基于非线性降维算法的人脸识别
首发时间:2008-07-08
摘要:人脸识别技术是模式识别与机器视觉领域最有挑战的研究课题之一。本文着重介绍它的核心部分——特征提取与识别技术。 本文在收集和分析大量近年来国内外相关学术论文及文献的基础上,实现了完整的基于PCA的线性识别算法和基于ISOMAP的非线性识别算法,并在UMIST人脸库中用“数据可视化”的方式分析了引入流形学习方法的必要性。文中采用ISOMAP算法,同时与线性算法PCA作对比,得到结论:当数据处于非线性嵌入子流形上时,基于流形学习算法的引入是必要且有效的。
For information in English, please click here
THE NONLINER DIMENSIONALITY REDUCTION METHODS FOR FACE RECOGNITION
Abstract:The technique of human face recognition is a challengeable subject in the field of pattern recognition and machine vision. This paper mainly focused on its core part—feature the extraction and human face recognition algorithms. The paper collected and analyzed mass of recent internal and foreign papers and literatures about human face recognition. According to the understanding to the relevant papers and literatures, I implemented a complete linear human face recognition algorithm base on PCA and a non-linear human face recognition algorithm base on ISOMAP. We tested our algorithms on the UMIST dataset. Finally, the necessity and validity are confirmed via data visualization technique.
Keywords: face recognition PCA manifold ISOMAP classifier
基金:
论文图表:
引用
No.2276327786712155****
同行评议
共计0人参与
勘误表
基于非线性降维算法的人脸识别
评论
全部评论0/1000