中国农机化学报2026，Vol.47Issue(1)：73-78,86,7.DOI:10.13733/j.jcam.issn.2095-5553.2026.01.011
基于YOLOv8的轻量化机收小麦杂质检测方法
Impurity detection method for lightweight machine-harvested wheat based on YOLOv8
摘要
Abstract
To achieve efficient detection of wheat impurities,a lightweight wheat impurity detection method based on YOLOv8 is proposed.Firstly,in order to reduce the computational load during the convolution process,the C2f module in the Backbone is replaced with the CPC module that introduces partial convolution(PConv).Then,the Advanced Screening Feature Fusion Pyramid Network(HS—FPN)is introduced to solve the problem of scale differences between the two types of impurities,namely straw and wheat ears.Finally,CIoU is replaced with EIoU to obtain a more realistic prediction box and accelerate the convergence speed of the model.The results show that the precision rate,recall rate and average precision of the improved YOLOv8 model under the test set are 94.8%,94.5%and 98.5%respectively.Compared with the original basic network YOLOv8n,the model weight is reduced by 47.71%,and the precision rate,recall rate and average precision are increased by 1.6%,0.9%and 1.1%respectively.Compared with YOLOv5,YOLOv7 and YOLOv7—Tiny,the improved YOLOv8 model has the least memory usage,only 3.1 MB,and the average precision rate is increased by 1.8%,1.9%and 1.1%respectively.关键词
小麦杂质/YOLOv8/轻量化模型/部分卷积/HS—FPNKey words
wheat impurity/YOLOv8/lightweight model/partial convolution/HS—FPN分类
农业科技引用本文复制引用
Qian Rui,Zhao Liqing,Yin Yuanyuan,Liu Chuang,Xia Junjie,Zhang Jingke..基于YOLOv8的轻量化机收小麦杂质检测方法[J].中国农机化学报,2026,47(1):73-78,86,7.基金项目
国家自然科学基金项目(32071911) （32071911）
国家重点研发计划(2023YFD2000404) （2023YFD2000404）
山东省现代农业产业技术体系小麦创新团队(SDIT—0-12) （SDIT—0-12）
