Printresults is used to produce a line-printer map using specially se 
lected alphanumeric symbols for each type of land use after the entire area is 
classified using the final selection of clusters and channels. This map is at 
a scale of 1:48,000 and shows the county boundaries with the + symbol. 
The aspect ratio of the classification is changed from the line-printer 
4X5 format to a 1 XI format for use on video display devices by producing 
a new tape with every fifth column of the original deleted. This is then 
reformatted for use on a Dicomed D47 film recorder. Colors are selected for 
each land use/cover class and the classification is plotted using a low 
resolution matrix where every pixel from the classification is plotted as 
16 spots on 4 X 5 inch color film. 
Enlargements are then made photographically to match the land use to 
varying scales. The final product is a color-coded land use map (figure 3). 
PROBLEM SOLUTION 
The method described above comprises all of the manipulations of the 
system which have been developed while working on urban land-use mapping for 
nearly a two year period at LARS, Purdue. While many of the steps are part 
of the basic pattern recognition series of programs (LARSYS), a number of 
refinements and modifications in the analysis procedure and employment of 
LARSYS have come in direct response to the special needs and situations 
encountered in examining dynamic and highly complex metropolitan areas. 
The problem of making a sharp distinction between urban and rural areas 
has, by itself, created a need for many modifications in the analysis procedure. 
Interestingly, while rural-urban separation (at least for large cities) is 
simple enough on the 1:1,000,000 scale visual ERTS-1 images,* some nearly 
identical sets of digital values between certain urban land uses and certain 
rural uses cause mis-classification when analyzing spectral information alone. 
While some degree of error is inherent in all computer-aided mapping, mis- 
classification in rural-urban distinctions is much more noticeable (and less 
tolerable) than, for example, mistaking a crop type or tree species in an 
area which is entirely rural. 
Probing the causes of the problem revealed that much of the urban scene 
was indeed spectrally similar to certain rural features. For instance, some 
residential areas, with their combinations of landscaping and certain degrees 
of weathering were very much like some rural cropland areas which had com 
binations of bare soils and some crop and other vegetative cover. A search 
for possible distinguishing characteristics led to the investigation of the 
use of multi temporal in conjunction with multispectral data. That the 
character of the agricultural landscape changes materially from one season to 
another while the urban scene -- with its large component of man-made cover -- 
remains relatively constant, was the working basis for making distinctions. 
* Several researchers have referred to the "electric blue" color of com 
mercial areas of cities on the color infrared composite images.