A POTENTIAL ALGORITHM FOR TARGET CLASSIFICATION IN BISTATIC SONAR GEOMETRIES

The recent evolution in oceanographic sensing and platforms, including the availability of Autonomous Underwater Vehicles (AUVs), is encouraging the investigation of new high-resolution sonar concepts based on multistatic geometries. The rationale behind this concept is that multistatic systems, in particular if the geometry can be adapted with the experiment, can be located in the regions where the scattered signal is carrying the most information. The work described in this paper is a follow-up of the research carried out by the authors as part of the European Union project SITAR (Seafloor Imaging and Toxicity: Assessment of Risks caused by buried waste), in which a Multiple Aspect Scattering measurement technique has been investigated with the aim of imaging buried objects of small dimensions. Within SITAR, a very rich data set has been acquired in the tank facilities of the University of Bath. The data set consisted in scattered signals from a set of targets, ensonified with a high frequency source (238 kHz) as a function of grazing, scattering and bistatic angle. Moreover, the targets were proud on the surface, semi-buried or flush-buried in different kinds of seabed sediments. The SITAR data set has been processed by a suite of algorithms, inspired from spectral distances often use