53
ischer Boden-
Digital -
sdaten vom
on Forschern
or fur
worden. Als
sowie Farb-
gsund Boden-
len Gebrauch
en. Einige
auf: 1)
nung von
statistische
ritik der ent-
des Systems
were found to comprise numerous spectral signatures, as supporting ground data
quickly confirmed. Spectral character was but a surrogate, if correctly iden
tified, for a composite of characteristics such as housing age and quality.
The map scale of 1:24,000, achieved in line-printer output, was also far larger
and with accordingly more detail -- than the 1:1,000,000 (and any further low
resolution enlargement) scale of the visual ERTS images. In a subsequent ex
periment, Ellefsen, Swain and Wray (1973) encountered similar situations in a
California urban area. Differences in results were but a product of the varying
environments. In yet another locality, Brian Erb and others (1973) probed the
prospect of urban land-use mapping in the Houston, Texas area. Dornbach and
McKain (1973) followed with a comparison of visual ERTS images and digital
tapes. Other centers in the' United States attempting the processing of scanner
data by computer are: Environmental Research Institute of Michigan (ERIM);
Pennsylvania State University; the University of California, Berkeley; the
Center for Advanced Computation at the University of Illinois (Urbana-Cham-
paign); the University of Kansas, Lawrence; the Georgia Institute of Technology;
and the Mississippi Test Facility of NASA. Some commercial research labor
atories have also been developing techniques. Earth Information Services
(1974) of McDonnel-Douglas, for example, conducted a test of machine processing
of ERTS digital data for the U.S. Army Corps of Engineers in mapping urban
land use in Oklahoma.
-1 and follow-
been made to
system. While
aqes recon-
in false color,
used. These
recoqnition
-drawn" maps
ion, land-
th rural and
ents.
While not all of these laboratories addressed themselves specifically
to urban land-use mapping, the results of those who did exposed a number of
common problems. It has been the express purpose of our work to develop
methodology which will consistently be able to solve these universal prob
lems. Specifically, these are: (1) making a sharp distinction between rural
and urban land uses; (2) achievement of the maximum possible number of cat
egories of urban land uses at; (3) levels of accuracy high enough to make the
product credible and useful to urban planners and; (4) to have the computer
aggregate the data by jurisdictional and other statistical units. Earlier
advances along these lines are presented in Ellefsen, Gaydos, and Wray (1974)
and Ellefsen (1974).
scover how
n be employed
els of corn-
satellite's
e highest
ical number
ds possible
Solutions to all of the problems have been sought throughout work on
each of the test sites -- Arizona, Missouri, and Indiana -- and gains have
been made in each. Since the experience has been cumulative, the highest
level has been reached in the recently completed Indianapolis work. For
purposes of clarity, considering the complexity of the analysis procedure,
a recitation of the methodological steps precedes a discussion of how
features of the method were directed toward solution of each of the stated
problems.
the last
h computer
ner infor-
hers at the
versity
, Wiscon-
concep-
as alike
METHOD
As a first step, ERTS computer compatible tapes from contrasting seasons
covering the area of interest are selected and reformatted for interaction
with the LARSYS package of pattern recognition programs. A quarter sampling
is taken of the total data (every other line and column). This proves to be