Comparisons of multi-source localization and spatial filtering techniques

Abstract

Localization of moving sources in a multiple source environment has been investigated with simulated annealing algorithms. Results obtained, both by classical annealing and by fast simulated annealing using a rotated coordinate basis, are presented. The fast annealing method varies from the classical method in that each parameter is varied only once per temperature reduction, resulting in many fewer iteration calls per inversion. Accuracy and efficiency for these methods as a function of signal-to-noise ratio are examined using simulated data containing a weak source in the presence of a loud interferer. A method for filtering element level data prior to inversion is compared to subarray processing for effectiveness in localizing the source of interest. The effect of varying the level of the weak source relative to the level of the interferer, as well as the sources azimuthal separation, is also presented. Comparisons between an equally spaced horizontal line array (HLA) and an HLA with nonuniform sensor spacing are made. Finally, comparisons of the inversion techniques for a real data set from a nonuniformly spaced HLA at a Gulf of Mexico site, where ground truth bottom properties have been established, are discussed. [Work sponsored by ONR.]