Knowledge‐based geologic mapping with imaging spectrometers

Abstract

A knowledge‐based expert system has been developed that allows automated mineralogical mapping using data from imaging spectrometers. An algorithm for extraction of absorption features from digital spectral libraries produces “fact tables” characterizing the individual absorption features. An “expert” (knowledgeable user) then interactively analyzes these results to determine key absorption features for mineral identification. These key features are in turn used in simple rules to examine each picture element (pixel) in an imaging spectrometer data set. An information data cube is produced that contains a measurement of the certainty of occurrence of specific materials in the library at each pixel. An image map is also produced showing the best mineral match for each pixel. Interactive programs for analysis and visualization of the results allow display of specific certainty thresholds and automatic building of spectral end member libraries to be used in quantitative procedures such as linear spectral unmixing. Analysis results demonstrate that mineralogical information can be automatically extracted from the imaging spectrometer data to produce detailed maps for geologic applications. It also appears that the techniques described are extensible to other Earth‐surface materials.