High Frequency Synthetic Aperture Sonar Acoustic Time Series Model Validation
Abstract
The Office of Naval Research (ONR) Mine Countermeasures (MCM) community is held back by the absence of a high frequency (HF) synthetic aperture sonar (SAS) simulation tool capable of generating realistic data. It is critical that data products produced by this model have been validated to guarantee sufficient fidelity to, among other applications, augment ML training datasets. The proposed effort will result in a description of a generalized validation strategy and development of specific model validation procedures applicable to a HF SAS time series simulator. The ability to validate simulated data will raise confidence in the fidelity of simulated SAS images and also the machine learning classification algorithms trained on those images. The unified approach proposed here will consider a variety of tests, from basic through advanced, and will organize these tests into a flow chart that illustrates the connections between validation tests and the simulator’s configuration and applications. Basic tests can be derived from fundamental acoustic concepts presented in common textbooks, for instance, ensuring that the mean squared acoustic pressure matches the sonar equation. More advanced tests should address a variety of calculations, which may vary depending on the subset of models used in a simulation, but may include validation of the bi-static scattering strength, comparing statistical moments of the scattered field to the Van Cittert-Zernike theorem, ensuring that given certain assumptions the SAS image statistics fit known distributions, and others. Although some basic validation tests will need to be passed for all applications, certain applications may require lower fidelity data which would allow for a faster computational time. Understanding the bounds on input parameters, such as scattering unit density, sonar frequency and others, will be important with respect to their effect on the resulting validity of the overall simulation for a given application. Ultimately, validation should be performed in a two-step process. First, individual models should be compared to the underlying physics, and given the assumptions made, the output of the model should be compared to analytical solutions or numerical approximations. Second, the synthetic data resulting from a realistic simulation should be compared to experimental sonar data. Detailed information about the test and environmental conditions will be needed to set up the simulation, so data from well-documented sonar experiments should be used. The proposed approach to model validation has been successful in many physics simulations tools and will raise the confidence in the fidelity of the simulated SAS images used by the ONR MCM research community.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 09, 2021
- Source ID
- N000142112438
Entities
People
- Jason Philtron
Organizations
- Office of Naval Research
- Pennsylvania State University
- United States Navy