Autonomous Detection of Near-Surface Marine Mammals

We propose to develop a system that automatically detects and locates marine mammals using sensors onboard a maritime vessel. We will begin by surveying candidate sensors and sensor combinations, but we expect that infrared cameras will be a key component of any system. Our approach for processing infrared imagery is to compute spatial as well as temporal signatures of whale blows and other exposures of marine mammals near the vessel. Our algorithms will benefit from years of prior work that we have performed in this area using months of data collected in California and Canada. Animals recorded in our test data have included Gray, Fin, Humpback, Beluga, and Minke whales, as well as Dalls and harbor porpoises, sea lions and sea otters. On this contract we propose to collect data using a camera that is mounted on a vessel-based mechanically-stabilized gimbal and process this data to automatically detect marine mammals using recorded video. We will simultaneously record observations of Marine Mammal Observers (MMOs) and subsequently compare the performance of the automated system to the performance of the human observers.