A maximum entropy approach to filtering and reconstructive imaging of the underwater environment

In this paper an approximate autoregressive model is developed to image a combination of point targets and extended objects in the far field, from the returns collected by an array sonar. The method operates on a single snapshot, allowing for super-resolution imaging through application to non-stationary returns derived from multiple range cells. Furthermore, a novel adaptive spatial filter is designed based on the polar decomposition of the more general estimator. This filter can be calculated on-line, suppressing the unwanted signals. Results are presented for data collected from simulation and from a 269 kHz forty-four channel linear array sonar.