Improving Hokkai shrimp tracking accuracy using YOLOv8 with reflection and parallax correction.

Area of Science:

• Marine biology
• Ecology
• Aquaculture technology

Background:

• Eelgrass ecosystems are vital habitats supporting biodiversity and fisheries, including the Hokkai shrimp (Pandalus latirostris).
• Accurate tracking of shrimp behavior is crucial for conservation but is hindered by reflections and parallax in controlled tank environments.

Purpose of the Study:

• To develop and evaluate a You Only Look Once (YOLO) tracking system for monitoring Hokkai shrimp movement in aquaria.
• To address challenges posed by surface reflections and parallax distortions in underwater tracking.

Main Methods:

• Implementation of a YOLOv8-based tracking system for Hokkai shrimp.
• Controlled tank experiments to assess system performance.
• Development of preliminary measures to mitigate reflection and parallax artifacts.

Main Results:

• The YOLOv8 system demonstrated robust detection performance in aquaria.
• High detection accuracy was achieved: 95.65% from the front view and 100% from the side view.
• The system successfully captured key movement metrics like distance, velocity, and angle.

Conclusions:

• The developed YOLOv8 tracking system provides a reliable method for high-precision measurement of shrimp movement in aquaria.
• This technology offers a strong foundation for future ecological and aquaculture studies requiring accurate behavioral data.