Underwater Minefield Detection in Clutter Data Using Spatial Point-Process Models

In this paper, we study the problem of detection of underwater minefields amidst dense clutter as that of statistical inference under a spatial point-process model. Specifically, we model the locations (mine and clutter) as samples of a Thomas point process with parent locations representing mines and children representing clutter. Accordingly, the parents are distributed according to a homogeneous Poisson process and, given the parent locations, the children are distributed as independent Poisson processes with intensity functions that are Gaussian densities centered at the parents. This provides a likelihood function for parent locations given the observed clutter (children). Under this model, we develop a framework for penalized maximum-likelihood (ML) estimation of model parameters and parent locations. The optimization is performed using a combination of analytical and Monte Carlo methods; the Monte Carlo part relies on a birth-death-move procedure for adding/removing points in the parent set. This framework is illustrated using both simulated and real data sets, the latter obtained courtesy of Naval Surface Warfare Center Panama City Division (NSWC-PCD), Panama City, FL, USA. The results, evaluated both qualitatively and quantitatively, underscore success in estimating parent locations and other parameters, at a reasonable computation cost.