The Research Of Lane Recognition Technology Based On Machine Vision

Lane recognition technology as an important part of the future intelligent transportation, and has played an important role in today’s road security. The use of modern computer vision, pattern recognition, automatic control technology. Lane recognition technology used modern computer vision, pattern recognition, automatic control technology. In the lane departure warning system(LDWS), autopilot system need to use the lane line detection. This paper research the lane detection and tracking algorithm based on machine vision.This paper proposes a modified channel priority dark haze cancellation algorithm, Improved the method of obtaining atmospheric light values. The improved algorithm is applied to the pre-treatment process of Lane recognition. In order to resolve the current frequent fog and haze in China. Solve the haze problem affecting image clarity. It also proposes a method, which is according to the speed of determining the number of frames fuzzy processing method. The method may use the minimum number of frames in the case of image mean processing. Making the intermittent lane road into a continuous lane, Facilitate follow-up lane recognition. Using Inverse Perspective Mapping to transform the Image with perspective into aerial View. making the non-parallel lane segments into parallel lane segments, Different from the previous methods of Looking Perspective Vanishing Point and World dividing line. Using median filter before and after the Bernsen adaptive binarization process, Improve the filtering effect. On Lane recognition stage, in order to prevent erroneous recognition across the lane, proposed a axis based method to solve the praoblem of Cross-lane recognition. Explores the using of RANSAC Algorithm on non-lane elimination. Get a clean lane line image. Finally using Kalman filter to track the lane and proposed method of determining the credibility forecast. Solve the problem of lane was shadowed.Through take video on the street, using MATLAB to recognition video frame by frame, to verify the feasibility of the proposed method to explore the optimal algorithm parameters. This paper presents an optimization algorithm lane recognition process.