Singular spectrum analysis for effective noise removal and improved data classification in hyperspectral imaging

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

2 Citations (Scopus)
9 Downloads (Pure)

Abstract

Based on the well-known Singular Value Decomposition (SVD), Singular Spectrum Analysis (SSA) has been widely employed for time series analysis and forecasting in decomposing the original series into a sum of components. As such, each 1-D signal can be represented with varying trend, oscillations and noise for easy enhancement of the signal. Taking each spectral signature in Hyperspectral Imaging (HSI) as a 1-D signal, SSA has been successfully applied for signal decomposition and noise removal whilst preserving the discriminating power of the spectral profile. Two well-known remote sensing datasets for land cover analysis, AVIRIS 92AV3C and Salinas C, are used for performance assessment. Experimental results using Support Vector Machine (SVM) in pixel based classification have indicated that SSA has suppressed the noise in significantly improving the classification accuracy.
Original languageEnglish
Title of host publication2014 6th Workshop on Hyperspectral Image and Signal Processing
Subtitle of host publicationEvolution in Remote Sensing (WHISPERS)
Place of PublicationPiscataway, NJ
PublisherIEEE
Number of pages4
ISBN (Electronic)9781467390125
DOIs
Publication statusPublished - 26 Oct 2017
Event 2014 6th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS) - Lausanne, Switzerland
Duration: 24 Jun 201427 Jun 2014

Conference

Conference 2014 6th Workshop on Hyperspectral Image and Signal Processing
CountrySwitzerland
CityLausanne
Period24/06/1427/06/14

Keywords

  • hyperspectral imaging
  • feature extraction
  • spectral analysis
  • support vector machines
  • image reconstruction
  • matrix decomposition

Fingerprint Dive into the research topics of 'Singular spectrum analysis for effective noise removal and improved data classification in hyperspectral imaging'. Together they form a unique fingerprint.

Cite this