Researching | Accurate and Fast Primitive Detection Method for 3D Point Cloud Data

Laser & Optoelectronics Progress, Volume. 61, Issue 4, 0415006(2024)

Accurate and Fast Primitive Detection Method for 3D Point Cloud Data

Min Shi¹, Shaoqing Zhou¹, Suqing Wang^1、*, and Dengming Zhu^2,3、**

Author Affiliations

¹School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China

²Taicang Institute of Information Technology, Taicang215400, Jiangsu, China

³Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China

show less

Abstract Get PDF(in Chinese)

Figures&Tables (19)

References (28)

Paper Information

Figures & Tables(19)

Fig. 1. The $l_{M D P}$ of the same size plane (the black solid line in the figure is the projection of the plane block on the plane where vector N and F are located) at different curvatures, as well as the angle between vector F and plane normal N, the smaller the curvature from left to right, the smaller the $l_{M D P}$ , the greater the angle

Download full size

Fig. 2. Overall flowchart

Download full size

Fig. 3. The octree structure storing point cloud

Download full size

Fig. 4. Void schematic, the colored areas in the figure are the identified surface blocks

Download full size

Fig. 5. Mechanical part model. (a) (b) (c) The mesh model of three mechanical parts; (d) (e) (f) point cloud model obtained from Poisson sampling of mesh model

Download full size

Fig. 6. Two models in the ABC dataset.(a) (b) The mesh model; (c) (d) the point cloud model

Download full size

Fig. 7. Plane detection optimization. (a) (b) (c) Before optimization; (d) (e) (f) after optimization

Download full size

Fig. 8. Cylindrical recognition results of each method on the BASE. (a) (c) (e) The render result of the recognized cylinder on the point cloud; (b) (d) (f) the color-distinguished cylinder recognition result and enlarged view of the dotted area

Download full size

Fig. 9. Cylindrical protrusions, red is the protrusion part

Download full size

Fig. 10. Grooves, as shown in red in the figure

Download full size

Fig. 11. Cylindrical recognition results of each method on the BLOCK. (a) (d) (g) The render results of the extracted cylinder; (b) (e) (h) the render results of the recognized cylinder; (c) (f) (i) the color-distinguished cylindrical face recognition results

Download full size

Fig. 12. Cylindrical recognition results of each method on the MOOV. (a)(c)(e) The render results of the recognized cylindrical face on point cloud; (b)(d)(f) the color-distinguished cylindrical face recognition results, and enlarged view of the dotted area

Download full size

Fig. 13. Cylindrical recognition results of each method on the ABC_03. (a) (d) (g) The recognized cylindrical faces on point cloud; (b) (e) (h) the color-distinguished cylindrical face front recognition results; (c) (f) (i) back

Download full size

Fig. 14. Cylindrical recognition results of each method on the ABC_27. (a) (d) (g) The recognized cylindrical faces on the full point cloud; (b) (e) (h) the color-distinguished cylindrical face front recognition results; (c) (f) (i) back

Download full size

Fig. 15. The recognition results of proposed method and eRANSAC on the enlarged cylindrical surface. (a) Results identified by the eRANSAC method; (b) results identified by proposed method

Download full size

Table 1. Data parameters and characteristics
View table
Table 1. Data parameters and characteristics
Data Points Planes Cylinders Features
BASE 1477125 35 42 Concentric
BLOCK 737843 14 13 Different orientations
MOOV 999984 22 14 Multi-cylindrical surface combinations
ABC_03 73211 6 8 Crossover
ABC_27 293098 22 20 Incomplete

Table 2. Parameter settings for each method
View table
Table 2. Parameter settings for each method
Method Threshold Value
CbCD α 3
θ /（°） 75
φ /（°） 15
eRANSAC ε 0.1
β 0.9
α /（°） 25
τ 300
ρ 0.01
Proposed method ε 0.0001
β cos15°
γ 50
K 50
η 0.95
θ /（°） 75
σ /（°） 5
σ_c /（°） 25

Table 3. Performance of different methods on each data

Table 3. Performance of different methods on each data

Dataset	Method	F1	$N_{m i s s}$	$N_{M}$	$D_{A R D}$ /mm	$D_{M R D}$ /mm	$D_{S C}$ /mm	T /ms
BASE	CbCD	0.8333	12	0	0.0216	0.0424	0.0275	18439
	eRANSAC	0.6667	20	2	0.0364	0.1317	0.0818	2258
	Proposed method	1.0000	0	0	0.0169	0.0298	0.0169	2681
BLOCK	CbCD	1.0000	0	0	0.0068	0.0208	0.0067	8906
	eRANSAC	0.9286	0	2	0.0118	0.0376	0.0118	1003
	Proposed method	0.9286	0	2	0.0043	0.0229	0.0043	8149
MOOV	CbCD	0.6667	7	0	0.0279	0.1042	0.066	15528
	eRANSAC	0.8235	0	6	0.0412	0.1716	0.0412	1572
	Proposed method	0.9333	0	2	0.0161	0.052	0.0161	5899
ABC_03	CbCD	0.7692	3	0	0.0372	0.0485	0.0372	957
	eRANSAC	0.8889	0	2	0.0447	0.0747	0.0447	208
	Proposed method	0.8421	0	3	0.0257	0.0565	0.0257	1071
ABC_27	CbCD	0.5714	6	3	0.0061	0.0079	0.0064	11572
	eRANSAC	0.9231	0	2	0.1038	0.6982	0.1038	640
	Proposed method	0.9600	0	1	0.0146	0.0074	0.0146	3870

Table 4. The influence of different parameter settings on recognition accuracy

Table 4. The influence of different parameter settings on recognition accuracy

Dateset	β	γ	$D_{A R D}$ /mm	$D_{M R D}$ /mm	$N_{m i s s}$
ABC_03	cos25°	70	0.0283	0.0563	0
	cos25°	50	0.0278	0.0445	1
	cos25°	30	0.0456	0.1356	0
	cos20°	70	0.0263	0.0446	1
	cos20°	50	0.0287	0.0563	0
	cos20°	30	0.0282	0.0562	0
	cos15°	70	0.0212	0.0446	1
	cos15°	50	0.0257	0.0565	0
	cos15°	30	0.0203	0.0446	1
	cos10°	70	0.0283	0.0446	1
	cos10°	50	0.0300	0.0446	1
	cos10°	30	0.0332	0.0569	0
	cos5°	70	0.0349	0.0446	1
	cos5°	50	0.0351	0.0446	1
	cos5°	30	0.0352	0.0446	1

Tools

Get Citation

Copy Citation Text

Min Shi, Shaoqing Zhou, Suqing Wang, Dengming Zhu. Accurate and Fast Primitive Detection Method for 3D Point Cloud Data[J]. Laser & Optoelectronics Progress, 2024, 61(4): 0415006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Jan. 30, 2023

Accepted: Apr. 3, 2023

Published Online: Feb. 26, 2024

The Author Email: Suqing Wang (wsq@ncepu.edu.cn), Dengming Zhu (mdzhu@tcict.ac.cn)

DOI:10.3788/LOP230549

CSTR:32186.14.LOP230549

Topics

laser devices and laser physics

Lasers and Laser Optics

laser manufacturing

Instrumentation, Measurement and Metrology