Answers are grounded in this paper and cite their sources.
The rapid advancement of cage aquaculture had necessitated precise monitoring of fish quantity, distribution, and behavioral patterns, yet this remained a significant challenge within the industry. In this study, Trachinotus ovatus cultured in offshore cages were selected as the research object. To improve the accuracy of fish quantification, a method integrating YOLOv8 and fish density estimation was developed. Firstly, this method commenced with the acquisition of image data using omnidirectional sonar. Subsequently, the YOLOv8 model was employed for object recognition, enabling accurate quantification of fish across specific water layers. Finally, the total fish quantities in the cage were estimated using a density-based methodology. Continuous monitoring experiments were conducted to observe fish quantity in the offshore cage. Data regarding variations in cage net volume, as well as fish density, quantity, and spatial distribution, were collected. Estimation accuracy rates for fish quantity at 9:00, 13:00, and 17:00 were 95.1%, 93.9%, and 96.5%, respectively. The overall estimation accuracy reached 95.3%, with absolute error values remaining within ±10%. The vertical distribution characteristics of Trachinotus ovatus at different times were also illustrated. Specifically, fish predominantly lived deep-water layers (4-6 m) at 9:00, migrated towards mid-water layers (2-4 m) by 13:00, and exhibited a more uniform distribution throughout the cage by 17:00. Moreover, feeding behavior significantly influenced vertical distribution. During feeding, fish rapidly ascended to the water surface, demonstrating a tendency to move towards food sources. The proposed method facilitated precise monitoring of fish quantity in cage aquaculture and provided reliable data on fish density and spatial distribution at various time intervals. Thus, the limitations of traditional aquaculture monitoring methods were addressed through the innovative application of acoustic monitoring and assessment techniques. This research contributed robust technical and scientific data supporting the intelligent and information-driven evolution of modern marine ranching. • An improved YOLO model detects fish in images collected by omnidirectional sonar. • A fish density-based approach is employed to estimate fish population size. • Fish density distributions are analyzed across different time periods and behavioral states (e.g., feeding).
No takes yet. Share an insight, caveat, or question.
Sun et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: