15 / 2025-03-28 10:52:00

Implementation of a cattle management system based on facial identification using an edge computing device

Cattle Management,Individual Identification,Precision Livestick Farming,Animal Welfare

3. 智慧牧场与动物福利 > （2）智慧养殖的大数据、机器学习和物联网应用

The agricultural sector faces substantial challenges due to labour shortages and escalating costs. Existing management strategies are inadequate in meeting precision livestock farming (PLF) demands. Contactless recognition, the most prevalent method in the application of machine vision within livestock farming, has seldom been applied to Jersey cattle. Consequently, this study introduces an innovative solution employing the Jetson Orin Nano platform and PyQt to create a cost-effective and user-friendly graphical interface for the facial identification of Jersey cattle, dynamic herd management, and behaviour monitoring. The proposed system utilises a two-stage network, integrating an enhanced YOLO11 and MobileNetV4, to facilitate the facial identification of multiple Jersey cows within the feeding area via cameras, achieving an identification accuracy of 97.9%. This system enables users to capture images of newly introduced cattle or remove those no longer present. The identification model is retrained on a local edge computing device to ensure seamless integration or removal of the cow. Furthermore, users can set update intervals, allowing the system to automatically refresh the identification model to accommodate changes in cattle facial features over time. For behaviour recognition, this study employs the K-Means algorithm to achieve long-term eating detection at the individual level as a case study. This comprehensive approach reduces operational costs and provides an integrated perspective on animal health, facilitating timely interventions. By addressing the critical needs of contemporary farm management, this research offers a sustainable and efficient solution to the challenges confronting the agricultural industry.