基于分块LBP和分块PCA的指静脉识别方法  

Finger vein recognition based on block LBP and block PCA

作　　者：杨文文[1] 毛建旭[1] 陈姜嘉旭 

机构地区：[1]湖南大学电气与信息工程学院机器人视觉感知与控制技术国家工程实验室,长沙410082

出　　处：《电子测量与仪器学报》2016年第7期1000-1007,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61175075;61271382);国家科技支撑计划(2015BAF13B00)资助项目

摘　　要：针对单独提取传统的局部二元模式(local binary pattern,LBP)指静脉特征识别率不高的问题,提出一种结合分块LBP和分块主成分分析(principal component analysis,PCA)的特征提取方法。首先对手指静脉感兴趣区域(region of interested,ROI)进行分块,提取分块LBP特征;然后,采用分块PCA对所提取特征进行降维除冗,得到一组新的降维后特征;最后计算欧氏距离并采用最近邻分类器进行分类。实验结果表明,论文提出方法识别率可达99.33%,等误率(equal error rate,EER)低至0.21%。与传统的3种算法进行比较,该方法的识别率有很大提高,且具有较好的稳定性和鲁棒性。Aiming at the problem of the traditional local binary pattern （LBP） of finger vein image may have low recognition rate, a method combined block LBP and block principal component analysis （PCA）is proposed in this paper. First, the region of interested （ROI） of finger vein image is dividedinto several blocks and the block LBP featureisextracted. Second, the dimension of extracted featureis reduced by using the block PCA. At last, Euclidean distance is calculatedandthe nearest neighbor classifier is takenfor verification. The results of the experiment show that the method can increase the recognition rate to 99.33% and decrease the equal error rate （EER） to 0. 21% . Compared with threetraditional algorithms, the method can greatly increase the recognition rate and has excellent stability and robust.

关 键 词：局部二元模式 主成分分析 特征提取 指静脉识别 

分 类 号：TN98[电子电信—信息与通信工程] TP391.4[自动化与计算机技术—计算机应用技术]