第45卷第1期2023年1月铁道学报JOURNALOFTHECHINARAILWAYSOCIETYVol.45No.1January2023文章编号:1001-8360(2023)01-0128-07基于三维重构的钢桥表面锈蚀提取方法研究倪有豪1,王浩1,茅建校1,贡照华2,惠卓1(1.东南大学混凝土及预应力混凝土结构教育部重点实验室,江苏南京211189;2.中国铁路上海局集团有限公司南京桥工段,江苏南京210011)摘要:为实现钢桥表面锈蚀的准确提取及可视化,提出基于三维重构的钢桥锈蚀提取方法。该方法利用无人机采集桥梁影像,基于倾斜摄影生成三维点云模型,通过点云模型进行RGB颜色分布直方图统计,获取模型初始聚类中心,运用K-means聚类算法结合最大类间方差法获取锈蚀最佳分割阈值,进而对锈蚀率及锈蚀等级进行评估,并结合点云模型实现锈蚀可视化。以某钢桁拱桥为背景,采用该方法提取锈蚀区域,基于视觉分析和定量分析进行精确度评价,验证方法的有效性。结果表明,该钢桥的表面锈蚀率1.3%~9.4%,锈蚀等级为3~4级;提出的钢桥锈蚀提取方法平均正确提取率可达90.1%,平均漏提取率、误提取率分别为9.9%、8.2%,可有效用于钢桥表面锈蚀提取。关键词:桥梁工程;锈蚀提取;倾斜摄影;点云模型;K-means聚类算法中图分类号:U216.41文献标志码:Adoi:10.3969/j.issn.1001-8360.2023.01.016StudyonExtractionMethodofSurfaceCorrosionofSteelBridgeBasedon3DReconstructionNIYouhao1,WANGHao1,MAOJianxiao1,GONGZhaohua2,XIZhuo1(1.KeyLaboratoryofC&PCStructuresofMinistryofEducation,SoutheastUniversity,Nanjing211189,China;2.NanjingBridgeDivision,ChinaRailwayShanghaiGroupCo.,Ltd.,Nanjing210011,China)Abstract:Inordertopreciselyextractandvisualizesteelbridgesurfacecorrosion,anextractionmethodofsteelbridgesurfacecorrosionwasproposedbasedon3Dreconstruction.Initially,basedonthebridgevideoacquiredbyunmannedaerialvehicle,athree-dimensionalpointcloudmodelofthebridgewasformedthroughtheobliquephotographytechnolo-gy,andtheanalysisofhistogramstatisticsofgrayvaluedistributionwasperformedinthreechannelsofRGBcolour.Subsequently,theinitialclusteringcenterwasobtained,whilethebestsegmentationthresholdwasobtainedbycombi-ningwithK-meansclusteringalgorithmandOtsu’smethod.Finally,thecorrosionrateandcorrosiongradewereevalua-ted,and...