Related Experiment Video
Updated: Aug 7, 2025
14:58
Behavioral Tracking and Neuromast Imaging of Mexican Cavefish
Published on: April 6, 2019
7.8K
Abnormal Behavior Monitoring Method of Larimichthys crocea in Recirculating Aquaculture System Based on Computer
Zhongchao Wang1, Xia Zhang2, Yuxiang Su1
1School of Marine Engineering Equipment, Zhejiang Ocean University, Zhoushan 316022, China.
Sensors (Basel, Switzerland)
|March 11, 2023
Summary
This study introduces an improved YOLOX-S model for real-time detection and tracking of abnormal behavior in high-density fish populations within recirculating aquaculture systems (RASs). The enhanced system achieves over 95% accuracy, improving fish welfare and production efficiency.
Area of Science:
- Aquaculture technology
- Computer vision
- Animal behavior monitoring
Background:
- Recirculating aquaculture systems (RASs) require continuous monitoring due to high fish densities and intensification.
- Existing object detection algorithms struggle with complex environments and dense fish populations, leading to potential losses.
- Monitoring abnormal fish behavior is crucial for preventing disease outbreaks and optimizing production.
Purpose of the Study:
- To develop an efficient method for detecting and tracking abnormal behavior in Larimichthys crocea within RASs.
- To improve the accuracy and robustness of object detection algorithms in challenging aquaculture conditions.
- To provide a foundation for automated treatment strategies by enabling reliable fish behavior monitoring.
Main Methods:
- An improved YOLOX-S object detection algorithm was developed by modifying the CSP module and incorporating coordinate attention.
- The algorithm was enhanced to address issues like object occlusion, deformation, and small object detection in dense fish environments.
- Bytetrack was employed for object tracking to maintain stable identification and prevent ID switching, crucial for visually similar fish.
Main Results:
- The improved YOLOX-S achieved a 98.4% AP50 and a 16.2% increase in AP50:95 compared to the original algorithm.
- Tracking performance metrics, MOTA and IDF1, exceeded 95% in real-world RAS environments, ensuring stable ID maintenance.
- The system demonstrated efficient real-time identification and tracking of abnormal fish behavior.
Conclusions:
- The proposed method significantly enhances the detection and tracking of abnormal behavior in Larimichthys crocea in RASs.
- This technology offers valuable data support for subsequent automated treatments, mitigating losses and boosting production efficiency.
- The advancements in computer vision are pivotal for the future of intelligent and sustainable aquaculture management.
