Optimized Dairy Cow Identification and Tracking with PTZ Camera Technology

The introduction of smart technologies for managing large herds with a small workforce is gaining attention. For example, methods that involve attaching activity monitors or IC tags to livestock and using the information obtained from these devices for livestock management are already in use. Some of these methods have issues such as high costs, unnecessary stress on the cow, and an increase in workload due to the introduction and maintenance of IC tags. In this study, a model has been proposed for identifying a specific individual cow and estimating their position in the barn, using only images obtained from cameras installed in the barn. The proposed methodology does not require equipment other than cameras to obtain information about the cows and captures images over a wide area of the barn with a single rotating PTZ camera, allowing for a low-cost setup, and the individual identification model created using YOLOv8 showed an overall accuracy of $\mathbf{9 6. 6 \%}$ and a recall rate of $\mathbf{9 5 \%}$, demonstrating that it is possible to practically identify and estimate the position of a specific individual.