Monocular Vision Structured Road Lane Detection And Tracking Technology

Autonomous navigation System has become the main research area of intelligent vehicle since the 1980's. Lane serves as one critical information in autonomous navigation, so the research of it has attracted great attention and many researchers devote their time and energy to develop efficient methods to improve the robustness and efficiency of the lane detection. This paper mainly researches the structured road lane detection and tracking technology based on the monocular vision. By combination of the theory analysis with specific experiments, the paper has confirmed a testing process of lane detection, which applies the technology of image preprocessing, line detection, vanishing point detection and lane extracting. The lane tracking is achieved by Kalman filtering theory.In image preprocessing, this paper proposes an optimal algorithm:G component of image graying, median filtering, Sobel edge detection, OSTU image binaryzation. Experimental results show that this process is effective in realizing the detection of lane border under good or poor illumination road conditions.In order to reduce the impact of the interference lines in vanishing point detection of the first road image, this paper adopts a weighted clustering method. Many lines with similar inclinations will be detected around the lane in traditional hough transform and these lines correspond to closely spaced dots in the parameter space. According this feature, central one of these lines can be obtained by weighted clustering. Then the center of these lines' intersections is taken as the vanishing point. Simulation results indicate that this method could rule out the effect of the interfering lines.In order to improve the real-time performance of the system, different methods are used in line and vanishing point detection for the first frame of the road image and the tracking process. For the first frame the traditional hough transform is used to find lines and weighted clustering is used to detecting vanishing point. Vanishing point constraint center lines to identify lane. In tracking process, Kalman filtering is used to fix interesting domain, and then vanishing point is detected by Linear neighborhood method, and then Hough transform based on vanishing point is used to find lines, at last lane is recognized according to the angle informations of the the lines.Simulation test of the collected road images results show that the lane detection method designed in this paper is stable enough to show the lane position for engineering application not matter in good or poor illumination road condition.