视觉约束的增强三角剖分指纹识别算法  

Enhanced triangulation fingerprint recognition based on visual constraint

作　　者：叶学义[1] 邹茹梦 应娜[1] 季毕胜 王鹤澎 Ye Xueyi;Zou Rumeng;Ying Na;Ji Bisheng;Wang Hepeng(University College of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学通信工程学院,杭州310018

出　　处：《电子测量与仪器学报》2021年第11期194-205,共12页Journal of Electronic Measurement and Instrumentation

摘　　要：针对低质量指纹的真实细节点缺失和伪细节点增加以及典型指纹识别算法对细节点准确性过于依赖的问题,提出了视觉约束的增强三角剖分指纹识别算法。首先根据所提取细节点利用三角形重构获得增强三角剖分集;然后计算三角形特征向量,使用递减验证进行三角形匹配确定匹配细节点对,并使用视觉约束优化;最后根据匹配点对的比例获得相似度从而完成识别。采用国际标准测试库FVC2000DB2、FVC2006DB2和FVC2006DB3进行综合性能比对实验,该算法等错误率(EER)分别为4.32%、2.64%和7.98%,相比改进前的Delaunay三角剖分降低了1.28%、1.71%和2.83%,相比改进前的扩展三角剖分降低了1.26%、0.52%和2.58%,相比尺度不变特征转换(SIFT)算法分别降低了0.89%、2.97%和0.03%。实验结果表明,所提算法无需校准且对低质量指纹导致的真实细节点缺失和增加的伪细节点有较好的适应能力。Aiming at the problems of missing real minutiae and increasing pseudo minutiae of low-quality fingerprint,and the typical fingerprint identification algorithm is too dependent on the accuracy of minutiae,a visually constrained enhanced triangulation fingerprint recognition algorithm is proposed.First,use triangle reconstruction to obtain the enhanced triangulation set according to the extracted minutiae points;then calculate the triangle feature vector,use the decrement verification for triangle matching to determine the matching minutiae pair,and use the visual constraint optimization;finally obtain the similarity according to the ratio of the matching point so as to complete the recognition.The international standard test libraries FVC2000 DB2,FVC2006 DB2 and FVC2006 DB3 were used for comprehensive performance comparison experiments,and the EER rates of the algorithm were 4.32%,2.64% and 7.98%,respectively.Compared with the Delaunay triangulation algorithm,the modified Delaunay triangulation algorithm can reduce the EER by 1.28%,1.71% and 2.83%,compared with the extended triangulation algorithm by 1.26%,0.52% and 2.58%,and compared with the SIFT algorithm by 0.89%,2.97% and 0.03%,respectively.The experimental results show that the proposed algorithm does not need calibration and has good adaptability to the loss of real fine nodes and the increase of pseudo fine nodes caused by low-quality fingerprints.

关 键 词：指纹识别 增强三角剖分 三角形递减验证 视觉约束 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]