基于YOLOv11的电厂工作服高精度检测算法改进

打开文本图片集

中图分类号：TP391.4 文献标识码：A

文章编号：1006-8228（2026）02-24-04

Abstract：Ahigh-precisiondetectionalgorithmforpowerplantworkclothesisproposedbasedonYOLOv11.Firstlyaglobal-local dual-streambranchatentionmechanismnamedSBCFormerisembeddedintothebackbonenetwork，whichefectivelyreducesthe numberofmodelparameterswhileenhancingtheglobalinformationagregationcapabiltySecondly，asmallobjectfeaturefusion modulecaledSDConvisaddedtoeicientlylearnthefeaturerepresentationsofsmallojectsunderdiferentresolutionsTidly theWIoUlossfunctionisadoptedtoreducetheinterferenceoflow-qualityanchorboxesonthealgorithmauracyFinalythe effectivenessoftheproposedalgorithmisverifiedonaself-builtpowerplantworkclothesdtaset.Experimentalresultsshowthat compared with the original algorithm，the mAP5O is increased by 4.8% ，theRecall（R）isimproved by 7.6% ，andtheFPS reaches 182.Moreovervisualizationomparisonresultsmonstratethatteproposedalgoritcanefectivelyidentifytargetsindfert scenarios.Thismethodnotonlyensureseficientandaccuratedetectionofworkclothesbutalsomeestherequirementofrealtime detection，showing strong robustness and good application prospects.

Keywords：PowerPlant;Work Clothes Detection;YOLOv11;Transformer;WIoU

0引言

电厂的生产环境危险性极高，为保障工人安全，施工人员规范佩戴防护装备至关重要，可减少意外事故带来的伤害。（剩余4934字）