Study On Lane Departure Warning System Based On Machine Vision

This dissertation concentrates on the lane departure warning system which is suitable forthe highway, this lane departure warning system is based on machine vision. The dissertationdiscusses the work of vision processing system, concentrating on how to obtain theinformation of road using computer vision technology for expert system to make warningdecision. The dissertation put forward the development of the science and technology of LaneDeparture Warning System and introduces some basic concepts of computer vision at first.And then, the model of monocular vision system is established to obtain the features of visionfield and to obtain the mapping connections between image coordinates and road coordinateswith relevant theories of perspective projection pinhole imaging model and the opticalgeometry. After that, a variety of image processing methods are studied to improve theadaptability and flexibility of vision system and compared to put forward the relativegood-effect ones for the specific road soundings. We study smoothing-filtering approaches,edge enhancing approaches and threshold segmentation approaches to the road image, thealgorithms this dissertation determines are median filtering, improved Sobel arithmeticoperator and the maximum between-class variance method to image segmentation, andcascaded Hough transform through lane combined with the edge of the lane width to get thelane. After the above processing, machine vision system can quickly and accurately obtainthe lane track line, laid the foundation for the warning decision-making. Then the dissertationproposed vehicle departure warning decision-making methods. Finally, we designed anembedded ARM-WINCE-based image processing system, then tested the algorithm, the testresults show that the proposed lane recognition algorithm and lane departure warningalgorithm is feasible and effective.