基于改进YOL0v11的籽棉异纤智能分选系统设计

打开文本图片集

关键词：籽棉异纤检测；智能分选系统；深度学习；注意力机制；多尺度特征融合中图分类号：S226.5；TP23 文献标识码：A 文章编号：2095-5553（2025）12-0162-09

Abstract：Toenhance thelevelofitellgence incotton processingand improve productquality，anintelligentforeign fiber sorting system for sed coton was designed based on an improved YOLOvl1 model.The system acquired images in real timeusinganindustrialcamera，andacombinationofanindustrialcomputerandareal-timecontrolboardwasemployed toprocess thedataanddrivehigh-speedelectromagneticvalves forforeignfiberremoval．Intermsofforeignfiber detection，gray-levelhistogramanalysiscombinedwiththresholdsegmentationwasfirstappliedtorapidlyidentifyforeign fibers with significantcolordiferences from sed cottn.However，since plastic film fragments exhibited small color diferencesfromseedcotton，andcottnhullsandstalkswereoftenconfusedwithdark-coloredforeignfibers，threshold segmentation frequentlymisclasifiedhullsandstalksandfailedtodistinguish plasticflmefectively.Toaddress these isues，an improved YOLOvll model was introduced torecognize plastic film，coton hull，and stalks.The improvementsincludedtheintegrationofaBRAatentionmechanismintothebackbonenetworktoenhancesmall-object detectioncapability，the adoptionof BiFPN in the neck structure for multi-scale feature fusion，and thereplacementof the originallossfunction with the DIoUlosstoacelerateconvergence.Experimentalresultsdemonstrated thattheimproved YOLOvll model achieved a detection precision of 94.2% ，representing an increase of 4.5% over the original model.Furthermore，multiple tests conducted under actual production conditions showed that the system achieved an average impurityremoval rateof 91.1% ，therebyverifying its stabilityand application value in real-world scenarios.

Keywors：seed cotton foreign fiber detection；intellgent sorting system；deep learning；atention mechanism；multi-scale feature fusion

0 引言

棉花对国家经济发展和人民生活极其重要，为我国农业现代化提供物质基础。（剩余13961字）