Revolutionizing Aquatic Ecosystem Monitoring: Autonomous Real-Time Fish Detection and Multi-Object Tracking Using YOLOv11, Norfair, and IoT-Enabled Edge Computing

Accurate identification and tracking of aquatic animals are crucial for aquaculture, marine biology, and environmental research. This study presents a real-time underwater monitoring system that combines deep learning and multi-object tracking within a compact mobile device. The system utilizes You Only Look Once version 11 (YOLOv11) for swift and accurate fish identification, while the Norfair library facilitates reliable tracking of multiple objects across continuous video streams. The hardware platform operates on a Raspberry Pi enclosed in a watertight casing and equipped with autonomous navigation capabilities, allowing for continuous visual documentation of target species and representing a significant advancement over traditional static monitoring methods. Although the primary goal of the current prototype focuses on real-time detection and monitoring of fish, the design anticipates the integration of Internet of Things (IoT)-enabled environmental sensors that can measure temperature, turbidity, pH, and distance. Additionally, the system architecture includes considerations for potential cloud-based services, such as Message Queuing Telemetry Transport (MQTT) and Firebase, to improve remote accessibility and facilitate large-scale deployment, although these components are currently conceptual. A Streamlit dashboard offers real-time interactive visualization of processed outputs, and experimental results confirm the system’s robustness even in challenging underwater conditions. By merging computer vision with an extensible IoT framework and a scalable embedded design, the proposed technology provides a versatile and efficient solution for ecological analysis and sustainable management of aquatic resources.