基于机器视觉的乡村道路跟踪方法研究

摘要提出了一种适用于半结构化和非结构化道路的自然边界检测方法。因为摄像机成像过程中发生了透视投影变换，自然道路边界在图像中总是相交于远处的消失点，该消失点对道路边界的定位能起到重要的作用，所以本论文首先采用了快速的Gabor方向滤波器和鲁棒的投票机制求出消失点位置，以确定道路主方向和道路兴趣区域；对于兴趣区域图像，采用了基于模糊C均值聚类（FCM）的分割方法分割出道路区域，继而得到道路的边界；最后结合梯度信息得出可信度较高的道路边界，即将分割得到的道路边界和Sobel算子得到的梯度图像相结合，使用Hough变换得出道路边界主方向直线，由主方向直线筛选出有效的边界点，再使用贝塞尔曲线方法将这些边界点拟合出最终道路边界。经过大量真实道路图像的测试表明，该方法具有一定的鲁棒性，但实时性尚有不足，主要是因为FCM聚类的速度较慢。8630
关键词自然道路边界检测消失点 Gabor滤波模糊C均值聚类 Hough变换
毕业设计说明书（论文）外文摘要
Title    The Research Of The Country Roads Following Methods       Based On Machine Vision
Abstract
This paper proposed a natural boundary detection method for semi-structured and unstructured road. Because of the perspective projection transformation occurred in the camera imaging process, the natural road boundary in the image is always intersect at the vanishing point in the distance which can play an important role in the positioning of the road boundary, the paper first uses a fast oriented Gabor filters and a robust voting mechanism to find the location of the vanishing point to determine the main road direction and the interested road area; for the interested road area, we use a segmentation method based on fuzzy C-means clustering(FCM) to segment the road area, then we can get the boundary of the road; finally, we can get the reliable road boundary to combine the gradient information, that is combining the segmented road boundary and the gradient image obtained by Sobel operator, using the Hough transform to draw the straight line of the main direction of the road boundary, screening out the effective border point, and then using the Bezier curve method to fit the final road boundary. After the testing of a large number of real road images show that this method has some robustness, but real time is still inadequate, because of the slower of the FCM clustering.
Keywords：Natural road boundary detection; Vanishing point;
          Gabor filtering; Fuzzy C-means clustering; Hough transform
目次
1   绪论   1
1.1   背景与意义   1
1.2   国内外的研究现状   2
1.3   本文的主要工作和创新点   4
1.4   本文的结构安排   5
2   消失点检测   6
2.1   Gabor变换   7
2.2   投票机制   9
2.3   消失点的确认   10
3   道路分割   12
3.1   FCM聚类   12
3.2   基于FCM的道路图像分割   14
4   道路边界提取   18
4.1   基于抽样的兴趣区域提取   18
4.2   基于小波变换的图像抽样   19
4.3   道路边界提取   20
5   道路跟踪系统   24
5.1   软硬件环境和系统结构   24 基于机器视觉的乡村道路跟踪方法研究:http://www.youerw.com/jisuanji/lunwen_7044.html