Large memory acousto-optically addressed pattern recognition

Abstract

A method of and apparatus for acousto-optically addressed pattern recognition using Vander Lugt matched filtering. In optical correlation a beam of light containing an image to be examined is directed to fall on a matched filter array that contains a representation of an object of interest. Critical alignment between the beam and a matched filter of the array is achieved by acousto-optical deflection of the light beam. A large number of matched filters can be addressed sequentially at very high sampling rates, thereby providing a large memory capability for coherent optical image processing. In the system an object beam containing an image of interest is directed to a plane of interest where the matched filter is located. An acousto-optical device located in the object beam path is driven by a signal generator at megahertz frequencies. The acousto-optical device converts the megahertz frequency signals to acoustic waves that affect the density of the optical path of the object wave, changing the index of refraction at a selected plane in this path. This spatial modulation of the index of refraction across the path of the wave causes the incident light to be deflected at selectable angles. By proper selection of the megahertz frequency driving signals these variable angles are selected to direct the Fourier transform of the object beam to selected ones of an array of matched filters at the focal plane, allowing rapid correlation of any incoming intelligence information on the object beam with prerecorded information stored in a filter array.