ICP parallel registration of multiple point clouds with MPI

doi:10.13474/j.cnki.11-2246.2021.0083

Abstract

Abstract: Iterative closest point (ICP) algorithm is a classic algorithm for precise point cloud registration. In this paper, in order to solve the problem of time-consuming and low efficiency in ICP registration of multiple point clouds, the message passing interface (MPI) is proposed to perform batch parallel registration for multiple point clouds. First, the adjacent transformation matrix of two adjacent point clouds is solved in parallel, then the local transformation matrix of each point cloud in the current batch is calculated, and finally the global transformation matrix of each point cloud is obtained. 65 chip point clouds with a total size of over 40 million points in space are registered in batch parallel with DELL PowerEdge R730 server as the computing platform. The results show that batch processing of multiple point clouds using MPI can significantly accelerate the registration speed, and the acceleration ratio is 5.3 when the optimal number of processes is the kernel of the computer.

Key words: MPI, ICP, multiple point clouds, parallel registration, transformation matrix

CLC Number:

P237

CUI Jiawu, ZHOU Boyang, ZHANG Xingfu, CHAI Xianghao, FAN Shuchun. ICP parallel registration of multiple point clouds with MPI[J]. Bulletin of Surveying and Mapping, 2021, 0(3): 87-90.

References

[1] 杨必胜, 梁福逊, 黄荣刚. 三维激光扫描点云数据处理研究进展、挑战与趋势[J]. 测绘学报, 2017, 46(10):1509-1516.
[2] 姜如波. 基于三维激光扫描技术的建筑物模型重建[J]. 测绘通报, 2013(S1):80-83.
[3] 宋宏. 地面三维激光扫描测量技术及其应用分析[J]. 测绘技术装备, 2008, 10(2):40-43.
[4] 宇超群, 门葆红, 王鑫. 海量点云数据分布式并行处理技术综述[J]. 信息工程大学学报, 2018, 19(5):611-615.
[5] BEDKOWSKI J, BRATUS' R, PROCHASKA M, et al. Use of parallel computing in mass processing of laser data[J]. Archiwum Fotogrametrii, Kartografii i Teledetekcji, 2015, 27:45-59.
[6] CONNOR M, KUMAR P. Parallel construction of k-Nearest neighbor graphs for point clouds[C]//Proceedings of the Eurographics/IEEE VGTC Workshop on Volume Graphics.[S.l.]:IEEE, 2008:25-31.
[7] TAMAKI T, ABE M, RAYTCHEV B, et al. Softassign and EM-ICP on GPU[C]//Proceedings of the First International Conference on Networking and Computing.[S.l.]:IEEE, 2010:179-183.
[8] 荆锐, 赵旦谱, 台宪青. 基于GPU的实时三维点云数据配准研究[J]. 计算机工程, 2012, 38(23):198-202.
[9] 孙晓鹏, 颜士新. 一种新的点云并行Softassign配准算法:中国, CN104463826A[P]. 2015-03-25.
[10] 刘忠建.基于OpenCL的ICP点云并行配准算法[J]. 计算机应用与软件, 2016, 33(11):185-187.
[11] 贾志成, 张希晋, 陈雷, 等. 基于并行粒子群优化的三维点云配准算法[J]. 电视技术, 2016, 40(1):36-41.
[12] BŁASZCZAK-BAK W, JANOWSKI A, SROKOSZ P. High performance filtering for big datasets from Airborne Laser Scanning with CUDA technology[J]. Survey Review, 2018, 50(360):262-269.
[13] 郑汉之. 基于并行计算的LiDAR数据滤波方法研究[D]. 成都:西南交通大学, 2011.
[14] HU X Y, LI X K, ZHANG Y J. Fast filtering of LiDAR point cloud in urban areas based on scan line segmentation and GPU acceleration[J]. IEEE Geoscience and Remote Sensing Letters, 2013, 10(2):308-312.
[15] 王宗跃, 马洪超, 徐宏根, 等. 多核CPU的海量点云并行kNN算法[J]. 测绘科学技术学报, 2010, 27(1):46-49.
[16] GVNTHER C, KANZOK T, LINSEN L, et al. A GPGPU-based pipeline for accelerated rendering of point clouds[J]. Journal of WSCG, 2013, 21:153-162.
[17] BESL P J, MCKAY N D. A method for registration of 3-D shapes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(2):239-256.