[1Q4-GS-10] AI application:

[1Q4-GS-10-02] Sequential Estimation of Fish Weight Distribution in Aquaculture Using Particle Filters

〇HIDEKI ASOH¹, MASASHI TSUBAKI¹, JUNYA KOBAYASHI², YUI MINESHITA², ICHIRO NAGANO² (1. National Institute of Advanced Industrial Science and Technology, 2. Nissui Corporation)

Keywords:Aquaculture, Fish Weight Distribution Estimation, Sequential Monte Carlo Method, Particle Filter

In aquaculture, understanding the condition of fish within net pens is essential for monitoring growth, detecting abnormalities, and planning shipments. Specifically, it is important to track not only the average fish weight but also changes in the weight distribution. Recent methods estimate weight distributions by measuring fish body shapes using stereo cameras submerged in the net pens and deriving weight estimates from these shapes. However, conducting daily measurements is often impractical. Furthermore, fish weight distributions are not always unimodal and may exhibit multimodal characteristics. In this study, we propose a sequential Monte Carlo method combining particle filter and smoother to interpolate fish weight distributions on unmeasured days and extrapolate distributions beyond the final measurement day. We report the results of experiments using data from real net pens.