Development of an artificial intelligence-based monitoring system for livestock animals

The project titled Blue Proton Initiative, launched by four 17-year-old students from Forlì, is aimed at automating the monitoring of livestock animals. The developers seek to solve the problem of inefficient manual recording, which often leads to the late detection of diseases and unjustified economic losses for farms.

The technical architecture of the system is built on the use of neural networks and computer vision technologies, which allow for real-time identification of individuals and tracking of changes in their behavior. To implement edge AI computing, the team selected the Arduino® UNO™ Q board. This choice is driven by the need to create a compact, energy-efficient, and affordable device capable of functioning in remote field conditions without a constant connection to a server.

The workflow is distributed among team members: Pietro Maria Piazza is responsible for the software, Alessandro Nesci handles the training of YOLO models, Matteo Angiolillo develops the backend in C, and Davide Santucci performs chassis design using CAD systems and 3D printing. Special attention is paid to the quality of training data: the developers use diverse scenarios, including differences in lighting and weather conditions, to ensure the reliability of the algorithms in real-world operation.

At the current stage, four months after the start of the work, the team has successfully implemented a system for the automatic counting of livestock. The developers' plans include expanding the functionality to monitor the health status of animals. The project authors note that using hardware with limited resources requires constant optimization of the program code and finding a balance between processor power and RAM consumption.

At the moment, the project is in the prototyping stage and is not a finished commercial product. Nevertheless, the work done demonstrates the feasibility of using accessible microcontrollers to solve applied agronomic tasks that require automatic data collection directly at livestock housing locations.