三维点云快速拼接方法研究  

Research on 3D point cloud quick registration

作　　者：黄潜 王泽勇[1] 李金龙[1] 孟倞安 高晓蓉[1] HUANG Qian;WANG Ze-yong;LI Jin-long;MENGJing-an;GAO Xiao-rong(School of Physical Science and Technology,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学物理科学与技术学院,成都610031

出　　处：《信息技术》2018年第7期77-80,共4页Information Technology

基　　金：国家自然基金资助项目(61471304)

摘　　要：三维点云数据的多视角拼接技术是三维测量领域的研究重点之一。文中设计实现了一种使用精简数据引导的迭代最近点(ICP)算法来实现对点云数据快速、高效的拼接。为了验证算法的有效性,本文整理实现了几种常见的快速拼接方案,使用bunny数据对它们进行了拼接对比实验。用作对比的方法包括使用K-d树加速的ICP算法,以及基于特征提取与匹配的拼接方法。实验结果表明,使用精简数据引导的拼接方法运算速度最快,效率最高,同时保持了与其它方法相近的拼接精度。Three-dimensional point cloud registration is one of the research focuses in three-dimensional measurement field. This paper designed an algorithm that uses simplified data to implement quick and efficient registration of point cloud data. In order to verify the effectiveness of the algorithm,the paper introduced several common quick register methods,using bunny data to perform register experiments.The methods used for comparison include the ICP algorithm with K-d tree to have an acceleration,and feature matching based register method. The experimental results show that the register method using the simplified data is the fastest and most efficient,and the register precision is similar to that of other methods.

关 键 词：三维点云 点云拼接 点云精简 K-D树 ICP 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]