中国农机化学报2026，Vol.47Issue(1)：79-86,8.DOI:10.13733/j.jcam.issn.2095-5553.2026.01.012
改进YOLOv8_obb的大豆主茎节点识别研究
Research on recognition of soybean stem nodes using improved YOLOv8_obb
摘要
Abstract
Plant type has a significant impact on soybean yield,and the number of main stem nodes in soybeans is a crucial trait in the formation of soybean plant type.To achieve the recognition and counting of the number of soybean main stem nodes under field conditions,taking soybeans in the Jiusan region of Heilongjiang Province as the research object,an improved soybean main stem node recognition method,YOLOv8_obb—AES,based on the YOLOv8_obb model was developed,the number of soybean main stem nodes is obtained by calculating them in soybeans.The refined model incorporated an efficient attention mechanism module to reduce the computational requirements,replaced the Path Aggregation Network(PANet)in the YOLOv8_obb network with a Progressive Feature Pyramid Network(PFPN)to enhance multi-scale fusion capabilities,and substituted the Intersection over Union(IoU)loss function to accelerate boundary regression and improve model convergence speed.The results demonstrated that the YOLOv8_obb—AES algorithm achieved a mean average precision of 89.45%and a detection speed of 78.8 frames/ms for identifying soybean main stem nodes in the field,representing a 8.45%increase in mAP and a 7.6 frames/ms boost compared to the original algorithm.Specifically,for the Jiuyan 17 soybean variety,the recognition accuracies for the six categories of main stem node counts were 85.4%,84.5%,87.6%,85.2%,81.6%,and 82.2%,respectively.This research provides technical support for exploring the correlation between soybean yield and the number of main stem nodes,and advance the understanding of soybean growth and yield potential.关键词
大豆/主茎节点/目标识别/渐进特征金字塔网络/高效注意力机制Key words
soybean/main stem nodes/object recognition/asymptotic feature pyramid network/efficient channel attention分类
农业科技引用本文复制引用
Yang Yanxu,Li Jinyang,Shi Wenqiang,Qi Liqiang,Zhang Wei..改进YOLOv8_obb的大豆主茎节点识别研究[J].中国农机化学报,2026,47(1):79-86,8.基金项目
国家现代农业产业技术体系(CARS—04—PS30) （CARS—04—PS30）
黑龙江八一农垦大学研究生创新科研项目(YJSCX2023—Y19) （YJSCX2023—Y19）
