Target/clutter discrimination using Bayesian active/passive data fusion.

Abstract

Discriminating targets of interest from background clutter is a key challenge in undersea active sonar. This is particularly true in littoral areas, where scattering from the bottom, motion in the water column, and activity on the surface all contribute to produce echoes which are difficult to distinguish from targets of interest. Statistically, these effects manifest themselves in non-Rayleigh, heavy-tailed amplitude distributions. An approach to target/clutter discrimination is described which uses the complementary information from both active and passive acoustic sensors to facilitate this task. The method uses an efficient, grid-based Bayesian track-before-detect scheme to combine data from the two types of sensors by carefully modeling the effects of array processing, replica correlation, normalization, and clutter statistics. Representing measurements and uncertainty in terms of likelihood functions then provides a common framework for fusion. In this manner, active returns with coincident and appropriate passive signals are given more credence, while the presence of only one or the other, while perhaps suggestive, is not as compelling. The presentation will give a general overview of the approach, and an example will be used to illustrate the potential power of using passive data to mitigate active clutter. [Work supported by the Office of Naval Research.]