Presentation Information
[A-14-22]General Behavior Analysis for Anomaly Detection Based on 3D Motion Measurement
– Behavioral Feature Analysis of Fish Swimming Motion –
◎Shoki Ishikawa1, Kazuhiro Tokunaga1 (1. National Fisheries University)
Keywords:
Behavior Analysis,Smart Sensor,YOLO,Anomaly detection,Aquaculture
Behavior analysis plays an important role in understanding the states of humans, animals, and biological systems, as well as in detecting abnormalities. It has been studied in fields such as livestock farming, biological experiments, aquaculture, and marketing. Recent advances in artificial intelligence and object recognition technologies enable quantitative capture of motion and behavior using cameras, allowing objective and automatic evaluation instead of human observation. This study aims to develop a smart sensor system for detecting abnormal behavior based on camera-based measurement and analysis. The proposed method is designed as a general-purpose behavior analysis framework independent of specific targets. As an application, we focus on fish behavior in aquaculture, where early detection of disease is important but currently relies on human observation. We propose a method to quantify swimming behavior and detect abnormalities. The process consists of: (1) constructing an imaging environment, (2) tracking fish positions using YOLO, (3) reconstructing 3D coordinates from multi-view images, and (4) calculating velocity and acceleration to quantify behavioral features and compare healthy and diseased fish.