Application Of Image Processing Technology In Vehicle Inspection At Intersection

In the current most of traffic lights in our country still uses a fixed period of traffic signal control method, When one direction have a lot vehicles for a red light, and with the direction of the intersection of the other have few vehicles for the green light,the fixed period control method is easy to lead the vehicle waiting for a long time,and the result is traffic jam.Therefore,it is one of the most pressing problems that how to achieve Intelligent Transportation Systems use of existing resources make the traffic signal light intelligent control cycle,to ease traffic congestion and improve traffic efficiency.To acquired vehicles flow from traffic intersection vehicle recognition as a basis for intelligent traffic lights control,have great significance to solve the above problems.A method was proposed,it was that Hough transform to determine lane endpoint to divide each lane.Then traffic intersection vehicle recognition use the background difference-based method.Aiming at the problem that background difference-based method need a stable and capable of handling complex scenes background model,proposes a improving Gaussian mixture model method to traffic intersection vehicle detection.Specifically,the main works of this thesis are as follows:(1) Deeply analyzes the basic technology of image processing and introduces the basic theory of gray-scale image, image binarization, morphological processing and so on; the thesis emphatically analyzes some frequently-used ways of filter. At last, bilateral filtering is used to eliminate the Noise-induce annoyance in the image.(2) Through analyzes all kinds of typical theories of edge detection operator and Hough transform line detection, and detecting lane line in the scene by Hough transform. The thesis puts forward a method of extracting the lane according the ends’ coordinates of lane line to eliminate the Noise-induce annoyance for others sample in the scene.(3) Three ways of vehicle recognition thesis was introduced,including optical flows, frame difference, and background difference-based method. Experimental simulation is carried out on these three methods, and their advantages and disadvantages is compared. Combined with traffics’conditions of intersection of urban traffic, the thesis uses simple principle, background difference-based method which has strong real-time to detect vehicle.(4)Aim at the issues of background modeling in background difference-based method, this thesis researches some kinds of frequently-used background modeling method. For traditional disadvantages of Gaussian mixture model, this thesis presents a method of Gaussian mixture model of adaptive update rate, it also as background modeling of background difference-based method. After simulation experiment is carried out, the thesis compares effects of vehicle detection base that two background modeling of background difference-based methods. For the effects of two methods processing extract the lane, work out the lane space occupancy ration of every lane, comparing the result of lane space occupancy ration and manual segmentation. And measurement traffic flow of every lane by the ration. All in all, it is confirmed that the methods that thesis presents not only has better detection effects, but also have real-time and accuracy.