AUTOMATED FISH DETECTION FOR PRECISE AQUACULTURE MONITORING

Khalfan Al-Rashdi, Ali Al-Mabsali, Sanaz Keshvari, Said Al-Abri, Rami Al-Hmouz, Hadj Bourdoucen

Automated fish detection represents a pivotal advancement in aquaculture management, enabling real-time monitoring that addresses key challenges in the industry. This work focuses on developing a robust detection system to accurately identify and track fish populations, which serves as the foundation for critical applications such as biomass estimation, behavioral analysis, optimized feeding strategies, and health monitoring. By providing continuous, non-invasive insights, real-time detection can significantly reduce operational costs by minimizing the need for manual labor and overcoming human limitations like fatigue and subjectivity. Furthermore, it facilitates precise control over individual parameters, including water purification levels, feeding amounts, disease detection, and other environmental factors, thereby optimizing overall fish farming efficiency and sustainability. Focusing on Tilapia nilotica (Nile tilapia), we implemented a machine vision-based approach to detect and quantify fish in real-time within commercial tanks.

The methodology utilized image processing algorithms to identify fish based on shape and motion characteristics under varying environmental conditions. Our approach leverages deep learning models to enhance detection reliability across diverse settings, such as fluctuating light and water turbidity. As illustrated in Figure 1, the system marks detected fish with red bounding boxes, demonstrating effective identification in a sample image.

Preliminary results show a fish detection accuracy exceeding 94% across diverse real-world conditions. We evaluated YOLOv5, YOLOv8, and YOLOv11 models on our dataset, with precision values depicted in Figure 2, demonstrating YOLOv5’s superior performance. These findings highlight the effectiveness of our automated detection system for aquaculture monitoring, offering a scalable solution for real-time fish tracking.