152 
clear that the potential exists to expand this narrow scope of 
Operation. 
PRINCIPAL OF INFORMATION PROCESSING 
Image classification isolates from the da*a set clusters of 
picture elements (pixels), which represent a certain thematic 
information or theme. These clusters are analogous to nolygons 
in a. vector oriented GIS, Further manipulation of these 
clusters is termed information processing. A large number of 
these themes (up to 32 in the system used by the authors) can 
be displayed colour coded, singley or in groups on a monitor 
for viewing. Programs are readily available in the software 
package to generate tabulated statistics of the area covered by 
particular themes within user specified geographic boundaries. 
Individual themes are identified by a numerical value assigned 
to all pixels belonging to the same theme. This means that a 
combination of disparted information layers can be achieved by 
simple arithmetic operations of addition, subtraction, 
multiplication and division, to create new layers of 
information. Weights can be assigned to themes by applying a 
constant multiplier. The theme composites can then be viewed 
on the monitor in pseudo-colour, which is formed by the 
addition of the original theme layer colours. This process is 
far less complex than polygon overlay in a vector based GIS, 
where the intersection and union of polygon webs superimposed 
upon another must be determined to create new polygon sets. 
A more sophisticated overlay operation can be performed by 
treating the theme files as ficticious image files and applying 
one of the classification aleorithms. This procedure allows 
the overlay of themes from many files, in various combination 
and in a selective manner. {Up to 16 files can be handled 
simultaneously in the system used by the authors). 
Finally, it should be pointed out that digital image analysis 
systems can process any data which are in raster format. Thus, 
maps digitized in this format can serve on the screen as 
background for thematic information. Furthermore, thematic 
information extracted from maps can be included as additional 
nata set in the classification or in the subsequent information 
processing operation. This option can substantially improve 
the classification accuracy and can provide an added dimension 
to information processinz. A proper geometric registration of 
all data sets is, of course, essential. 
A case study is now presented to illustrate the principle of 
information processing in digital image analysis systems. 
EXAMPLE: CROP ROTATION MONITORING 
  
The objective of this project is to ascertain the frequency at 
which farmers alternate the agricultural crops grown in 
individual farm fields. Four subsequent growing seasons are 
considered and geometrically corrected Landsat MSS data are 
used.