Autonomous anti-submarine warfare (ASW) sonars require robust automatic target classification algorithms. In conventional systems with human operators, the main role of such algorithms is to simplify the work of the sonar operator, while in autonomous systems, automatic target classification is crucial for the operative value of the systems. The emergence of the autonomous underwater vehicle (AUV), coupled with ongoing increase in computational power allowing more advanced real-time processing, has increased the interest in automatic target classification in the naval community. Detailed knowledge of the environment and an acoustic model may be used to estimate the probability that contacts are generated due to the signal processing induced phenomenon called false alarm rate inflation (FARI). This is a phenomenon often encountered in the littorals in presence of bathymetric features such as sea mounts and ridges. In this paper, we propose combining FARI information with track information, using two different machine learning techniques, k-Nearest neighbours and ID3.
展开▼
