An unfortunate legacy of former military activities at sites designated for base realignment and closure and at Formerly Used Defense Sites is the contamination of aquatic environments with military munitions. In the United States, more than 400 underwater sites, spanning an area in excess of 10 million acres, potentially contain such munitions. The presence of these munitions is a serious threat to both humans and the environment, so remediation is necessary. But the return of these contaminated waters to public use is contingent upon the analysis and assessment of wide-area and detailed underwater surveys. Therefore, the Department of Defense has an express need for the development of technologies that will enable the detection and classification, at high probability, of military munitions found at underwater sites.

The primary objective of this project was to develop novel unexploded ordnance (UXO) detection and classification algorithms specifically for volumetric sonar data from two experimental systems, the Sediment Volume Search Sonar (SVSS) and the Multi-Sensor Towbody (MuST). Because no automatic target recognition algorithms previously existed for these two new systems, the methods developed here addressed a capability gap. The general-purpose detection algorithm that was created exploited the concept of integral images to flag suspicious regions in a given data volume in a fast, computationally efficient manner. The follow-on classification algorithm leveraged a flexible deep-learning framework developed during a previous Phase I SEED project. That framework, based on deep convolutional neural networks, was then extended in the present effort to function with three-dimensional (i.e., volumetric) input data cubes. The developed algorithms were assessed using large sets of SVSS data, and they were also applied to modest amounts of data from the MuST system.

Preliminary results showed the promise of the approaches for detecting and classifying both proud and buried targets in measured volumetric sonar data. This project covered one-year but had been envisioned as a four-year project until it ended prematurely (due to an organization change of the Principal Investigator). Nevertheless, the progress made during this abbreviated period provides a solid foundation from which to further this line of research. Because the algorithms were purposely developed to be functional with measured data from existing systems, they should be readily deployable in a short time frame for use in actual remediation efforts. This result can be achieved by executing the remainder of the original project plan, which includes rigorous testing at new SERDP UXO test-bed sites.

The successful culmination of the project’s approach should enable the attainment of higher probabilities of detection and classification, at much lower false alarm rates, than is possible with existing approaches. As a result, the application of these machine-learning algorithms to sonar data collected at potentially contaminated underwater sites can guide remediation efforts to effect savings.