790 
Supervised training program dat 
Pinal processing program anc 
th€ 
As indicated by many authors, man-machine interaction greatly 
facilitates all of these processes. However, the speed and 
accuracy of the interactive regime is not restricted to those to 
with access to single-purpose machinery; any computer system gra 
which has terminal capabilities is amenable to this type of a h 
work, those emphasizing public magnetic disk storage being of 
most convenient. the 
gra 
With the interactive system, the user can extract particu- all 
lar study areas from the CCT to be put on magnetic disk, which lev 
is a more easily accessible form of storage on interactive pro 
systems, and more efficient in terms of computer time for re- mat 
peated use of the same data set. He may then use various one 
statistical programs conversationally, to "interrogate" the data gra 
set as to its characteristics, producing enhanced images using 
the statistics derived in minutes, without leaving the terminal. 
If automated classification is desired, training areas may be by 
selected directly from the terminal output, and their coordinates res 
supplied to further statistical programs, the output of these his 
being then used in a final classification routine. шар 
cou 
When the iteration these processes typically require re- etc 
suits in an acceptable product, output may then be directed to hel 
the usual hard copy devices. All of these procedures may be res 
carried out on readily available terminals by the use of alpha- is 
numeric maps and gray-level maps as picture representations, dec 
and while the pushbutton convenience of the specialized equip- hig 
ment is enviable, the alternative of well-designed software can thi 
also be comprehended quite easily by those with little experience mor 
in quantitative methods. As the dedicated machine, the terminal sc к 
interactive user, and the batch (non-conversational) user can pos 
all employ the same techniques, the real obstacle in each case Thi 
is awareness of the significance of the procedures involved, pho 
not hardware availability. The following three sections attempt to 
to explain the basis for some of the simpler quantitative methods des 
for image enhancement and analysis. 
rou 
FREQUENCY DISTRIBUTION AND MAPPING the 
gro 
In theory, separation of various classes of land use, and a d 
many other surficial features, can be performed by analyzing a ana 
summation of the responses in the various ERTS bands, which gre 
together approximate a spectral signature for each object or the 
cell in the image. In practice, this requires a fairly spphis- pha 
ticated statistical treatment of the data. If, however, the int