extraction procedures successfully produced
continuum-removed spectra that show this offset.
Spot checking of the imaging spectrometer data for
areas of known mineralogy shows a good match
between extracted absorption features, laboratory
measurements, and the expert system's automated
identification
The final step in the analysis of the AVIRIS
data was to map the spatial distribution of the
minerals using the expert system. Figure 8 outlines
the actual procedures used to implement the single
spectrum analysis on an entire imaging spectrometer
cube. The analysis was segmented, with production
of several intermediate image cubes to allow
evaluation of each step. Interactive viewing of the
derived cubes showed that each contained valuable
C. Step 3. Binary encoding of library and image
cube.
CUBE OF
BINARY
MATCHES
information. The final image cube containing the
automated mineral identifications and probabilities
appears to match known mineralogy fairly well,
however, its overall accuracy is still being evaluated.
D. Step 4. Expert system analysis.
BAND PARAMETERS
push ion
DEPTH
WIDTH
ASYMMETRY
g[f 1QNS
MINERAL B
MINERAL C
REFLECTANCE
CUBE
CONTINUUM
REMOVED
CUBE
Figure 8.
Procedures for analysis of Imaging
Spectrometer data.
CUBE OF
ABSORPTION —
BAND
EXPERT
A. Step 1
Continuum removal.
PARAMETERS
ANALYSIS
MINERAL Z
VEGETATION
50 nm
^0.40 |im
INFORMATION
CUBE
PROBABILITIES
NEARLY CERTAIN
VERY HIGH
HIGH
MEDIUM HIGH
MEDIUM
MEDIUM LOW
LOW
VERY LOW
NOT LIKELY
CONCLUSIONS
An expert system has been developed that
allows automated identification of Earth surface
materials based on their spectral characteristics in
imaging spectrometer data. Automated, techniques
have been developed for the extraction and
characterization of absorption features by analyzing a
suite of laboratory spectra of some of the most
common minerals, vegetation types, and vegetation
components. Critical absorption band characteristics
for identification have been defined and these have
B. Step 2 Feature extraction and determination of
absorption band parameters.
BAND PARAMETERS
POSITION
BAND
PARAMETERS
been used to develop facts and rules defining a
generalized knowledge base for analysis of reflectance
spectra.
A tree hierarchy is being used to implement
the facts and rules in a logical fashion that allows the
computer to make decisions similar to those that
would be made by an experienced analyst. Each
spectrum is broadly classified based on its strongest
absorption features. Primary band characteristics and
secondary/tertiary absorption bands are used to
progress through the tree until an identification is
made. If the decision process fails at any point
because there is insufficient information to identify a
specific material, then the last classification is used to
give the best possible answer.
The feature extraction procedures and the
expert system have been successfully used to analyze
lab and field spectra of unknown materials and