基于BSLDP和典型相关分析的掌纹掌脉融合识别

Palmprint and Palm Vein Feature Fusion Recognition Based on BSLDP and Canonical Correlation Analysis

针对非接触采集图像时容易出现模糊,导致识别系统稳健性下降、识别效果不佳的问题,提出一种基于分块增强局部方向模式(BSLDP)和典型相关分析的掌纹掌脉融合识别方法。首先对传统局部方向模式(LDP)进行了改进,提出BSLDP来获取掌纹掌脉图像的纹理方向特征,然后基于典型相关分析法实现掌纹掌脉特征的有效融合,最后根据融合特征向量间的卡方距离进行匹配识别,并在CASIA-M图库及自建非接触图库上进行实验测试,最小等误率分别为0.63%和1.21%。结果表明与其他传统及最新算法相比,本文方法能够消除冗余信息、准确地保留掌纹掌脉的有效特征信息,提高系统识别性能。

Aiming at the problems of non-contact images acquisition such as blur phenomenon, poor system identification systems and poor recognition effect, a palmprint and palm vein feature fusion recognition method based on block strengthened local directional pattern（BSLDP） and canonical correlation analysis is proposed. Firstly, we improve the traditional local directional pattern（LDP）, and proposed the BSLDP algorithm to obtain the texture direction feature of palmprint and palm vein images. Secondly, the palmprint and palm vein feature fusion is realized effectively based on the canonical correlation analysis. Finally, the match identification is realized based on the chi- square distance. The experimental results show that the equal error rate is only 0.63~ and 1.21% in the CASIA-M and the self-built non-contact image database. The results indicate that compared with other traditional and newest algorithms, the proposed method can eliminate redundant information, retain accurate feature inforination of palmprint and palm vein and improve system identification performance.