Research On Vehicle Line Pressing Judgment Method Based On Machine Vision

Traffic is the support of urban development,which relates to every aspect of people's lives.The traffic rules is fundamental in the development of transportation.With the development of science and technology,intelligent transportation enters into people's life.On the basis of the development of intelligent transportation system,this paper carried on the thorough research of vehicle line pressing violation in traffic scene.Combined with the current mature technology of image processing and machine learning.Study on the separation of vehicle prospects in video,road line detection,vehicle line pressing judgment and license plate capture technology in traffic scene.Research with the aid of Gaussian background modeling theory,CodeBook modeling method,Hough transform for line detection,Support vector machine(SVM)method,etc.Studied the traffic line detection and vehicle information access.The main research contents are as follows:(1)On the image preprocessing,analysis of the HSV color model;Improved image equalization technology,For different image settings for different equilibrium ranges;Studied filter denoising technology,using effective bilateral filtering on traffic image;Study on edge detection technology,selection algorithm for edge detection of Canny operator for image segmentation.(2)For the selection of image binarization threshold,using the two methods were studied,The method of using a combination of local threshold segmentation for image block and image binarization in the paper;For vehicle video foreground segmentation,respectively on Gaussian mixture background modeling and CodeBook algorithms are studied and the analysis of utility,Finally using CodeBook algorithms as the segmentation method of experiment.And to deal with the result,using the expansion algorithm with flood water filling technology,gets the scope of vehicles the prospects for effective.(3)Using the Hough transform to detect the lane line.Dealing with the result and get the efficient Lane line position.Studied the image moment feature and the quadrilateral surrounded by technology,used to get the vehicle outline the effective range;Using vehicle characteristics in the judgement of vehicle line pressing;Summarizes the process of Vehicle line pressing.(4)For the wide range of traffic visual range,Using support vector machine(SVM)and image HOG feature to test the license plate area in the image.Through a lot of vehicle image data,using the above algorithm to obtain an efficient classification model.Implementation of license plate location search in the broad vision area.