Spatial-spectral contextual image analysis of hyperspectral data to aid in the characterisation of hydrothermal alteration in epithermal gold deposits

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

13 Downloads (Pure)

Abstract

Hyperspectral remote sensing deals with instruments that sample the EM spectrum at high spectral resolution and with high spectral sampling intervals to produce surface reflectance data that can be readily compared to spectral signatures contained in spectral libraries. In geology, this technology has been used for mineral mapping and geobotanical studies to aid in mineral prospecting, environmental studies and petroleum exploration. Most image processing techniques that are available to analyse hyperspectral data by-pass three important aspects of remote sensing: Neighbourhood information is not incorporated and hence the common knowledge formulated in geostatistics that nearby control points are more likely to contain information more similar than far apart control points, surface information is not coupled to shallow subsurface information although many problems are in need of a 3 dimensional approach of study and the change over time of the spectral signature of certain earth surface materials that contains information on the type (and use) of the material at hand that may allow a better classification and or discrimination (from other) of materials is not used in sub pixel classification. This paper addresses our efforts in designing spatial-spectral contextual image analysis approaches, thus tackling some of the aspects mentioned under 1. To combine both spectral as well as spatial information in the analysis of hyperspectral imagery we have developed a spatial spectral algorithm; the template matching algorithm. A template consists of a one- or two-dimensional array that is filled on both sides of a central pixel with spectra (or spectral derivatives such as ratios etc) that characterize certain cover types of interest. The template is matched with an image by moving the kernel or a filter over the image and calculating parameters at each pixel. At each pixel the template is also rotated and a number of parameters are calculated (e.g., best template fit, worst template fit, mean template fit, variance in template fit, and optimal scan angle). Examples are given to the application of template matching as a spatial-spectral contextual image processing technique dealing with the detection of mineral assemblage coexistence in relation to hydrothermal alteration in a epithermal gold deposit.
Original languageEnglish
Title of host publicationNew developments and challenges in remote sensing
Subtitle of host publicationProceedings of the 26th Annual Symposium of the European Association of Remote Sensing Laboratories (EARSeL), Warsaw, Poland, 29 May - 2 June 2006
EditorsZ. Bochenek
Place of PublicationRotterdam, The Netherlands
PublisherMillpress
Pages275-285
ISBN (Print)978-90-5966-053-3
Publication statusPublished - 2007
Event26th Annual Symposium of the European Association of Remote Sensing Laboratories (EARSeL) 2006: New Developments and Challenges in Remote Sensing - Warsaw, Poland
Duration: 29 May 20062 Jun 2006
Conference number: 26

Conference

Conference26th Annual Symposium of the European Association of Remote Sensing Laboratories (EARSeL) 2006
Country/TerritoryPoland
CityWarsaw
Period29/05/062/06/06
Other

Keywords

  • ADLIB-ART-224
  • ESA
  • 2023 OA procedure

Fingerprint

Dive into the research topics of 'Spatial-spectral contextual image analysis of hyperspectral data to aid in the characterisation of hydrothermal alteration in epithermal gold deposits'. Together they form a unique fingerprint.

Cite this