二维线性鉴别分析和协同表示的面部识别方法  

Face Recognition Method of Two-dimensional Linear Discriminant Analysis and Cooperative Representation

作　　者：林克正[1] 邓旭 张玉伦 LIN Ke-zheng;DENG Xu;ZHANG Yu-lun(School of Computer Science&Technology,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《小型微型计算机系统》2021年第8期1688-1693,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(62071157)资助;黑龙江省自然科学基金项目(2015040)资助。

摘　　要：针对图像特征提取方法提取单一特征不能很好地表示图像的问题,提出了二维线性鉴别分析和协同表示的面部识别方法.该方法首先通过二维线性鉴别分析(Two-Dimensional Linear Discriminant Analysis,2DLDA)分别对训练样本的类间散布矩阵和类内散布矩阵提取特征,之后利用得到的特征重建图像,包括类间虚拟图像和类内虚拟图像.其次,将类间虚拟图像、类内虚拟图像和原始图像利用协同表示(Collaborative Representation,CR)算法进行得分.最后,采用加权得分融合算法将上述得分进行融合以获得最终得分,并利用最终得分进行图像识别.该方法不仅有效的抑制了光照和表情对面部识别的影响,同时根据获得的类间虚拟图像、类内虚拟图像与原始图像互补,有效的提高面部图像识别的性能.实验结果表明,该方法在不同的数据库下(ORL、AR、GT)具有较好的识别精度.Aiming at the problem that the image feature extraction method can not represent the image perfectly,this paper points out a face recognition method of Two-Dimensional Linear Discriminant Analysis(2DLDA)and Collaborative Representation.The method extracts features from the inter-class scatter matrix and the intra-class scatter matrix of the training samples by 2DLDA firstly.And using the features reconstruct images,including the inter-class virtual images and intra-class virtual images.Then,the inter-class virtual image,the intra-class virtual image,and the original image are scored by Collaborative Representation(CR)algorithm.Finally,the above scores are fused using a weighted score fusion algorithm to obtain the final score,and the final score is used for image recognition.The method not only effectively suppresses the influence of illumination and expression on facial recognition but also enhances the performance of facial image recognition based on the obtained inter-class virtual image and intra-class virtual image complementary to the original image.Experimental results on different datasets(ORL,AR,and GT)show that the proposed method has better recognition accuracy.

关 键 词：图像识别 二维线性鉴别分析 协同表示 得分融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]