Acoustic Response of Underwater Munitions Near a Water-Sediment Boundary
Dr. Steven Kargl | University of Washington
The project objective is to investigate the use of sonar in the detection and classification of underwater munitions. This research extends previous efforts (SERDP projects MR-1665 and MR-2231) that measured acoustic responses from a collection of inert munitions, scientific targets, and clutter items. The central hypothesis is that the environment and the scattering geometry within that environment can alter an acoustic response of an object. The target-in-the-environment response (TIER) must be taken into account during the development of robust detection and classification strategies. This research effort will use validated models to construct large data sets, which are required for the development, training, and testing of classification algorithms. The previous efforts concentrated on low-frequency (LF) sonar (i.e., < 50 kHz). This research will also investigate fusing LF sonar results with high-frequency (HF) imaging sonar (i.e., 100-300 kHz) to enhance the detection and classification of munitions and provide guidance in operational protocols for a towed system or remotely-operated vehicle.
Acoustic scattering data collected during the Target and Reverberation Experiment 2013 (TREX13) and Bay Experiment 2014 (BAYEX14) will be used to: (1) continue the validation of models via data-model comparisons and (2) provide data for development, training, and testing of detection and classification algorithms. Once validated, models provide the ability to produce large data sets to augment the experimental data needed for development of detection and classification algorithms. TREX13 was an at-sea experiment, where the sandy seafloor was relatively flat exhibiting only minor sand ripple and small-scale roughness. Diver surveys during site selection for BAYEX14 suggest that the water-mud interface is also relatively flat. The robustness of trained classification algorithms will be tested by an at-sea rail experiment in an acoustically hard environment during the second year. This type of environment is characterized by high clutter. The general concept for this experiment is to include natural (e.g., rocks, shells) and man-made (e.g., tires, barrels, lobster or crab traps) clutter in close proximity to targets. The experiment would be performed in conjunction with an Office of Naval Research-funded program and in collaboration with the Naval Surface Warfare Center, Panama City Division. LF and HF broadband sources and receivers, scanned along the straight rail, will be used to collect scattered acoustic signals suitable for synthetic aperture sonar processing and acoustic color template processing. The LF band covers approximately 1-31 kHz and the HF band will be within the 100-200 kHz range. The targets will be placed at horizontal ranges from 5 to 50 meters from the rail. During the third year, the performance of the trained classification algorithms against targets in an acoustically hard environment will be assessed, and retrained on the expanded set of data if needed.
The research will provide validated models for a selection of targets, trained detection and classification algorithms, and acoustic data on targets within an acoustically hard environment. Validated models can provide simulated data to (re)train the detection and classification algorithms. The data collected in the acoustically hard environment will be used to test the robustness of the detection and classification algorithms, and possibly retrain the algorithms prior to merging the operations/classification package with an applied system. Given a priori environmental information about an underwater munitions site, the package can be used to determine pre-mission operational protocols such as track spacing in a wide-area survey and predict expected TIER data. After acquisition of data at a munitions site, post-mission classification can then be performed to identify the locations of likely munitions. An operations/classification package to simulate sonar performance allows for informed decisions on the relative merits of various remediation actions (e.g., leave in place and monitor, or an expensive removal process). (Anticipated Project Completion - 2018)