Low-frequency synthetic aperture sonar (LF-SAS) in tandem with magnetic sensing techniques have successfully demonstrated the ability to detect buried underwater munitions; however, these sensors also detect an inordinate amount of buried clutter. Processing techniques are needed to discriminate buried munitions from clutter.

The long-term objective of this joint research effort between the U.S. Army Engineer Research and Development Center (ERDC) and the Naval Research Laboratory (NRL) is to develop automated methods to discriminate buried underwater munitions from buried clutter in the 3-D sub-bottom environment. This SERDP Exploratory Development (SEED) project centered on showing the practicality of an at-sea test range, bringing together the various detectors into the NRL Environmental Post Mission Analysis (EPMA) software framework, developing a detection scheme for 3-D acoustic imagery, and designing a first phase clutter classifier that uses characteristics of buried munitions and clutter derived from modeled acoustic and magnetic signatures. Validation of the discrimination methods will be performed in a controlled environment at an environmentally representative field site in follow-on research.

Technical Approach

The prior art of 2-D seafloor clutter discrimination detects mine-like objects in acoustic imagery and classifies them as mines or clutter based on object dimensions, acoustic shadows, brightness, and shape. To achieve the objectives of this project, distinguishable characteristics between munitions and clutter were discovered through modeling and controlled experiments. The 3-D Munitions and Clutter Classifier (MACC), which will be finalized in a follow-on study, will rely on NRL’s automated techniques, including derived bottom clutter, roughness, 2-D side scan detection, 3-D sub-bottom detection, and magnetic detection. Bayesian inference will be used to fuse the various detection sensors. The detections will be passed to a Support Vector Machine (SVM) classifier, which will examine feature vectors in - or derived from - the detection sensors. The classifier will separate unexploded ordnance (UXO) from UXO-like targets. UXO signatures will be obtained, and MACC will be calibrated at the Army’s underwater test facility at Duck, North Carolina. Parametric sonar and magnetic surveys will be conducted over inert munitions and clutter placed in different sediment types and at different sub-bottom depths to test the methods.


For this SEED project, previously collected data was examined to help select test sites. A survey was conducted in Duck, North Carolina, and three sites were selected to reflect the desired environmental and operational variability. Side scan Automated Target Recognition (ATR) algorithms were improved and applied to the problem, as were volumetric ATRs. Magnetic ATRs were discussed for future implementation. A preliminary design and revised technical approach to MACC was developed using Bayesian Inference to determine when the various sensors would collectively come to a detect decision and using SVM classifiers to classify the detects.


The improved discrimination techniques developed through this effort will significantly reduce the time, effort, and thus operational costs associated with typical underwater UXO remediation efforts. By more accurately identifying clutter, the false detection rate can be reduced allowing for more efficient recovery of munitions. New sub-bottom sensors are capable of improved detection of UXO; however, they also detect increased amounts of clutter, driving the need for improved clutter discrimination techniques.