Unsupervised classification of hyperspectral images on spherical manifolds

Dalton Lunga; Okan Ersoy

doi:10.1007/978-3-642-23184-1_11

Unsupervised classification of hyperspectral images on spherical manifolds

Dalton Lunga, Okan Ersoy

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

5 Scopus citations

Abstract

Traditional statistical models for remote sensing data have mainly focused on the magnitude of feature vectors. To perform clustering with directional properties of feature vectors, other valid models need to be developed. Here we first describe the transformation of hyperspectral images onto a unit hyperspherical manifold using the recently proposed spherical local embedding approach. Spherical local embedding is a method that computes high-dimensional local neighborhood preserving coordinates of data on constant curvature manifolds. We then propose a novel von Mises-Fisher (vMF) distribution based approach for unsupervised classification of hyperspectral images on the established spherical manifold. A vMF distribution is a natural model for multivariate data on a unit hypersphere. Parameters for the model are estimated using the Expectation-Maximization procedure. A set of experimental results on modeling hyperspectral images as vMF mixture distributions demonstrate the advantages.

Original language	English
Title of host publication	Advances in Data Mining
Subtitle of host publication	Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings
Pages	134-146
Number of pages	13
DOIs	https://doi.org/10.1007/978-3-642-23184-1_11
State	Published - 2011
Externally published	Yes
Event	11th Industrial Conference on Data Mining, ICDM 2011 - New York, NY, United States Duration: Aug 30 2011 → Sep 3 2011

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	6870 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th Industrial Conference on Data Mining, ICDM 2011
Country/Territory	United States
City	New York, NY
Period	08/30/11 → 09/3/11

Funding

Acknowledgements. The authors would like to thank the reviewers for helpful comments that improved the final draft of this paper. We also acknowledge inputs and insights from Dr Sergey Kirshner on related topics during the course of the study. Dalton has previously been supported by Fulbright, the National Research Foundation of South Africa and The Oppenheimmer Memorial Trust. The authors would like to thank the reviewers for helpful comments that improved the final draft of this paper. We also acknowledge inputs and insights from Dr Sergey Kirshner on related topics during the course of the study. Dalton has previously been supported by Fulbright, the National Research Foundation of South Africa and The Oppenheimmer Memorial Trust.

Keywords

directional data
hyperspectral image clustering
mixture models
spherical manifolds

Access to Document

10.1007/978-3-642-23184-1_11

Cite this

Lunga, D., & Ersoy, O. (2011). Unsupervised classification of hyperspectral images on spherical manifolds. In Advances in Data Mining: Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings (pp. 134-146). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6870 LNAI). https://doi.org/10.1007/978-3-642-23184-1_11

Lunga, Dalton ; Ersoy, Okan. / Unsupervised classification of hyperspectral images on spherical manifolds. Advances in Data Mining: Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings. 2011. pp. 134-146 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{dabfcdc7cb794a2b93492ac36d865a91,

title = "Unsupervised classification of hyperspectral images on spherical manifolds",

abstract = "Traditional statistical models for remote sensing data have mainly focused on the magnitude of feature vectors. To perform clustering with directional properties of feature vectors, other valid models need to be developed. Here we first describe the transformation of hyperspectral images onto a unit hyperspherical manifold using the recently proposed spherical local embedding approach. Spherical local embedding is a method that computes high-dimensional local neighborhood preserving coordinates of data on constant curvature manifolds. We then propose a novel von Mises-Fisher (vMF) distribution based approach for unsupervised classification of hyperspectral images on the established spherical manifold. A vMF distribution is a natural model for multivariate data on a unit hypersphere. Parameters for the model are estimated using the Expectation-Maximization procedure. A set of experimental results on modeling hyperspectral images as vMF mixture distributions demonstrate the advantages.",

keywords = "directional data, hyperspectral image clustering, mixture models, spherical manifolds",

author = "Dalton Lunga and Okan Ersoy",

year = "2011",

doi = "10.1007/978-3-642-23184-1\_11",

language = "English",

isbn = "9783642231834",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "134--146",

booktitle = "Advances in Data Mining",

note = "11th Industrial Conference on Data Mining, ICDM 2011 ; Conference date: 30-08-2011 Through 03-09-2011",

}

Lunga, D & Ersoy, O 2011, Unsupervised classification of hyperspectral images on spherical manifolds. in Advances in Data Mining: Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6870 LNAI, pp. 134-146, 11th Industrial Conference on Data Mining, ICDM 2011, New York, NY, United States, 08/30/11. https://doi.org/10.1007/978-3-642-23184-1_11

Unsupervised classification of hyperspectral images on spherical manifolds. / Lunga, Dalton; Ersoy, Okan.
Advances in Data Mining: Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings. 2011. p. 134-146 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6870 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Unsupervised classification of hyperspectral images on spherical manifolds

AU - Lunga, Dalton

AU - Ersoy, Okan

PY - 2011

Y1 - 2011

N2 - Traditional statistical models for remote sensing data have mainly focused on the magnitude of feature vectors. To perform clustering with directional properties of feature vectors, other valid models need to be developed. Here we first describe the transformation of hyperspectral images onto a unit hyperspherical manifold using the recently proposed spherical local embedding approach. Spherical local embedding is a method that computes high-dimensional local neighborhood preserving coordinates of data on constant curvature manifolds. We then propose a novel von Mises-Fisher (vMF) distribution based approach for unsupervised classification of hyperspectral images on the established spherical manifold. A vMF distribution is a natural model for multivariate data on a unit hypersphere. Parameters for the model are estimated using the Expectation-Maximization procedure. A set of experimental results on modeling hyperspectral images as vMF mixture distributions demonstrate the advantages.

AB - Traditional statistical models for remote sensing data have mainly focused on the magnitude of feature vectors. To perform clustering with directional properties of feature vectors, other valid models need to be developed. Here we first describe the transformation of hyperspectral images onto a unit hyperspherical manifold using the recently proposed spherical local embedding approach. Spherical local embedding is a method that computes high-dimensional local neighborhood preserving coordinates of data on constant curvature manifolds. We then propose a novel von Mises-Fisher (vMF) distribution based approach for unsupervised classification of hyperspectral images on the established spherical manifold. A vMF distribution is a natural model for multivariate data on a unit hypersphere. Parameters for the model are estimated using the Expectation-Maximization procedure. A set of experimental results on modeling hyperspectral images as vMF mixture distributions demonstrate the advantages.

KW - directional data

KW - hyperspectral image clustering

KW - mixture models

KW - spherical manifolds

UR - https://www.scopus.com/pages/publications/80052219032

U2 - 10.1007/978-3-642-23184-1_11

DO - 10.1007/978-3-642-23184-1_11

M3 - Conference contribution

AN - SCOPUS:80052219032

SN - 9783642231834

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 134

EP - 146

BT - Advances in Data Mining

T2 - 11th Industrial Conference on Data Mining, ICDM 2011

Y2 - 30 August 2011 through 3 September 2011

ER -

Lunga D, Ersoy O. Unsupervised classification of hyperspectral images on spherical manifolds. In Advances in Data Mining: Applications and Theoretical Aspects - 11th Industrial Conference, ICDM 2011, Proceedings. 2011. p. 134-146. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-23184-1_11

Unsupervised classification of hyperspectral images on spherical manifolds

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this