Related Experiment Video
Updated: Oct 4, 2025
05:57
Long-term Video Tracking of Cohoused Aquatic Animals: A Case Study of the Daily Locomotor Activity of the Norway Lobster Nephrops norvegicus
Published on: April 8, 2019
7.0K
A low-cost, long-term underwater camera trap network coupled with deep residual learning image analysis
Stephanie M Bilodeau1,2, Austin W H Schwartz1,2, Binfeng Xu3
1Department of Biology, Wake Forest University, Winston-Salem, NC, United States of America.
Plos One
|February 2, 2022
Summary
New underwater camera traps and machine learning accurately monitor marine ecosystems. This technology enables broad-scale, long-term ecological studies of fish behavior and distributions.
Area of Science:
- Marine ecology
- Ecological monitoring
- Behavioral ecology
Background:
- Traditional diver-based surveys have limitations for long-term, large-scale marine ecosystem monitoring.
- Previous attempts at long-term underwater camera systems faced environmental challenges.
Purpose of the Study:
- To develop a cost-effective, scalable system for long-term underwater camera trapping.
- To apply machine learning for automated image classification in marine environments.
- To assess fish behavior and distribution patterns in relation to benthic features.
Main Methods:
- Developed Dispersed Environment Aquatic Cameras (DEACs) for remote, long-term deployment.
- Utilized a ResNet-50 deep learning model for image classification.
- Deployed cameras for over five months, collecting over 100,000 images.
- Validated camera data against diver surveys.
Main Results:
- Cameras operated continuously underwater for extended periods (up to 5 months) without maintenance.
- The deep learning model achieved 92.5% accuracy in identifying images with fish.
- Camera data accurately reflected local fish communities compared to diver surveys.
- Successfully documented fish movement and feeding behavior related to benthic halos.
Conclusions:
- The DEAC system represents the first successful broad-scale underwater camera trapping method.
- This technology offers a viable solution for long-term, large-scale marine ecological research.
- The system has significant potential for studying marine animal behavior, distributions, and spatial patterns.
