文档介绍：沈阳理工大学
硕士学位论文
基于双目视觉的图像匹配算法研究
姓名:李志
申请学位级别:硕士
专业:计算机软件与理论
指导教师:王琰
20090301
沈阳理工大学硕士学位论文
摘要
双目视觉中的一个重点和难点就是图像匹配,图像匹配也是计算机视觉和图像处
理领域的一个非常重要的研究方向。本文针对图像匹配方法进行了深入的研究,并
提出了一种高效的图像匹配算法。
特征点提取之后,首先用相关法进行初始匹配,匹配集合中会出现很多有歧义
的匹配点,其中含有大量的误匹配点,这样的匹配点对集合不利于对基本矩阵进行
有效估计。提出一种消除匹配歧义的算法,求一个称为匹配支持强度的函数值,这
个函数描述的是一个匹配点对被附近其他匹配点对支持的程度,然后用某种策略,
将匹配支持强度适当的匹配点对提取出来放入一个集合,以达到消除匹配歧义的目
的。
消除歧义以后,以 8 点算法和最小中值区域算法为基础来估计基本矩阵。最后
应用得到的精确的基本矩阵即外极几何来重新引导匹配。
实验证明,本文的算法较好地解决了双目视觉中的图像匹配问题。
关键词:双目视觉;特征点提取;特征点匹配;外极几何
沈阳理工大学硕士学位论文
Abstract
An important and main part of binary vision is image matching. It is also an important
research topic in the fields puter vision and image processing. This paper has done
a great work to research image matching algorithms deeply and a highly effective and
image matching algorithm has been developed.
After corner detection, correlation method is used for initial matching. After that,
many ambiguous matching points will be in the matching set. And a large amount of
outliers will be taken into account. This kind of feature points set can not be used for
estimating the fundamental matrix. A disambiguating algorithm has been proposed using a
so-called strength of matching function which describes the extent of one pair of feature
points being supported by other pairs nearby. After that, a strategy will be used to extract
pairs of feature points with the proper value of strength of match in order to disambiguate.
After disambiguating, we use 8-points algorithm and least-median-of-squares
algorithm to estimate the fundamental matrix. At last, we can solve image matching
problem with the help of the precise fundamental matrix which we have got just now.
It has been proved by experiment that the algorithm in this paper has solved image
matching in binary vision well.
Key words:binary vision; corner detection; matching of feature points; epipolar geometry
沈阳理工大学
硕士学位论文原创性声明
本人郑重声明