一种新的形变指纹匹配方法  

A New Algorithm for Distorted Fingerprint Matching

作　　者：曹国[1] 孙权森[1] 毛志红[2] 梅园[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,南京210094 [2]中山大学计算机系,广州510275

出　　处：《中国图象图形学报》2010年第4期645-649,共5页Journal of Image and Graphics

基　　金：国家自然基金项目(60773172);中国博士后基金项目(20070411055);江苏省博士后基金项目(0701037B)

摘　　要：提出了一种基于细节点局部配准的形变指纹匹配方法。首先,结合细节点的纹理信息以及结构信息获取多个参照点;然后依据选取的多参照点实现模板指纹图像与输入指纹图像的全局配准从而获得指纹之间的公共区域;将公共区域内的细节点与它们最近的参照点聚类组合,形成多个分组,并将各分组内的细节点以对应的参照点为极点转化到极坐标系下建立指纹的局部对应性;最后,采用界限盒约束条件实现指纹匹配。实验结果表明,基于局部配准的指纹匹配方法对形变指纹匹配具有较好的鲁棒性,能较大提升指纹的识别性能。A novel minutiae-based method using local alignment to match the deformed fingerprints is proposed in this paper. We apply texture-based and structure-based minutiae information to obtain multiple reference minutiae at first, and then globally and evenly align two sets of minutiae to obtain the common overlapping region based on these reference minutiae. Next, we use the minutiae and their closest reference minutia to establish the local correspondence. After the registration of the fingerprints according to the local correspondence, the number of matching minutiae can be counted using bounding box constraints. Experimental results confirm that the proposed algorithm which is based on local correspondence is reliable for fingerprint matching with nonlinear distortions and leads to improvement in identification performance.

关 键 词：指纹匹配 局部配准 细节点 多参照点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]