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

Eelgrass ecosystems provide essential habitats for species, such as the Hokkai shrimp (Pandalus latirostris), supporting biodiversity and fisheries. Understanding shrimp behavior in these environments is vital for conservation efforts, yet accurately tracking shrimp movement in controlled tank settings is challenging because of surface reflections and parallax distortions. This study primarily focuses on developing and evaluating a You Only Look Once (YOLO) tracking system for Hokkai shrimp in aquaria. We implemented preliminary measures to address reflection artifacts and parallax distortions; however, our core contribution is the robust detection performance of YOLOv8. Through controlled tank experiments, the system demonstrated high detection accuracy and captured metrics such as distance, velocity, and angle. The results showed a high detection accuracy, with success rates of 95.65% and 100% from the front and side views, respectively, ensuring reliable data on shrimp movement in aquaria. These outcomes provide a robust foundation for high-precision measurements in future ecological or aquaculture studies.