基于改进ResNet50与Outlook注意力的番茄病虫害识别方法

打开文本图片集

中图分类号：S641.2 文献标志码：A 文章编号：2096-9902（2026）05-0012-0

Abstract：Tomatoisanimportantcashcrop，anditsdiseaseandinsectconditionsseriouslyaffctitsyieldand quality. Traditionalmanualidentificationmethodsfordiseasesandpestsreineficientandhighlysujective，makingitdifulttomeet monitoringrequirements.Thispaperproposesamethodfortomatodiseaseandpestrecogntionandclasificationbasedonan improvedResNet5Omodel.Firstly，ResNet5Oisusedasthebasenetwork，withtheintegrationofcontextualconvolutionalblocks toexpandthereceptivefieldandincorporatecontextualinformation.Secondly，anovelOutlookatentionmechanismisintroduced toenhancethemodel'sabilitytocapturelocallysignificantfeatures，whiletransferlearningisutilizedtoaceleratemodel trainingandimprovegeneralizationcapability.TheexperimentswereconductedundertheLinuxsystem，withthemodel implementedusingthePyTorchframework．Acuracywasusedastheevaluationmetricfortheexperiments.Theresults demonstrate thattheimproved modelachievesanacuracyof98.74%intomatodiseaseandpestidentificationtasks，whichisa 3.31percentagepointhigherthanthestandardResNet5Omodel.Additionall，themodelexhibitsfasterconvergenceandlower lossvalues.Theproposedmethodefectivelyimprovestheaccuracyandefciencyoftomatodiseaseandpestrecognitionand holds significant practical value for real-world agricultural production.

Keywords：tomatodiseaseandpestidentification；ResNet5Omodel;contextconvolution;Outlookatentionmechanism；transfel learning

番茄是全球范围内广泛种植的重要经济作物，在传统的农业生产中有着非常重要的地位。（剩余6724字）