基于改進(jìn)YOLOv8的無(wú)人機(jī)圖像玉米幼苗檢測(cè)
打開(kāi)文本圖片集
中圖分類號(hào)：S513 文獻(xiàn)標(biāo)識(shí)碼：A 文章編號(hào)：1000-4440（2025）06-1179-09
Abstract：Unmannedaerial vehicletechnology，withits higheficiencyandprecision，hasbeen playingan importantrole intheagricultural fieldandhasben widelyappedinfarmlandmonitoring，precisionfertilization，andpestanddiseasecontrol. However，inthecomsedingimagescolectedbyunmannederialvehicle，weedsndotherinterferentsinthefieldhavecolors similar tocomseedlings，whichcaneasilylead tofalsedetectionand mised detectionintheYOLOv8 model.Additionally，the occlusionbetweencornsedlingscanalsoafectthedetectionaccuracyof the model.Inresponsetotheseproblems，this study proposed the YOLOv8+MultiSEAM + MetaNeXtStag + WIoU（YOLOv8-MMW）model. Based on the YOLOv8 model architecture， this modelfirstintroducedthe MultiSEAMatentionmechanismintothenecknetwork，whichefectivelyenhancedthemodel's featureextractionabilityincomplexscenes.SecondlyitincorporatedtheMetaNeXtStagemodulefromInceptionNeXt.Onhisbasis，theWise-IoUlossfunctionwasadoptedtoimprovemodelaccuracy.Theexperimentalresultsonthetestsetshowedthatthe accuracy and the mean average precision at an intersection-over-union threshold of O.50（ mAP50 ）ofthe YOLOv8-MMW model
reach 98.9% and 89.6% ，respectively，which were 6.1 percentage points and 2.4 percentage points higher than those ofthe original YOLOv8n model.The YOLOv8-MMW model proposed in this study demonstrates stronger robustnessin complex farmland environmentsand can effectively improve the detection accuracy of corn seedling images captured by inmanned aerial vehicle，and can provide technical support for agricultural management and monitoring
Key words：corn seedling；unmanned aerial vehicle；YOLOv8；MultiSEAM；MetaNeXtStage；lossfunctior
玉米是中國(guó)種植面積1和產(chǎn)量均居首位的作物2，也是重要的經(jīng)濟(jì)作物，對(duì)國(guó)家糧食安全具有重大戰(zhàn)略意義[3]。（剩余12959字）