字典学习优化结合二维KCM选择的人脸识别方法  

On the Selection of 2D Krawtchouk Moments for Face Recognition

作　　者：唐守军 吴洪武 TANG Shoujun;WU Hongwu(Department of Basic Teaching, The Open University of Guangdong, Guangzhou 510091;School of Mathematics, South China University of Technology, Guangzhou 510641 China)

机构地区：[1]广东开放大学基础教学部,广东广州510091 [2]华南理工大学数学学院,广东广州510641

出　　处：《湘潭大学自然科学学报》2017年第4期85-88,共4页Natural Science Journal of Xiangtan University

基　　金：广东省高校自然科学研究课题(KF1420)

摘　　要：针对人脸识别中区域的高阶隐藏非线性结构发现问题,提出了字典学习优化结合2D Krawtchouk矩(KCM)选择的人脸识别方法.首先,利用二维KCM选择提取特征向量;然后,利用字典学习优化得到最优特征矩阵,并将特征进行组合;最后,使用最近邻分类器完成分类.实验结果表明,相比其他几种方法,该方法获得的平均精度高且鲁棒性更好.For the problems of capturing region based higher order hidden nonlinear structures on face recognition,a new face recognition method based on dictionary learning optimization and2D KCM is proposed.Firstly,2D KCM selection is used to extract feature vectors.Then,dictionary learning optimization is used to get the optimal matrix,and features are combined.Finally,nearest neighbor classifier is used to finish the classification.The experimental results show that the proposed method has higher average precision and better robustness than several other methods.

关 键 词：字典学习优化 人脸识别 正交Krawtchouk矩 最近邻分类器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]