Visual Detection of Ratoon-Emerged Sugarcane Seedling Hill Using an Improved YOLOv8n Model

文献类型： 外文期刊

第一作者： Mo, Yongmei

作者： Mo, Yongmei;Li, Hongwei;Lai, Xindong;He, Deqiang;Zhang, Shunsheng;Wu, Tao;Li, Hongwei;He, Deqiang

作者机构：

关键词： Ratoon-emerged sugarcane seedling hill; Multi-scale target; Visual detection; YOLOv8n

期刊名称：SUGAR TECH （ 影响因子：2.0； 五年影响因子：2.0 ）

ISSN： 0972-1525

年卷期： 2025 年

页码：

收录情况： SCI

摘要： Accurate visual detection of ratoon-emerged sugarcane seedlings hill (RSSH) is a key technology for improving intelligent replanting during the ratoon seedling stage. The unique growth characteristics of RSSH in complex sugarcane field environments, including irregular growth morphology and disordered distribution, present technical challenges for visual detection. To address these challenges, an improved YOLOv8n model is proposed in this study to achieve efficient detection. Firstly, a P6 detection layer was added to the backbone network to expand the receptive field, enabling better detection of RSSH with multi-scale target sizes. Subsequently, the SPDConv modules and a BiFPN network incorporating a high-resolution P2 feature layer were introduced into the neck network, effectively enhancing the model detection precision. To reduce the number of model parameters and computational complexity, GhostConv and C3 modules were respectively incorporated into the backbone and neck networks. Finally, the loss function was modified to a combined Focaler-CIoU loss function, further enhancing the model detection capability. The improved YOLOv8n model was compared with mainstream object detection models, including YOLOv5s, YOLOv6n, YOLOv7-tiny, YOLOv8n, YOLOv8s, YOLOv9s, YOLOv10n, and YOLOv11n. It achieved the highest precision and F1 score of 95.5% and 92.93%, respectively. The mAP@0.5 reached 96.2%, while the mAP@0.5-0.95 reached 84.4%, with a model size of only 12 MB. These results demonstrate that the improved YOLOv8n model has significant advantages in detecting RSSH in complex sugarcane field environments. This study provides technical support for intelligent ratoon sugarcane replanting.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序