0228008-1第60卷第2期/2023年1月/激光与光电子学进展研究论文基于邻域曲率改进的迭代最近点激光雷达目标点云配准李艳红1，2*，闫建国2，王晓燕31咸阳师范学院物理与电子工程学院，陕西咸阳712000；2西北工业大学自动化学院，陕西西安710072；3西安建筑科技大学机电工程学院，陕西西安710055摘要为解决激光雷达目标点云配准技术中精确配准步骤中所存在的匹配速度慢和匹配误差大的问题，提出了一种基于邻域曲率改进的迭代最近点（ICP）精准化匹配算法。初始配准采用传统的主成分贴合法，给精确配准找到一个较好的初始位置，精配准采用基于领域曲率改进的ICP算法。以斯坦福兔子和场景点云作为实验研究对象，配准结果和数值分析共同表明，基于邻域曲率改进的ICP算法在点云配准中的可行性，且与其他算法相比，所提算法的配准速度更快、匹配精度更高，为三维数据重建和目标识别技术提供一种更高效的新方法。关键词遥感；激光雷达；邻域曲率；精确配准；迭代最近点算法；点云数据重建中图分类号TP391.9文献标志码ADOI：10.3788/LOP212521LidarTargetPointCloudAlignmentBasedonImprovedNeighborhoodCurvaturewithIterationClosestPointAlgorithmLiYanhong1,2*,YanJianguo2,WangXiaoyan31SchoolofPhysics&ElectronicEngineering,XianyangNormalUniversity,Xianyang712000,Shaanxi,China;2SchoolofAutomation,NorthwesternPolytechnicalUniversity,Xi’an710072,Shaanxi,China;3SchoolofElectricalandMechanicalEngineering,Xi’anUniversityofArchitectureandTechnology,Xi’an710055,Shaanxi,ChinaAbstractTosolvetheproblemsofslowmatchingspeedandlargematchingerrorintheprecisealignmentstepoflidartargetpointcloudalignmenttechnology,aniterationclosestpoint(ICP)precisionmatchingalgorithmbasedonneighborhoodcurvatureimprovementisproposed.Theregistrationprovidesagoodinitialposition;theneighborhoodcurvatureisintroducedintothetraditionalICPalgorithmtoachievethefineregistration.PerformregistrationandnumericalanalysisexperimentsontheStanfordBunnyandthescenepointcloud.TheexperimentalresultsdemonstratethattheimprovedICPalgorithmbasedontheneighborhoodcurvaturecanefficientlyperformthepointcloudalignment,andcomparedwithotheralgorithms,thealignmentspeedoftheproposedalgorithmisbetterthanthealignmentmatchinga...
