高光谱遥感影像核最小噪声分离变换特征提取

基金项目: 国家自然科学基金资助项目(41071265); 高等学校博士学科点专项科研基金资助项目(20105122110006); 重庆市自然科学基金资助项目(cstc2012jjA40055); 国土资源部地学空间信息技术重点实验室开放基金资助项目(KLGSIT2013-03)

详细信息

Hyperspectral Image Feature Extraction via Kernel Minimum Noise Fraction Transform

¹School of Civil Engineering & Architecture,Chongqing Jiaotong University, 66 Xuefu Road,Chongqing 400074,China;²Key Laboratory of Geoscience Spatial Information Technology,Ministry of Land and Resources, Chengdu University of Technology,1 Dongsan Road,Chengdu 610059,China;³Chongqing Geomatics Center,62 Xingguang Road,Chongqing 401121,China

Funds: 国家自然科学基金资助项目(41071265); 高等学校博士学科点专项科研基金资助项目(20105122110006); 重庆市自然科学基金资助项目(cstc2012jjA40055); 国土资源部地学空间信息技术重点实验室开放基金资助项目(KLGSIT2013-03)

摘要: 针对高光谱遥感影像线性特征提取方法易导致信息丢失和失真的特点,在最小噪声分离(MNF)变换的基础上,引入核方法,提出核最小噪声分离(KMNF)变换高光谱影像非线性特征提取方法。Cuprite矿区AVIRIS数据实验结果表明,样本个数对KMNF特征提取的结果影响很小,较少的样本即可达到较多样本时特征提取的效果;KMNF特征提取体现了高光谱影像的非线性特征,KMNF特征提取后的影像可获得优于MNF特征提取的端元提取效果。
- 高光谱遥感 /
- 特征提取 /
- 核最小噪声分离变换 /
- 核方法
Abstract: Hyperspectral image linear feature extraction methods often cause information loss and distortion.In view of this,a new kernel minimum noise fraction(KMNF) transform hyperspectral image nonlinear feature extraction method is proposed that introduces a kernel method to minimum noise fraction(MNF) transform.Hyperspectral image KMNF feature extraction experiments were carried out.CUPRITE AVIRIS data experimental results show that sample number influences KMNF slightly,a small number of samples can get almost the same result as a large number of samples;KMNF feature extraction reflects the nonlinear characteristics of hyperspectral images,and endmember extraction effects based on KMNF images outweigh MNF images.
- hyperspectral remote sensing /
- feature extraction /
- kernel minimum noise fraction transform /
- kernel method