers flow information for supermarket,hospital,bank or the other public places,making the management and decision more scientific.Firstly,this paper puts forward the overall design idea,including the software module and the hardware device.Then the video frames were pretreated,including frames gray scale,binaryzation,median filtering and mathematical morphology processsing.Secondly,several current commonly moving target detection algorithms were researched,including optical flow method,frames difference method and background modeling method.The advantages and disadvantages of the different methods were analyzed and compared.This paper focused on mixture gaussian background modeling method,and used this method as the moving target detection algorithm for the system.This artitle used Canny edge detection for moving object segmentation.Then,some current commonly moving target tracking algorithms were researched,including Mean Shift algorithm,Particle Filter algorithm and contour centroid tracking algorithm.The performances of several tracking algorithms were compared by the simulation experiments,and the contour centroid tracking method was used in this design system.Finally,the virtual count line was set,the two-way(in and out)passengers flow were counted.Then this system calculated the contour areas to remedy system error,further improving the accuracy of detection.In order to verify the system,s performance,passengers flow in real-time and non-real-time videos were counted,the experimental results show that this system can effectively count passenger flow information,meet expectations.