An optical decision filter

This paper presents the basic concepts, design details, and construction techniques used in the development of an advanced type of self-organised image filter. The filter has the capability of recognizing and classifying three-dimensional objects regardless of their size and orientation with respect to the viewing mechanism. The applications of the techniques include reconnaissance, photo interpretation, and star field recognition for space probes. Self-organization was used as a design tool only; the decision filter itself is not adaptive. Design principles described here are directly applicable to an even wider variety of problems in the broad areas of communication and data processing. A natural selection process of eliminating elements of marginal utility enables a significant reduction of network size without adversely affecting its classification capability. Full advantage was taken of the self-organization process to minimize the engineering design task involved. Many of the features described have not previously been reported in the literature.