| 基于图像矩阵的核主分量分析技术 |
| |
| 引用本文： | 高秀梅. 基于图像矩阵的核主分量分析技术[J]. 淮阴师范学院学报(自然科学版), 2010, 9(5) |
| |
| 作者姓名： | 高秀梅 |
| |
| 作者单位： | 淮阴师范学院,计算机科学与技术学院,江苏,淮安,223300 |
| |
| 基金项目： | 淮安市科技支撑计划(工业)项目 |
| |
| 摘 要： | 作为一种新的特征抽取算法,核主分量分析(KPCA)已经广泛应用于人脸等图像识别任务中.但是该方法的时间复杂度依赖于训练样本的数目N,当N很大时,算法所耗费的时间是相当可观的.本文提出了基于图像矩阵的核主分量分析技术(I-KPCA),解决了上述核方法普遍存在的问题.在CENPARM I数据库上的实验结果验证了本文方法的有效性.
|
| 关 键 词： | 核主分量分析 图像矩阵 特征抽取 手写字符识别 |
Image Matrix-based Kernel Principal Component Analysis Technique |
| |
| GAO Xiu-mei. Image Matrix-based Kernel Principal Component Analysis Technique[J]. Journal of Huaiyin Teachers College(Natrual Science Edition), 2010, 9(5) |
| |
| Authors: | GAO Xiu-mei |
| |
| Abstract: | As a novel feature extraction method,kernel principal component analysis technique(KPCA) has been applied to image recognition tasks such as human faces.There are a key problems: whose time complexity depends on the number of the training samples N.When N is very big,time the KPCA consumes is considerable large.This paper proposes image matrix-based kernel principal component analysis technique(I-KPCA) for solution of the above problems in kernel methods.Finally,the experimental results on Concordia University CENPARMI handwriting numeral database indicate that the proposed method is effective and more efficient than KPCA. |
| |
| Keywords: | principal component analysis image matrix feature extraction handwriting numeral recognition |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
