If the 
bute- 
ts of 
ed to 
ploit 
class 
tored 
their 
V, O, 
,V,0) 
) and 
  
  
CROP CLASSES: 
AG: START WITH AGRICULTURAL LANDS. 
AG1: NON PRODUCTIVE LANDS. 
AG211: TIMBERLAND. 
AG2121: BRUSHLAND. 
AG2122:GRASSLANDS. 
AG221: FALLOW LANDS. 
AG22211 : ORCHARDS. 
AG22212:VINE AND BUSH CROPS. 
AG22221:ROW CROPS. 
AG222221:IRRIGATED PASTURE LANDS. 
AG AG2 s AG212 
AG22 
AG222^ AG2221 
    
AG2121 
AG2122 
  
AG2222 
AG22222 
    
  
    
AG222221 
AG222222 
AG222222:CONTINUOS COVER CROPS, SMALL GRAINS, HAY ETC. 
CLUES FOR INTERPRETATION: 
AG1:VEGETATION & SOIL ABSENT OR OBSCURED BY ROCK ETC. 
AG2:VEGETATION CLEARLY DISCERNIBLE ON PHOTOGRAPH. 
AG21:CULTIVATION PATTERN ABSENT; FIELD BOUNDARIES IRREGULARLY SHAPED. 
AG22:CULTIVATION PATTERN ABSENT; FIELD BOUNDARIES REGULARLY SHAPED. 
AG211: TREES PRESENT, COVERING MOST OF THE GROUND. 
AG212: TREES ABSENT OR WIDELY SCATTERED, GROUND COVERED BY LOW LYING 
VEGETATION. 
AG2121:CROWNS OF INDIVIDUAL PLANTS DISCERNIBLE, TEXTURE COARSE AND 
MOTTLED. 
«ETC. 
Figure 6: A Decision Tree Representation of the Dichotomous Photo Interpretation 
Key for Crop Classification (Source, E.G. Mtalo, 1990, page 188). 
S VO 
(s, V, o) ————M (V, O) (s.v,0) — (S) 
V SO 
(s, v, o) ——— Bs (S, O) (s,v,0) — (V) 
O sv 
(v0  — y (S, V) (svo — pp (0) 
SVO 
(s, v, o) — (True, False) 
Figure 7: Basic Operators of the LEARN Subsystem. 
A typical query is processed as illustrated below: 
Choose Option: 
Enter Subject: BAND I 
BAND 1 has spectral wavelength 0.45 to 0.52 
BAND 1 has blue nominal spectral location. 
BAND 1 application is sensing in the chlorophyll 
absorption region. 
BAND 1 has spectral wavelength 0.45 to 0.52 or 
blue nominal spectral location. 
Complex queries involving class relationships and limited 
inheritance of class attributes by class members can also be 
processed by the LEARN sub-system. During knowledge 
compilation multiple assignments of attributes(on the same 
relationship) to the same object are generalized into a single fact 
203 
, that is the statements, "plot A56 has area-50 ha"; "plot A56 
has average slope-15 degrees"; "plot A56 has curve 
number-80" ; are replaced by "plot A56 has area-50 ha or 
average slope-15 degrees or curve number-80". Conversely, 
occurrences of multiple facts bearing the same relationship to a 
single attribute are replaced by a single expression in which the 
"subject" consists of individual subjects concatenated by the 
"or" operator. 
3.3 Processing and Manipulation of Fuzzy 
Knowledge. 
It is an established fact that human beings often express 
knowledge in fuzzy terms(Zadeh ,1989). In certain fields of 
application experts rely to a large extent on fuzzy, non-precise 
facts and information during the solution of complex problems. 
This is also true in the soil erosion domain where crucial inputs, 
such as, soil characteristics, plant cover, topography, erosion 
state etc. are often expressed in fuzzy non-precise terms. 
Processing of fuzzy knowledge by computer systems poses 
three important problems, specifically, the representation and 
storage of non-precise (fuzzy) information and data, 
representation of the uncertainty inherent in fuzzy information 
and data, and retrieval and analysis of fuzzy information and 
data. Conventional databases cannot store or query fuzzy 
information since they impose a strict format for data entry and 
query. Partial solution to the problem of fuzzy queries has 
however been achieved by the extension of the relational data 
model as in Kandel (1986). Although such solutions enable 
processing of fuzzy queries they still do not permit the storage 
of fuzzy data such as for example "terrain slope-slightly less 
than 15 degrees".