Thesis

Feature extraction and data reduction for hyperspectral remote sensing Earth observation

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2015
Thesis identifier
  • T14171
Person Identifier (Local)
  • 201278738
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • Earth observation and land-cover analysis became a reality in the last 2-3 decades thanks to NASA airborne and spacecrafts such as Landsat. Inclusion of Hyperspectral Imaging (HSI) technology in some of these platforms has made possible acquiring large data sets, with high potential in analytical tasks but at the cost of advanced signal processing. In this thesis, effective/efficient feature extraction methods are proposed. Initially, contributions are introduced for efficient computation of the covariance matrix widely used in data reduction methods such as Principal Component Analysis (PCA). By taking advantage of the cube structure in HSI, onsite and real-time covariance computation is achieved, reducing memory requirements as well. Furthermore, following the PCA algorithm, a novel method called Folded-PCA (Fd-PCA) is proposed for efficiency while extracting both global and local features within the spectral pixels, achieved by folding the spectral samples from vector to matrix arrays. Inspired by Empirical Mode Decomposition (EMD) methods, a recent and promising algorithm, Singular Spectrum Analysis (SSA), is introduced to hyperspectral remote sensing, performing extraction of features in the spectral (1D-SSA) and also the spatial (2D-SSA) domain. By successfully suppressing the noise and enhancing the useful signal, more effective feature extraction and data classification are achieved. Furthermore, a fast implementation of the SSA methods is also proposed, leading to reduction of computational complexity. In addition, combination of both spectral- and spatial-domain exploitation is also included, comprising data reduction. Finally, promising Deep Learning (DL) approaches are evaluated by the analysis of Stacked AutoEncoders (SAEs) for feature extraction and data reduction, introducing a method called Segmented-SAE (S-SAE), working in local regions of the spectral domain. Preliminary results have validated its great potential in this context.
Resource Type
DOI
Date Created
  • 2015
Former identifier
  • 1239089

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T14171 2021-07-02 Public Download