三维点云初始配准方法的比较分析  

Comparison and analysis of initial registration methods of 3D point cloud

作　　者：王佳婧 王晓南 郑顺义[1,3] 朱锋博 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079 [2]南京中观自动化科技有限公司,武汉430079 [3]地球空间信息技术协同创新中心,武汉430079

出　　处：《测绘科学》2018年第2期16-23,共8页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41171357);湖北省科研条件与资源研究开发项目(2015BCE080)

摘　　要：针对多种多样的三维点云自动初始配准算法,该文依照配准过程中侧重点的不同将其分为基于局部特征描述、基于全局搜索策略和基于统计学概率3个算法类别,对其中具有代表性的FPFH、4PCS、NDT 3种算法进行了配准效率和数据适应性的比较分析。采用4组数据分别使用这3种算法进行配准实验,提出了算法中相关参数的设置方法,验证了它们用于最近点迭代精配准的适用性,并记录相应的配准耗时。实验结果表明,不同算法适用的数据不同,初始配准时可以依据数据的特点选取合适的方法。There are kinds of methods for automatic initial registration,which can be categorized into three major types by different emphasis：local feature description method,global search strategy,and statistical likelihood method.In this paper,three representative registration methods including FPFH,4 PCS and NDT algorithm were chosen to compare.Each algorithm were evaluated using four sets of experimental data;the setting method of related parameters in each algorithm were proposed;the efficiency and data adaptability applied to ICP precise registration were analyzed and the registration time was recorded.The results demonstrated that different methods are suitable for different data.

关 键 词：三维重建 点云配准 初始配准 算法比较 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]