Improved ICP Point Cloud Registration Algorithm Based on Fast Point Feature Histogram

Liu Yuzhen; Zhang Qiang; Lin Sen

doi:10.3788/LOP202158.0611003

Laser & Optoelectronics Progress, Volume. 58, Issue 6, 611003(2021)

Improved ICP Point Cloud Registration Algorithm Based on Fast Point Feature Histogram

Liu Yuzhen¹, Zhang Qiang^1、*, and Lin Sen²

¹College of Electronic and Information Engineering, Liaoning University of Engineering and Technology, Huludao, Liaoning 125105, China

²College of Automation and Electrical Engineering, Shenyang Ligong University, Shenyang, Liaoning 110159, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Normal vectors on different regions. (a) Plane; (b) surface

Download full size

Fig. 2. Neighborhood features of FPFH algorithm. (a) Principle of FPFH; (b) single neighborhood of FPFH

Download full size

Fig. 3. Comparison of weight coefficients

Download full size

Fig. 4. Flow chart of algorithm

Download full size

Fig. 5. Comparison of ICP algorithms. (a) Registration time comparison; (b) registration error comparison

Download full size

Fig. 6. Registration results of Bunny. (a) ICP; (b) MR-TICP; (c) NICP; (d) RICP; (e) proposed algorithm

Download full size

Fig. 7. Registration results of Dragon. (a) ICP; (b) MR-TICP; (c) NICP; (d) RICP; (e) proposed method

Download full size

Fig. 8. Point clouds of physical objects. (a) Bottle; (b) lion

Download full size

Fig. 9. Registration results of point clouds of physical objects. (a) ICP; (b) MR-TICP; (c) NICP; (d) algorithm of this paper

Download full size

Table 1. Point cloud registration results of Bunny model

View table

Table 1. Point cloud registration results of Bunny model

Algorithm	Registration error /(10^-4 mm)			Registration time /s
Algorithm	25 dB	30 dB	35 dB	25 dB	30 dB	35 dB
ICP	64.8910	61.9821	65.8834	0.28	0.36	0.43
MR-TICP	7.6345	6.6025	5.3034	41.95	38.48	30.28
NICP	0.1612	1.1372	0.0597	112.04	92.54	56.89
RICP	0.1615	2.3085	0.0597	114.25	11.97	30.18
Proposed method	0.1612	0.1051	0.0597	147.21	88.34	56.35

Table 2. Registration error of Dragon model
View table
Table 2. Registration error of Dragon model
Algorithm Registration error /(10^-4 mm)
10% 20% 30% 50%
ICP 77.12 76.73 76.69 76.65
MR-TICP 3.70 3.32 3.74 4.09
NICP 0.02 0.31 0.02 0.54
RICP 0.02 0.02 0.70 0.64
Proposed algorithm 0.02 0.02 0.02 0.03

Table 3. Registration time of Dragon model
View table
Table 3. Registration time of Dragon model
Algorithm Registration time /s
10% 20% 30% 50%
ICP 2.05 0.68 0.74 0.55
MR-TICP 74.61 60.06 55.74 53.28
NICP 82.91 137.44 244.57 106.84
RICP 320.29 178.41 71.38 92.96
Proposed algorithm 77.19 121.92 36.84 73.53

Table 4. Comparison of registration for point clouds of physical objects
View table
Table 4. Comparison of registration for point clouds of physical objects
Algorithm Registration error /(10^-4 mm) Registration time /s
Bottle Lion Bottle Lion
ICP 1.21×10⁴ 205.20 1.28 10.69
MR-TICP 3.79 357.60 25.53 116.42
NICP 0.04 11.87 9.57 5.04
Proposed method 0.04 2.29 7.50 66.33

Tools

Get Citation

Copy Citation Text

Liu Yuzhen, Zhang Qiang, Lin Sen. Improved ICP Point Cloud Registration Algorithm Based on Fast Point Feature Histogram[J]. Laser & Optoelectronics Progress, 2021, 58(6): 611003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites