Automated buried object classification with a broadband autonomous underwater vehicle sonar system

Abstract

The classification of buried objects in the littoral ocean is a challenging task for the underwater community. The high attenuation rate of the seabed imposes the use of a low frequency sonar system, which leads to an increased aperture length requirement to achieve high resolution. A simple classification algorithm is proposed in which a broadband source insonifies the target area while multiple receiver vehicles process the target scattering strengths over various bistatic angles to reconstruct a planar cut of the scattered field of the target. The advantages of this technique are the low rate of information exchange between vehicles, the simplicity of the processing algorithm, and less demanding processing requirements compared to the image-based techniques. The algorithm is based on multistatic measurements of the target spatial acoustic signatures, which permits a parameterization of the identification problem to a decreased number of degrees of freedom, with the consequence of decreased processing and implementation difficulty. The use of a broadband sonar system splits the processing in parallel tasks, leading to an overall classification improvement. The GOATS experiment series is used to demonstrate the classification performance of the algorithm for various targets and conditions. The results compare favorably with the state-of-the-art imaging techniques.