Scholarlink

<p>In precision livestock farming, traditional sheep management relies primarily on ear tagging. However, this conventional approach has several limitations, including high implementation costs, significant labor requirements, and tag retention issues. These challenges have created an urgent need for automated individual sheep identification systems. The development of such systems faces significant technical difficulties due to complex real-world conditions such as variable lighting, different sheep poses, and frequent occlusions. To address these challenges, this study makes two major contributions: First, we develop an automated pipeline for collecting and annotating sheep face datasets. Second, we propose an improved recognition model by integrating the <i>Content-Aware Reassembly of Features (CARAFE)</i> operator into the YOLOv5 architecture. Finally, we implement a comprehensive training and validation protocol to optimize model parameters and improve robustness. Our experimental results show that the proposed framework achieves a recognition accuracy of 97.87%, outperforming recent canonical methods. This advancement contributes to the development of a practical solution for precision breeding applications, offering reliable identification without the drawbacks of physical tagging systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Deep Detection and Identification Framework for Smart Sheep Farm

  • 期刊论文

  • 发表日期：

    2025年6月26日

查看全文
  • Xianjun Fu,
  • Wei Guo,
  • Chunzhi Li,
  • Xiaohua Chen,
  • Huimeng Zheng

摘要

In precision livestock farming, traditional sheep management relies primarily on ear tagging. However, this conventional approach has several limitations, including high implementation costs, significant labor requirements, and tag retention issues. These challenges have created an urgent need for automated individual sheep identification systems. The development of such systems faces significant technical difficulties due to complex real-world conditions such as variable lighting, different sheep poses, and frequent occlusions. To address these challenges, this study makes two major contributions: First, we develop an automated pipeline for collecting and annotating sheep face datasets. Second, we propose an improved recognition model by integrating the Content-Aware Reassembly of Features (CARAFE) operator into the YOLOv5 architecture. Finally, we implement a comprehensive training and validation protocol to optimize model parameters and improve robustness. Our experimental results show that the proposed framework achieves a recognition accuracy of 97.87%, outperforming recent canonical methods. This advancement contributes to the development of a practical solution for precision breeding applications, offering reliable identification without the drawbacks of physical tagging systems.