基于数字图像处理的车道线检测

发布时间：2018-01-16 11:17

本文关键词：基于数字图像处理的车道线检测　出处：《新疆大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着国民经济的快速发展，汽车的数量也随之迅速增加。然而于此同时，人、车、路之间的矛盾也日益加深。为了解决这一矛盾，智能交通系统(ITS)概念就应运而生。本文选取智能交通系统(ITS)中最为关键问题之一的车道线检测作为研究对象。本文针对结构化道路图像中车道线检测所要面对的问题，提出了基于特征的两种不同检测算法。 1.基于边缘检测与类内最小方差法相结合的算法来进行对车道线的检测。该方法可以快速有效的检测不同情况下的直车道线。其主要思想是直车道线上各像素点的梯度幅值与梯度方向角内类方差最小。该算法首先利用边缘检测方法计算出边缘像素点的梯度幅值和梯度方向角，得到道路图像的边缘；然后将二值化处理过的边缘图像进行连通域的标记；在提取直车道线时，利用类内最小方差法计算各个连通域边界上点的梯度方向角最小方差，最后设定阈值分离出直线车道线，达到车道线检测的目的。 2.基于改进的K均值聚类算法和自定义边缘检测相融合的方法来检测车道线。该方法可以有效克服光照不均，雨雪天气等的影响。首先利用平滑直方图方法确定K均值聚类数目K，使得车道线能够被准确的聚为相同类，再将用自定义算子提取的图像边缘与聚类图像进行融合，充分利用了车道线的边缘特征，得到了车道线区域。通过经典提取直线方法Hough变换就能准确定位车道线，完成车道线的检测。最后，，通过对两种检测算法进行了大量仿真实验，证实了上述两种算法的有效性。
[Abstract]:With the rapid development of the national economy, the number of cars also increases rapidly. However, at the same time, the contradiction between people, cars and roads is also deepening day by day. In order to solve this contradiction. The concept of Intelligent Transportation system (ITS) arises at the historic moment. In this paper, lane detection, one of the most important problems in Intelligent Transportation system (ITS), is selected as the research object. Aiming at the problems of lane detection in structured road images, two different detection algorithms based on features are proposed in this paper. 1. The method is based on the combination of edge detection and intraclass minimum variance method to detect lane lines. This method can quickly and effectively detect straight lane lines under different conditions. The main idea of this method is that all images on the straight lane line can be detected quickly and effectively. The gradient amplitude and gradient direction angle of prime point are the smallest in class variance. Firstly, the edge detection method is used to calculate the gradient amplitude and gradient direction angle of edge pixel. Get the edge of the road image; Then the binary edge image is marked to the connected domain. In the extraction of straight lane, the minimum variance of gradient direction angle of points on each connected domain boundary is calculated by using the method of minimum variance within class, and the threshold value is set to separate the lane line to achieve the purpose of lane line detection. 2. Based on the improved K-means clustering algorithm and the self-defined edge detection fusion method to detect the lane line, this method can effectively overcome the uneven illumination. The influence of rain and snow weather. Firstly, the number of K-means clustering is determined by using smooth histogram method, so that the lane can be accurately clustered into the same class. Then the image edge extracted by the custom operator is fused with the clustering image, which makes full use of the edge features of the lane line. The lane line area is obtained and the lane line can be accurately located by Hough transform, which can be used to detect the lane line. Finally, a large number of simulation experiments are carried out to verify the effectiveness of the two algorithms.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;TP391.41

【参考文献】