Slim-YOLOv8n：基于改进YOLOv8n的轻量级棉铃检测模型

打开文本图片集

Abstract：[Objective]Cottonyield predictionisanimportantpartofcotonproductionmanagement inourcountry.The accuracyofbolldetectionduringthebollopeningstagedirectlyafects theprecisionofyieldestimation.Inordertosolvethe problemthat thecomputationalcomplexityofexistingcottonbolldetectionmodels increasessignificantlyduetothepursuitof detectionaccuracythisstudyproposesalightweightcotonbolldetectionmodel，Slim-YOLOv8n，basedonimproved YOLOv8n.Methods]ThismodeltakesYOLOv8nasitsmainframework.Ononehand，itintegratesalightweghtcros-scale featurefusionnetworkintheneckstructuretoefectivelyreducethedimensionalityofmulti-scalefeaturefusionandlower computationalcomplexityOntheotherhand，itreconstructsthedetectionheadthroughreparameterizedconvolutionandthe ideaofsharing，designingareparameterizedheadandatwo-stage featureprocesingstream tomaintainaccuracy while achieving modellightweighting.[Results]Experimentalresults showthatthismodelachievesadetectionaccuracyofupto 98.20% .Compared with the YOLOv8nmodel，it reduces the number of parameters by 44.84% ，computational cost by 39.51% ， and model size by 43.34% ，verifying the superiority of the model improvement.[Conclusion] Slim-YOLOv8n fully meets the dualdemandsofhighaccuracyandlightweightingforbolldetectiontasks，providingstrong technicalsupportfortheeicient and precise detection ofbolls at the boll-opening stage in cotton yield prediction.

Keywords： cotton boll detection; YOLOv8n; deep learning

棉花作为全球纺织工业的重要原料，其产量预测是农业生产管理中的重要环节[]。（剩余16251字）