一种基于修正的最小生成树及其邻接谱的特征匹配算法  

A Feature Matching Algorithm Based on Adjacent Spectrum of Modificatory Minimize Spanning Tree

作　　者：宣善立[1,2] 梁栋[1] 朱明[1] 范益政[1,3] 王年[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039 [2]合肥工业大学计算机与信息学院,安徽合肥230009 [3]安徽大学数学科学学院,安徽合肥230039

出　　处：《电子学报》2010年第2期269-274,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.70772121;No.10601001);安徽省自然基金(No.070412065);安徽大学211工程学术创新团队

摘　　要：提出一种基于修正的最小生成树及其邻接谱的特征匹配算法.该算法利用两幅图像的特征点分别构造最小生成树,并对最小生成树进行修正,然后对修正的最小生成树的赋权邻接矩阵进行SVD分解,获得点的特征表示,进而利用特征值及特征向量来构造匹配矩阵,实现特征匹配.该算法的优点在于采用图的最小生成树(而不是整个图),可以减少多余信息的干扰,提高匹配精度,实验结果表明,该算法具有较高的匹配精度.Based on adjacent spectrum of modificatory minimize spanning tree,a new feature matching algorithm was proposed in this paper. According to the feature points of two related images, two minimize spanning trees were found and modified. The weighted adjacent matrices of the modificatory minimal spanning trees were submitted to singular-value decomposition（SVD）, and then the characteristics of the feature points were obtained. The matching was completed by constructing matching matrix with eigenvalues and eigenvectors. The advantage of this algorithm is that it can reduce the intrusion of the extra information and improve matching accuracy by using the minimal spanning tree of the graph. Experimental results show that the algorithm has a higher accuracy.

关 键 词：特征点 匹配 最小生成树 邻接谱 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]