A Research Of Lane Detection And Tracking Based On Machine Vision

In recent decades, with the development of science and technology, especially therapid development of computer technology、 information technology、 artificialintelligence, and electronic technology, laid the solid material base for thedevelopment of intelligent vehicle research, intelligent vehicle into a deep, system,large-scale research phase. The key technology of intelligent vehicle is auxiliarydriving vehicle system, this system using image processing and computer visiontechnology testing environmental conditions (road pavement, traffic signs, othervehicles, pedestrians and road traffic accidents) to ensure that the vehicle driving inthe right lane with the safety car distance between the others and the right speed from,and we can handle in time, even if there are some abnormal situation.Our paper had explored some positive and beneficial methods which stand thepoint of view on the vehicle active safety, for safety technology in the security areasof intelligent vehicle, the purpose of which is to our country vehicles auxiliary drivingsystem provide the realistic theoretical and technical support of the application ofauxiliary. The first step of vehicles driving system realizes the navigation auxiliary islane detection and recognition. Lane detection and recognition is the aim of intelligentvehicle analyze image, and detect the deviation of vehicles relative to the lane, thengive this information to the vehicles auxiliary driving system, so as to realize the"anti-eccentrically safe driving".Our method based on extract of the line’s characteristics, the image preprocessingaspects (including image processing, image filtering of gray, and custom differentialoperator, etc.) were all improved in corresponding, and then use condition constraintRadon transform to detect the lane marker’s boundary on the image which after imagepreprocessing, finally, through the prediction parameters of Kalman filter on the frame beside of the first frame to establish dynamic ROI (Region of Interest), i.e. theinterested region, and then give the line parameters information which get from theRadon transform into the Kalman predictors in this area, then we can predict the lineparameters in the next frame.The experimental results showed that the proposedmethod is effectively,because it not only improved the accuracy of lane detection, butalso saved the system running time.