基于低空热红外影像的城市轨道交通风险安全监护关键技术

打开文本图片集

中图分类号：P631.7;U293.6 文献标志码：A 文章编号：2095-1329（2025）02-0065-07

Abstract：Ispeoofailtansitsetyprotectioonsisiticalmeasuefoesuringurbanailtasitsaety.Howeveincoplex noctumalenvironents，traditioalmanalinspectiosandUAVvisiblelightechnoloiesstruggletoetispctionqts.o adresthesecallnges，tispaperproposesaintellgentmoitoringmetodintegratigteralinfraredimageryechnologywiththe YOLOv5 deep learing algorithm.The approach involves equippng unmanned aerial vehicles （UAVs）with thermal infrared cameras for nocturnalimageacquisitioninprotectionzones.TheYOLOv5 modelisemployedfortarget detectionofconstructionequipment intheralinfraredimagery，wilethermalanomalyfeatureextractio，dynamicbackground modeling，andamulti-ruleclasifier arecomprehensivelyappliedtoachieveautomaticidentificationand judgmentofconstruction equipment tatus.This studyselects theLiuchenRoadDepotofShanghaiMetroLine21astheexperimentalarea，conducting testresearchthroughroutedesign，dataset preparation，oeltngdpfoaceaatExpetalsultsemostratecuracipententti statusmonitoring，with theoverall model algorithmprecision reaching 94.33% andaKappacoefficientofO.89.Thismethod effectively fulfillsthedemandsofoctualailtransitispectios，enablingprciserecogitionofnightieconstructionequipmentsafety protectionzones，earlywamingofabnoralenvironmentaltemperaturechanges，anddiscriminationofequpmentoperatingstates.It provides a feasible technical solution for implementing nocturnal inspections of urban rail transit systems.

Keywords：rail tansitsfetyprotectionzoneralinfraredimageryUAV（unmanederialeicle）deepleaingiteliet monitoring

截至2024年底，上海轨道交通已形成896公里运营里程、日均客运量突破1027万人次的超大规模系统[，线路两侧50米（地下段）范围为轨道交通安全保护区，保护区内的违规施工、卸载、加载均会造成隧道变形、漏水甚至破损[2。（剩余9101字）