Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization

Jing Qin; Thomas Laurent; Kevin Bui; Ricardo Vicente R. Tan; Jasmine Dahilig; Shuyi Wang; Jared Rohe; Justin Sunu; Andrea L. Bertozzi

doi:10.1117/12.2177342

Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization

Jing Qin, Thomas Laurent, Kevin Bui, Ricardo Vicente R. Tan, Jasmine Dahilig, Shuyi Wang, Jared Rohe, Justin Sunu, Andrea L. Bertozzi

Mathematics

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

3 Scopus citations

Abstract

We consider the problem of identifying chemical plumes in hyperspectral imaging data, which is challenging due to the diffusivity of plumes and the presence of excessive noise. We propose a robust nonnegative matrix factorization (RNMF) method to segment hyperspectral images considering the low-rank structure of the noisefree data and sparsity of the noise. Because the optimization objective is highly non-convex, nonnegative matrix factorization is very sensitive to initialization. We address the issue by using the fast Nyström method and label propagation algorithm (LPA). Using the alternating direction method of multipliers (ADMM), RNMF provides high quality clustering results effectively. Experimental results on real single frame and multiframe hyperspectral data with chemical plumes show that the proposed approach is promising in terms of clustering quality and detection accuracy.

Original language	English
Title of host publication	Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI
Editors	Miguel Velez-Reyes, Fred A. Kruse
ISBN (Electronic)	9781628415889
DOIs	https://doi.org/10.1117/12.2177342
State	Published - 2015
Event	Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI - Baltimore, United States Duration: Apr 21 2015 → Apr 23 2015

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	9472
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI
Country/Territory	United States
City	Baltimore
Period	4/21/15 → 4/23/15

Bibliographical note

Publisher Copyright:
© 2015 SPIE.

Keywords

Data analysis
Hyperspectral images
Image processing
Label propagation
Non-negative matrix factorization
Nyström extension
Robust principal component analysis
Spectral clustering

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Condensed Matter Physics
Computer Science Applications
Applied Mathematics
Electrical and Electronic Engineering

Access to Document

10.1117/12.2177342

Cite this

Qin, J., Laurent, T., Bui, K., Tan, R. V. R., Dahilig, J., Wang, S., Rohe, J., Sunu, J., & Bertozzi, A. L. (2015). Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization. In M. Velez-Reyes, & F. A. Kruse (Eds.), Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI Article 94720V (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 9472). https://doi.org/10.1117/12.2177342

Qin, Jing ; Laurent, Thomas ; Bui, Kevin et al. / Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization. Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI. editor / Miguel Velez-Reyes ; Fred A. Kruse. 2015. (Proceedings of SPIE - The International Society for Optical Engineering).

@inproceedings{020f5d8d415946e7a85fc203d08b99a9,

title = "Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization",

abstract = "We consider the problem of identifying chemical plumes in hyperspectral imaging data, which is challenging due to the diffusivity of plumes and the presence of excessive noise. We propose a robust nonnegative matrix factorization (RNMF) method to segment hyperspectral images considering the low-rank structure of the noisefree data and sparsity of the noise. Because the optimization objective is highly non-convex, nonnegative matrix factorization is very sensitive to initialization. We address the issue by using the fast Nystr{\"o}m method and label propagation algorithm (LPA). Using the alternating direction method of multipliers (ADMM), RNMF provides high quality clustering results effectively. Experimental results on real single frame and multiframe hyperspectral data with chemical plumes show that the proposed approach is promising in terms of clustering quality and detection accuracy.",

keywords = "Data analysis, Hyperspectral images, Image processing, Label propagation, Non-negative matrix factorization, Nystr{\"o}m extension, Robust principal component analysis, Spectral clustering",

author = "Jing Qin and Thomas Laurent and Kevin Bui and Tan, {Ricardo Vicente R.} and Jasmine Dahilig and Shuyi Wang and Jared Rohe and Justin Sunu and Bertozzi, {Andrea L.}",

note = "Publisher Copyright: {\textcopyright} 2015 SPIE. Copyright: Copyright 2015 Elsevier B.V., All rights reserved.; Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI ; Conference date: 21-04-2015 Through 23-04-2015",

year = "2015",

doi = "10.1117/12.2177342",

language = "English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

editor = "Miguel Velez-Reyes and Kruse, {Fred A.}",

booktitle = "Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI",

}

Qin, J, Laurent, T, Bui, K, Tan, RVR, Dahilig, J, Wang, S, Rohe, J, Sunu, J & Bertozzi, AL 2015, Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization. in M Velez-Reyes & FA Kruse (eds), Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI., 94720V, Proceedings of SPIE - The International Society for Optical Engineering, vol. 9472, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI, Baltimore, United States, 4/21/15. https://doi.org/10.1117/12.2177342

Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization. / Qin, Jing; Laurent, Thomas; Bui, Kevin et al.
Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI. ed. / Miguel Velez-Reyes; Fred A. Kruse. 2015. 94720V (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 9472).

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

TY - GEN

T1 - Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization

AU - Qin, Jing

AU - Laurent, Thomas

AU - Bui, Kevin

AU - Tan, Ricardo Vicente R.

AU - Dahilig, Jasmine

AU - Wang, Shuyi

AU - Rohe, Jared

AU - Sunu, Justin

AU - Bertozzi, Andrea L.

PY - 2015

Y1 - 2015

N2 - We consider the problem of identifying chemical plumes in hyperspectral imaging data, which is challenging due to the diffusivity of plumes and the presence of excessive noise. We propose a robust nonnegative matrix factorization (RNMF) method to segment hyperspectral images considering the low-rank structure of the noisefree data and sparsity of the noise. Because the optimization objective is highly non-convex, nonnegative matrix factorization is very sensitive to initialization. We address the issue by using the fast Nyström method and label propagation algorithm (LPA). Using the alternating direction method of multipliers (ADMM), RNMF provides high quality clustering results effectively. Experimental results on real single frame and multiframe hyperspectral data with chemical plumes show that the proposed approach is promising in terms of clustering quality and detection accuracy.

AB - We consider the problem of identifying chemical plumes in hyperspectral imaging data, which is challenging due to the diffusivity of plumes and the presence of excessive noise. We propose a robust nonnegative matrix factorization (RNMF) method to segment hyperspectral images considering the low-rank structure of the noisefree data and sparsity of the noise. Because the optimization objective is highly non-convex, nonnegative matrix factorization is very sensitive to initialization. We address the issue by using the fast Nyström method and label propagation algorithm (LPA). Using the alternating direction method of multipliers (ADMM), RNMF provides high quality clustering results effectively. Experimental results on real single frame and multiframe hyperspectral data with chemical plumes show that the proposed approach is promising in terms of clustering quality and detection accuracy.

KW - Data analysis

KW - Hyperspectral images

KW - Image processing

KW - Label propagation

KW - Non-negative matrix factorization

KW - Nyström extension

KW - Robust principal component analysis

KW - Spectral clustering

UR - http://www.scopus.com/inward/record.url?scp=84946062882&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84946062882&partnerID=8YFLogxK

U2 - 10.1117/12.2177342

DO - 10.1117/12.2177342

M3 - Conference contribution

AN - SCOPUS:84946062882

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI

A2 - Velez-Reyes, Miguel

A2 - Kruse, Fred A.

T2 - Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI

Y2 - 21 April 2015 through 23 April 2015

ER -

Qin J, Laurent T, Bui K, Tan RVR, Dahilig J, Wang S et al. Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization. In Velez-Reyes M, Kruse FA, editors, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI. 2015. 94720V. (Proceedings of SPIE - The International Society for Optical Engineering). doi: 10.1117/12.2177342

Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this