摘要
为了可靠地实现点模式匹配,提出了一种基于谱图理论与几何相容性分析的点模式匹配算法。利用拉普拉斯矩阵的特征向量获得待匹配点集间谱匹配代价的表示;结合以邻近关系表示的几何相容性,定义了一种混合形式的匹配目标函数;给出了基于松弛迭代的求解算法。仿真数据和真实图像上的比较实验表明所给出的方法具有较好的精度与时间性能。
To match point-sets reliably, an algorithm for point pattern matching based on spectral graph theory and the analysis of geometric consistency is presented. The cost of spectral correspondences between the matched point- sets is obtained by means of eigenvectors of Laplacian matrix. An object function with hybrid form is defined by incorporating geometric consistency represented by neighborhood relationship. The given object function is solved by utilizing iterative relaxation method. Comparative experiments applied to synthetic data and real-world images demonstrate the proposed method possesses better precision and time performance.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2012年第7期161-166,共6页
Acta Optica Sinica
基金
国家自然科学基金(11071002
61172127)
安徽省教育厅自然科学研究项目(KJ2011A008)
安徽大学211工程学术创新团队资助课题
关键词
机器视觉
匹配
谱图理论
几何相容性
machine vision
matching
spectral graph theory
geometric consistency
