Target Recognition Using Associative Parallel Processors

SDIO HAS THE DIFFICULT PROBLEM OF FINDING TARGETS IN CLUTTER WITH PRACTICALLY NO TOLERANCE FOR ERROR BECAUSE OF THE AUTOMATED NATURE OF RESPONSE AND THE SHORT TIMES INVOLVED. TARGET RECOGNITION IN CLUTTER REQUIRES HIGH SPEED COMPUTATION AND THE OPTIMUM USE OF PARALLEL PROCESSING TO PRODUCE RESULTS IN A TIMELY WAY. ATLANTIC AEROSPACE HAS DEVELOPED ON-LINEAR ALGORITHMS BASED ON GRAY SCALE MORPHOLOGICAL SIGNAL PROCESSING THAT HAVE DEMONSTRATED IMPORTANT TARGET RECOGNITION CAPABILITY. THE DEPARTMENT OF DEFENSE HAS SUPPORTED THE DEVELOPMENT OF MANY GENERATIONS OF NON VON NEUMANN COMPUTER ARCHITECTURES. AMONG THESE, THE SINGLE INSTRUCTION MULTIPLE DATA (SIMD) ARCHITECTURES ESPECIALLY EXPLOIT THE PARALLEL NATURE OF THE ALGORITHMS USED IN IMAGE PROCESSING. AMONG THE SIMD MACHINES, WE BELIEVE THAT ASSOCIATIVE PARALLEL PROCESSORS WITH CONTENT ADDRESSABLE ASSOCIATIVE MEMORY ARE WELL MATCHED TO THIS APPLICATION. HOWEVER, EXPLOITATION OF THE EXISTING ASSOCIATIVE PARALLEL PROCESSORS REQUIRES CAREFUL ANALYSIS OF THEIR SPECIFIC IMPLEMENTATION AS APPLIED TO A SPECIFIC CLASS OF APPLICATIONS. WE PROPOSE TO MAP THE VARIOUS COMPONENTS OF THE MORPHOLOGY-BASED TARGET DETECTION/RECOGNITION INTO THE ARCHITECTURE OF THE CURRENTLY AVAILABLE ASSOCIATIVE PARALLEL PROCESSORS WITH CONTENT ADDRESSABLE MEMORY TO OPTIMIZE EXECUTION TIME OF A CLASS OF ALGORITHMS USED FOR TARGET RECOGNITION.