Sparse representations for hyperspectral data classification

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

7 Citations (Scopus)

Abstract

We investigate the use of sparse principal components for representing hyperspectral imagery when performing feature selection. For conventional multispectral data with low dimensionality, dimension reduction can be achieved by using traditional feature selection techniques for producing a subset of features that provide the highest class separability, or by feature extraction techniques via linear transformation. When dealing with hyperspectral data, feature selection is a time consuming task, often requiring exhaustive search of all the feature subset combinations. Instead, feature extraction technique such as PCA is commonly used. Unfortunately, PCA usually involves non-zero linear combinations or 'loadings' of all of the data. Sparse principal components are the sets of sparse vectors spanning a low-dimensional space that explain most of the variance present in the data. Our experiments show that sparse principal components having low-dimensionality still characterize the variance in the data. Sparse data representations are generally desirable for hyperspectral images because sparse representations help in human understanding and in classification.

Original languageEnglish
Title of host publication2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings
Volume2
Edition1
DOIs
StatePublished - 1 Dec 2008
Event2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings - Boston, MA, United States
Duration: 6 Jul 200811 Jul 2008

Other

Other2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings
CountryUnited States
CityBoston, MA
Period6/07/0811/07/08

Fingerprint

Feature extraction
Linear transformations
imagery
Experiments
experiment

Keywords

  • DSPCA
  • Hyperspectral data
  • PCA
  • Sparse representation
  • SPCA

Cite this

Siddiqui, S., Robila, S., Peng, J., & Wang, D. (2008). Sparse representations for hyperspectral data classification. In 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings (1 ed., Vol. 2). [4779058] https://doi.org/10.1109/IGARSS.2008.4779058
Siddiqui, Salman ; Robila, Stefan ; Peng, Jing ; Wang, Dajin. / Sparse representations for hyperspectral data classification. 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. Vol. 2 1. ed. 2008.
@inproceedings{c281b29a89c345f68c6217c1fd6d2bd7,
title = "Sparse representations for hyperspectral data classification",
abstract = "We investigate the use of sparse principal components for representing hyperspectral imagery when performing feature selection. For conventional multispectral data with low dimensionality, dimension reduction can be achieved by using traditional feature selection techniques for producing a subset of features that provide the highest class separability, or by feature extraction techniques via linear transformation. When dealing with hyperspectral data, feature selection is a time consuming task, often requiring exhaustive search of all the feature subset combinations. Instead, feature extraction technique such as PCA is commonly used. Unfortunately, PCA usually involves non-zero linear combinations or 'loadings' of all of the data. Sparse principal components are the sets of sparse vectors spanning a low-dimensional space that explain most of the variance present in the data. Our experiments show that sparse principal components having low-dimensionality still characterize the variance in the data. Sparse data representations are generally desirable for hyperspectral images because sparse representations help in human understanding and in classification.",
keywords = "DSPCA, Hyperspectral data, PCA, Sparse representation, SPCA",
author = "Salman Siddiqui and Stefan Robila and Jing Peng and Dajin Wang",
year = "2008",
month = "12",
day = "1",
doi = "10.1109/IGARSS.2008.4779058",
language = "English",
isbn = "9781424428083",
volume = "2",
booktitle = "2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings",
edition = "1",

}

Siddiqui, S, Robila, S, Peng, J & Wang, D 2008, Sparse representations for hyperspectral data classification. in 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. 1 edn, vol. 2, 4779058, 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings, Boston, MA, United States, 6/07/08. https://doi.org/10.1109/IGARSS.2008.4779058

Sparse representations for hyperspectral data classification. / Siddiqui, Salman; Robila, Stefan; Peng, Jing; Wang, Dajin.

2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. Vol. 2 1. ed. 2008. 4779058.

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

TY - GEN

T1 - Sparse representations for hyperspectral data classification

AU - Siddiqui, Salman

AU - Robila, Stefan

AU - Peng, Jing

AU - Wang, Dajin

PY - 2008/12/1

Y1 - 2008/12/1

N2 - We investigate the use of sparse principal components for representing hyperspectral imagery when performing feature selection. For conventional multispectral data with low dimensionality, dimension reduction can be achieved by using traditional feature selection techniques for producing a subset of features that provide the highest class separability, or by feature extraction techniques via linear transformation. When dealing with hyperspectral data, feature selection is a time consuming task, often requiring exhaustive search of all the feature subset combinations. Instead, feature extraction technique such as PCA is commonly used. Unfortunately, PCA usually involves non-zero linear combinations or 'loadings' of all of the data. Sparse principal components are the sets of sparse vectors spanning a low-dimensional space that explain most of the variance present in the data. Our experiments show that sparse principal components having low-dimensionality still characterize the variance in the data. Sparse data representations are generally desirable for hyperspectral images because sparse representations help in human understanding and in classification.

AB - We investigate the use of sparse principal components for representing hyperspectral imagery when performing feature selection. For conventional multispectral data with low dimensionality, dimension reduction can be achieved by using traditional feature selection techniques for producing a subset of features that provide the highest class separability, or by feature extraction techniques via linear transformation. When dealing with hyperspectral data, feature selection is a time consuming task, often requiring exhaustive search of all the feature subset combinations. Instead, feature extraction technique such as PCA is commonly used. Unfortunately, PCA usually involves non-zero linear combinations or 'loadings' of all of the data. Sparse principal components are the sets of sparse vectors spanning a low-dimensional space that explain most of the variance present in the data. Our experiments show that sparse principal components having low-dimensionality still characterize the variance in the data. Sparse data representations are generally desirable for hyperspectral images because sparse representations help in human understanding and in classification.

KW - DSPCA

KW - Hyperspectral data

KW - PCA

KW - Sparse representation

KW - SPCA

UR - http://www.scopus.com/inward/record.url?scp=66549117642&partnerID=8YFLogxK

U2 - 10.1109/IGARSS.2008.4779058

DO - 10.1109/IGARSS.2008.4779058

M3 - Conference contribution

SN - 9781424428083

VL - 2

BT - 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings

ER -

Siddiqui S, Robila S, Peng J, Wang D. Sparse representations for hyperspectral data classification. In 2008 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. 1 ed. Vol. 2. 2008. 4779058 https://doi.org/10.1109/IGARSS.2008.4779058