第40卷第3期计算机应用与软件Vol.40No．32023年3月ComputerApplicationsandSoftwareMar．2023一种新的三维点云兴趣点提取算法郭建华吕常魁(南京航空航天大学机电学院江苏南京210016)收稿日期:2020－06－26。国家自然科学基金项目(61671240)。郭建华，硕士生，主研领域:机器视觉及其工业应用，三维点云数据分析。吕常魁，副教授。摘要现有的三维点云兴趣点提取算法容易漏检和误检兴趣点，针对该问题，提出一种新的三维点云兴趣点提取算法。假设锥体为三维物体边角基元特征，根据各点与其k个近邻点的差向量集，构建突出度特征值描述点的局部锥度特征。基于点云突出度特征值的全局阈值得到初始兴趣点集，按照局部最大原则获取候选兴趣点集，依据每个候选兴趣点被重复选中的次数进行投票，获取最终兴趣点。在单位圆上模拟点云的突出度相关参数特征，检验了算法的鲁棒性。以人工标注统计确定的兴趣点作为真实值评估算法的性能，结果表明，该算法能准确提取到大部分真实兴趣点，整体性能优于传统算法。关键词点云兴趣点锥体突出度特征中图分类号TP391文献标志码ADOI:10．3969/j．issn．1000-386x．2023．03．038ANOVELALGOＲITHMFOＲINTEＲESTPOINTEXTＲACTIONOF3DPOINTCLOUDSGuoJianhuaLüChangkui(CollegeofMechanicalandElectricalEngineering，NanjingUniversityofAeronauticsandAstronautics，Nanjing210016，Jiangsu，China)AbstractTheexistinginterestpointextractionalgorithmsof3Dpointcloudsarepronetomisdetectionandomission．Tosolvethisproblem，weproposedanovelalgorithmforinterestpointextractionof3Dpointclouds．Undertheassumptionthatconewastheprimitivefeatureoftheedgesandcornersofa3Dobject，saliencedegreewasconstructedtodescribethelocalconefeatureoftheobjectbasedonthedifferencevectorssetcalculatedbyeachpointanditsknearestneighbors．Theinitialsetofinterestpointswasobtainedbasedontheglobalthresholdofthesaliencedegreeofpointclouddata．Thesetofcandidateinterestpointswasobtainedbasedonthelocalmaximumprinciple．Thefinalinterestpointswereselectedbyvotingaccordingtothenumberoftimesthateachcandidateinterestpointwasrepeatedlyselected．Theparametersrelatedtosaliencedegreeofthepointcloudsweresimulatedontheunitcircletotesttherobustnessofthealgorithm．Theperformanceoftheproposedalg...
