WHAT: Adaptive Underwater Object Recognition System



WHY: Hostile forces use low-cost sea mines to delay, damage, or destroy ocean-going vessels and equipment. Detecting, classifying, and locating such mines in dark, densely packed ocean environments require trained dolphins with human handlers. In 2007, tests of a prototype remote robotic mine-hunting system aboard a U.S. Navy destroyer failed. Now, new adaptive robotic systems promise the ability to learn from their mistakes.



HOW: Second-generation computer algorithms automatically detect, classify, and target sea mines in fast-changing underwater environments, dramatically reducing false positives with a variety of sensor types and modalities. Technical innovations include new processes to extract and process images, identify different types of marine terrain, distinguish dangerous objects from mere debris, and improve real-time recognition of any “targets of interest.” The United States Department of Defense will supply initial datasets and funding for five projects, including research collaboration between Rochester, NY-based Impact Technologies and The Georgia Institute of Technology.



CAVEATS: In 2003, a Navy spokesperson told Smithsonian Magazine that no machine is likely to match all the capabilities of trained dolphins.



QUOTE: Dr. M.R. Azimi-Sadjadi, Colorado State University: “The unique advantage of our proposed solutions is the ability to offer system flexibility while preserving the stability of the previously-learnt information.”



MORE INFORMATION: Department of Defense solicitation selection abstracts

http://www.dodsbir.net/selections/abs091/dodabs091.htm