BIOLOGICAL SIGNAL PROCESSING: MULTIACUITY VISION

A FUNDAMENTAL PROBLEM IN MACHINE VISION FOR DYNAMIC SCENARIO APPLICATIONS IS THE TRADE-OFF BETWEEN SENSOR FIELD-OF-VIEW (FOV), SENSOR ACUITY, AND THE OVERWHELMING AMOUNT OF SENSOR DATA WHICH MUST BE PROCESSED, LIMITING OVERALL SYSTEM, PERFORMANCE. TYPICALLY, FEATURES TO BE RESOLVED IN A SCENE ARE LOCALIZED WITHIN THE FOV. UNIFORM SAMPLING WITHIN THE FOV IS THUS INAPPROPRIATE; REGIONS WITH LITTLE OR NO RELEVANCE TO THE TASK ARE SAMPLED AT THE SAME RESOLUTION AS KEY FEATURES, OCCUPYING VALUABLE SIGNAL BANDWIDTH AND COMPUTATIONAL RESOURCES, AND INCREASING SYSTEM LATENCIES. THE DEVELOPMENT OF A NEW CLASS OF MACHINE VISION SYSTEMS, CALLED FOVEAL SYSTEMS, IS PROPOSED. FOVEAL SYSTEMS FEATURE IMAGING SENSORS AND SIGNAL PROCESSING WITH GRADED ACUITY COUPLED WITH CONTEXT SENSITIVE SENSOR GAZE CONTROL, ANALOGOUS TO THAT PREVELANT THROUGHOUT VERTEBRATE VISION. FOVEAL SYSTEMS OPERATE MORE EFFICIENTLY THAN UNIFORM ACUITY SYSTEMS BECAUSE RESOLUTION IS TREATED AS A DYNAMICALLY ALLOCATABLE RESOURCE. TI-IIS PERMITS A SIMULTANEOUS WIDE FOV AND HIGH ACUITY, WHILE MINIMIZING SENSOR DATA TO ONLY THAT WHICH IS RELEVANT. THE DEVELOPMENT OF ADVANCED FOVEAL MACHINE VISION TECHNIQUES IS PROPOSED FOR PHASE I BY IMPLEMENTING PREATTENTIVE AND ATTENTIVE BIOLOGICAL TECHNIQUES, AND ADAPTING UNIFORM ACUITY MACHINE TECHNIQUES. FOVEAL TECHNIQUES INCLUDE HIERARCHICAL KNOWLEDGE REPRESENTATION AND DATA STRUCTURE-MULTIPROCESSOR DESIGN, ADVANCED GAZE CONTROL, AND VARIABLE ACUITY FEATURE EXTRACTION. FOVEAL SYSTEMS OFFER IMPROVED VISION PERFORMANCE, GREATER PLATFORM INTELLIGENCE AND AUTONOMOUS OPERATION, AND LOWER SYSTEM COST. FOVEAL SYSTEMS INTERROGATE TARGETS WITH HIGH RESOLUTION TO REDUCE AMBIGUITIES AND CLASSIFICATION FALSE ALARM RATES, WHILE SUPPORTING A WIDE FOV WHICH IMPROVES MULTITARGET DETECTION AND TRACK CORRELATION. FOVEAL MACHINE VISION SYSTEMS FEATURE ORDERS OF MAGNITUDE LESS COMPUTATIONAL LATENCY AND HARDWARE THAN UNIFORM ACUITY SYSTEMS WITH THE SAME MAXIMUM RESOLUTION AND FOV.