A critical need of the U.S. Navy is the development of a reliable, efficient and robust underwater target detection and classification system that can operate in real-time with various sonar systems and in different environmental and operating conditions. To maintain performance in such difficult conditions, new solutions are needed to update the parameters of the detection and classification systems in-situ in response to environmental and operational changes. The main goal of this Phase II research is to develop robust automatic target recognition (ATR) systems for MCM applications that offer in-situ learning ability for classification and possible identification of the underwater targets. The system will be able to provide: (a) a supervised in-situ learning using expert operatorsÂ’ high-level concepts via an online relevance feedback mechanism, (b) a robust decision-making rule that uses multiple metrics such as belief and information content to decide whether or not a pattern should be learned in an unsupervised learning, (c) flexibility in the new environment to learn new patterns while maintaining the stability of the previous training for life-long in-situ learning, and (d) ability to incorporate operatorsÂ’ proficiency and confidence in scoring as well as methods for conflict resolution. This Phase II research will lead to the development of a complete system that will be tested and evaluated on many sonar imagery databases. The system will be transitioned for inclusion in the Navy testbed systems.
Keywords: In-Situ Learning, In-Situ Learning, Underwater Target Detection And Classification, Sparse Representation, Relevance Feedback, Belief Theory, Sonar Imagery., Fisher Informatio