2024A-GS11-3 Fish Fin Kinematics: Deep Learning Approach

－Tracking fin motion using the YOLOv8 Pose Detection Model－

Abstract

The locomotion of aquatic organisms has long fascinated biologists and engineers alike, with fish fins serving as a prime example of nature's remarkable adaptations for efficient underwater propulsion. This paper presents a comprehensive method focused on the hydrodynamic analysis of fish fin kinematics, employing an innovative approach that combines machine learning and image processing techniques. Through high-speed videography and advanced computational tools, we gain new insights into the complex and dynamic motion of the fins of a Tilapia (Oreochromis Niloticus) fish. This study was initially done by experimentally capturing videos of the various motions of a Tilapia in a custom-made setup. Using deep learning and image processing on the videos, the motion of the Caudal and Pectoral fin was extracted. This motion included the fin configuration (i.e. the angle of deviation from the mean position) with respect to time. The key objectives include mathematical modelling of motion of a flapping fin at different naturally occurring frequencies and amplitudes. This work aims to cut down on time required and improve on research that has been done in the past on similar topics. Also, the method can help in the better and more efficient study of the propulsion systems for biomimetic underwater vehicles that are used to study aquatic ecosystems, explore uncharted or challenging underwater regions, do ocean bed modelling, etc.