Research On The Algorithms Of Lane Recognition Based On Machine Vision

Today, with the development of technology and the improvement of people’s living standards, the number of cars is increasing in a straight line, resulting in a series of problems such as traffic congestion and traffic accidents. Therefore, the research of intelligent transportation system has emerged, and the auxiliary driving system is one of the main contents of the research, and the key to the research of the auxiliary driving system is the lane departure warning system.The detection and recognition of lane line is the basis and the key of the lane departure warning system. Whether detect and identify the lane quickly and accurately impacts the real-time and stability of the whole system. On the basis of studying the lane line identification algorithms of normal road environment, the paper puts forward the road image recognition algorithms, which are based on the situations of fog, low light and shadow. So it makes a far-reaching analysis of road image recognition algorithms based on multiple operating conditions.In this paper, we take the road images taken by a CCD camera mounted on a test vehicle as the research objects. Firstly, the road images under the normal illumination should be pretreated. Preprocessing steps of the image are generally divided into: graying, filtering and edge detecting. Secondly, for the road images in the fog and haze weather conditions, the Retinex algorithm is used to enhance the images in view of the traditional gray scale histogram equalization method. Pavement contrast under the condition of weak illumination or shading is low, so the improved coefficient Sobel algorithm is used to edge detection. If the road is covered by rain water and road brightness is inhomogeneous, the steerable filters can be used to enhance the lane line.When pretreatment on multiple working conditions is finished, the images are segmented by threshold, extracting the lane feature. This paper mainly studies four kinds of threshold segmentation algorithms and contrasts the segmentation effect by experiment. Finally, it uses Hough transform to detect and fit the lane marking lines under various conditions, and using relevant the continuity between the image sequence of road images track the lane marking lines, reducing the scanning range, improving the real-time for the tracking and detection.The experimental results show that these algorithms solve the problems of various road image conditions preprocessing. During the image segmentation process, OTSU algorithm and local threshold segmentation algorithm have achieved good experimental effects. In the lane detection stage, Hough transform algorithm can fit the lane marking line under various conditions accurately. What’s more, its robustness is strong. Finally, it uses the method of setting the dynamic interesting region to realize lane tracking, which lays the foundation of lane departure warning system.