基于法向特征直方图的点云配准算法

沈萦华; 李卓嘉; 杨成; 李昊勇

doi:10.3788/OPE.20152313.0592

您当前的位置：

首页 >

文章列表页 >

基于法向特征直方图的点云配准算法

更新时间：2020-08-13

- 基于法向特征直方图的点云配准算法
- Point cloud registration with normal feature histogram
- 光学精密工程 2015年23卷第10z期页码：591-598
- 作者机构：
  
  中国传媒大学信息工程学院, 北京 100024
- 作者简介：
- 基金信息：
  
  国家科技支撑计划资助课题(No.2014BAH10F02)()
- DOI：10.3788/OPE.20152313.0592
  中图分类号： TP242.6;TP391.41
- 收稿日期：2015-05-18，
  
  修回日期：2015-06-10，
  
  纸质出版日期：2015-11-14
- 稿件说明：
移动端阅览
沈萦华, 李卓嘉, 杨成等. 基于法向特征直方图的点云配准算法[J]. 光学精密工程, 2015,23(10z): 591-598

SHEN Ying-hua, LI Zhuo-jia, YANG Cheng etc. Point cloud registration with normal feature histogram[J]. Editorial Office of Optics and Precision Engineering, 2015,23(10z): 591-598
沈萦华, 李卓嘉, 杨成等. 基于法向特征直方图的点云配准算法[J]. 光学精密工程, 2015,23(10z): 591-598 DOI： 10.3788/OPE.20152313.0592.

SHEN Ying-hua, LI Zhuo-jia, YANG Cheng etc. Point cloud registration with normal feature histogram[J]. Editorial Office of Optics and Precision Engineering, 2015,23(10z): 591-598 DOI： 10.3788/OPE.20152313.0592.

摘要

为了提高点云配准速度并减少描述子的维度

本文首先提出了一种查找点云重叠区域的预处理办法。该方法使用基于八叉树的区域生长K-means聚类算法对点云进行分块

通过近似三角形查询点云的重叠区域。另外

在关键点描述中提出了基于点云特征直方图的低维度描述子邻域点积直方图(LDFH)算法。提出的预处理方法可以使初始配准的点云数据量减少10%~20%

去除了不必要的冗余运算过程。与快速点特征直方图(FPFH)描述子的点云配准算法相比

提出的邻域点积直方图算法将维度降低至24维

同时使描述速度提高了15%左右。利用本文算法对实际扫描获取到的点云数据进行配准时

可以在5 min内完成小于1 m

数据的准确配准。本文算法减少了配准耗时

降低了描述子维度

在实际的点云配准中有好的效果。

Abstract

To reduce the time cost of point cloud registration and to decrease the dimension of a descriptor

this paper proposes a pre-processing method to look up the overlap region of a point cloud. This method uses the region growing variant K-means clustering based on octree structure to block the point cloud

and then get the overlap region of point cloud by the triangle & point number decision formula. Moreover

a lower dimension descriptor named Local Dot Feature Histogram(LDFH) is also created based on a point cloud feature histogram in the key point description. The pre-processing method decreases about 10%-20% data volume of point cloud and removes some unnecessary redundant operation. As compared with the Fast Point Feature Histogram(FPFH) descriptor

the proposed LDFH algorithm just has 24-dimension and takes the computation time by 15%. When the methods proposed in this paper are used to register point cloud data in practice

the proposed method can complete small geometry solid registration for one cubic meter in less than five minute. The proposed algorithm achieves the goals of reducing the cost time

lowering descriptor dimension

and has a good effect in actual registration.

关键词

Keywords

references

BESL P J, MCKAY N D. Method for registration of 3-D shapes[C].Robotics-DL tentative. International Society for Optics and Photonics, 1992:586-606.

RUSU R B, COUSINS S. 3d is here:Point cloud library(pcl)[C].2011 IEEE International Conference on Robotics and Automation(ICRA), 2011:1-4.

MITRA N J, NGUYEN A, GUIBAS L. Estimating surface normals in noisy point cloud data[J]. International Journal of Computational Geometry & Applications, 2004, 14(4-5):261-276.

孙军华,谢萍,刘震,等.基于分层块状全局搜索的三维点云自动配准[J]. 光学精密工程,2013,21(1):174-179. SUN J H, XIE P, LIU ZH, et al.. Automatic 3D point cloud registration based on hierarchical block global search[J]. Opt. Precision Eng., 2013,21(1):174-179(in Chinese).

王欣,张明明,于晓,等. 应用改进迭代最近点方法的点云数据配准[J]. 光学精密工程,2012,20(9):2068-2077. WANG X, ZHANG M M, YU X, et al.. Point cloud registration based on improved iterative closest point method[J]. Opt. Precision Eng., 2012,20(9):2068-2077.(in Chinese)

孟禹. 基于采样球和ICP算法的点云配指标准方法研究[D].北京:清华大学硕士论文,2012. MENG Y. Registration of Point Clouds Basing on Sample-Sphere and ICP Algorithm[D].Beijing:Tsinghua University,2012.(in Chinese)

肖慧敏. 点云数据的配指标准算法[D].西安:西安电子科技大学硕士论文.2012. XIAO H M.A New Registration Method of Point Cloud Data[D]. Xi'an:Xidian University,2012.(in Chinese)

FOX S D, LYONS D M. An approach to stereo-point cloud registration using image homographies[C].IS&T/SPIE Electronic Imaging. International Society for Optics and Photonics, 2012:830108-6.

张晓,张爱武. 基于图像的点云初始配准[J]. 计算机工程与设计, 2014,35(10):3507-3512. ZHANG X, ZHANG A W. Initial registration of point clouds based on images[J]. Computer Engineering and Design, 2014,35(10):3507-3512.(in Chinese)

葛宝臻,彭博,田庆国. 基于曲率图的三维点云数据配准[J]. 天津大学学报,2013,46(2):174-180. GE B ZH, PENG B, TIAN Q G. Registration of three-dimensional point-cloud data based on curvature map[J].Journal of Tianjing University,2013,46(2):174-180.(in Chinese)

张学昌,习俊通,严隽琪. 基于扩展高斯球的点云数据与CAD模型配准[J]. 机械工程学报, 2007,43(6):142-148. ZHANG X CH, XI J T, YAN J Q. Registration of point cloud and CAD model based on extended Gaussian sphere[J]. Chinese Journal of Mechanical Engineering, 2007,43(6):142-148.(in Chinese)

黄戈,李晓峰. 基于CEGI和Fourier变换的全自动点云配准算法[J]. 四川大学学报,2014,46(5):104-109. HUANG G, LI X F. Fully automatic point cloud registration algorithm based on CEGI and Fourier transformation[J]. Journal of Sichuan University, 2014,46(5):104-109.(in Chinese)

PRICE M, GREEN J, DICKENS J. Point-cloud registration using 3D shape contexts[C].Robotics and Mechatronics Conference of South Africa(ROBOMECH), 2012:1-5.

XIAO J, ADLER B, ZHANG H. 3D point cloud registration based on planar surfaces[C].2012 IEEE Conference on Multisensor Fusion and Integration for Intelligent Systems,2012:40-45.

陶海跻,达飞鹏. 一种基于法向量的点云自动配准方法[J]. 中国激光, 2013,40(8):0809001-0809001-6. TAO H J, DA F P. Automatic registration algorithm for the point clouds based on the normal vector[J]. Chinese Journal of Lasers, 2013,40(8):0809001-6.(in Chinese)

陆军,彭仲涛,董东来,等.点云FPFH特征提取优化配准算法[J]. 新型工业化,2014,7:75-81. LU J, PENG ZH T, DONG D L, et al.. The registration algorithm of point cloud based on optimal extraction of FPFH feature[J]. The Journal of New Industrialization, 2014,7:75-81.(in Chinese)

孙经先. 关于点集拓扑学中的一个定理[J]. 系统科学与数学,1987,7(2):148-150. SUN J X. A thedrem in point set topology[J]. Journal of System Science and Math Science,1987,7(2):148-150.(in Chinese)

张大中. 点集拓扑的40个问题[J]. 辽宁大学学报,2004,31(4):314-315. ZHANG D ZH. Forty Problems of set topology[J]. Journal of Liaoning University, 2004,31(4):314-315.(in Chinese)

浏览量

378

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

LightDiffu-DCE：基于光照强度扩散的低光照图像增强

多级图特征融合引导相机位姿回归

复杂空心涡轮叶片点云快速配准

基于机载的红外动态目标视频实时超分辨率重建

基于知识蒸馏的Transformer视觉跟踪器