Efficient registration algorithm for models and indoor and outdoor point clouds

Fig. 5. Registration results of Dragon model with different precision registration algorithms. （a） Initial pose of point cloud；（b） GICP algorithm；（c） NDT algorithm^［20］；（d） Improved ICP^［21］；（e） Proposed precise registration algorithm.

Download full size

View in Article

Fig. 6. Registration results of small scale point clouds with noise in different registration algorithms. （a），（b） Initial pose of point cloud；（c），（d） Ref.［20］ algorithm；（e），（f） Ref.［21］ algorithm；（g），（h） Proposed point cloud registration algorithm.

Download full size

View in Article

Fig. 7. Registration results of Park point cloud. （a） Initial pose of point cloud；（b） Ref.［20］ algorithm；（c） Ref.［21］ algorithm；（d） Proposed point cloud registration algorithm.

Download full size

View in Article

Fig. 8. Registration results of Residence point cloud. （a） Initial pose of point cloud；（b） Ref.［20］ algorithm；（c） Ref.［21］ algorithm；（d） Proposed point cloud registration algorithm.

Download full size

View in Article

Fig. 9. Registration results of Heritage Building point cloud. （a） Initial pose of point cloud；（b） Ref.［20］ algorithm；（c） Ref.［21］ algorithm；（d） Proposed point cloud registration algorithm.

Download full size

View in Article

Table 1. Registration time and registration error of different precise registration algorithms
View table
View in Article
Table 1. Registration time and registration error of different precise registration algorithms
Registration algorithm Registration time/s Registration error/m
GICP 7.049 2.242 6×10^-5
NDT^［20］ 0.941 5.318 0×10^-5
Improved ICP^［21］ 10.550 5.761 8×10^-6
Proposed algorithm 0.292 3.545 7×10^-8

Table 2. Registration time and registration error of 0° and 24° dragon model with noise in different registration algorithms

View table

View in Article

Table 2. Registration time and registration error of 0° and 24° dragon model with noise in different registration algorithms

Registration method for 0° and 24° dragon model		Coarse registration	Precise algorithm	Coarse align error/m	Precise align error/m	Total time/s
Ref.［20］	Method	NDT	VGICP	9.783 0×10^-5	5.965 0×10^-5	0.232
Ref.［20］	Time/s	0.154	0.078	9.783 0×10^-5	5.965 0×10^-5	0.232
Ref.［21］	Method	ISS+FPFH+SAC-IA	Improved ICP	4.078 7×10^-4	6.438 8×10^-5	0.565
Ref.［21］	Time/s	0.065	0.500	4.078 7×10^-4	6.438 8×10^-5	0.565
Proposed algorithm	Method	ISS+FPFH+RANSAC	VGICP-CUDA	2.662 2×10^-4	5.457 9×10^-5	0.118
Proposed algorithm	Time/s	0.068	0.050	2.662 2×10^-4	5.457 9×10^-5	0.118

Table 3. Registration time and registration error of indoor point cloud with noise in different registration algorithms

View table

View in Article

Table 3. Registration time and registration error of indoor point cloud with noise in different registration algorithms

Registration method for indoor point cloud		Coarse registration	Precise algorithm	Coarse align error/m	Precise align error/m	Total time/s
Ref.［20］	Method	NDT	VGICP	0.003 14	0.003 63	1.740
Ref.［20］	Time/s	1.527	0.213	0.003 14	0.003 63	1.740
Ref.［21］	Method	ISS+FPFH+SAC-IA	Improved ICP	0.037 40	0.003 16	1.444
Ref.［21］	Time/s	0.185	1.259	0.037 40	0.003 16	1.444
Proposed algorithm	Method	ISS+FPFH+RANSAC	VGICP-CUDA	0.014 78	9.004 9×10^-4	0.306
Proposed algorithm	Time/s	0.196	0.110	0.014 78	9.004 9×10^-4	0.306

Table 4. Registration time and registration error of Park point cloud with different registration algorithms

View table

View in Article

Table 4. Registration time and registration error of Park point cloud with different registration algorithms

Registration method for park point cloud		Coarse registration	Precise algorithm	Coarse align error/m	Precise align error/m	Total time/s
Ref.［20］	Method	NDT	VGICP	18.928 20	19.262 70	40.003
Ref.［20］	Time/s	36.091	3.912	18.928 20	19.262 70	40.003
Ref.［21］	Method	ISS+FPFH+SAC-IA	Improved ICP	0.091 896	1.346 4×10^-6	111.546
Ref.［21］	Time/s	9.311	102.235	0.091 896	1.346 4×10^-6	111.546
Proposed algorithm	Method	ISS+FPFH+RANSAC	VGICP-CUDA	0.000 394	1.327 4×10^-6	9.846
Proposed algorithm	Time/s	7.013 2	2.833	0.000 394	1.327 4×10^-6	9.846

Table 5. Registration time and registration error of Residence point cloud with different registration algorithms

View table

View in Article

Table 5. Registration time and registration error of Residence point cloud with different registration algorithms

Registration method for residence point cloud		Coarse registration	Precise algorithm	Coarse align error/m	Precise align error/m	Total time/s
Ref.［20］	Method	NDT	VGICP	99.573 30	99.144 70	23.156
Ref.［20］	Time/s	20.332	2.824	99.573 30	99.144 70	23.156
Ref.［21］	Method	ISS+FPFH+SAC-IA	Improved ICP	9.597 27	5.212 60	118.767
Ref.［21］	Time/s	9.570	109.197	9.597 27	5.212 60	118.767
Proposed algorithm	Method	ISS+FPFH+RANSAC	VGICP-CUDA	8.291 29	5.707 04	6.517
Proposed algorithm	Time/s	5.342	1.175	8.291 29	5.707 04	6.517

Table 6. Registration time and registration error of Heritage Building point cloud with different registration algorithms

View table

View in Article

Table 6. Registration time and registration error of Heritage Building point cloud with different registration algorithms

Registration method for heritage point cloud		Coarse registration	Precise algorithm	Coarse align error/m	Precise align error/m	Total time/s
Ref.［20］	Method	NDT	VGICP	152.652 0	156.444 0	5.528
Ref.［20］	Time/s	2.077	3.451	152.652 0	156.444 0	5.528
Ref.［21］	Method	ISS+FPFH+SAC-IA	Improved ICP	54.491 5	55.632 7	45.646
Ref.［21］	Time/s	0.484	45.162	54.491 5	55.632 7	45.646
Proposed algorithm	Method	ISS+FPFH+RANSAC	VGICP-CUDA	146.128 0	156.496 0	0.648
Proposed algorithm	Time/s	0.296	0.352	146.128 0	156.496 0	0.648

Tools

Get Citation

Copy Citation Text

Junjie LI, Chen LEI, Weicheng LI, Xiaohui YU, Yuhan YANG, Wenli ZHU. Efficient registration algorithm for models and indoor and outdoor point clouds[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(3): 493

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Jun. 28, 2024

Accepted: --

Published Online: Apr. 27, 2025

The Author Email: Wenli ZHU (zwl829@126.com)

DOI:10.37188/CJLCD.2024-0184

Topics

Table 1. Registration time and registration error of different precise registration algorithms

Table 1. Registration time and registration error of different precise registration algorithms

Table 2. Registration time and registration error of 0° and 24° dragon model with noise in different registration algorithms

Table 2. Registration time and registration error of 0° and 24° dragon model with noise in different registration algorithms

Table 3. Registration time and registration error of indoor point cloud with noise in different registration algorithms

Table 3. Registration time and registration error of indoor point cloud with noise in different registration algorithms

Table 4. Registration time and registration error of Park point cloud with different registration algorithms

Table 4. Registration time and registration error of Park point cloud with different registration algorithms

Table 5. Registration time and registration error of Residence point cloud with different registration algorithms

Table 5. Registration time and registration error of Residence point cloud with different registration algorithms

Table 6. Registration time and registration error of Heritage Building point cloud with different registration algorithms

Table 6. Registration time and registration error of Heritage Building point cloud with different registration algorithms