第60卷第2期/2023年1月/激光与光电子学进展0211001-1研究论文提取桥梁点云的组合滤波算法谷帆1，张长伦1*，郭志光2，王恒友1，何强1，安彤11北京建筑大学北京未来城市设计高精尖创新中心，北京100044；2中国建筑土木建设有限公司，北京100071摘要在土木桥梁施工场景中，点云数据包含大量植被、地面和桥梁建筑点云。现有滤波算法对桥梁建筑的提取完整度仍存在一些问题，因此提出一种基于组合滤波的桥梁点云提取算法。首先，该算法根据植被点云分布发散这一特征运用分散度法对植被进行粗滤除；其次，基于半径滤波与充分利用颜色与高程特征思想进行半径滤波算法的改进，对剩余植被点云进行精细滤除；最后，利用法向滤波模型对地面点云进行滤除。实验结果表明，所提算法提取桥梁点云的完整率为99.3%，误差为0.73%，对比现有滤波算法，可更完整地提取桥梁点云，准确率更高。关键词成像系统；桥梁建筑提取；分散度；半径滤波；法向滤波中图分类号P2文献标志码ADOI：10.3788/LOP212709CombinedFilteringAlgorithmforExtractingBridgePointCloudGuFan1,ZhangChanglun1*,GuoZhiguang2,WangHengyou1,HeQiang1,AnTong11BeijingAdvancedInnovationCenterforFutureUrbanDesign,BeijingUniversityofCivilEngineeringandArchitecture,Beijing100044,China;2ChinaConstructionCivilEngineeringCo.,Ltd.,Beijing100071,ChinaAbstractThepointclouddatainthecivilbridgeconstructionsceneincludesalargenumberofvegetation,ground,andbridgeconstructionpointclouds.Theextractionintegrityofbridgebuildingsisstillanissuefortheexistingfilteringalgorithms.Thisstudyproposesabridgepointcloudextractionalgorithmbasedoncombinedfiltering.First,theproposedalgorithmappliesthedispersionmethodtocoarselyfilterthevegetationbythefeatureofdispersionofvegetationpointclouddistribution.Second,theradiusfilteringalgorithmisimprovedtofinelyfiltertheresidualvegetationpointcloudsbasedontheideaofradiusfilteringandmakingfulluseofcolorandelevationfeatures.Finally,thegroundpointcloudisfilteredusingthenormalfilteringmethod.Theexperimentalresultsdemonstratethattheproposedalgorithmforextractingbridgepointcloudshasa99.3%integrityrateand0.73%errorrate.Whencomparedtoexistingfilteringalgorithms,theproposedalgorithmextractsthebridgepo...
