基于改进YOLOv8n的苹果叶片病害检测

打开文本图片集

中图分类号：TP391.4；S-3 文献标识码：A 文章编号：2095-5553（2026）04-0147-09

Appleleafdiseasedetection based on improved YOLOv8n

Zhang Hequn，Lin Suyin，Chu Yuntong，Fan Mingchao，Gao Xinfeng （College of Animation and Media，Qingdao Agricultural University，Qingdao，2661O9，China）

Abstract：Toimprove thedetectionacuracyofapple leaf diseases incomplex backgroundsand mitigate their impacton production，an improved YOLOv8n model was developed.A comprehensive dataset was constructed，consisting of 4 686 images depicting 6 common appe leaf diseases acrossvarious scenarios. Several enhancements were introduced to the model.TheOptimized—CBAMattentionmechanism was incorporated tostrengthenthe model’sabilityto focus on key features.Theoriginal feature fusion module was replacedwith themore lightweightand eficient SimSPPFmodule. Additionally，theFocal—EIoUlossfunction thatcombines theadvantagesofCIoUandEIoUwasimplemented to further enhance feature extractionandimprove theprecision of bounding box localization.Experimental results showed thatthe improved YOLOv8n model achieved an accuracy of 94.16% ，a recall rate of 84.16% ，and a mean average precision （204号 （mAP） of 87.4% .The model size was 6.15MB ，and the detection speed reached 37O.4 FPS.Compared to the baseline YOLOv8n model，these values represented increases of 1.02% in accuracy， 4.07% in recall， 2.8% in mAP ，a slight size increase of 0.03MB ，and a 14.8% improvement in speed.Overall，the enhanced YOLOv8n model significantly improvedthedetection performanceofappleleaf diseases，particularlyincomplex backgrounds.Itsoptimized accuracy and robustness offers reliable support for intellgent agricultural disease monitoring and management.

Keywords：apple leaf；plant disease；YOLOv8n;deep learning；object detection

0 引言

叶片部位发生的病害，对苹果质量和产量产生直接影响。（剩余12037字）