Introduction
Efficiency and cost in aquaculture depend largely on feeding, making it crucial to optimize feeding strategies to reduce feed loss and improve fish health and welfare. One way to achieve this is by monitoring individual and group swimming patterns, especially in response to external factors like feeding. Intelligent feeding control that utilizes behavioral changes and growth status is gaining attention as a useful tool for improving husbandry practices. Efforts are underway to identify behavioral indicators that can detect satiation levels and regulate feeding for different species. In this study, we investigated two behavioral-based indicators, speed and the feeding behavioral index (a newly defined metric), across different feeding scenarios for E. seabass. Our findings suggest that fish exhibit distinct behavior patterns in response to various feeding situations, and both the speed and feeding behavior index can identify threshold values that correspond to satiation levels, facilitating controlled feeding.
Materials and methods
A group of E. seabass fish of 300 g body weight was reared in a circular polyester cage (40 m diameter, 9 m depth) located at the pilot scale netpen cage farm of HCMR at Souda bay, Crete (certified as an aquaculture facility from the national veterinary authority; code GR94FISH0001). A submerged network camera (Fyssalis V3.1) capturing at 10 fps was used for monitoring and video recording during daylight hours . The camera was positioned at 4 m depth using a gyroscopic gimbal stabilizer to ensure it pointed upwards. We trained YOLOv5 (a machine learning model for object detection) to detect fish and adapted Deepsort (a model for tracking people) to track fish individually (using OPENCV/Python) and extract their speed and direction. In addition, we used computer vision techniques that detect changes in the crowding behavior of the fish, and we called the parameter feeding behavior index. To detect behavioral changes as response to satiation levels we modified the feeding schedule of the fish by varying three crucial feeding parameters: the feeding frequency (once, twice and three times a day), the feeding time (morning, afternoon and evening) and the feeding quantity (normal, reduced, overfeeding and no feeding). The experimental trial lasted from June 2022 until April 2023.
Results
Our preliminary results show that the f eeding behavioral index shows different qualitative and quantitative evolution across the three different feeding quantities. The e xcitation s tep at the start of the feeding is significantly larger when fish are underfed than when overfed or fed normally (Figure 1). In addition, the duration of the excitation is longer. The speed on the other hand shows a significantly different qualitative behavior (Figure 2) . There is a change in activity relative to the start of the feeding. When feeding is normal, fish show symmetry in the activity pattern relative to start of the feeding . They gradually increase their activity before feeding and gradually decrease it after feeding. In contrast, during reduced feeding fish appear to show increased activity before feeding, while during overfeeding periods, fish show more increased activity as response to the start of the feeding.
Conclusion
Our results indicate that fish show distinct patterns of behavior under different feeding situations, and both the speed and feeding behavior index can be used to detect threshold values that correspond to satiation levels and facilitate controlled feeding in marine cages. Under reduced feeding, fish show an early and prolonged excitation before feeding and a slow relaxation to the baseline activity after feeding times indicating strong anticipation for feeding and increased appetite. The sudden increase in the feeding behavior index also contains important information about the fish appetite, as larger activation steps in the feeding index suggest lower satiation levels. Last, the duration of excitation is also an important parameter that can reveal the satiation levels of the fish, suggesting that fish with higher satiation levels show longer excitation times in their feeding behavior index. Further studies are required to help us understand the contribution of other factors such as the human presence, or the internal circadian rhythm to the variation of the activity.
Acknowledgments
This work was partly supported by the European Union Horizon 2020 Project iFishIENCi (818036).
References
Antonucci, Francesca, and Corrado Costa. "Precision aquaculture: a short review on engineering innovations." Aquaculture International 28.1 (2020): 41-57.
Wojke, Nicolai, Alex Bewley, and Dietrich Paulus. "Simple online and realtime tracking with a deep association metric." 2017 IEEE international conference on image processing (ICIP). IEEE, 2017.