Scholarlink

<p>Real-time acquisition of the husking rate is crucial for adjusting and optimizing the husker parameters. To address the issues in existing rice husking rate identification models, such as large computational demands, excessive parameters, and challenges in deploying on embedded platforms, a lightweight algorithm based on YOLOv5s is proposed. First, the image dataset is augmented to improve the detection capability of the YOLOv5_Get Small (YOLOv5s_GS) model under various conditions. Then, the structure of the YOLOv5s_GS model is optimized by reducing the high-resolution detection head and adjusting the feature fusion mechanism to adapt to the resource constraints of embedded platforms. Finally, the YOLOv5s_GS model is tested and compared with the YOLOv5s model. The results show that the YOLOv5s_GS model demonstrates significant improvements over the original YOLOv5s model in terms of model parameters, model size, and inference time, specifically in the context of rice husking rate detection. The inference time is reduced by 2.9 ms per image, the number of parameters has been reduced by approximately 73.13%, and the model size is reduced by 10.3&#xa0;MB, verifying the effectiveness of the improved algorithm in resource-constrained environments.</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 lightweight method for evaluating the husking rate of rice using an optimized YOLOv5 model

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Shixiong Zhang,
  • Xingchong Li,
  • Jianxin Ren,
  • Ang Li

摘要

Real-time acquisition of the husking rate is crucial for adjusting and optimizing the husker parameters. To address the issues in existing rice husking rate identification models, such as large computational demands, excessive parameters, and challenges in deploying on embedded platforms, a lightweight algorithm based on YOLOv5s is proposed. First, the image dataset is augmented to improve the detection capability of the YOLOv5_Get Small (YOLOv5s_GS) model under various conditions. Then, the structure of the YOLOv5s_GS model is optimized by reducing the high-resolution detection head and adjusting the feature fusion mechanism to adapt to the resource constraints of embedded platforms. Finally, the YOLOv5s_GS model is tested and compared with the YOLOv5s model. The results show that the YOLOv5s_GS model demonstrates significant improvements over the original YOLOv5s model in terms of model parameters, model size, and inference time, specifically in the context of rice husking rate detection. The inference time is reduced by 2.9 ms per image, the number of parameters has been reduced by approximately 73.13%, and the model size is reduced by 10.3 MB, verifying the effectiveness of the improved algorithm in resource-constrained environments.