Many active and former United States military installations are contaminated with munitions, including underwater sites. Of particular concern are shallow sites with exposed munitions. Optical imaging techniques offer promise for detecting these munitions: the imaging process is less affected by clutter; information such as optical contrast, color, and exact geometry is preserved, providing valuable inputs for automatic target recognition; imaging resolution is high enough to assess the integrity of the munition; and optical images are natural and intuitive for users to assess. Here the project team discusses the development of a proof-of-concept Optical Munition Detector (OMD). The OMD uses optical imaging techniques to survey the seafloor, extracting information useful for munition detection and classification.

The overall objective of this effort was to develop an OMD that uses optical imaging techniques to detect and classify munitions underwater. The OMD is designed to close operational gaps with existing remedial investigation technologies. The specific objective of this limited scope project was to design, fabricate, and demonstrate a proof-of-concept OMD.

Technical Approach

The OMD uses two mature and complementary optical metrology techniques: structured light and structure from motion. The first approach uses triangulation of a coherent light source and a camera to reconstruct the three-dimensional scene; the second extracts matched features in multiple views over a scene to determine relative camera positions so the three-dimensional scene can be reconstructed via stereo vision techniques.


The project team demonstrated the feasibility of the OMD by designing and fabricating a proof-of-concept system and collecting and analyzing data of inert munitions in a local lake. Image quality was sufficient to construct three-dimensional scenes from both optical modalities. The project team investigated and demonstrated a number of promising techniques for automatic munition detection from OMD data. In particular the team developed a deep learning approach that does not require the extensive and expensive training typically necessary. The team identified a number of technical areas with room for improvement in a next generation design. Finally, the team investigated how the OMD may integrate with and improve existing munition remediation activities.


The OMD is designed to close operational gaps in the remedial investigation of underwater sites suspected of munition contamination. Specifically, the OMD is designed to provide information on munition location, type, and integrity under conditions that are difficult for currently fielded technologies. This additional data will improve the performance of automatic target recognition and will improve the overall quality of data site managers have when making plans and decisions related to remedial action.