基于改进协同表示的二级分类人脸识别方法  

A Two-Stage Face Recognition Method Based on Improved Collaborative Representation Classification

作　　者：施志刚[1] 

机构地区：[1]南通航运职业技术学院管理信息系,江苏南通226010

出　　处：《西南大学学报（自然科学版）》2017年第1期172-178,共7页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金项目(61473159);南通航运职业技术学院科技基金重点资助项目(HYKJ/2016A02)

摘　　要：基于协同表示分类(CRC)算法在实际应用中的效果,考虑到样本局部相似性先验信息对分类识别的不同贡献,构建加权矩阵,并嵌入到CRC中,称为加权CRC.为进一步改善人脸识别的性能,设计了如下算法:一种将加权CRC重复两次;另一种则将其与线性表示分类(LRC)结合.两种方法的共同特点为:首先基于主成分分析(PCA)对所有图像样本进行降维,以降低计算的复杂度;其次都是在第一阶段的加权CRC中根据重构残差排序保留相关性较大的训练样本用于下一阶段的分类识别.这种缩小分类目标的做法,使识别更精确.在ORL,FERET及AR人脸数据库上通过仿真验证了本文所提方法的有效性.Based on the effect of collaborative representation classification（CRC） algorithm in practical application and taking into account the different contributions of the samples＇ local similarity prior information in classification and identification,a weighted matrix is constructed in this article and embedded in CRC,called weighted CRC.In order to improve the performance of face recognition,two algorithms are designed：one is that the weighted CRC is repeated two times,and the other is that it combines with linear representation classification（LRC）.What is common for the two methods lies in that firstly.they both reduce the image dimension based on principal component analysis（PCA）,so as to reduce the complexity of the computation,and secondly,according to the reconstruction residuals based on weighted CRC in the first stage,large relatively training samples are sorted out for classification and recognition of the next stage.These approaches to reduce the classification targets make the recognition more accurate.The effectiveness of the proposed methods is verified by simulation on ORL,FERET and AR face database.

关 键 词：协同表示分类 样本局部相似性 加权矩阵 线性表示分类 人脸识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]