图像点模式匹配的一种凸包序列的图谱方法  

Convex hull sequence method using graph spectra in image point pattern matching

作　　者：刘文静[1] 田铮[1,2] 张朝阳[1] 

机构地区：[1]西北工业大学理学院,西安710129 [2]中国科学院遥感应用研究所国家遥感科学重点实验室,北京100101

出　　处：《计算机工程与应用》2011年第23期197-200,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.60972150;No.10926197;西北工业大学科技创新基金(2007KJ01033)~~

摘　　要：利用点集的凸包具有仿射不变性和局部可控性,针对图谱方法难以精确匹配旋转角度较大图像的问题,提出了图像点模式匹配的一种凸包序列的图谱方法,使得匹配在图像旋转角度较大的情形下仍具有稳定性。构建图像特征点集新的图模型(凸包),利用改进的图谱方法对凸包进行匹配,并减小原始特征点集,迭代上述过程,通过构造凸包序列,自特征点集的外围到内部逐步匹配,得到较精确的匹配对。实现基于凸包序列的图谱方法的图像点模式匹配。实验结果表明,该方法不但能精确匹配旋转角度较小的图像,而且对于旋转角度大的图像以及多光谱图像匹配精度也较高。Aiming at the exact matching problem of larger rotation image,which is difficult to be solved by graph spectra method,a convex hull sequence method using graph spectra in the image point pattern matching is presented,which makes use of the local controllability and affine invariance of the convex hull.It makes the matching still stable in the case of larger image rotation angles.The new graph model of the image feature point set（convex hull） is built,the graph spectral method to match the convex hull is used,and the original feature point set is reduced.In order to get more accurate matching pairs, apply the iterative method,from the exterior of the feature point set to the interior to match through construction of the convex hull sequence.It achieves image point pattern matching using the graph spectral method based on convex hull sequence.The experimental results show that this method not only can accurately match the images of the smaller rotation,but also its matching accuracy is high for the images of a larger rotation angle and multi-spectral images.

关 键 词：点模式匹配 凸包序列 图谱 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]