一种基于非线性特征提取的数据降维算法  

A Reduced Dimension Algorithm Based on Nonlinear Feature Extraction Data

作　　者：侯远韶 

机构地区：[1]郑州工业应用技术学院信息工程学院,河南郑州451100

出　　处：《平顶山学院学报》2016年第5期70-73,共4页Journal of Pingdingshan University

基　　金：河南省科技攻关计划(0721002210032)

摘　　要：为了获得良好的图像分类效果,需要采集尽可能多的图像数据特征,进而使得图像原始特征空间的维数越来越高,造成维数灾难.特征提取是通过线性或者非线性映射,将高维特征空间映射到低维空间,从而降低数据维数.现有的特征提取算法往往忽视了数据本身特有的复杂结构以及非线性因素,造成映射方向的模糊以及分类精确度的缺失.充分考虑了图像数据本身的二维特性,通过改进的非线性特征提取方法、流形学习方法来提取图像特征.实验表明,该算法在不影响图像分类效果的前提下可以大大降低数据维数,减少计算复杂度.In order to obtain good image classification results,we need to collect as many image data features as possible,which makes the dimension of the original feature space higher and higher,resulting in curse of dimensionality. Feature extraction is to map the high-dimensional feature space to low-dimensional space through linear or nonlinear mapping,thus reducing the data dimension. Existing feature extraction algorithms often ignore the complex structure and nonlinear factors inherent in the data,resulting in ambiguity in the direction of the mapping and lack of classification accuracy. In this paper,the two-dimensional characteristic of the image data itself is fully considered. The improved nonlinear feature extraction method and the manifold learning method are used to extract the image features. Experimental results show that the algorithm can greatly reduce the data dimension and reduce the computational complexity without affecting the image classification performance.

关 键 词：特征提取 流形学习 非线性特征提取 数据维数 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]