imns
and
B.
ance
las-
T.
less
ing
Lass
owed
tly
less
RR IR
2 of
242
racy
| to
‚ing
[ to
be
ent
Iden
ion
the
WEB
ral
not
ets
nce
the
90)
ept
nds
ept
ure
lly
ges
age
was
ned
lts
"in
acy
of
UR
AG
DE
CO
WA
WE
BA 46
Figure 5: Confusion matrix for generalized
categories, with columns representing actual
categories of pixels and rows representing
classifications by SX-WEB.
spectral values were used for the clas-
sification process.
SX-WEB is currently running on an Intel (TM)
80386-based machine with a clock speed of 33
mhz and without a math coprocessor. The times
needed to run the experiments specified in this
paper were 10 to 35 minutes. This was largely
dependent on the size of the training set. It
is assumed that running SX-WEB on an Intel (TM)
80486-based machine with a higher clock speed
would significantly improve performance,
resulting from both the increased clock Speed
and the integrated math coprocessor. Once this
type of machine is available to the authors, a
much larger data set will be used to
empirically evaluate the time requirements of
the system.
In addition, PC Scheme has a fairly high
overhead for "garbage collection," and it
should be possible to rewrite the program to
minimize this, or to implement SX-WEB in
another language, such as C.
The authors are currently preparing a data set
from Landsat MSS data acquired over southern
Minnesota in July of 1988. The limitation of
the input to SX-WEB to the four spectral bands
(0.5-0.6 um, 0.6-.07 um, 0.7-0.8 um, 0.8-1.1
Hum) should be instructive.
Another area of endeavor will be to utilize
EX-WEB's abilities to perform incremental
learning and unsupervised classification.
SX-WEB will be trained to perform clas-
sification into two categories (Water/Wetland
and Other), using a small subset of the
southern Minnesota data set. The resultant
classification tree will then be used to
extract all Water/Wetland pixels from the
entire data set. These pixels will then be the
input for an unsupervised incremental
classification using EX-WEB. This will done in
order to further differentiate between
Water /Wetland types.
This classification of Water/Wetland types is
currently being performed manually in the Water
Resources Center at Mankato State University,
and it would appear that automation may be
possible.
REFERENCES
Cantor, N. and Mischel, W. 1979. Prototypes in
person perception. Advances in Experimental
Social Psychology, 12:3-52.
Civco, D.L. 1991. Landsat TM land use and land
cover mapping using an artificial neural
network. In: Proc. of the 1991 Annual Meeting
of the American Society for Photogrammetry and
Remote Sensing, Baltimore, MD. 3:67-77.
Civco, D.L. 1992a. AI and the environment. PC
AI, 6(1):43-45,56.
655
Civco, D.L. 1992b. Artificial neural networks
for land cover classification and mapping.
International Journal of Geographic Information
Systems. (in press)
Keil, F.C. 1987. Conceptual development and
category structure. In: U. Neisser (Ed.),
Concepts and Conceptual Development: Ecological
and Intellectual Factors in Categorization.
Cambridge University Press, New York.
Lillesand, T.M. and Kiefer, R.W. 1987. Remote
Sensing and Image Interpretation, Second
Edition. John Wiley and Sons, New York.
Porter, B.W., Bareiss R. and Holte R.C. 1990.
Concept learning and heuristic classification
in weak-theory domains. Artificial Intel-
ligence, 45:229-263.
Roiger, R.J. 1991. Incremental Concept
Formation In An Instructional Environment.
Unpublished doctoral dissertation, University
of Minnesota.
Rosch, E., and Mervis, C.B. 1975. Family resem-
blances: Studies in the internal structure of
categories. Cognitive Psychology: 573-603.
Rosch, E., Mervis, C.B., Gray, W., Johnson, D.,
and Boyes-Braem, P. 1976. Basic objects in
natural categories, Cognitive Psychology,
8:382-439.
Smith, E.E., and Medin, D.L. 1981. Categories
and Concepts. Harvard University Press,
Cambridge, MA.
Tversky, A. 1977. Features of similarity,
Psychological Review 84:327-352.
