New approaches for feature extraction in hyperspectral imagery

Stefan Robila, Lukasz Maciack

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

5 Citations (Scopus)

Abstract

In this paper we introduce a novel feature extraction method based on Nonnegative Matrix Factorization (NMF) for hyperspectral image processing. Given the large size of the hyperspectral imagery, feature extraction plays an important role in producing fast and accurate results. Traditional approaches such as Principal Component Analysis and Independent Component Analysis generate the features as a linear combination of the hyperspectral bands emphasizing on the decorrelation or independence of the features. Compared to this, NMF offers a decomposition solution that is less restrictive requiring only the positivity of the features and the associated linear transform. Such scenario has a natural meaning in hyperspectral imagery where each pixel observation is thought to be formed as a linear positive mixture of reflectance values of the materials in the scene (endmembers) covered by the pixel. With hyperspectral imagery spatial resolution ranging from millimeters to kilometers, it likely that the data observed are formed as a mixture. In this case, the linear transform used to generate the features would be associated to the endmembers and the resulting features would be associated to the abundance of each endmember in the pixels. We present our results on using NMF for feature extraction by performing experiments with hyperspectral digital imagery collection experiment (HYDICE) data as well as in-house imagery collected with a SOC 700 hyperspectral camera. The experiments suggest that NMF outperforms PCA in feature and endmember extraction.

Original languageEnglish
Title of host publication2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT
DOIs
StatePublished - 1 Dec 2006
Event2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT - Long Island, NY, United States
Duration: 5 May 20065 May 2006

Publication series

Name2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT

Other

Other2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT
CountryUnited States
CityLong Island, NY
Period5/05/065/05/06

Fingerprint

Factorization
Feature extraction
Pixels
Experiments
Independent component analysis
Principal component analysis
Image processing
Cameras
Decomposition

Cite this

Robila, S., & Maciack, L. (2006). New approaches for feature extraction in hyperspectral imagery. In 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT [4302652] (2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT). https://doi.org/10.1109/LISAT.2006.4302652
Robila, Stefan ; Maciack, Lukasz. / New approaches for feature extraction in hyperspectral imagery. 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT. 2006. (2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT).
@inproceedings{6ac25e19098644c199a284fca36b10fa,
title = "New approaches for feature extraction in hyperspectral imagery",
abstract = "In this paper we introduce a novel feature extraction method based on Nonnegative Matrix Factorization (NMF) for hyperspectral image processing. Given the large size of the hyperspectral imagery, feature extraction plays an important role in producing fast and accurate results. Traditional approaches such as Principal Component Analysis and Independent Component Analysis generate the features as a linear combination of the hyperspectral bands emphasizing on the decorrelation or independence of the features. Compared to this, NMF offers a decomposition solution that is less restrictive requiring only the positivity of the features and the associated linear transform. Such scenario has a natural meaning in hyperspectral imagery where each pixel observation is thought to be formed as a linear positive mixture of reflectance values of the materials in the scene (endmembers) covered by the pixel. With hyperspectral imagery spatial resolution ranging from millimeters to kilometers, it likely that the data observed are formed as a mixture. In this case, the linear transform used to generate the features would be associated to the endmembers and the resulting features would be associated to the abundance of each endmember in the pixels. We present our results on using NMF for feature extraction by performing experiments with hyperspectral digital imagery collection experiment (HYDICE) data as well as in-house imagery collected with a SOC 700 hyperspectral camera. The experiments suggest that NMF outperforms PCA in feature and endmember extraction.",
author = "Stefan Robila and Lukasz Maciack",
year = "2006",
month = "12",
day = "1",
doi = "10.1109/LISAT.2006.4302652",
language = "English",
isbn = "1424403006",
series = "2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT",
booktitle = "2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT",

}

Robila, S & Maciack, L 2006, New approaches for feature extraction in hyperspectral imagery. in 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT., 4302652, 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT, 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT, Long Island, NY, United States, 5/05/06. https://doi.org/10.1109/LISAT.2006.4302652

New approaches for feature extraction in hyperspectral imagery. / Robila, Stefan; Maciack, Lukasz.

2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT. 2006. 4302652 (2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - New approaches for feature extraction in hyperspectral imagery

AU - Robila, Stefan

AU - Maciack, Lukasz

PY - 2006/12/1

Y1 - 2006/12/1

N2 - In this paper we introduce a novel feature extraction method based on Nonnegative Matrix Factorization (NMF) for hyperspectral image processing. Given the large size of the hyperspectral imagery, feature extraction plays an important role in producing fast and accurate results. Traditional approaches such as Principal Component Analysis and Independent Component Analysis generate the features as a linear combination of the hyperspectral bands emphasizing on the decorrelation or independence of the features. Compared to this, NMF offers a decomposition solution that is less restrictive requiring only the positivity of the features and the associated linear transform. Such scenario has a natural meaning in hyperspectral imagery where each pixel observation is thought to be formed as a linear positive mixture of reflectance values of the materials in the scene (endmembers) covered by the pixel. With hyperspectral imagery spatial resolution ranging from millimeters to kilometers, it likely that the data observed are formed as a mixture. In this case, the linear transform used to generate the features would be associated to the endmembers and the resulting features would be associated to the abundance of each endmember in the pixels. We present our results on using NMF for feature extraction by performing experiments with hyperspectral digital imagery collection experiment (HYDICE) data as well as in-house imagery collected with a SOC 700 hyperspectral camera. The experiments suggest that NMF outperforms PCA in feature and endmember extraction.

AB - In this paper we introduce a novel feature extraction method based on Nonnegative Matrix Factorization (NMF) for hyperspectral image processing. Given the large size of the hyperspectral imagery, feature extraction plays an important role in producing fast and accurate results. Traditional approaches such as Principal Component Analysis and Independent Component Analysis generate the features as a linear combination of the hyperspectral bands emphasizing on the decorrelation or independence of the features. Compared to this, NMF offers a decomposition solution that is less restrictive requiring only the positivity of the features and the associated linear transform. Such scenario has a natural meaning in hyperspectral imagery where each pixel observation is thought to be formed as a linear positive mixture of reflectance values of the materials in the scene (endmembers) covered by the pixel. With hyperspectral imagery spatial resolution ranging from millimeters to kilometers, it likely that the data observed are formed as a mixture. In this case, the linear transform used to generate the features would be associated to the endmembers and the resulting features would be associated to the abundance of each endmember in the pixels. We present our results on using NMF for feature extraction by performing experiments with hyperspectral digital imagery collection experiment (HYDICE) data as well as in-house imagery collected with a SOC 700 hyperspectral camera. The experiments suggest that NMF outperforms PCA in feature and endmember extraction.

UR - http://www.scopus.com/inward/record.url?scp=47849113913&partnerID=8YFLogxK

U2 - 10.1109/LISAT.2006.4302652

DO - 10.1109/LISAT.2006.4302652

M3 - Conference contribution

SN - 1424403006

SN - 9781424403004

T3 - 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT

BT - 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT

ER -

Robila S, Maciack L. New approaches for feature extraction in hyperspectral imagery. In 2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT. 2006. 4302652. (2006 IEEE Long Island Systems, Applications and Technology Conference, LISAT). https://doi.org/10.1109/LISAT.2006.4302652