摘要
目的针对玉米茎腐病的精准识别问题,提出一种基于特征融合结构改进的YOLO v11n-RCA模型,解决玉米生长期因茎腐病导致的产量损失问题.方法在YOLO v11n模型的主干网络中加入RCM模块,增强上下文特征的捕捉能力,并在颈部引入RCS-OSA模块替换部分C2k3模块,提升特征提取能力.采用ATFL函数替代原损失函数,优化模型的多尺度特征融合能力,进一步提高检测精度.结果改进后的YOLO v11n-RCA模型在精度、召回率、mAP50及F1分数上均有显著提升.精准度达到94.7%,召回率为0.734,mAP50为0.818,F1分数为0.829,相较于原YOLO v11n模型分别提升了3.8%、12.2%、12.1%、18.5%、11.5%.结论改进后的YOLO v11n-RCA模型具备优异的综合性能,能够准确识别玉米茎腐病,为后续灌溉防治提供有效支持.
Objective An improved YOLO v11n-RCA model with enhanced feature fusion architecture was developed for accurate detection of corn stalk rot in this study,aiming to address yield loss caused by stalk rot during the growing season of corn.Methods Initially,the model enhances the original YOLO v11n by integrating an RCM module into the backbone network to improve contextual feature capture.Additionally,an RCS-OSA module is introduced at the neck,replacing part of the C2k3 modules,which enhances feature extraction capabilities.The ATFL function is then used to replace the original loss function,optimizing the model's multi-scale feature fusion ability and further enhancing detection accuracy.Results Experimental results demonstrate significant improvements in precision,recall,mAP50,and F1 score with the improved YOLO v11n-RCA model.Specifically,accuracy reached 94.7%,recall was 0.734,mA P50 was 0.818,and F1 score was 0.829,showing improvements of 3.8%,12.2%,12.1%,18.5%,and 11.5%,respectively,compared to the original YOL0 v11n.Conclusion The improved YOL0 v11n-RCA model exhibits excellent overall performance and accurately identifies corn stalk rot disease,providing effective support for subsequent irrigation control and disease management.
作者
张平川
李珊
张彩虹
马泽泽
杨莹
ZHANG Pingchuan;LI Shan;ZHANG Caihong;MA Zeze;YANG Ying(School of Computer Science and Technology,Henan University of Science and Technology,Xinxiang 453003,China)
出处
《河南科技学院学报(自然科学版)》
2026年第1期52-64,共13页
Journal of Henan Institute of Science and Technology(Natural Science Edition)
基金
河南省科技攻关(222102210116)。
