An Efficient Point Cloud Registration Algorithm Based on Principal Component Analysis

Yi Chen; Yong Wang; Jinlong Li; Dengzhi Liu; Xiaorong Gao; Yu Zhang

doi:10.3788/LOP222075

Laser & Optoelectronics Progress, Volume. 60, Issue 14, 1428008(2023)

An Efficient Point Cloud Registration Algorithm Based on Principal Component Analysis

Yi Chen, Yong Wang^*, Jinlong Li, Dengzhi Liu, Xiaorong Gao, and Yu Zhang

Author Affiliations

School of Physical Science and Technology, Southwest Jiaotong University, Chengdu 610031, Sichuan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. Stanford dataset. (a) bunny model; (b) armadillo model; (c) bunny registration initial pose; (d) armadillo registration initial pose

Download full size

Fig. 2. Registration of bunny. (a) SAC-IA algorithm; (b) 4PCS algorithm; (c) PCA algorithm; (d) improved PCA algorithm

Download full size

Fig. 3. Registration of armadillo. (a) SAC-IA algorithm; (b) 4PCS algorithm; (c) PCA algorithm; (d) improved PCA algorithm

Download full size

Fig. 4. Train components. (a) Standard rail model; (b) bolt point cloud after denoising; (c) brake cylinder; (d) gearbox

Download full size

Fig. 5. Registration results of proposed algorithm for train components. (a)-(d) Initial positions of point clouds of each component; (e)-(h) coarse registration results by improved PCA algorithm

Download full size

Fig. 6. Registration of rail model by 4PCS algorithm. (a) 4PCS coarse registration results; (b) ICP fine registration results

Download full size

Fig. 7. Point cloud data of sand-spreading pipe. (a) Point cloud data of sand pipe after discrete points removed; (b) initial pose

Download full size

Fig. 8. Point cloud registration effect of sand-spreading pipe. (a)-(d) Coarse registration results of SAC-IA, 4PCS, PCA, and improved PCA respectively; (e)-(h) results of Fig.8 (a)-(d) by ICP fine registration

Download full size

Fig. 9. Point cloud data of brake cylinder. (a) Brake cylinder 1; (b) two point clouds collected from perspective 1; (c) brake cylinder 2; (d) two point clouds collected from perspective 2

Download full size

Fig. 10. Registration of two point clouds from perspective of brake cylinder 1. (a)-(d) coarse registration results of SAC-IA, 4PCS, PCA, and improved PCA respectively

Download full size

Fig. 11. Registration of two point clouds from perspective of brake cylinder 2. (a)-(d) coarse registration results of SAC-IA, 4PCS, PCA, and improved PCA respectively

Download full size

Table 1. Experimental indicators of bunny registration
View table
Table 1. Experimental indicators of bunny registration
bunny $Δ R$ $Δ T$ R_RMSE t /s
SAC-IA 16.5739 0.0530608 0.00293281 1.29
4PCS 15.0242 0.0548299 0.00111164 1.659
PCA 0.000126358 7.57978×10^-7 5.27204×10^-7 23.873
Improved PCA 0.000126358 7.57978×10^-7 5.27204×10^-7 2.571

Table 2. Experimental indicators of armadillo registration
View table
Table 2. Experimental indicators of armadillo registration
Armadillo $Δ R$ $Δ T$ R_RMSE t /s
SAC-IA 4.71442 4.04072 1.5096 12.131
4PCS 1.09063 9.86902 1.95982 4.9
PCA 0.000122943 0.00036171 0.000413456 27.007
Improved PCA 0.000122943 0.00036171 0.000413456 2.751

Table 3. Experimental indicators of train components registration

View table

Table 3. Experimental indicators of train components registration

Train components	$Δ R$	$Δ T$	R_RMSE	t /s
Rail	0.000273208	0.00802271	0.00698371	3.511
Bolt	0.000105868	0.00193686	0.0016199	5.329
Brake cylinder	8.19623e-5	0.00218411	0.000987644	2.03
Gearbox	0.000122943	0.00036171	0.000413456	2.751

Table 4. Comparison of registration time of train components
View table
Table 4. Comparison of registration time of train components
t /s SAC-IA 4PCS PCA Improved PCA
Rail 98.837 1.791 11.066 3.511
Bolt 6.209 7.148 23.976 5.329
Brake cylinder 41.156 4.09 23.325 2.03
Gearbox 22.019 1.435 22.948 2.751

Table 5. Experimental indicators of sand-spreading pipe registration

View table

Table 5. Experimental indicators of sand-spreading pipe registration

Algorithm	$Δ R$	$Δ T$	R_RMSE	t /s
SAC-IA	3.95155	33.0466	1.63944	5.37
4PCS	1.20092	16.9889	1.28889	5.656
PCA	0.266965	2.00767	0.522078	23.934
Improved PCA	0.000109283	0.000448835	0.000605929	4.427
SAC-IA+ICP	0.194783	3.17572	0.409967	1.103
4PCS+ICP	0.649496	16.1011	0.653928	1.376
PCA+ICP	0.00464453	0.026082	0.30726	0.682
Imp-PCA+ICP	0.000341509	0.00201196	0.000995685	0.464

Table 6. Coarse registration time of brake cylinder
View table
Table 6. Coarse registration time of brake cylinder
t /s SAC-IA 4PCS PCA Improved PCA
Brake cylinder 1 27.434 1.722 23.393 4.939
Brake cylinder 2 26.911 48.525 96.748 1.34

Tools

Get Citation

Copy Citation Text

Yi Chen, Yong Wang, Jinlong Li, Dengzhi Liu, Xiaorong Gao, Yu Zhang. An Efficient Point Cloud Registration Algorithm Based on Principal Component Analysis[J]. Laser & Optoelectronics Progress, 2023, 60(14): 1428008

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites