基于改进YOLO11模型的棉花叶片病害检测

打开文本图片集

，.基于改进YOLO11模型的棉花叶片病害检测[J].江苏农业学报，2026，42（3）：521-530.doi：10.3969/j.issn.1000-4440.2026.03.010

关键字：棉花；叶片病害；YOLO11模型；多尺度特征融合；注意力机制；动态特征融合中图分类号：TP391.41 文献标识码：A 文章编号： 1000-4440（2026）03-0521-10

Cotton leaf disease detection based on an improved YOLO11 model

FENG Wei， ZHAO Xia （Collegeof Information ScienceandTechnology，Gansu Agricultural University，Lanzhou 73oo7o，China）

Abstract：Inorder toimprove the performanceof theYOLO11 model indetecting cotton leaf diseases，this study usedthe YOLO11 modelas the benchmark model，introduced a multi-scale edge enhancement module （MEEM）to replace theoriginal C3k2 module，and added a condensed atention （CA）mechanism in the C2PSA module.The improved YOLO11 model was constructed byreplacingthe feature fusion module（Contat）with dynamic feature fusion（DFF），thereby enhancing its ability in extractingcomplex detailsand edge features of thetarget，recognizing smallobjects，and fusing multi-scale features.Thecoton disease imagedataobtainedbyKaggleplatform，Googleand Baidusearchtools wereused to comparethemodel detectionperformance.The resultsshowedthatthedetectionaccuracy，recallrate，mean averageprecision （ mAP50 ）and mAP50：95 of the optimized model for cotton leaf diseases were 6.1 percentage points，2.5 percentage points，4.5percentage pointsand2.5percentage points higher thanthoseofthebenchmark modelrespectively.Thefloating point operations and parameter count increased by 34.92% and 26.74% ，respectively. Compared with YOLOv5，YOLOv8n，YOLOv1On and other models，although the improved YOLO11 model had higher parameter count and floating point operations，its detection accuracy，recall rate and average accuracy were higher. The improved model significantlyimproved thedetectionperformance ofcotton leaf curl，gray mold，leaf spot and wilt diseases，and thedetectionaccuracyofFusariumwiltandhealthyleavesremainedhigh.Theresultsofthisstudyplayanimportantroleinthe automaticandaccurate detection，early warning and control of cotton diseasesand thesafe production of coton.

KeyWords：coton；leaf disease；YOLOl1 model；multi-scale feature fusion；atention mechanism；dynamic feaure fusion

棉花是重要的经济作物之一，在纺织、化工等诸多领域有着广泛的应用[1]。（剩余14037字）