Detection of Transient Burst Swimming Scenes Focused on Displacement of Farmed Bluefin Tuna Juveniles

This paper presents a method for detecting transient burst swimming behavior in farmed bluefin tuna juveniles using video analysis. Although the juveniles are raised in indoor aquaculture tanks, they often succumb to injuries caused by collisions with the tank wall or with each other because they exhibit the burst swimming of suddenly increasing their swimming speed together due to external stimuli. The proposed method aims to analyze the environmental factors around the tanks that provoke the burst swimming, ultimately contributing to a reduction in juvenile mortality. It is suggested that this data‐driven approach could contribute to improving the quality of production management in juvenile aquaculture. In the proposed method, the juvenile swimming in a land‐based tank are monitored using a video camera, and the video scenes of the burst swimming are detected by a discriminant machine with relevant features extracted from the video footage. The features represent displacement of the juvenile shoal and are extracted by identifying the juvenile regions in every frame and tracking their movement across sequential frames. Preparing videos which include the burst swimming scenes by the juveniles, the performance of the proposed method was examined by trials for detecting the scenes using multiple discriminant machines. The experimental results confirmed that the F‐score for the discrimination of the burst swimming and regular scenes exceeded 80% in every of the discriminant machines. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.