Akademik Veri Yönetim Sistemi
31st International Conference on Artificial Life and Robotics, ICAROB 2026, Oita, Japonya, 29 Ocak - 01 Şubat 2026, ss.184-189, (Tam Metin Bildiri)
Automatic feeding systems in aquaculture are vital for feed optimization and reducing labor. However, operational reliability is critical, as malfunctions cause significant economic loss and compromise stock health. This study proposed a computer vision system to continuously monitor feeder performance by analyzing real-time fish movement patterns. Three behavioral metrics were analyzed: fish recognition to detect and localize fish, fish density to describe spatial aggregation, and group disorder to represent irregular collective movement. The findings reveal that each successful feeding produces a clear, repeatable behavioral signature, such as a simultaneous spike in fish density and group disorder as fish actively respond to feed. This signature serves as an early warning system for confirming effective feeding. The absence of this signature indicates abnormal feeding conditions, including feeder malfunction, insufficient feed release, or reduced fish responsiveness. Collectively, these capabilities enable a robust monitoring strategy. Overall, this monitoring approach ensures reliable automation, enhances management efficiency, and protects both profitability and stock health.