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

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-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 languageEnglish
Title of host publicationAlgorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI
EditorsMiguel Velez-Reyes, Fred A. Kruse
ISBN (Electronic)9781628415889
DOIs
StatePublished - 2015
EventAlgorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI - Baltimore, United States
Duration: Apr 21 2015Apr 23 2015

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume9472
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceAlgorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XXI
Country/TerritoryUnited States
CityBaltimore
Period4/21/154/23/15

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

Fingerprint

Dive into the research topics of 'Detecting plumes in LWIR using robust nonnegative matrix factorization with graph-based initialization'. Together they form a unique fingerprint.

Cite this