would also show up within large fields. Lodging and moisture stress 
could impact the homogeneity of field signatures, and often cause the 
spectral response to vary within fields, 
A special case of non-homogeneous fields is strip cropping, fig 10 a,b. 
Within a particular area there will be an organized structure between 
two or more surface conditions. The field itself can be 1dentified by 
strip widtn and length, orientation of strips and strip width and 
lengtn, orientation of strips and strip offsets from other fields. 
Inspite of the many fields which can exist in an area they can contain 
the same crop. If there are many different crops, the fields will 
have a multitude of different spectral responses. When the fields 
are similar, 40-50 distinct and homogeneous fields can easily be 
distributed between 5-8 crops, e.g wheat, barley, flax, corn, and 
rye. What cannot be seen with only one acquisition is the field 
history. The acquisitions might have to cover several crop years 
before the field history is identified. Crop analysts are particularly 
interested in which crops are grown in the fields during previous 
years. This will give a clue as to what will be planted in the 
current year and the following year. 
A.2 Processing Considerations 
There exists certain limitations on the range of characteristics 
which the different fields can take. The decision three Figure |, 
gives several paradoxial modes. There are only three conflicting 
situations. If there is a field structure the fields have to be 
few or many, but there cannot be no fields. The terms few and many 
are completely subjective, the author will consider few as béing less 
than 20 fields in a sample segment. If there are many fields within 
a fixed size area then the fields cannot be large. The converse can 
be true, there can be few small fields, but if the fields are of 
pixel size a definite field structure is lost in the resolution. 
This is true even if there are a larger number of pixels which 
were not grouped together, but are interspersed with native vegetation. 
In the case where there are many pixel size fields and they dominate 
small pockets in the segment scene, then one can consider it to be 
similar to a no field structure situation. 
In Figure 1 there are 52 possible decisions of which 2 equal branches 
are identical, and that is the two ways of processing a segment, 
highly automated or a more manually oriented process. There exists 
26 different field characteristic combinations in each of the two 
primary branches. Each of the 52 ways of processing have been given 
a maximum 7 letter code, Each letter corresponds to a level of a 
decision point or a processing consideration. Figure 2 outlines the 
coding method. The first and left most letter reflects if the area 
~~ 0D UI FF On 
TST +X 
vo 
— 
m 
— hHMO£E VO wuwmamccrumocrmpmr-mo 
© 
CI 
%-HO hAADOOAKZC