DOI: 10.3724/SP.J.1010.2013.00569

Journal of Infrared and Millimeter Waves (红外与毫米波学报) 2013/32:6 PP.569-575

Dimensionality reduction and classification based on lower rank tensor analysis for hyperspectral imagery

Sub-tensor based lower rank tensor analysis used for dimensionality reduction and classification in hyperspectral imagery was proposed.The method aims at raising classification accuracy by representing the hyperspectral image as a tensor.The tensor is divided into sub-tensors,wherein,dimensionality reduction and pixel classification were performed.Benefiting from the sub-tensors,the method capitalizes on local spatial correlation,exploits interaction between spatial and spectral dimensions,and maintains hyperspectral data structure with 3D tensor.Compared with existing theories based on tensor analysis,the proposed method eliminates the negative impacts of poor subspace dimension estimation and low global spatial correlation,which might seriously degrade performances of dimensionality reduction.Moreover,as long as subspace dimensions are smaller than sub-tensor dimensions,the method with sub-tensors achieves much higher classification accuracy than the method without sub-tensors.Therefore,for the proposed method,it is not necessary to estimate the subspace dimension.Experimental results of both simulated and real hyperspectral data demonstrated that sub-tensor based lower rank tensor analysis gives better performance in dimensionality reduction and brings higher classification accuracy than existing methods do.

Key words:hyperspectral imagery,lower rank tensor analysis,sub-tensor,spatial correlation,subspace dimension,dimensionality reduction,classification

ReleaseDate:2015-05-04 09:28:35

[1] Chang C-I.Hyperspectral imaging:techniques for spectral detection and classification [M].New York:Plenum, 2003.

[2] Renard N, Bourennane S., Dimensionality reduction based on tensor modeling for classification methods [J].IEEE Trans.on Geosci.Remote Sens., 2009, 47(4):`801-805.

[3] Cichochi A, Zdunek R, Phan A H, et al.Nonnegative matrix and tensor factorizations [M].New York:Wiley, 2009.

[4] Wax M, Kailathl T.Detection of signals by information theoretic criteria[J].IEEE Trans.on Geosci.Remote Sens.1985, 33(2):387-392.

[5] Bioucas-Dias J, Nascimento J.Hyperspectral subspace identification [J].IEEE Trans.on Geosci.Remote Sens.2008, 46(8):2435-2445.

[6] Chang C, Du Q.Estimation of number of spectrally distinct signal sources in hyperspectral imagery [J].IEEE Trans.on Geosci.Remote Sens.2004, 42(3):`608-619.

[7] Letexier D, Bourennane S, Blanc-Talon J.Nonorthogonal tensor matricization for hyperspectral image filtering [J].IEEE Geosci.Remote Sens.Letters, 2008, 5(4):`3-7.

[8] Renard N, Bourennane S, Blanc-Talon J.Multiway filtering applied on hyperspectral images [J].ACIVS 2006, LNCS2006, 4179:127-137.

[9] Letexier D, Bourennane S.Estimation of n-mode ranks of hyperspectral images for tensor denoising[J].EUSIPCO 2009:2594-2597.

[10] Huck A, Guillaume M.Estimation of the hyperspectral tucker ranks [J].ICASSP 2009:1281-1284.

[11] Melgani F, Bruzzone L.Classification of hyperspectral remote sensing images with support vector machines [J].IEEE Trans.on Geosci.Remote Sens.2004, 42(8):1778-1790.

[12] Cover T, Hart P.Nearest neighbor pattern classification [J].IEEE Trans.on Inf.Theory, 1967, 13(1):21-27.