改进YOLOX算法的葡萄病害识别

打开文本图片集

中图分类号：S663.1；TP391.41 文献标识码：A 文章编号：2095-5553（2025）08-0121-07

Abstract：Grape diseases pose a significant chalenge to grape growers worldwide.To met the dems intellgent agricultureaddressthelimitations loweficiencyreliabilityin grape disease detection，thisstudy propoeda real-time，high-performance detectionmodel basedonan improvedYOLOX algorithm.The model was improved by integratinganoptimized Spatial PyramidPooling（SPP）layer，which efectively extractedrelevant featuresat multiple scalesbyconcatenating multi-level features fromsmallto largescales.Onthetesteddatasets，theimproved model achieved an average accuracy 98.59% ，marking a 6.72% improvement over the original YOLOX model. Furthermore， comparative evaluations with five clasic object detection algorithms （YOLOv4，YOLOv5，YOLOv7，YOLOv8 YOLOX）revealed that the improved model achieved mean Average Precision （mAP ）improvements 23.33% ， 19.78% ， 7.62% ， 5.96% 6.62% ，respectively. This approach demonstrated strong potential for improving grape diseasemanagement，reducing economic losses，ering valuable insights forthe developmentautomated disease detection systems for other crops.

Keywords： grape；YOLOX model; disease detection； image classification；deep learning

0 引言

葡萄病害是全球葡萄种植者面临的重大问题，会对经济造成巨大损失[1]。（剩余10695字）