【作者】 杨万扣；

【导师】 杨静宇；

【作者基本信息】 南京理工大学 ， 模式识别与智能系统， 2009， 博士

【摘要】 在生物识别问题中,特别是人脸识别领域,由于原始图像的维数相当高,直接在原始图像的基础上进行处理,将加大算法的复杂度,并且对计算机的硬件性能也是一个挑战,因此特征抽取成为该领域最基本的问题之一,抽取有效的鉴别特征是解决该问题的关键。特征抽取的基本思想是将原始样本映射(或变换)到某一低维特征空间,得到最能反映样本本质的低维样本特征,这样能有效地减少样本的存储量和处理速度,实现人脸的自动分类。到目前为止,人们已给出了许多线性特征抽取方法,如主成分分析(PrincipalComponent Analysis,PCA或称K-L变换),Fisher线性鉴别分析(Linear DiscriminantAnalysis,LDA),独立成分分析(Independent Component Analysis,ICA)是特征抽取的几种经典和广泛使用的方法。本文研究工作主要如下:(1)在二维主成分分析的基础上,我们利用人脸图像的对称性,提出了基于对称二维主成分分析的特征提取方法;(2)在线性鉴别分析的基础上,我们利用模糊集理论,提出了基于完备模糊LDA的特征提取方法;(3)在间距最大准则的基础上,我们考虑了样本分布的潜在流形结构,提出了基于拉普拉斯间距最大准则的特征提取方法;(4)在非监督鉴别投影的基础上,我们利用核技巧,将非监督鉴别投影推广到核空间,提出了基于核非监督鉴别投影的特征提取方法。更多还原

【Abstract】 Face recognition is one of the hot topics in the field of pattern recognition, and it belongs to biometrics. In this field, feature extraction is one of the key steps. In the passed decade years, many correlated algorithms have been proposed to solve the problem. For example, linear discriminant analysis (LDA), principal component analysis (PCA) and independent component analysis (ICA) are developed to solve linear problem, and kernel methods based on support vector machine (SVM)) are proposed to solve nonlinear problem.The work in the paper includes:(1) In this paper, a new algorithm, called feature extraction based on symmetrical 2DPCA, is proposed. The algorithm is based on the theory of function decomposition in algebra and mirror symmetrical in geometry and 2DPCA.(2) In this paper, a new algorithm, called feature extraction based on complete fuzzy LDA, is proposed. The algorithm redefines the fuzzy between-class scatter matrix and fuzzy within-class scatter matrix that make fully of the distribution of sample and simultaneously extract the irregular discriminative information and regular discriminative information.(3) In this paper, a new algorithm, called feature extraction based on laplacian MMC, is proposed. The algorithm defines the total laplacian matrix, within-class laplacian matrix and between-class laplacian matrix using the samples similar weighting to capture the scatter information of samples. Lapalcian MMC gets the discriminant vectors by maximizing the difference between between-class laplacian matrix and within-class laplacian matrix.(4) In this paper, a new algorithm, called feature extraction based on kernel unsupervised discriminant projection (Kernel UDP). We formulate the Kernel UDP theory and develop a two-stage method to extract Kernel UDP features.更多还原