基于改进BEMD的Gabor-2DPCA掌纹识别  

Gabor-2DPCA palmprint recognition based on improved BEMD

作　　者：张雪锋 吕冰姿 ZHANG Xuefeng, LYU Bingzi(ColJege of Conmmunication and Information Engineering,Xi＇an Universily of Posls and Telecommunications, Xi＇an 710061 ,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710061

出　　处：《西安邮电大学学报》2018年第2期40-48,共9页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金资助项目(61301091)

摘　　要：为了提高掌纹识别的准确率,提出一种基于二维经验模式分解(bidemensional empirical mode decomposition,BEMD)的掌纹重构改进算法。该算法首先对预处理掌纹图像采用改进二维经验模式分解算法分解,提取前4个本征固有模态分量重构掌纹;其次,将重构掌纹通过二维Gabor滤波器分解成20个特征子图,利用二维主成分分析(two-dimensional principal component analysis,2DPCA)算法数据降维,提取掌纹特征;最后,计算样本的欧氏距离,实现掌纹识别。采取香港理工大学的PolyU掌纹数据库中的600张掌纹图像实验,结果表明,该算法重构掌纹含有更多纹理特征,在训练样本不同的情况下识别率均有提升,最高可达99.33%。提出的改进算法可以应用于掌纹识别。In order to improve the accuracy of palmprint recognition,an improved palmprint reconstruction algorithm based on biemensional empirical mode decomposition（BEMD）is proposed.Firstly,the preprocessed palmprint image is decomposed by the improved two-dimensional empirical mode decomposition algorithm to extract the first four intrinsic eigenmode components and reconstruct the palmprint.Secondly,the reconstructed palmprint is decomposed into two-dimensional Gabor filter Feature subgraphs are extracted,and the palmprint features are extracted by using two-dimensional principal component analysis（2 DPCA）algorithm.Finally,the Euclidean distance of the samples is calculated to realize the palmprint recognition.Taking 600 palmprint images in PolyU palmprint database of Hong Kong Polytechnic University,the results show that the palmprint reconstructed contains more texture features,and the recognition rate increases with different training samples,up to 99.33%.The proposed improved algorithm can be applied to palmprint recognition.

关 键 词：二维经验模式分解 重构掌纹 GABOR变换 二维主成分分析 掌纹识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]