Research On Highway Traffic Flow Detection Technology Based On Machine Vision

Highway vehicle flow detection based on machine vision technology research is under the condition of without human intervention,from obtaining the initial video frame,moving objects from the background segmentation method for camera recorded image sequence were analyzed,and then to the whole video car count,the counting system for the car in the scene,basic get cars and road congestion.The function of detecting and counting moving objects is realized in the terminal of video surveillance system.This paper mainly conducts theoretical research on two detection methods,which are as follows:(1)Frame difference method and Gaussian mixture model method: When processing video data,frame difference method selects representative frames from the video and develops algorithms on the frames.The algorithm can be applied to the processing of all frames in the video and check the frame containing light and dark cars.Typically,a number of techniques are required to remove irrelevant objects.The Im Extended Max feature is a way to remove dark cars from video frames.It also USES the shape function imopen to remove the lane markers and finally regionprops to look up the object’s center of mass in the no Small Structures(a light-colored car).(2)Background difference method and Gaussian mixture model method: For complex scenes,the single Gaussian model cannot well estimate the background,because the background pixel value is multi-modal.According to the idea of single mode,the collection of multiple single modes can be used to describe the change of image pixel value in complex scenes.The principle of background difference method is to subtract the currently acquired image frame from the background frame,so as to filter out the same background area in the image and leave the area where the moving target is located,so as to propose the moving target.The advantages of background subtraction method are simple and easy to implement,insensitive to vehicle motion rate,accurate extraction of moving target contour,and small calculation amount.At last,the application software of the above two detection methods is designed.In this paper,we study how to improve the intelligent capability of the terminal and realize moving target detection and target tracking in the terminal of the system.Its main advantage is embodied in the following two aspects: reduce manpower,improve the accuracy of counting,improve reliability.