Discrimination of poorly exposed lithologies in imaging spectrometer data

Abstract

High spectral resolution imagery produced by imaging spectrometers enables the discrimination of geologic materials whose surface expression is subpixel in scale. Moreover, the use of such data makes it possible to distinguish materials which are characterized only by subtle differences in the spectral continuum. We define the “continuum” as the reflectance or radiance spanning the space between spectral features. The capability to distinguish subpixel targets will prove invaluable in studies of the geology of the Earth and planets from airborne and spaceborne imaging spectrometers. However, subpixel targets can only be uniquely identified in a truly optimal sense through the application of data reduction techniques that model the spectral contribution of both target and background materials. Two such techniques are utilized herein. They are a spectral mixture analysis approach and a low probability detection routine based on orthogonal subspace projection. These techniques were applied to the problem of detecting two different volcanic tuff units, one basaltic and one rhyolitic, in two different scenes of data measured by the airborne visible/infrared imaging spectrometer (AVIRIS). These tuff units have limited exposures from an overhead perspective and have spectral signatures which differ from those of background materials only in terms of subtle slope changes in the reflectance continuum. Of the two approaches, it was found that the low probability detection algorithm was more effective in highlighting those pixels that contained the target tuff units while suppressing the response of undesired background materials.

Detection Algorithms for Hyperspectral Imaging Applications

“…Background characterization for the detection of low-probability targets can be done using the eigenvectors [11,56] of the HSI cube correlation matrix R x = X 1 X/N or equivalently the singular vectors [61] of the data matrix X T . In the first case, matrix S b is formed by the first Q significant eigenvectors of R x .…”

Ronda peridotite massif: Methodology for its geological mapping and lithological discrimination from airborne hyperspectral data

“…DiOE erent analytical techniques have been used by several authors (see the review article of Cloutis 1996 ) to deal with hyperspectral geological remote sensing data and overcome its inherent detection limits. With the same goal as previous works that aim at discriminating subtle sub-pixel lithological variations within hyperspectral images (Mustard 1993, Farrand andHarsanyi 1995), spectral mixing analysis (SMA) was applied with a diOE erent approach (Chabrillat et al 1994 ) to airborne hyperspectral data produced by an Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) survey over the Ronda peridotite massif in southern Spain. The emphasis, beyond a primary identi cation of the major surface constituents on the basis of their spectral properties, was to propose a methodology which focuses on the subtle spectral variations within the Ronda ultrama c body that are related to bedrock and soil lithologies.…”

Thermal infrared spectral band detection limits for unidentified surface materials

“…Several remote sensing studies of solid materials have examined detection limits based on spectral signature mapping, such as linear spectral mixing models of the full spectrum using laboratory spectra, 1,2,3 or optimal matched filters. 4,5,6 Spectral signature mapping techniques have in common the same first step, which is to define the spectral signatures to search for using a spectral library or from within the scene. The successful application of these methods to identifying mineral signatures in remotely sensed spectra is affected in main by two issues.…”