565 
ARTIFICIAL INTELLIGENCE FOR ANALYSIS OF 
IMAGING SPECTROMETER DATA 
F. A. Kruse 
Center for the Study of Earth from Space (CSES) 
Cooperative Institute for Research in Environmental Sciences (CIRES) 
University of Colorado, Boulder, CO 80309 
ABSTRACT 
The Earth Observing System (EOS) scheduled to be launched in the late 1990s will carry high spectral 
resolution sensors called "imaging spectrometers" into orbit to begin observing in detail the geology and 
ecosystems of our planet. Imaging spectrometers measure near-laboratory-quality spectra in narrow spectral 
bands with a corresponding spatial image for each band. While they provide us with the means to improve the 
geologic mapping procedure by collecting detailed information about the Earth's surface, the major difficulty 
confronting scientists is that the immense volume of data collected by these systems prohibit detailed manual 
analysis. 
An expert system has been developed that allows automated identification of Earth surface materials 
based on their spectral characteristics in imaging spectrometer data. Field and laboratory spectral reflectance 
measurements were used to develop a generalized knowledge base for analysis of the visible and infrared 
reflectance spectra. The knowledge base allows the computer to make decisions similar to those made by an 
experienced analyst. Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data were used as an analog to 
the EOS High Spectral Resolution Imaging Spectrometer (HIRIS). A complete reflectance spectrum for each 
picture element (pixel) was derived directly from the imaging spectrometer data and automated techniques 
were used to extract and characterize absorption feature positions and shapes. Each spectrum was classified 
using a tree hierarchy emphasizing groups of materials with similar spectral characteristics. The final product 
of the automated analysis was an "image cube" showing the location and probability of occurrence of specific 
materials. This provides the starting point for detailed scientific analysis. 
KEY WORDS: Imaging spectrometers, Expert system, spectral 
analysis, Artificial intelligence, AVIRIS, HIRIS 
INTRODUCTION 
Many naturally occurring materials can be 
identified and characterized based on their reflected- 
light spectral characteristics. In geology, the exact 
positions and shapes of visible and infrared 
absorption bands are different for different minerals, 
and reflectance spectra allow direct identification 
(Hunt and Salisbury, 1970; Hunt et al., 1971; Hunt, 
1977). Vegetation also has distinct absorption 
features caused by pigments, cell morphology, 
internal refractive index discontinuities, and water 
content (Gates et al., 1965; Knipling, 1970). Studies of 
plant constituents such as chlorophyll, lignin, sugar, 
starch, and protein have demonstrated that 
reflectance measurements can be used to obtain 
quantitative information about plant biochemistry, 
health, and productivity (Thomas and Oerther, 1972; 
Gausman, 1978; Peterson et al., 1988). 
Imaging spectrometers acquire continuous, 
near-laboratory-quality reflectance spectra in narrow 
bands. This makes possible direct identification of 
surface materials based on spectral characteristics 
(Figure 1) and presentation of the results as images. 
A complete spectrum for each picture element (pixel) 
can be derived from the data to allow quantification 
of physical parameters. 
Figure 1. The imaging spectrometer concept 
(from Vane, 1985). 
carried on the polar platform "Earth Observing 
System" (EOS) will provide high spatial and spectral 
resolution measurements of the Earth's surface for 
geologic and ecosystems investigations (NASA, 1987). 
The Moderate Resolution Imaging Spectrometer 
(MODIS) will make multispectral measurements on 
the continental scale (NASA, 1986). Simultaneous 
use of the the HIRIS and MODIS sensors will provide 
opportunities for nested measurements of high 
spatial and spectral resolution HIRIS data for selected 
sites on the regional scale within the lower 
resolution, synoptic scale MODIS images. These 
measurements will provide new information about