摘要
针对单一描述符无法准确获取有效掌纹特征导致识别率低的问题,提出一种基于子空间与纹理特征融合的掌纹识别方法。利用稳健线性判别分析和局部方向二值模式分别获取掌纹图像的子空间特征和纹理特征;基于加权串联方法实现子空间特征与纹理特征的有效融合;根据融合特征向量间的卡方距离进行匹配识别。在PolyU图库和自建非接触图库上的实验结果表明,识别时间分别为0.3069s和0.3127s,最低等误率分别为0.3440%和1.4922%;与其他方法相比,所提方法在保证实时性的前提下,能够准确提取掌纹图像的有效特征信息,提高系统识别性能。
Aiming at the problem of low recognition rate because the single descriptor cannot accurately obtain the effective palmprint features,apalmprint recognition method is proposed based on subspace and texture feature fusion.The subspace feature and texture feature of a palmprint image are obtained by robust linear discriminant analysis and local direction binary pattern,respectively.The weighted concatenation method is used for the subspace and texture feature fusion.The chi-square distance among the fused feature vectors is used for identification matching.The experimental results on the PolyU and the self-built non-contact databases show that the recognition time is 0.3069 sand 0.3127 s,respectively,and the lowest equal error rate is only 0.3440% and 1.4922%,respectively.Compared with other methods,the proposed method can accurately obtain the effective feature information of a palmprint image and improve the system recognition performance under the premise that the realtime performance is ensured.
作者
李新春
马红艳
林森
Li Xinchun;Ma Hongyan;Lin Sen(School of Electronic and Inform ation Engineering,Liaoning Technical University,Huludao,Liaoning 125105,China;Postgraduate College,Liaoning Technical University,Huludao,Liaoning 125105,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2019年第7期139-147,共9页
Laser & Optoelectronics Progress
基金
辽宁省教育厅科学研究一般项目(L2014132)
辽宁省自然科学基金面上项目(2015020100)
关键词
图像处理
稳健线性判别分析
局部方向二值模式
融合
等误率
image processing
robust linear discriminant analysis
local direction binary pattern
fusion
equal error rate
