Monitoring of Animal Movement using Computer Vision

This paper presents a novel approach for identifying individual pigs and monitoring the movement of individual pigs using computer vision, designed explicitly for non-intrusive and real-time tracking in farm environments. Identifying individual pigs is the first step taken to track their movement. Accurately identifying individual pigs enables real-time monitoring of the pigs in order to get essential parameters such as their movement. Monitoring the movement of individual pigs is very important as it allows farmers to track their health. Pig movements such as walking, rolling, and laying down give the farmer a clear indication of how healthy each pig is. In this study, the focus is on tracking the movement of individual pigs when they move in the pen. Leveraging a Raspberry Pi 4 equipped with a Raspberry Pi Camera Module 3, the system integrates multiple deep learning models, including YOLOv8n, MobileNetSSD, and EfficientNet-B0, to achieve high-accuracy breed identification and motion detection. Among these, YOLOv8n emerged as the most effective, reaching 97-98% accuracy. The project involves a centroid-based tracking algorithm that measures movement by calculating Euclidean distances between centroids across frames. The data gathered on the individual pig movements in meters is analysed and visualised through time series graphs, providing valuable insights into the motion of individual pigs. This paper discusses the system’s design, implementation, and performance, highlighting its potential applications in precision livestock farming.