To main content

Quantifying fish behavior in aquaculture using visual open-world tracking models

Abstract

Monitoring and analyzing fish behaviors is critical for advancing aquaculture management and understanding fish welfare. Traditional methods, such as manual observation and sonar tracking, face significant limitations including inefficiency, low accuracy, and high dependence on experimental conditions. In this paper, we explore the use of the open-world tracking model for fish tracking and behavior analysis based on camera-based observation. We propose a unified framework that integrates fish position extraction and statistical behavior analysis to enable quantitative assessment of fish activities. The framework is validated with real-world fish recordings, where it successfully detects sudden behavioral changes caused by external disturbances. Experimental results demonstrate the potential of our approach to provide robust, scalable, and efficient fish behavior analysis. This work highlights the feasibility of leveraging an open-world tracking model for practical aquaculture monitoring applications.
Read the publication

Category

Academic article

Language

English

Author(s)

  • Feng Xiao
  • Peihua Han
  • Houxiang Zhang
  • Jingwen Ding
  • Lars Christian Gansel
  • Yanran Cao

Affiliation

  • SINTEF Ocean / Aquaculture
  • Norwegian University of Science and Technology

Date

01.06.2026

Year

2026

Published in

Applied Ocean Research

ISSN

0141-1187

Volume

171

Page(s)

105076 - 105076

View this publication at Norwegian Research Information Repository