Many active and former United States military installations are impacted with munitions, particularly unexploded ordnance (UXO) and discarded military munitions (DMM). The impacted locations often include underwater sites such as ponds, lakes, and coastal ocean areas. Of particular concern are munitions in shallow waters that are proud of the seafloor because the public is most likely to use these waters for recreation. Site managers need tools to assess the presence, location, type, and condition of munitions. The objective of this effort is to develop an Optical Munition Detector (OMD) that uses optical imaging techniques to detect and classify munitions underwater. The OMD will close operational gaps with existing remedial investigation technologies. The specific objectives of this project are to (1) reconfigure the system components used in the proof-of-concept OMD to improve optical imaging quality and to support deployment from an unmanned surface vessel; (2) integrate the system with a surface vessel; (3) perform a “tabletop demonstration” with the prototype OMD in a staged environment in the presence of clutter and biofouling; and (4) continue to investigate automatic target recognition algorithms best suited to exploit OMD data.

Technical Approach

The OMD uses two mature and complementary optical metrology techniques—structured light and structure from motion—to survey the bottom of the water column. It generates high-fidelity, high-resolution, three-dimensional (3D) topology (point clouds) that preserves optical contrast and color. Automatic target recognition algorithms use these features to detect and classify UXO and DMM. The OMD has several advantages over traditional underwater acoustic imaging: it is lower power and higher resolution, and its data products (colored 3D scenes) are naturally and intuitively human-readable.

In the SERDP Limited Scope project, MR19-1423, the project team demonstrated the ability to generate 3D representations of replica munitions on a lakebed. The project team also demonstrated the potential of several automatic target detection algorithms. Most promising was a curvature-based approach well suited for most unexploded ordnance (as they often have known and consistent calibers) and the transfer-of-learning of a preexisting deep convolution neural network. This project will execute the recommendations made at the conclusion of the previous limited scope effort: the project team will upgrade system components to improve the clarity of the optical subsystems, demonstrate the technology at a SERDP test bed, and use the resulting data to improve the automatic target recognition algorithms.


The OMD will close operational gaps in the remedial investigation of underwater sites with known or suspected munition contamination. Specifically, the OMD will provide information on location, type, and condition of proud munitions in shallow water environments (the area of most concern to the public). The OMD will reduce the schedule and costs of remedial investigation while OMD data products will improve the overall quality of data that site managers have when making plans and decisions related to remedial action.

  • Optical Sensors,

  • Platforms and sensors to detect and classify munitions,