Monitor fish feeding activity through computer vision

Concept

Fish feeding activity has a direct link to fish health, but it can be difficult to measure in large rearing systems, particularly in absence of fish keepers. Toward this end, I have developed a computerized device that can automatically feed fish and measure their feeding activity by counting the number of fish appearing in a feeding area. Using time-series analyses and machine-learning algorithms, the machine can further determine if fish feeding activities have deviated from normal patterns, and subsequently send alerts to fish keepers when necessary.

---

Fish feeding model: In a feeding event, the number of fish under the camera view (i.e., feeding zone) drastically increases right after fish food is released, and reaches to its peak, and then decreases after they consumed the fish food.

Hypothetical normal (blue) and abnormal (red and orange) feeding patterns. The change from blue to red indicates that fish feeding activity has become slow - they spend more time to reach the peak; from blue to orange, the number of feeding fish decreases - some fish did not enter the feeding zone to eat.

Learn more

Results

Applications

• Laboratory Research - Fish FeederBot can help understand how fish feeding activities change under various conditions. Feeding parameters (e.g., feed quantity and timing) are programmable, and data stored in databases can be used for further statistical analyses or visualization. Other electronic components, such as temperature and pH sensors, can be integrated into the system; therefore, it can fit a variety of research needs.

• Aquaculture & Aquarium - Through computationally monitoring fish feeding activity, aquaculture farmers or fish keepers can be notified if the fish have decreasing appetite and possibly suffer from environmental changes. As such, it helps farmers take necessary actions in advance of worse conditions. For valuable and unique aquarium fish, modified computer vision programs allow it to track individual fish and monitor its feeding activity to reflect the health status.

• Internet of Things (IoT) - Raspberry Pi has a built-in Ethernet port and also can connect to the Internet wirelessly through USB WiFi dongle (or built-in WiFi in model 3). This makes the FeederBot an IoT device and enable other functionalities, such as wirelessly real-time control and monitor.

About