计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (10): 153-156.DOI: 10.3778/j.issn.1002-8331.2010.10.049

• 图形、图像、模式识别 • 上一篇    下一篇

高光谱数据分类新方法研究

王祥涛,冯 燕,吴 政   

  1. 西北工业大学 电子信息学院,西安 710072
  • 收稿日期:2008-09-24 修回日期:2008-12-24 出版日期:2010-04-01 发布日期:2010-04-01
  • 通讯作者: 王祥涛

Research on new classification method for hyperspectral data

WANG Xiang-tao,FENG Yan,WU Zheng   

  1. School of Electronics and Information,Northwestern Polytechnical University,Xi’an 710072,China
  • Received:2008-09-24 Revised:2008-12-24 Online:2010-04-01 Published:2010-04-01
  • Contact: WANG Xiang-tao

摘要: 传统的独立分量分析(ICA)算法无法确定高光谱数据中独立分量的个数,利用概率神经网络(PNN)训练时间短的优点,根据分类精度可以较快地确定出独立分量的个数。提出了一种在确定高光谱数据的维数之后利用支持向量机(SVM)分类的新算法思想,首先利用ICA对高光谱数据降维,并利用PNN确定出独立分量的个数,而后对降维后的数据利用SVM作交叉验证,并采用混合核函数进行分类的算法思想。通过仿真实验表明,该算法可以在保证分类精度的同时大大减少分类的时间。

关键词: 独立分量分析, 支持向量机, 高光谱, 概率神经网络, 混合核函数

Abstract: The traditional Independent Component Analysis(ICA) method can’t determine the number of Independent Components (ICs) in hyperspectral data,but it can quickly determine the number of ICs by using the advantage of short training time in Probability Neural Network(PNN).A combined method of ICA,PNN and Support Vector Machine(SVM) is proposed for hyperspectral data classification.Use ICA to do dimensionality reduction for hyperspectral dada at first,and then use PNN to determine the number of ICs,at last use SVM with mixture kernels to do classification for the dimensionality reduction data.By the experiments,this method can insure the classification accuracy while reducing the time of classification.

Key words: independent component analysis, support vector machine, hyperspectral, probability neural network, mixture kernels

中图分类号: