基于改进YOLOv8n的谷子谷瘟病检测方法

打开文本图片集

中图分类号：S435.15TP39 文献标识码：A 文章编号：1671-8151（2025）04-0078-10

Abstract：[Objective]Intellgent detectionoffoxtailmillt blast disease in field environments faces challnges suchasstrong background interference，varyinglesionscales，andlimitedadaptabilityofexisting modelsinreal-worldapplications.Methods] Toaddress these isses，this study leveragedalarge-scale foxtail milltblastdataset and proposed an improveddetection model，YOLOv8-SDL，basedonYOLOv8n.[Results]The model enhanced detection stabilityand accuracy incomplex backgrounds through three key modifications： strengthening backbone network feature extraction，introducinglightweight and eficientupsampling to improvefeature fusion，and incorporating anattentionmechanism toenhance critical feature selection. First，the Switchable Atrous Convolution （SAC） structure was optimized by adding a 3×3 convolution layer before the 1 ×1 convolution initscontext structure.Combinedwith the Squeee-and-Excitation（SE）chanelatention mechanism，the SESAC module was constructed and embedded into theC2f moduleofthebackbone network to enhance multi-scale feature extractionof lesionareas.Second，DySampleupsamplingreplaced nearest-neighborinterpolation.Byemployingadynamicpointsampling strategy，DySamplereduced computational overhead whileminimizing feature information loss during upsampling， thereby mproving feature fusionandsmallesionlocalizationaccuracy.Third，theLargeSeparableKernelAtention（LSKA） mechanism wasintegratedintotheneck network.Itsseparableconvolutiondesignenhanced themodel'sabilitytofocus ondis ease lesion features incomplex backgrounds.Experimentalresults showedthatYOLOv8-SDLachieved adetectionaccuracyof 91.5% ，mAP @0.5 of 93.9% ，and mAP@0.5～0.95 of 62.0% ，outperforming the original model by 4.4% ， 1.4% ，and 2.2% ，respectively.[Conclusion]This model provided robustand reliable technical support foracurate foxtail miletblastdetection inchallengingfield environments.

Keywords：Deeplearning，Target detection，Foxtail milletblast，YOLOv8r

谷子（SetariaitalicaL.），一年生，种子脱皮后俗称小米，是我国北方传统的粮食作物，因其耐旱、耐瘠、适应性强等特点，是发展有机旱作农业的重要作物[1]。（剩余14046字）