Many unexploded ordnance (UXO) in the aquatic environment are buried under seafloor sediment, making them difficult to detect and remediate. Though buried, these UXO can resurface after storms or be encountered during dredging. Low frequency, sediment penetrating acoustics has the potential to improve the ability to localize, detect and classify buried UXO, aiding in remediation. The aim of this project was to improve the UXO detection and classification capabilities of the Buried Object Scanning Sonar (BOSS), a low frequency, sediment penetrating sonar, by improving the beamforming, target response extraction, and data visualization methods used with this sonar.

Technical Approach

Research initially focused on improving the beamforming methodology associated with data collected by the eBOSS. Back-projection of planar array data to a volumetric image can be a computationally burdensome task. A goal of this project was to enable the generation of complex three-dimensional imagery from which spectral response information of targets could be extracted and retain the generality and flexibility of time-domain beamforming while still remaining computationally efficient enough to be useful both to researchers and the end users of the system. The beamformer, which leveraged aperture factorization and graphics processing unit acceleration to achieve these goals, was initially developed using a basic point-scatterer simulator created as part of this project. Following testing on simulated data, the beamformer was tested with real data captured by two BOSS systems, one an autonomous underwater vehicle mounted system and the other a towed system which was actively being developed over the course of this project under ESTCP project MR18-5004 (“Multi-Sensor Towbody (MuST) for Detection, Classification, and Geolocation of Underwater Munitions”).


The original project plan called for extensive research focusing on developing data-driven autofocusing procedures to improve the acoustic data products and augment the navigation. Due to multiple factors, however, it was found that the data-driven autofocusing was unnecessary, and research was diverted to feature-based mosaic alignment and multi-aspect data fusion.


Following successful beamformer development and testing, research branched into several related threads: 1) improving the signal-to-noise ratio of target spectral features by reducing the influence of volume scattering and specular glints from sediment interfaces, 2) interfacing with current SERDP-funded UXO classification efforts, and 3) user-oriented data processing, organization and data visualization methods for improving data interpretability, a large task that included everything from live data visualization for decision making, to post-processing tools for beamforming, filtering, viewing, combining and exporting data products. The final product resulting from these research threads is the “MuST toolbox,” software library which greatly improves the UXO-remediation capabilities of the BOSS system.