摘要
将两枚指纹相似性的判断归为一个二分类问题,介绍了一种基于支持模型和SVM的指纹匹配方法,采用了基于支持模型的细节特征匹配求取匹配的特征点集合,并利用匹配细节特征点所包含的匹配信息,对训练样本构造匹配向量进行分类训练。将训练后的分类器应用于FVC2002的测试数据库,实验结果表明该方法优于传统的计算方法,在度量指纹相似性方面是有效而且可靠的,且具有较强的推广性。
This paper introduces a fingerprint matching algorithm based on the support model and support vector machine where calculating the similarity is considered as a general classification problem. The algorithm establishes the minutiae correspondence between two fingerprint images based on the support model, and constructs the feature vectors of a SVM using the minutiae＇s matching relationships. The similarity of these two fingerprints is then measured by the distance between the feature vector and the decision hyper-plane of the SVM. Experimental results prove the method is more efficient than traditional methods. And it is robust and has strong generalization ability.
出处
《计算机工程》
CAS
CSCD
北大核心
2006年第15期168-170,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60472069)
