基于YOLOv8n的罂粟识别改进算法研究

打开文本图片集

中图分类号：TN248.4;TN918.91 文献标识码：A文 章 编 号：1673 9868（2025）06 0201 12

Abstract：Aimin atthe roblemsofsmalobectreconition，comlexbackround，andrelativescaleof targetobjectchangeinpopydetectiontaskmodel，apopyrecognitionalgorithmimprovementstrategy basedonYOLOv8nwasproposed.Inimprovedalgorithm，firstly，theiRMBatentionmechanism was introducedtoenhancetheabilitytodetectsmaltargetsofpopy.Secondly，theupsamplemodulewas replacedwiththeCARAFEupsamplingoperatortoimprovetherecognitionrateofpopyincomplexbackgrounds.Finaly，thelosfunctionwasreplacedfromCIoUtoMPDIoUtocopewiththerelativescalevariationofremotesensingtargetsofpopy.Thestudyfoundthattheimprovedpopyrecognitionalgorithm basedonYOLOv8ncanincreasethemAPoftraditionalYOLOv8nalorithmfrom 83.1% to86. 6% ，of which the AP of popy fruit recognition increased by 1.0 percentage point， and AP of popy stamen recognitionincreasedby6.1percentagepoints.TheimprovedYOLOv8nrecognitionalgorithmenhancedthe comprehensiveperformanceofpopyrecognition.

Keywords：popyrecognition;algorithmimprovement;YOLOv8nalgorithm;atentionmechanism

国家高度重视毒品预防、管控和打击工作。（剩余10901字）