Full text: Transactions of the Symposium on Photo Interpretation

56 
SYMPOSIUM PHOTO INTERPRETATION, DELFT 1962 
carried out using a general-purpose digital computer (I.B.M. 7090), and are 
showing promising results. We will consider now, very quickly, what occurs. 
Unknown patterns are presented to the computer as a series of zeros and 
ones on a 20x20 matrix, as seen in fig. 1. Next, the program generates a 
series of “operators”. These operators are local 5x5 matrices of zeros and 
ones which have been chosen from random portions of the unknown input 
pattern. They might then be thought of as imitating, or extracting, information 
from these unknown patterns. The operators are then swept across the pattern 
to generate a list of characteristics of the pattern. The list is actually a record 
of the number of pattern matches and the location of pattern matches for a 
given operator. These characteristics are then compared with other lists of 
characteristics in the machine memory; one for each type of pattern previously 
processed. As a result of similarity tests, the name of the list most similar to 
the list of characteristics just computed is chosen as the name of the input 
pattern. The operators involved are then examined by the program and, 
depending on whether they individually contributed success or failure in 
identifying the inputs, amplifiers for each are then automatically turned up 
or down. This adjustment leads eventually to an automatic discarding of 
operators which produce poor characteristics as indicated by low amplifier 
settings, and to their replacements by new computer-generated operators. 
This system differs from other pattern recognition devices in that it can 
generate and test its own recognition criteria, and continually improve upon 
i i 
i 
i 
i 
i 
i 
i 
i 
ili 
ili 
i 
i 
i 
i 
i 
i 
i 
i i 
i i 
i 
i 
i 
i 
i 
i 
ili 
ili 
i 
i 
i 
ili 
i i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
liii 
ii i 
i i 
i i 
i i 
i i 
i liiiii 
i ili 
ili 
ii i 
ili i 
i liii 
i ili 
i ii 
i i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
i 
Fig. 2. Program flexibility 
that criteria to raise the effectiveness of the program. Fig. 2 illustrates the 
system’s capability for flexibility. The program is capable of calling both of 
these patterns a “chair” although training has been accomplished only on 
one of them.
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.