摘要
为提高指纹匹配的正确率,综合局部细节匹配算法和全局匹配算法,提出一种将两者相融合的二次匹配方法。在提取指纹细节特征信息并去除伪特征点后,首先利用k-近邻法进行局部细节特征的一次匹配,获得局部特征之间的匹配分数;然后根据匹配分数对指纹图像进行旋转校正,进而对全局特征进行二次匹配,计算匹配向量,并利用匹配向量获得匹配率决定最终匹配结果。实验结果表明:在不同质量的多个指纹数据库上测试,算法最高正确率达到错误拒绝率为2.5%,错误接受率为0.22%,说明了该方法的有效性。
A minutiae matching approach based on re-matching was developed to improve the accuracy of fingerprint verification. After the basic minutiae are extracted, the k-neighbors algorithm was used to get local matching scores through the first local-matching step. Then the local matching scores were used to adjust the two images for a second global-matching step. The final decision was made from the matching rate of the matching vectors in the second matching step. Tests using three test collections, including both international fingerprint verification competition databases and the database collected by the on-line FPS200 capture system, show that the approach is insensitive to noise with accurate results.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2006年第10期1776-1779,共4页
Journal of Tsinghua University(Science and Technology)
关键词
指纹识别
二次匹配
K-近邻法
fingerprint verification
re-matching method
k-neighbors algorithm
