摘要：20世纪80年代以来，自主导航已经成为了智能车领域的主要研究，车道线是自主导航中的一项重要信息，因此需要开发一种鲁棒性较高的车道线检测方法已经成为智能车领域中的研究热点。本文主要研究了结构化道路中车道线的检测方法。通过理论分析及仿真结果确定了一种包括了图像预处理、直线检测、消失点检测、车道线提取的算法来进行检测。

在图像预处理中，本文中给出了一个最佳的算法流程：G分量灰度化、中值滤波、Sobe1边缘检测。为了减少干扰线对第一帧图像消失点检测的影响，本文提出了权值聚类法。根据传统的Hough变换会在车道线附近检测出多条倾角相近的直线，而这些直线对应参数空间中一组密集点的特点，则先采用权值聚类法得到直线簇的中心直线，然后再利用权值聚类法得到这些中心直线交点的中心点，最后将得到的中心点作为最终的消失点。仿真结果也表明该方法可以很好地排除干扰线的影响。74586

通过对采集的道路图像和视频进行离线仿真测试，结果表明本文所设计的车道线检测算法在光照较好和光照较恶劣的道路环境下都能稳定实时地识别出车道线，能满足工程应用的需求。

毕业论文关键词:车道线，图像预处理，消失点，改进霍夫变换，权值聚类

Abstract： Since the 80s of the 20th century, autonomous navigation has become the field of intelligent vehicle, the main research, lane is autonomous navigation in an important information。 Therefore, it is necessary to develop a robust lane detection method has become a research hotspot in the field of intelligent vehicles。 The purpose of this paper is to study the structural road lane detection method。 Through theoretical analysis and simulation results identified a including image preprocessing, line detection and disappeared point detection, lane line extraction algorithm to detect。

In image pre-processing, this paper gives an optimal algorithm procedure: G component of image graying, median filter, Sobe1 edge detection。 In order to reduce the interference lines of the first frame of the image disappear point detection, this paper proposed a weighted clustering method。 According to the traditional Hough transform will in the lane line detection near the many lines with similar inclinations, and the characteristics of these lines correspond to the parameter space in a set of closely spaced points, is used to weighted clustering is used to get the linear cluster at the center of a straight, then the weighted clustering is used to get the center of the center point of intersection of a straight line, the most will be the center as the ultimate vanishing point。 The simulation results It shows that this method can eliminate the influence of interference line。

Through the collection of road image and video are off-line simulation test, the results show that the design of the lane detection algorithm in good illumination and light is bad road conditions under can stable real-time identify lane, can meet the demand of engineering application。

Keywords：the lane line, image preprocessing, the vanishing point, improved hough transform, weights of clustering

目   录

1 绪论 4

1。1 本课题研究的目的和意义 4

1。2 国内外技术发展现状 4

1。3 本文主要研究内容及章节安排 4

2 道路图像预处理 5

2。1 道路图像灰度化算法分析 5

2。2 道路图像的滤波算法分析 6