False alarm suppressing for passive underwater acoustic target detecting with computer visual techniques

Given the complexity of marine environments, the detection of underwater acoustic targets frequently relies on manual visual interpretation of the imagery displayed on monitoring screens, which limits the application of related technologies on unmanned platforms. To replace human visual observation, understanding, and reasoning within underwater unmanned equipment, this study explores an intelligent detection method for Bearing Time Records based on computer vision techniques, drawing inspiration from human visual perception in detecting targets on BTR images. Firstly, an unsupervised learning method is employed to extract the bases representing various signal patterns from BTR images, with clear physical meanings. Subsequently, we utilize both algorithm-driven and data-driven methods for automated base classification, and both methods achieve a remarkable 100% accuracy rate in automatic base classification. In comparison to human visual detection, this automated approach exhibits a false alarm rate of less than 3%.