Publication Date:
2019-07-17
Description:
Detecting whales at sea by visual observation for mitigation purposes is inherently difficult
and personal intensive while restricted to daylight hours. These caveats are overcome by the
system described herein, which employs a state-of-the-art thermographic infrared scanner in
conjunction with a learning computer algorithm to automatically and reliably detect whale
blows. The stand-alone system provides detection, verification and documentation of each
ship-whale encounter, allowing a retrospective review of every mitigation decision taken
aboard. The system has been developed over the course of 5 years and was thoroughly tested
in polar waters during 7 expeditions to the Arctic and Antarctic, accumulating 5871 hours of
operation. Of these, 3472 hours, were analyzed with various learning automatic detection
algorithms, which discovered about 4500 whale blows in this data. Direct comparisons of
auto-detections with concurrent whale sightings by visual observers (double blind setup)
confirm a very high degree of detection reliability within 2-3 nautical miles from the ship in
subpolar and polar environments. The system, when used as a an “assistant”, allows a single
marine mammal observer to efficiently monitor the ships entire surroundings and to take
instantly and retrospectively verifiable decisions regarding the use of airguns, as all relevant
data is automatically stored.
Repository Name:
EPIC Alfred Wegener Institut
Type:
Article
,
notRev
Format:
application/pdf
Permalink