|
|||||
|
|
| 基于正交化的无监督鉴别投影映射（UDP）特征提取的人脸识别方法 | |
| 引用本文： | 刘粉香,李勇智,吴松松.基于正交化的无监督鉴别投影映射（UDP）特征提取的人脸识别方法[J].计算机与现代化,2010(9):16-20,26. |
| 作者姓名： | 刘粉香 李勇智 吴松松 |
| 作者单位： | 1. 三江学院计算机基础教学部,江苏,南京,210012 2. 南京林业大学理学院,江苏,南京,210037 3. 南京理工大学计算机系,江苏,南京,210094 |
| 摘 要： | 基于UDP（Unsupervised Discriminant Projection）特征提取方法的研究,本文提出改进算法：基于正交化的UDP算法。正交化的UDP算法是基于样本的局部散度、非局部散度及正交化的条件建立准则函数。准则的目的是在满足共轭正交条件下,寻求一组投影轴使得投影后的样本特征的非局部散度最大,局部散度最小。通过在YALE库和AR库人脸库上进行实验,结果表明本文提出的正交化的UDP算法在识别率方面整体上要好于UDP算法。 |
| 关 键 词： | 人脸识别 特征提取 流形学习 局部散度 非局部散度 正交化 |
A Method of Face Recognition Based on Orthogonal UDP Feature Extraction |
|
| LIU Fen-xiang,LI Yong-zhi,WU Song-song.A Method of Face Recognition Based on Orthogonal UDP Feature Extraction[J].Computer and Modernization,2010(9):16-20,26. | |
| Authors: | LIU Fen-xiang LI Yong-zhi WU Song-song |
| Affiliation: | 1.Department of Computer Basic Education of Sanjiang University,Nanjing 210012,China;2.College of Science,Nanjing Forestry University,Nanjing 210037,China;3.Department of Computer Science,Nanjing University of Science and Technology,Nanjing 210094,China) |
| Abstract: | This paper presents a modified unsupervised discriminant projection algorithms called orthogonalized UDP.A general rule for feature extraction is to extract features as uncorrelated as possible.In the paper,an optimal set of uncorrelated discriminant features is presented,and some formula to compute it is given.This paper proves feature equation of these modified UDP algorithms and applies them to face recognition.Through doing experiment on the different face bases,this paper verifies that they can reach better recognition performance comparing with UDP. |
| Keywords: | face recognition feature extraction manifold learning local scatter nonlocal scatter orthogonal |
| 本文献已被 维普 万方数据 等数据库收录！ | |