华中农业大学学报2026，Vol.45Issue(3)：45-55,11.DOI:10.13300/j.cnki.hnlkxb.2026.03.004
基于YOLOv8n改进的玉米幼苗杂草识别模型
An improved YOLOv8n-based model for identifying weeds and seedling of maize
摘要
Abstract
A lightweight model named as YOLOv8n-DSSW was developed based on YOLOv8n to effectively control the damage of weeds in maize seedlings,improve the accuracy of identifying weeds,and meet the requirements for mobile deployment.The C2f_Dual module was integrated to achieve initial light-weight.The SPDConv module was used to enhance the capability of identification for small targets and low-resolution images while further reducing the complexity of model.The SPPELAN module was incorporated to strengthen the perception of multi-scale feature.The regression loss function used Inner WIOU to assist in optimizing the accuracy of bounding box localization and allocating weights to anchor boxes of different qualities,thereby overall improving the accuracy and robustness of identification.The results showed that the proposed model achieved an increase of 3.4 percentage points in precision,1.9 percentage points in re-call,and 2.4 percentage points in mean average precision(mAP)in identifying weeds and seedling of maize.The weight of model was reduced by 15.9%and floating-point operations(FLOPs)decreased by 14.8%.This model maintained high performance of identification in complex field environments including weak features of weeds,overlapping weeds with maize seedlings,and similar morphology.Its compact size and computational efficiency make it suitable for the deployment in mobile devices for controlling weeds dur-ing the stage of seedling.关键词
玉米/杂草识别/轻量化/YOLOv8n/目标检测Key words
maize/identification of weeds/lightweight/YOLOv8n/object detection分类
信息技术与安全科学引用本文复制引用
路京奥,顾文辉,郑纪业,位国建,史嵩,郑世玲,张晓艳..基于YOLOv8n改进的玉米幼苗杂草识别模型[J].华中农业大学学报,2026,45(3):45-55,11.基金项目
山东省现代耕作制度技术体系农业灾害预警与防控岗位建设任务(SDAIT-31-05) （SDAIT-31-05）
山东省重点研发计划项目(2024CXGC010901) （2024CXGC010901）
山东省农业科学院农业科技创新工程(CXGC2025A05) （CXGC2025A05）
