In recent years, fish welfare is emphasized in aquaculture. Swimming speed of farmed fish is one of the important indicators of fish behavior. A powerful tool to measure the fish swimming speed is Biologging, which requires attaching data-logger to fish. Therefore, there are two concerns- the fish swimming speed may change due to drag force, as well as the fish is injured. On the other hand, Stereo Camera is a tool to measure the fish swimming speed, without surgical procedures. However, it has not been widely employed yet as a huge time and effort for detecting and measuring fish position from video analysis by manual operation is required. Furthermore, comparative evaluation study of stereo camera with existing methods is necessary. The purpose of this study is to evaluate the precision of measuring method of fish swimming speed using a stereo camera, where the video analysis is processed under automatic operation. A propeller-logger was attached to a single yellowtail for the experiment, and then released into an experimental fish tank. The experimental measurement duration was 2 hours. The fish swimming speed was measured by the logger every second. We also employed a stereo camera equipped with AI, which consists of image recognition, 3D measurement, and state-space model tracking, to obtain the fish speed. Result of fish swimming speed measured by stereo camera was 0.52±0.14 ms−1 (Mean±SD). Comparison of the fish swimming speed of these two methods shows that the difference is within 0.1 ms−1. Although this study was carried out in a fish tank, the future study is expected to be carried out even under wave condition, by compensating for fluctuations with the inertial measurement unit inside the stereo camera. This study indicates that the stereo camera is an efficient tool for measuring the fish swimming speed.
View full abstract