Aggregate Filter

Optical pattern recognition is extremely effective for well defined targets, and naturally discriminates against small changes in the target shape. Attempts to reduce the discriminatory power of the Fourier plane filter to permit some adjustments such as size or perspective changes while still identifying the 'overall shape' have been generally unsuccessful. These attempts usually involve reducing the information content in the detection filter through various forms of low pass filtering such as circular or radial smearing. Our research is directed at increasing the information content to obtain optical pattern recognition of distorted or rotated or multiple targets. This is accomplished by combining multiple Fourier plane filters which are non- overlapping in the Fourier plane and therefore do not interfere with each other. This system fills the Fourier plane with useful information for detecting multiple targets. It is most effective when the targets are most dissimilar, therefore having the least overlap in the Fourier plane . This is opposite from past approaches aimed at smearing together several similar targets. (RRH)