基于快速协同表示分类和组内预测重构系数向量l2范数的人脸识别算法  

Face recognition algorithm based on fast collaborative representation classification with intra-group prediction reconstruction coefficient l norm

作　　者：梅伟健 裘国永[1] Mei Weijian;Qiu Guoyong(School of Computer Science,Shaanxi Normal University,Xi’an 710119,China)

机构地区：[1]陕西师范大学计算机科学学院,西安710119

出　　处：《计算机应用研究》2020年第7期2196-2201,共6页Application Research of Computers

摘　　要：针对协同表示分类和规则化最小二乘算法(CRCRLS)在人脸识别应用中识别速率缓慢的问题,通过研究预测重构系数与人脸图片分类的关系,提出一种基于快速协同表示分类和组内预测重构系数向量l2范数算法(FCRCL2N)。与其他协同表示算法的不同之处在于,改进的FCRCL2N算法没有计算残差的过程,通过引入组内预测重构系数向量的l2范数直接对图片进行分类。标准数据集上的实验结果显示了该算法的高效性。Aiming at the problem that the collaborative representation classification with regularized least squares algorithm(CRCRLS)recognition execution was slow in face recognition applications,this paper developed an algorithm based on fast collaborative representation classification with intra-group prediction reconstruction coefficient l2-norm(FCRC L2N)by studying the relationship between predictive reconstruction coefficients and face image classification.Different from other cooperative representation algorithms,the FCRCL2N algorithm didn’t have a process of calculating the residual,and directly classified the picture by introducing the l2 norm of the intra-group prediction reconstruction coefficient vector.The experimental results on the standard data set show the efficiency of the algorithm.

关 键 词：协同表示 重构系数 l2范数 残差 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]