Automatic signal processing of forward looking surveillance sonar data in low signal-to-noise ratio conditions

In recent years, several surveillance sonar systems have been developed to detect intruders in difficult conditions of coastal waters like ports and harbors. Because man-made noise and reverberation from the bottom and surface are significant in shallow water, the detection process must remove this influence to be robust. Moreover, underwater static objects have strong reflection characteristics, and their echo signal can vary over time. These static objects behave as if they were low-speed, maneuvering, targets like divers. Therefore, they can cause numerous false alarms. These challenges make it difficult to realize an effective sonar-ADT (Automatic target Detection and Tracking) for shallow-coastal harbours. Starting in 2009, our group at the University of Tokyo has developed advanced ADT-techniques that reduces false alarms and improves detection performance using an interferometric method, and delivers improved target-tracking performance under difficult SNR conditions using an original stochastic framework In April 2010, RESON and the University of Tokyo started a collaborative study for automatic data processing techniques dedicated to surveillance sonars. In this paper, we present an overview of the collaborative research project and the current R&D status. Based on the result of our studies and development, we present the results of newly developed ADT techniques that significantly decrease the false alarm rate. We also present improvement of the tracking performance which is influenced by the fluctuations of the underwater target signal using our R&D techniques.