沈阳农业大学学报2026,Vol.57Issue(5):104-116,13.DOI:10.3969/j.issn.1000-1700.2026.05.010
基于田间复杂背景的改进YOLOv11n黄瓜霜霉病分级检测研究
Enhanced YOLOv11n-based Methodology for Cucumber Downy Mildew Severity Grading and Detection in Complex Field Environments
摘要
Abstract
[Objective]This study aimed to accurately identify and grade cucumber downy mildew disease in complex field environments,and then carry out precise prevention and control and scientific pesticide application.[Method]Based on the YOLOv11n framework,WTConv was introduced to enhance lesion feature extraction,CARAFE was adopted to improve detail restoration,and the SEAM module was integrated to optimize the channel-spatial attention mechanism,thereby improving the recognition of mild and complex lesions.[Result]The experimental results demonstrated that,compared with the baseline model,WCS-YOLOv11 achieved consistent improvements in four key metrics:precision,recall,mAP@50,and mAP@50-95 increased by 3.4%,6.4%,4.7%,and 4.4%,respectively,while maintaining a compact model weight size of only 5.5 MB.Visualization and confusion matrix analysis further demonstrate the advantages of the proposed model in terms of sensitivity for mild lesion recognition and boundary segmentation accuracy for moderate-to-severe lesions,effectively reducing the cross grade misclassification rate.Compared with mainstream models such as YOLOv5,YOLOv8,YOLOv10,YOLOv12,RT DETR,and Vision Transformer,WCS YOLOv11 achieves a mAP@50 of 82.6% and a recall of 80.4%,indicating its optimal performance in both accuracy and robustness.Finally,the model is successfully deployed on a LubanCat 5 development board,and a cucumber downy mildew grading detection system is constructed using the Qt5 platform.When processing 640×640 input images on the LubanCat 5 board,the end to end average inference speed reaches 22.5 f·s-1,enabling real time detection in field environments.[Conclusion]The proposed method provides a feasible technical pathway for intelligent severity grading of cucumber downy mildew and offers valuable insights for precision disease management and the digital development of protected agriculture.关键词
黄瓜霜霉病/分级检测/YOLOv11n/轻量化/田间复杂背景Key words
cucumber downy mildew/severity grading/YOLOv11n/lightweight/complex field environments分类
农业科技引用本文复制引用
程斌,李冉,陈恒..基于田间复杂背景的改进YOLOv11n黄瓜霜霉病分级检测研究[J].沈阳农业大学学报,2026,57(5):104-116,13.基金项目
西南林业大学智能装备研发项目(2025) (2025)