Deep learning methods applied to electronic monitoring data: automated catch event detection for longline fishing. (27th December 2020)

• Record Type: Journal Article
• Authors: Qiao, Maoying; Wang, Dadong; Tuck, Geoffrey N; Little, L Richard; Punt, Andre E; Gerner, Mike
• Editors: Beyan, Cigdem
• Is Part Of: ICES journal of marine science. Volume 78:Number 1(2021)
• Page Start: 25
• Page End: 35
• Publication Date: 2020-12-27
• Subjects: artificial intelligence -- artificial neural networks -- deep learning -- fisheries management -- machine learning; Ocean -- Periodicals; Fisheries -- Periodicals; Fishes -- Periodicals; Marine biology -- Bibliography -- Periodicals; 551.4605
• DOI: 10.1093/icesjms/fsaa158
• Languages: English
• ISSNs: 1054-3139

Abstract

Abstract: Electronic monitoring (EM) systems have become functional and cost-effective tools for the conservation and sustainable harvesting of marine resources. EM is an alternative to on-board observers, which produces video segments that can subsequently be reviewed by analysts. It is currently used in a range of fisheries. There are two major challenges to the widespread adoption of EM. One is the large storage requirement for the video footage recorded and the other is the long time required by analysts to review the video footage. We propose an automated catch event detection framework to address these challenges. Our solution, based on deep learning techniques, automatically extracts video segments of catch events, which substantially reduces storage space and review time by analysts. Here, we demonstrate the framework using video footage from three longline fishing trips. The system recalled nearly 100% of the catch events across all trips.