The Algorithm Research And Implementation On The Lane Detection Based On Binocular Vision

Binocular vision is the interdisciplinary cross field of pattern recognition, imageprocessing and artificial intelligence, it’s one of the main sources of computer access toexternal environmental information. In recent years, the rapid development of computertechnology and electronics, optics has provided the powerful condition for the research anddevelopment of binocular vision technology, and it makes the binocular vision technologybecomes a new research hotspot and the key problem in the field of machine vision.The principle of binocular stereo vision is based on the parallax, it uses imagingequipment from the space of different location to get the two images of the object,Calculated by the position deviation between the points corresponding to the image, toobtain the three-dimensional geometry information of space objects. Binocular visionsimulates the human vision system to perceive a three-dimensional world directly, with alarge amount of information, high efficiency, reliable performance, low noise, widedynamic range, low cost, appropriate precision and other significant advantages, inautomatic driving, quality control and testing of non-contact products, biomedical,three-dimensional measurement, aerospace, intelligent robot control and lane departurewarning systems and other areas of applied research have a great value.In allusion to the problems of traffic safety become more and more serious. In thispaper, based on the theory of binocular vision, artificial intelligence and image processing,proposes a method to detect the lane using the binocular vision technology. Firstly,describes the overall design of the lane detection system; details analysis of the principle ofbinocular vision and the main content. Secondly, a camera calibration method with a signof circular feature points in the target circle is given; and then processes the lane imagesusing the binarization, region of interest, edge detection and Hough transform. At last, onthe basis of the detected lane, combined with the internal and external parameters ofcamera calibration matrix, using the IPM puts the lane two-dimensional coordinateconversion into3D world coordinates. The experimental results show that, the systemdetection range is within6meters, the measurement error is below10%.Research done in this paper by binocular vision technology applications for the nextstep in the car driving safety and obstacle detection do some fundamental work, and for thefuture establish and improve the lane departure warning system based on binocular stereo vision system try to do some practical attempts.