Point Cloud Registration Algorithm Based on Angle Constraint and Maximal Clique

Yaochang Tan; Junwei Yang; Kunyang Li; Lixin Tang

doi:10.3788/LOP242367

Laser & Optoelectronics Progress, Volume. 62, Issue 12, 1215012(2025)

Point Cloud Registration Algorithm Based on Angle Constraint and Maximal Clique

Yaochang Tan, Junwei Yang, Kunyang Li, and Lixin Tang^*

School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuhan 430074, Hubei , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Point cloud registration algorithm flowchart based on angle constraint and maximal clique

Download full size

Fig. 2. Schematic diagrams of the first-order spatial compatibility graph and its adjacency matrix

Download full size

Fig. 3. Schematic diagram of spatial three-point angle invariability constraint

Download full size

Fig. 4. Schematic diagrams of the second-order spatial compatibility graph and its adjacency matrix

Download full size

Table 1. Comparison of registration accuracy of different algorithms on KITTI dataset

View table

Table 1. Comparison of registration accuracy of different algorithms on KITTI dataset

Algorithm	FPFH			FCGF
Algorithm	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑
RANSAC	0.87	43.74	74.21	0.76	26.84	80.45
FGR	0.84	42.58	16.79	0.48	25.46	89.24
TEASER++	1.03	17.98	91.17	0.39	13.58	94.89
SC²-PCR	0.41	8.60	96.40	0.41	9.68	97.12
MAC	0.41	8.62	97.66	0.36	8.10	97.25
DGR	1.59	33.25	77.25	0.37	21.68	94.92
PointDSC	0.43	8.66	96.42	0.59	13.39	96.39
Proposed algorithm	0.39	8.57	98.56	0.34	8.06	97.66

Table 2. Comparison of time and memory usage of KITTI dataset
View table
Table 2. Comparison of time and memory usage of KITTI dataset
Algorithm FPFH FCGF
Memory /kB↓ Time /s↓ Memory /kB↓ Time /s↓
MAC 11104244 3.95 508608 2.97
Proposed algorithm 576444 1.56 419064 0.51

Table 3. Comparison of registration accuracy of different algorithms on 3DMatch dataset

View table

Table 3. Comparison of registration accuracy of different algorithms on 3DMatch dataset

Algorithm	FPFH			FCGF
Algorithm	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑
RANSAC	3.96	11.12	66.13	2.69	8.41	91.34
FGR	4.95	10.23	40.94	2.93	8.47	78.89
TEASER++	2.49	7.29	75.52	2.73	8.64	85.78
SC²-PCR	2.19	6.74	83.72	2.09	6.51	93.16
MAC	2.11	6.79	83.80	2.03	6.55	93.71
3DRegNet	3.75	9.63	27.33	2.78	8.14	77.81
DGR	2.44	7.53	33.22	2.34	7.06	88.76
PointDSC	2.18	6.69	73.02	2.10	6.54	91.89
Proposed algorithm	2.09	6.67	83.61	2.00	6.46	93.47

Table 4. Comparison of time and memory usage of 3DMatch dataset
View table
Table 4. Comparison of time and memory usage of 3DMatch dataset
Algorithm FPFH FCGF
Memory /kB↓ Time /s↓ Memory /kB↓ Time /s↓
MAC 5274144 4.18 6519116 4.06
Proposed algorithm 4066500 2.50 3007944 1.66

Table 5. Comparison of registration accuracy of different algorithms on 3DLoMatch dataset

View table

Table 5. Comparison of registration accuracy of different algorithms on 3DLoMatch dataset

Algorithm	FPFH			FCGF
Algorithm	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑	$E_{R} / (°)$ ↓	$E_{T} / c m$ ↓	$R_{R} / %$ ↑
RANSAC	10.58	15.12	0.57	6.91	15.63	10.35
TEASER++	4.37	10.96	35.15	4.11	12.87	46.78
SC²-PCR	4.03	10.46	38.59	3.81	10.51	58.73
MAC	4.08	10.69	41.32	3.75	10.57	59.85
DGR	5.04	13.56	19.87	4.19	10.86	43.78
PointDSC	4.07	10.43	20.35	3.89	10.53	56.18
Proposed algorithm	3.90	10.33	40.99	3.71	10.49	59.74

Table 6. Comparison of time and memory usage of 3DLoMatch dataset
View table
Table 6. Comparison of time and memory usage of 3DLoMatch dataset
Algorithm FPFH FCGF
Memory /kB↓ Time /s↓ Memory /kB↓ Time /s↓
MAC 7595672 4.14 9461572 4.18
Proposed algorithm 4654516 1.73 4052644 1.78

Table 7. 3DMatch dataset ablation experimental results
View table
Table 7. 3DMatch dataset ablation experimental results
Algorithm FPFH FCGF
$E_{R} / (°)$ ↓ $E_{T} / c m$ ↓ $R_{R} / %$ ↑ $E_{R} / (°)$ ↓ $E_{T} / c m$ ↓ $R_{R} / %$ ↑
MAC 2.11 6.79 83.80 2.03 6.55 93.71
Proposed algorithm 2.11 6.76 84.10 2.01 6.52 93.77

Table 8. 3DLoMatch dataset ablation experimental results
View table
Table 8. 3DLoMatch dataset ablation experimental results
Algorithm FPFH FCGF
$E_{R} / (°)$ ↓ $E_{T} / c m$ ↓ $R_{R} / %$ ↑ $E_{R} / (°)$ ↓ $E_{T} / c m$ ↓ $R_{R} / %$ ↑
MAC 4.08 10.69 41.32 3.75 10.57 59.85
Proposed algorithm 4.06 10.68 41.38 3.73 10.57 60.08

Table 9. Average maximal clique number ablation experimental results
View table
Table 9. Average maximal clique number ablation experimental results
Algorithm FPFH ↓ FCGF ↓
KITTI 3DMatch 3DLoMatch KITTI 3DMatch 3DLoMatch
MAC 171875 35292 32363 6182 135741 154872
Proposed algorithm 171860 35271 32348 6172 135726 154863

Tools

Get Citation

Copy Citation Text

Yaochang Tan, Junwei Yang, Kunyang Li, Lixin Tang. Point Cloud Registration Algorithm Based on Angle Constraint and Maximal Clique[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1215012

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Dec. 3, 2024

Accepted: Feb. 17, 2025

Published Online: Jun. 25, 2025

The Author Email: Lixin Tang (lixintang@mail.hust.edu.cn)

DOI:10.3788/LOP242367

CSTR:32186.14.LOP242367

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of registration accuracy of different algorithms on KITTI dataset

Table 1. Comparison of registration accuracy of different algorithms on KITTI dataset

Table 2. Comparison of time and memory usage of KITTI dataset

Table 2. Comparison of time and memory usage of KITTI dataset

Table 3. Comparison of registration accuracy of different algorithms on 3DMatch dataset

Table 3. Comparison of registration accuracy of different algorithms on 3DMatch dataset

Table 4. Comparison of time and memory usage of 3DMatch dataset

Table 4. Comparison of time and memory usage of 3DMatch dataset

Table 5. Comparison of registration accuracy of different algorithms on 3DLoMatch dataset

Table 5. Comparison of registration accuracy of different algorithms on 3DLoMatch dataset

Table 6. Comparison of time and memory usage of 3DLoMatch dataset

Table 6. Comparison of time and memory usage of 3DLoMatch dataset

Table 7. 3DMatch dataset ablation experimental results

Table 7. 3DMatch dataset ablation experimental results

Table 8. 3DLoMatch dataset ablation experimental results

Table 8. 3DLoMatch dataset ablation experimental results

Table 9. Average maximal clique number ablation experimental results

Table 9. Average maximal clique number ablation experimental results