Cross-Source Point Cloud Registration Algorithm Based on Angle Constraint

Xiangxin Yan; Zheng Jiang; Bin Liu

doi:10.3788/LOP230478

Laser & Optoelectronics Progress, Volume. 60, Issue 22, 2215004(2023)

Cross-Source Point Cloud Registration Algorithm Based on Angle Constraint

Xiangxin Yan¹, Zheng Jiang^1,2、*, and Bin Liu^1,2

¹Engineering Research Center for Metallurgical Automation and Measurement Technology of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081, Hubei , China

²School of Information Science and Engineering, Wuhan University of Science and Technology, Wuhan 430081, Hubei , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(15)

Fig. 1. Examples of cross-source point clouds coming from different sensors. (a) Kinect point cloud; (b) SFM point cloud

Download full size

Fig. 2. Influence region of the FPFH algorithm

Download full size

Fig. 3. Matching point pairs generated by two different weight coefficients. (a) Matching point pair produced by the original FPFH; (b) matching point pair produced by the improved FPFH

Download full size

Fig. 4. Comparison of two and three sets of matching point pairs. (a) Both sets of matching point pairs are correct; (b) outliers can be detected

Download full size

Fig. 5. Undirected graph

Download full size

Fig. 6. Compatibility triangle generation

Download full size

Fig. 7. Cross-source point cloud data

Download full size

Fig. 8. Registration effect of each algorithm with the same scale

Download full size

Fig. 9. Registration effect of each algorithm with scale factor s=0.5 (s=2)

Download full size

Table 1. Cross-source point cloud data information

View table

Table 1. Cross-source point cloud data information

Point cloud	Image to be registered	Number of points
chair	chair_kinect	10849
chair	chair_sfm	41650
dustbin	dustbin_kinect	25396
dustbin	dustbin_sfm	66795
lab1	lab1_kinect	15031
lab1	lab1_sfm	65470
lab2	lab2_kinect	12863
lab2	lab2_sfm	87808
lab3	lab3_kinect	18669
lab3	lab3_sfm	189946
corner	corner_kinect	70819
corner	corner_sfm	19474

Table 2. RMSE of each algorithm with the same scale

View table

Table 2. RMSE of each algorithm with the same scale

Point cloud	ICP	CPD	K4PCS	SDRSAC	SAC-IA+NDT	Proposed algorithm
chair	0.05644	0.05313	0.04074	0.02688	0.05434	0.02662
dustbin	0.05690	0.03142	0.03947	0.03066	0.04719	0.02888
lab1	0.04480	0.02381	0.02563	0.02255	0.02914	0.02175
lab2	0.05082	0.05315	0.03130	0.03168	0.04090	0.02998
lab3	0.04208	0.03100	0.04123	0.02653	0.04794	0.02720
corner	0.05858	0.05817	0.05357	0.04067	0.05761	0.04110

Table 3. Time of each algorithm with the same scale

View table

Table 3. Time of each algorithm with the same scale

Point cloud	ICP	CPD	K4PCS	SDRSAC	SAC-IA+NDT	Proposed algorithm
chair	0.11	4.26	0.97	126.69	16.40	0.68
dustbin	0.089	3.61	0.61	313.06	10.11	0.58
lab1	0.17	13.08	1.94	50.02	28.21	2.86
lab2	0.14	12.83	0.47	81.32	25.96	1.65
lab3	0.41	32.29	0.92	182.62	51.52	2.66
corner	0.23	17.88	0.32	184.14	31.14	8.68

Table 4. Scale factor of each algorithm with different scales

View table

Table 4. Scale factor of each algorithm with different scales

Point cloud	Scale	Scale-ICP	CPD	Proposed algorithm
chair	0.5	0.4439	0.5767	0.5103
chair	2	1.7754	2.3068	2.0629
dustbin	0.5	0.4798	0.5180	0.5171
dustbin	2	1.9193	2.0833	2.0685
lab1	0.5	0.4526	0.4759	0.4960
lab1	2	1.8103	1.9037	2.0353
lab2	0.5	0.4676	0.5988	0.5206
lab2	2	1.8702	2.3878	2.0687
lab3	0.5	0.4894	0.5263	0.4922
lab3	2	1.9576	1.9601	1.9932
corner	0.5	0.4848	0.5618	0.5046
corner	2	1.9393	2.2467	1.9150

Table 5. RMSE of each algorithm with different scales

View table

Table 5. RMSE of each algorithm with different scales

Point cloud	Scale	Scale-ICP	CPD	Proposed algorithm
chair	0.5	0.03234	0.05719	0.02708
chair	2	0.03234	0.05719	0.02725
dustbin	0.5	0.03501	0.03090	0.02900
dustbin	2	0.03501	0.03090	0.02900
lab1	0.5	0.02580	0.02392	0.02229
lab1	2	0.02580	0.02392	0.02358
lab2	0.5	0.03289	0.05310	0.03043
lab2	2	0.03289	0.05310	0.03031
lab3	0.5	0.02748	0.03112	0.02718
lab3	2	0.02748	0.03112	0.02715
corner	0.5	0.04861	0.06131	0.04038
corner	2	0.04861	0.06131	0.04070

Table 6. Time of each algorithm with different scales unit: s

View table

Table 6. Time of each algorithm with different scales unit: s

Point cloud	Scale	Scale-ICP	CPD	Proposed algorithm
chair	0.5	1.96	21.01	1.38
chair	2	2.01	20.93	1.13
dustbin	0.5	0.79	10.01	0.51
dustbin	2	0.66	9.95	0.53
lab1	0.5	1.56	64.29	2.89
lab1	2	1.54	63.73	2.34
lab2	0.5	1.67	75.18	1.73
lab2	2	1.41	74.63	1.46
lab3	0.5	1.61	98.78	2.32
lab3	2	1.66	98.82	2.60
corner	0.5	1.38	81.23	8.96
corner	2	1.37	79.92	8.82

Tools

Get Citation

Copy Citation Text

Xiangxin Yan, Zheng Jiang, Bin Liu. Cross-Source Point Cloud Registration Algorithm Based on Angle Constraint[J]. Laser & Optoelectronics Progress, 2023, 60(22): 2215004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Jan. 9, 2023

Accepted: Mar. 1, 2023

Published Online: Nov. 6, 2023

The Author Email: Jiang Zheng (zjiangmail@126.com)

DOI:10.3788/LOP230478

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology