GEOSCAN Search Results: Fastlink

GEOSCAN Menu


TitleAdaptive Threshold for Spectral Matching of Hyperspectral Data
DownloadDownloads (Preprint)
 
LicencePlease note the adoption of the Open Government Licence - Canada supersedes any previous licences.
AuthorSchwarz, J W; Staenz, K
SourceCanadian Journal of Remote Sensing vol. 27, issue 3, 2001 p. 216-224, https://doi.org/10.1080/07038992.2001.10854938
Image
Year2001
Alt SeriesEarth Sciences Sector, Contribution Series 20042733
PublisherInforma UK Limited
Documentserial
Lang.English
Mediapaper; on-line; digital
File formatpdf
Subjectsremote sensing; Spectral Angle Mapper (SAM); Compact Airborne Spectrographic Imager (CASI); hyperspectral data; technique analysis; spectral classification; class membership; classification accuracy
Illustrationsgraphs; satellite images; histograms; tables
Released2014 07 28
AbstractSpectral matching is one of several techniques that derives information which may be used to classify hyperspectral data. In this paper an analysis technique is presented which supplements the typical spectral matching algorithms and facilitates the exploitation of the information they provide. This technique may be used to adaptively set a threshold on the similarity measure and classify spectra. This adaptive threshold may be set automatically, or the information from this analysis may be used to permit the manual selection of a threshold in a manner which is more intuitive that the direct specification of a similarity measure. The technique is demonstrated by applying the Spectral Angle Mapper (SAM) to simulated and actual imaging spectrometer data. The proposed technique does not provide additional information about the similarity of two spectra, but is does provide information which is valuable for deciding class membership. While analyzing the variability of a class, in the similarity space, it became obvious that analysis of the similarity measure for the entire scene has the potential of providing valuable information for classification purposes.
GEOSCAN ID219535

 
Date modified: