Computer Engineering and Applications ›› 2011, Vol. 47 ›› Issue (4): 165-167.DOI: 10.3778/j.issn.1002-8331.2011.04.045

• 图形、图像、模式识别 • Previous Articles     Next Articles

Probabilistic kernel PCA with application

ZHANG Jiulong,DENG Xiaonan,ZHANG Zhiyu   

  1. School of Automation Information and Engineering,Xi’an University of Technology,Xi’an 710048,China
  • Received:2009-07-30 Revised:2009-10-13 Online:2011-02-01 Published:2011-02-01
  • Contact: ZHANG Jiulong

概率核主成分分析及其应用

张九龙,邓筱楠,张志禹   

  1. 西安理工大学 自动化信息与工程学院,西安 710048
  • 通讯作者: 张九龙

Abstract: The PCA,Kernel PCA(KPCA) and Probabilistic PCA(PPCA) are popular feature extraction methods.This paper proposes a method to detect bright spots on LCD screen using Probabilistic Kernel Principal Component Analysis(PKPCA).PKPCA is considered a relatively new extraction technique which is a nonlinear extension of a PPCA.It illustrates the potential of PKPCA on detection compared with PCA and PPCA algorithms.The experiments indicate that PKPCA is superior to the PCA and PPCA in some aspects.

Key words: Principal Component Analysis(PCA), Kernel Principle Component Analysis(KPCA), Probabilistic Principal Component Analysis(PPCA), bright spot detection, Probabilistic Kernel Principal Component Analysis(PKPCA)

摘要: 主成分分析(PCA)、核主成分分析(KPCA)和概率主成分分析(PPCA)是已经取得广泛应用的特征提取方法。提出一种基于概率核主成分分析(PKPCA)的检测液晶屏幕亮点的方法。作为对PPCA的一种非线性扩展,PKPCA在PPCA的基础上引入了核函数方法,因而其捕获模式非线性特征的能力更强。在KPCA和PPCA的基础上推导了PKPCA过程公式,并在检测液晶屏幕亮点的应用中将PKPCA、PPCA、PCA算法进行比较。实验结果表明,PKPCA的检测率和局部信噪比优于其他两者。

关键词: 主成分分析, 核主成分分析, 概率主成分分析, 亮点检测, 概率核主成分分析

CLC Number: