基于HRM-YOLO的交通标志检测算法

打开文本图片集

DOI：10.16652/j.issn.1004-373x.2026.07.027引用格式：，，，等.基于HRM-YOLO的交通标志检测算法[J].现代电子技术，2026，49（7）：190-198.

关键词：交通标志检测；YOLOv1O；空洞卷积；注意力机制；损失函数；深度学习中图分类号：TN911.73-34；TP391.4 文献标识码：A 文章编号：1004-373X（2026）07-0190-09

Traffic sign detection algorithm based on HRM-YOLO

YANG Luxia1²， CUI Shuo1²， ZHANG Hongrui1²， MA Yongjie （1.College ofComputer Scienceand Technology，TaiyuanNormal University，JinzhongO3o619，China; 2.ShanxiProvincialKeyLaboratoryofIntellgentOptimizationComputingandBlockchainTechnology，nzhongO3069，a; 3.Collge ofPhysicsandElectronic Engineering，NorthwestNormal University，Lanzhou 73oo7o，China）

Abstract：Trafficsigndetectiontechnologyisanimportantaplicationinautonomousdrivingandinteligenttraffic management.However，miseddetectionsandfalsedetectionsarepronetooccurincomplexenvironments.Inviewoftheabove， this paper proposes an HRM-YOLO traffc sign detectionalgorithmbasedon improvedYOLOv10.Ahybrid poling enhancement moduleHPE-SPPFisdesignedinordertoimprovetheperceptionabilityofthemodelfordiferentscalesoftheobject.A receptivefieldinformationfusionmoduleRFIFisdesigned，andthemult-granularityinformationcaptureabilityisenhancedby theinteractionmechanismofcross-scalereceptivefields.AlightweightfeatureaggregationmoduleC2f-OMSAisdesignedto providecross-levelmulti-viewobjectclueswhileefectivelyretainingoriginalinformation.ThelossfunctionWise-IoU v3 is introducedtoreplacethelossfunction CIoUtoimprovetheaccuracyofboundingboxpositioning，soastoimprovetherobustness anddetectionaccuracyof themodel.Experimentsonthetraficsign datasetsCCTSDB2O21andTT10OKshowthat the precision，recall rate and mAP@0.5 of the HRM-YOLO are improved in comparison with those of the basic model，which verifies the effectiveness of the algorithm in practical applications.

Keywords：traffc sign detection; YOLOv1O; dilated convolution; attention mechanism; lossfunction; deep learning

0 引言

近年来，随着经济和科技的快速发展，城市化进程加快，车辆保有量持续增长，交通压力日益增大。（剩余14441字）