基于改进YOLOv5的多种植物叶片病虫害分类视觉模型研究

打开文本图片集

中图分类号：S435；TP391 文献标识码：A 文章编号：2095-5553（2026）04-0115-1

Abstract：Traditionalmonitoringofcroppestsanddiseasesreliesonmanualobservation，whichisdificult tomettheneeds of modern agricultural production，and there is anaggravating trendof major pests and diseases of major grain crops， vegetables，aswellasfruitsinthecountry，itisofgreat significancetostrengthenthemonitoringof pestsanddiseasesin farmland.Firstly，thefeaturerepresentationabilityof themodelisenhancedbyimproving theYOLOv5modeland introducing the SPD一Conv module.Secondly，thecontext enhancement module isadded toenrich the upper and lower information links of thefeature pyramid network of the image.Thirdly，the Context Aggregation and Coord Conv modules areaddedtoenhancethespatialperceptionabilityoftheimageandimprovethelocalisationabilityof thepestsanddisease.In ordertoverifytheefectivenessof theproposed model，the“AIChallenger 2O18Pestand Disease Classfication Dataset"and the“IP102Dataset”wereusedasthebasedata fordata integrationand expansion，and finally，9Otypes ofpestsand diseases withdiferent verification degrees of 15 plants such as rice，potato，apple andsoon were identified for experiments.The experimental results showed that the improved model achieved the optimal effect，and the mAP@0.5 and （20 both reached 90.4% . This study can provide a reference for the identification and application of plant leaf pests and diseases.

Keywords：plant leaf；agricultural decisionsupport；diseases and insect pests；image clasification；data augmentation improved YOLOv5

0 引言

我国是农业生产大国，在现代农业中，随着现代农业技术的发展，精准和高效的植物健康管理成为提高农业生产效率和作物品质的关键。（剩余17211字）