基于细颈结构的密集小目标车辆检测研究

——以LSGC-YOLOv8为例

打开文本图片集

中图分类号：TP391.4；TP181 文献标识码：A 文章编号：2096-4706（2025）11-0114-08

Research on Dense Small Target Vehicle Detection Based on Thin Neck Structure -Taking LSGC-YOLOv8 as an Example

ZHOU Zijian， LIU Haizhong （School ofMathematicsand Physics，Lanzhou Jiaotong University，Lanzhou 73oo70，China）

Abstract：The traditional monitoring systemrelieson manual identification of targets，which is inefficient and easy to misjudge.This paper improves and researches the series of YOLOalgorithms.Although theYOLO algorithm has been widely usedinmanyfelds，itsillfacestheechallengesinactualtraffcscenarios.Weatherchangesandhardwarelimitationscaeasily leadtoadeclieindataqualityThedetectionabilityofnarrowareatarget，occlusiontargetandincompletetargetisinsuient. Inadition，inomplexesdetaileatureetractionaslitatios.Tothisd，tispperproposanimproedodetat integrates lightweight designandstructuraloptimization.The model contains threeinnovations.Firstly，the backbone network introducesalightweightarchitecture，whichsignificantlyreduces thecomputationalcomplexityunderthepremiseofcontolable accuracylossSecondlythedetectionheadisreconstructedbyVoVGSCSPmodule tonhancetheabilityoffe-grainedfeature capture.ThirdlytheintegratedAtentionMechanismimproves thequalityoffeaturefusion.Theexperimentalresultsshowthat theimprovedmodelachievesthebalanceoptimizationofaccuracyandefciencyincomplextraffcscenarios，ndhasapication value.

Keywords：TargetDetection;intelligenttransportation; image recognition;DeepLearning;vehicledetection

0 引言

随着机动车的增加，实时反馈交通流量和监测追踪流动车辆在城市智慧交通系统中发挥着愈发重要的作用[1。（剩余10830字）