and modification into two light levels of value 1 and 0, a binary 
matrix is formed. Under normal circumstances the tree signal (of 
coffee) would be either 0 or i and that of the background soil 1 
or O0. The actual output would depend upon the part of the spectrum 
used (tube sensitivity) and filter combinations. 
The characteristics of the distributions of the 
binary figure within the matrix provide a simple basis for classi 
fication by automatic means. Several bases exist for automatic - 
pattern recognition some of which are outlined below: 
  
Linear Counting: The repetitive counting of the size of the comple e - 
te binary words within each scan line of the television (a word re 
presents a continuous series of Os (zeros) followed by a continuous 
series of ls (ones), or vice-versa) provide estimates of crop spa- 
cing (tree plus soil). Coffee, as related to actual spacing in the 
field, has a specific expected word-length range. Those signals - 
which respectively fall into the range are classified as coffee, 
The maximum tree signals which constitute the binary word indica- 
te the diameters involved. This technique can be operated in re- 
al time. 
Matrix Displacement: The Displacement of the binary lines (TV li- 
  
nes) so as to maximise the pairing of like units (Os and 1s) betw + -- 
een the lines provides a means of estimating the crop spacing and 
also the orientation angle of the TV scan to the pattern direction 
of the plantation. This technique could either be performed in a 
computer or in a minicomputer. 
Statistical Association: The association of various spacings (word 
  
lengths) can indicate the basic orientation and basic crop spacing 
involved (based upon statistical probabilities of word length asso 
ciation) with real time countings,