Point Cloud Registration Algorithm Based on Canonical Correlation Analysis

Zhirong Tang; Mingzhe Liu; Yue Jiang; Feixiang Zhao; Chengqiang Zhao

doi:10.3788/CJL201946.0404006

Chinese Journal of Lasers, Volume. 46, Issue 4, 0404006(2019)

Point Cloud Registration Algorithm Based on Canonical Correlation Analysis

Zhirong Tang^1、*, Mingzhe Liu^1、*, Yue Jiang², Feixiang Zhao¹, and Chengqiang Zhao¹

¹ State Key Laboratory of Geohazard Prevention and Geoenvironment Protection, Chengdu University of Technology, Chengdu, Sichuan 610059, China

² School of Electrical Engineering and Information, Sichuan University, Chengdu, Sichuan 610065, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(14)

Fig. 1. Flow chart of algorithm

Download full size

Fig. 2. Initial state of point clouds. (a) Armadillo; (b) Cat

Download full size

Fig. 3. Registration effect of each algorithm. (a) CCA; (b) GA+ICP; (c) GA+Scale-ICP ;(d) CPD; (e) Go-ICP

Download full size

Fig. 4. Registration effects of Bunny point cloud. (a) 15% data loss; (b) 20% data loss; (c) 25% data loss

Download full size

Fig. 5. Registration effects of Dragon point cloud. (a) 15% data loss; (b) 20% data loss; (c) 25% data loss

Download full size

Fig. 6. Registration error and registration time of 5 algorithms. (a) Registration error of Bunny point cloud; (b) registration error of Dragon point cloud; (c) registration time of Bunny point cloud; (d) registration time of Dragon point cloud

Download full size

Fig. 7. Initial state of point clouds. (a) Elephant; (b) Horse

Download full size

Fig. 8. Affine registration effects of two algorithms. (a) CCA; (b) Scale-ICP

Download full size

Fig. 9. Two sets of physical pictures. (a) A; (b) B

Download full size

Fig. 10. Initial state of two sets of point clouds. (a) A point cloud; (b) B point cloud

Download full size

Fig. 11. Registration effects of five algorithms. (a) CCA; (b) CCA+ICP; (c) CCA+Scale-ICP; (d) CPD; (e) Go-ICP

Download full size

Table 1. Registration time and error of 5 algorithms
View table
Table 1. Registration time and error of 5 algorithms
Algorithm Time /s RMSE /mm
Armadillo Cat Armadillo Cat
CCA 3.7 1.1 2.0332 8.5744×10^-4
GA+ICP 223.8 16.4 2.0327 8.5718×10^-4
GA+Scale-ICP 7.2 1.9 2.0337 0.0016
CPD 215.9 32.4 3.4036 0.0169
Go-ICP 31.2 26.6 2.1488 0.0326

Table 2. RMSE and registration time of two registration algorithms
View table
Table 2. RMSE and registration time of two registration algorithms
Algorithm Time /s RMSE /mm
Elephant Horse Elephant Horse
CCA 2.6 4.2 0.0203 5.0824×10^-4
Scale-ICP 2.8 6.8 0.0369 0.0034

Table 3. Registration error and time of 5 algorithms

View table

Table 3. Registration error and time of 5 algorithms

Algorithm	Time /s		RMSE /mm
Algorithm	A pointcloud	B pointcloud	A pointcloud	B pointcloud
CCA	1.9	3.8	4.1680	0.4251
CAA+ICP	317.8	375.8	4.1447	0.3755
CAA+Scale-ICP	10.3	26.7	5.5035	3.2537
CPD	43.0	62.3	5.9309	1.3431
Go-ICP	27.6	27.7	4.9031	0.4706

Tools

Get Citation

Copy Citation Text

Zhirong Tang, Mingzhe Liu, Yue Jiang, Feixiang Zhao, Chengqiang Zhao. Point Cloud Registration Algorithm Based on Canonical Correlation Analysis[J]. Chinese Journal of Lasers, 2019, 46(4): 0404006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites