加权估计纹理分析结合高斯黎曼流形的人脸识别方法  

A Face Recognition Method Based on Fusion of WETA and Gaussian Riemannian Manifold

作　　者：梁传君[1] 卜宇[1] 王红梅[1] 

机构地区：[1]新疆工程学院计算机工程系,乌鲁木齐830011

出　　处：《微型电脑应用》2017年第11期15-19,共5页Microcomputer Applications

基　　金：新疆维吾尔自治区高校科研计划青年教师科研启动基金项目(XJEDU2016S085);新疆工程学院科研基金项目(2015xgy101712)

摘　　要：针对图像集人脸识别中的子空间模型限制问题,提出了加权估计纹理分析结合高斯黎曼流形的人脸识别方法(WETA-GRMD)。使用样本图像和从样本获得的仿射包模型联合表示一幅图像。加权估计纹理分析进行人脸匹配,并解决权值最优化问题。利用高斯黎曼流形计算高斯分量具有识别能力的信息,并通过寻找最大判别分量识别人脸。在两个具有一定挑战性的数据集YouTube Celebrities(YTC)和YouTube Face(YTF)上的实验验证了提出方法的有效性,结果表明,相比其他几种较新的方法,提出的方法具有更高的识别率。In allusion to the subspace model limitation problem in the image set based face identification,the face identification method based on weighted estimation for texture analysis-Gaussian Riemann manifold（WETA-GRMD）is proposed in this article.Firstly,the sample image and the affine hull model obtained from the sample are combined to represent an image;then,weighted estimation for texture analysis（WETA）is adopted to execute the face matching operation and solve the weight optimization problem;finally,Gaussian Riemann manifold（GRMD）is adopted to calculate the information with identification capability in Gaussian component in order to find the maximum discriminant component for face identification.Meanwhile,the effectiveness of the proposed method is verified by the experiment in two challenging data sets YouTube Celebrities（YTC）and YouTube Face（YTF）,and the result shows that compared with several other new methods,the proposed method has higher identification rate.

关 键 词：人脸识别 高斯黎曼流形 加权估计 纹理分析 仿射包模型 特征提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]