Polar bears and Ice-associated seals (bearded, ringed, spotted, and ribbon seals) are key components of Arctic marine ecosystems and are important resources for coastal Alaska Native communities. Reliable abundance estimates for ice seals are needed for developing sound management decisions under the Marine Mammal Protection Act and extinction risk assessments under the Endangered Species Act. The animals’ broad and patchy geographic distributions and rapidly changing sea ice habitat make these species particularly challenging to study.
An autonomous payload is required to integrate UAS into surveys of ice-associated mammals, in order to improve the efficiency and human safety in gathering essential data for NOAA stewardship. Moving from occupied aircraft to long-range UAS operations will require an efficient and “smart” payload to collect images needed for abundance estimation and habitat analysis while providing situational awareness to the pilot in command.
The Alaska Fisheries Science Center’s Marine Mammal Lab is developing a system that can run advanced machine learning algorithms on-board the aircraft to process multispectral imagery in real time, minimizing the collection of extraneous imagery that requires burdensome data storage, management, and processing. Algorithm development is utilizing a neural network approach known as YOLO, which processes imagery at a rate of 60-100 frames per second. Over 1.8 million color and thermal images are being used to train YOLO to detect animals on the sea ice and classify detections to species. This algorithm will be tested in-flight during April 2019.