MULTISENSOR ACQUISITION AND TRACKING USING LIKELIHOOD STATISTICS

OPTIMUM MULTISENSOR PROCESSING IS ACCOMPLISHED BY FORMING A LARGE, DISTRIBUTED SENSOR ARRAY COMPOSED OF SUBARRAYS THAT MAY BE SEPARATED BY LARGE DISTANCES. ACQUISITION IS ACCOMPLISHED BY IMPLEMENTING A HYPOTHESIS TEST FOR AN OBJECT IN EACH VOLUME ELEMENT OF THE SPACE TO BE SEARCHED. SUCH TESTS PRODUCE LIKELIHOOD STATISTICS THAT CAN BE USED FOR DETECTION AND POSITION/VELOCITY ESTIMATION. THE SIZE AND SHAPE OF A VOLUME ELEMENT IS DETERMINED BY SENSOR BANDWIDTH AND SUBARRAY POSITIONS, AND BY PREDICTABLE MOVEMENT OF OBJECTS (SYNTHETIC APERTURE PROCESSING). CLOSELY SPACED OBJECTS CAN BE SEPARATED IF VOLUME ELEMENTS ARE SUFFICIENTLY SMALL; SMALL VOLUME ELEMENTS ARE ASSOCIATED WITH A HIGHLY DISTRIBUTED ARRAY, LARGE BANDWIDTH, AND PREDICTABLE MOTION OBSERVED OVER A LONG TIME. TRACKING IS ACCOMPLISHED BY COMPARING THE HYPOTHESIS TEST RESPONSES IN OVERLAPPING VOLUME ELEMENTS AND CORRECTING A TRAJECTORY HYPOTHESIS ACCORDINGLY. COMMERCIAL APPLICATIONS INCLUDE ROBOT VISION, MEDICAL IMAGING, AND AUTOMATIC INSPECTION SYSTEMS FOR QUALITY CONTROL.