NOVEL FILTER FABRICATION METHOD FOR OPTICAL PATTERN RECOGNITION

OPTICAL PATTERN RECOGNISTION BY MATCHED FILTERING IS IMPORTANT FOR MILITARY TARGET IDENTIFICATION AND INDUSTRIAL ROBOT VISION. LOW COST SPATIAL LIGHT MODULATORS HAVE BROUGHT THE TECHNIQUE CLOSER TO PRACTICAL USE BUT A BOTTLENECK REMAINS IN METHODS TO COMPUTER SYNTHESIZE HIGH QUALITY FOURIER PLANE MATCHED FILTERS EFFICIENTLY AND CHEAPLY. RECENT RESEARCH HAS DEMONSTRATED THAT THE BINARY PHASE-ONLY FILTER IS THE METHOD OF CHOICE. WE PROPOSE A NOVEL METHOD TO PRODUCE THE MASKS TO FABRICATE SUCH FILTERS BY MEANS OF LOW COST LASER PRINTERS WHICH HAVE RECENTLY BECOME AVAILABLE FOR MICROCOMPUTERS. PHASE I RESEARCH WILL DEMONSTRATE THAT PRODUCTION OF BINARY FILTERS WITHE 2400 X 2400 PIXELS OF FEATURE SIZE 5 MICROMETER IS FEASIBLE USING EQUIPMENT COSTING ONLY A FEW THOUSAND DOLLARS AND 20,000 X 20,000 PIXEL FILTERS OR COMPUTER GENERATED HOLOGRAMS WITH FEATURE SIZE 1-5 MICROMETERS ARE POSSIBLE WITH WIDELY AVAILABLE INDUSTRIAL DEVICES. THE LASER PRINTERSTO BE USED ARE DRIVEN BY A UNIVERSAL GRAPHIC DESCRIPTION COMPUTER LANGUAGE, POSTSCRIPT. PHASE I RESEARCH WILL EXAMINE THE BEST WAYS TO UTILIZE POSTSCRIPT AND POSTSCRIPT COMPATIBLE LASER PRINTERS FOR COMPUTER SYNTHESIS OF MATCHED FILTERS AND OTHER NOVEL APPLICATIONS IN OPTICAL SIGNAL PROCESSING.