叶学义,邹茹梦,应娜,季毕胜,王鹤澎.视觉约束的增强三角剖分指纹识别算法[J].电子测量与仪器学报,2021,35(11):194-205 |
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 |
|
DOI: |
中文关键词: 指纹识别 增强三角剖分 三角形递减验证 视觉约束 |
英文关键词:fingerprint recognition enhanced triangulation triangle bidirectional verification calibration free visual constraint |
基金项目: |
|
Author | Institution |
Ye Xueyi | University College of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China |
Zou Rumeng | University College of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China |
Ying Na | University College of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China |
Ji Bisheng | University College of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China |
Wang Hepeng | University College of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China |
|
摘要点击次数: 1551 |
全文下载次数: 880 |
中文摘要: |
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432%、264%和798%,相比改进前的Delaunay 三角剖分降低了128%、171%和283%,相比改进前的扩展三角剖分降低了126%、052%和258%,相比尺度不变特征转换(SIFT)算法分别降低了089%、297%和003%。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。 |
英文摘要: |
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 FVC2000DB2, FVC2006DB2 and FVC2006DB3 were used for comprehensive performance comparison experiments, and the EER rates of the algorithm were 432%, 264% and 798%, respectively. Compared with the Delaunay triangulation algorithm, the modified Delaunay triangulation algorithm can reduce the EER by 128%, 171% and 283%, compared with the extended triangulation algorithm by 126%, 052% and 258%, and compared with the SIFT algorithm by 089%, 297% and 003%, 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. |
查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|