MULTISCALE TARGET DETECTION IN SAR IMAGERY

THERE ARE STRONG MOTIVATIONS FOR THE DEVELOPMENT OF NEW SIGNAL, IMAGE, AND DATA PROCESSING ALGORITHMS THAT ARE WELL-MATCHED TO ATR PROBLEMS AND THUS OFFER PROMISE IN EXPLOITING THE SPECIFIC SPATIO-TEMPORAL CHARACTERISTICS THAT DISTINGUISH TARGETS OF INTEREST FROM ENVIRONMENTAL CLUTTER. IN THIS PROPOSAL WE DESCRIBE AN EFFORT AIMED AT ACCOMPLISHING THIS TASK BY EXPLOITING THE RECENT ADVANCES IN WAVELET TRANSFORMS, MULTIRESOLUTION SIGNAL AND IMAGE ANALYSIS, AND FRACTAL IMAGE PROCESSING. IN THE PAST FEW YEARS THERE HAVE BEEN MAJOR DEVELOPMENTS, MOTIVATED BY THE WAVELET TRANSFORM, IN OPTIMAL MULTISCALE IMAGE AND SIGNAL PROCESSING, PROVIDING A NEW FRAMEWORK FOR IMAGE SEGMENTATION AND FEATURE EXTRACTION, CLUTTER REMOVAL, AND MULTISENSOR FUSION THAT HOLD THE PROMISE OF SUBSTANTIAL IMPROVEMENTS IN BOTH PERFORMANCE AND COMPUTATIONAL EFFICIENCY. WE DESCRIBE HOW THIS FRAMEWORK CAN BE USED FOR TARGET DETECTION, FEATURE EXTRACTION, AND CLUTTER REJECTION BASED EITHER ON SPATIAL OR SPATIO-TEMPORAL FEATURES. WE ALSO DESCRIBE HOW THE WAVELET TRANSFORM LEADS NATURALLY TO THE IDEA OF A WAVELET NETWORK, A NONLINEAR ADAPTIVE LEARNING FRAMEWORK THAT APPEARS TO HOLD CONSIDERABLE PROMISE FOR TARGET FEATURE EXTRACTION AND CLASSIFICATION. ANTICIPATED BENEFITS: RESULTS OF THIS EFFORT SHOULD LEAD TO NEW AND EFFECTIVE METHODS FOR THE DETECTION AND CLASSIFICATION OF SPATIO-TEMPORALLY LOCALIZED PHENOMENA IN COMPLEX ENVIRONMENTS, AND FOR MOTION DETECTION AND ESTIMATION. SUCH METHODS WILL BE OF VALUE IN BOTH MILITARY APPLICATIONS AS WELL AS IN SUCH CIVILIAN/COMMERCIAL AREAS AS REMOTE ENVIRONMENTAL MONITORING AND BIOMEDICAL IMAGING TECHNOLOGIES.