| 摘 要: 为了适应智能农业的发展需求,并解决杂草检测过程中常见的效率低和可靠性不足的问题,文章提出一种基于改进YOLOv8(You Only Look Once)算法的轻量化高性能杂草检测模型。基于Ghost Conv(Ghost卷积)设计了全新的轻量化C3Ghost(Cross Stage Partial Network)模块以取代C2f(CSPDarknet53 to 2-Stage FPN),此外在模型的Neck部分引入CBAM(Convolutional Block Attention Module)注意力机制模块。通过构建专门的杂草数据集检测模型性能,发现该改进模型在杂草检测任务中的mAP-0.5达到了90.5%,mAP-0.5∶0.9评价标准达到54.4%。与官方发布的YOLOv8-s模型相比,分别提高了0.8%和0.3%,改进后的模型参数量减少了46.18%,计算量降低了42.81%。以上结果证明了所提改进策略在模型轻量化方面的有效性,为智能农业中的杂草管理提供了理论和技术支持。 |
| 关键词: 杂草识别;目标检测;YOLOv8;轻量化;注意力机制 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 国家自然科学基金(61905219);国家重点研发计划(2023YFD1401100);现代农业产业技术体系(CARS-01) |
|
| Lightweight Weed Detection Model Based on Improved YOLOv8 |
|
SHU Aijing, ZHANG Yanchao
|
(School of Inf ormation Science and Engineering, Zhejiang Sci-Tech University, Hangzhou 310000, China)
202130605274@mails.zstu.edu.cn; yczhang@zstu.edu.cn
|
| Abstract: To meet the demands of intelligent agriculture and address common issues of low efficiency and insufficient reliability in weed detection, this paper proposes a lightweight high-performance weed detection model based on the improved YOLOv8 (You Only Look Once) algorithm. A novel lightweight C3Ghost (Cross Stage Partial Network) module is designed based on Ghost Conv (Ghost Convolution) to replace the C2f (CSPDarknet53 to 2-Stage FPN). Additionally, a Convolutional Block Attention Module (CBAM) is introduced in the Neck part of the model. By constructing a dedicated weed dataset to evaluate model performance, it is found that the improved model achieves a mAP-0.5 of 90.5% and a mAP-0.5: 0.9 of 54.4% in weed detection tasks. Compared with the officially released YOLOv8-s model, improvements of 0.8% and 0.3% are observed, while the number of the parameters of the improved model are reduced by 46.18% , and computational complexity is reduced by 42.81% . These results demonstrate the effectiveness of the proposed improvements in model lightweighting, providing theoretical and technical support for weed management in intelligent agriculture. |
| Keywords: weed identification; target detection; YOLOv8; lightweight; attention mechanism |
