基于改进的LSN-YOLOv8模型和无人机遥感图像的水稻稻曲病检测方法

打开文本图片集

图分类号：S435.115 文献标识码：A 文章编号： 1000-4440（2025）05-0905-11

Abstract：Toaddress thechallenges ofcomplex backgrounds，smallesion targets，and the similaritybetween lesion targetsand backgroundfeatures inrice falsesmut imagescollcted byunmannedaerial vehicles（UAVs），we proposed the LSN-YOLOv8detection model.Themodel was basedon theYOLOv8 framework，andthe largeselective kernel network （LSKNet）was incorporated into the backbone network.Bydynamicallyadjusting thereceptive field range，the model enhanceditsabilitytoextractfeaturesofsmalltargets.Aditionally，acoordinateatention mechanism（CA）modulewas inte

grated into the backbone network to combine the spatial location information of lesionswith channel attention， thereby enhancingthe model's focusonkeyregionswhile reducing background interference.The detection process was visualized and analyzed using the gradient-weighted class activation mapping（Grad-CAM）technique，thereby

providingintuitive explanationsfor the model’sdecision-making.To verifythe model’s performance，ricefalse smut images captured by UAVsat diffrent disease stages andundervariousbackgroundconditionswereused toconstructarice false smutdataset.Thisdatasetwasutilizedfor modeltrainingand testing.Theexperimentalresultsindicatedthattheprecision， recall，and mean average precision at an intersection over union threshold of 0.50（ mAP50 ）of the LSN-YOLOv8 model proposed in this study were 94.8% ， 87.3% ，and 92.3% ，respectively. These indices were all higher than those of classic object detection models such as YOLOv5，YOLOv7，YOLOv8 and Faster R-CNN.The visualization analysis results using Grad-CAM technology indicated thatthe LSN-YOLOv8 model wascapableof moreaccurately focusing onthediseased regions in the images.TheLSN-YOLOv8 model proposed inthis studycan provide technical supportforthemonitoring of rice false smut，disease control and prevention，and the identification of rice disease resistance.

KeyWords：ricefalsesmut；disease identification；unmannedaerial vehicle；YOLOv8model；largeselective ker-nel network（LSKNet）；coordinate attention mechanism（CA）

水稻是全球最重要的粮食作物之一[1]，其产量与质量的提升是水稻生产的核心目标。（剩余13689字）