Page 390 - Introduction to Information Optics

7.3. Polychromatic Pattern Recognition

Fig. 7.21. (a) Reference patterns with a noisy input, (b) Obtained from the CJTC, (c) Output
correlation after first iteration, (d) Output correlation after second iteration.

8. If the correlation output requirement is not met, update weight coeffi-
cients for reference and filters and repeat steps 3 to 7, and so forth.

Let us assume the input to a nonzero JTC represents a set of letters; "B",
"C", "O," and "Q" are the reference targets with a noisy target "C," as shown
in Fig. 7.2la. The corresponding output correlation using a CJTC is shown in
Fig. 7.2 Ib, in which we see that CJTC fails to detect the target. Figures 7.2 Ic
and 7,2Id show the results obtained with the first and second iterations,
respectively, in which we see that detection of letter "C" is rather visible.
By exploiting the flexibility operation of computers, signal detection reli-
ability can be improved by continuous iteration.

7.3. POLYCHROMATIC PATTERN RECOGNITION

There are essentially two approaches to polychromatic correlation detec-
tion; one uses the VLC and the other uses JTP. We will describe these two
approaches independently.