Acoustic modeling for sea-mine CAD/CAC development

Computer aided detection/classification (CAD/ CAC) algorithms have been developed to help mine warfare operators correctly identify targets within vast quantities of synthetic aperture sonar/side look sonar (SAS/SLS) imagery. Typically, large, and thus expensive, sets of target and clutter exemplars are required for the evaluation and training of these algorithms. To expedite the assessment and preliminary training of CAD/CAC algorithms for systems under development, physics-based acoustic models can be utilized to generate synthetic data-sets. Model-based recognition algorithms that predict exemplars for a given search hypothesis may use acoustic models to reduce dependence on exemplar databases. Acoustic models predicting the scatter from proud and buried targets are developed and described in this paper. A rigid-target model, using boundary integral equations for the target and a sediment-scatter model, is applied for proud and partially buried targets. This model is suitable for high-frequency, narrowband sonars. An elastic-target model using a spheroidal T matrix approach is employed for buried targets. This model captures the responses appropriate for systems that scan the sediment volume. The synthetic, acoustic datasets produced are processed into SAS images and subsequently used to investigate candidate feature sets and for CAD/ CAC algorithm development.