Multiview Point Cloud Registration Method Based on Laser Radar

Lei Geng; Chunpeng Cao; Zhitao Xiao; Fang Zhang

doi:10.3788/LOP202259.1228004

Laser & Optoelectronics Progress, Volume. 59, Issue 12, 1228004(2022)

Multiview Point Cloud Registration Method Based on Laser Radar

Lei Geng¹, Chunpeng Cao², Zhitao Xiao^1、*, and Fang Zhang¹

¹Tianjin Key Laboratory of Optoelectronic Detection Technology and System, School of Life Sciences, TianGong University, Tianjin 300387, China

²Tianjin Key Laboratory of Optoelectronic Detection Technology and System, School of Electrical and Electronic Engineering, TianGong University, Tianjin 300387, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(14)

Fig. 1. Point cloud collection

Download full size

Fig. 2. Single frame point cloud data

Download full size

Fig. 3. Schematic diagram of radius filtering

Download full size

Fig. 4. Pre-processing effect diagram

Download full size

Fig. 5. Two adjacent point clouds

Download full size

Fig. 6. Block diagram of proposed method

Download full size

Fig. 7. Registration renderings. (a) Pfh-ICP; (b) Fpfh-ICP; (c) Super-4PCS; (d) NDT-ICP; (e) IMU-ICP

Download full size

Fig. 8. Registration effect of markers in truck

Download full size

Table 1. IMU Quaternion
View table
Table 1. IMU Quaternion
Collection time x y z w
1 0.706451 0.596095 0.152989 0.349563
2 0.677581 0.638681 0.244056 0.270937
3 0.637251 0.670253 0.330727 0.187861
4 0.586133 0.690527 0.411504 0.101411
5 0.525522 0.699104 0.484654 0.013803

Table 2. Data preprocessing results
View table
Table 2. Data preprocessing results
Frame Points Points after processing Percentage reduction /%
1 22390 4473 80
2 23360 4067 83
3 23765 3468 86
4 24024 3469 86
5 13433 2378 83

Table 3. Average rotation error of each method
View table
Table 3. Average rotation error of each method
Method X Y Z
Pfh-ICP 24.471 49.192 5.861
Fpfh-ICP 2.589 28.874 9.349
Super-4PCS 2.654 25.584 8.561
NDT-ICP 2.234 48.912 0.057
IMU-ICP 0.092 0.075 0.006

Table 4. Average translation error of each method
View table
Table 4. Average translation error of each method
Method X Y Z
Pfh-ICP 0.069 0.096 0.086
Fpfh-ICP 0.589 0.872 0.345
Super-4PCS 0.435 0.584 0.264
NDT-ICP 0.011 0.074 0.021
IMU-ICP 0.003 0.005 0.008

Table 5. Sum of squared distances between original ICP algorithm and improved ICP algorithm

View table

Table 5. Sum of squared distances between original ICP algorithm and improved ICP algorithm

Times	Original ICP	Improved ICP
1	0.7025643	0.7015648
2	0.3125624	0.2314568
3	0.0165345	0.0134564
4	0.0075612	0.0026429
5	0.0026564	0.0023536
6	0.0025644	0.0022203
7	0.0024652	0.0021132
8	0.0022356
9	0.0022334
10	0.0022267
11	0.0021469

Table 6. Point cloud continuous registration completion time
View table
Table 6. Point cloud continuous registration completion time
Method 1 /s 2 /s 3 /s 4 /s 5 /s Time consuming /s
Pfh-ICP 10.56 8.46 8.46 8.07 35.55
Fpfh-ICP 9.61 7.72 7.27 7.95 32.55
Super-4PCS 1.54 1.74 1.19 1.11 5.58
NDT-ICP 1.12 1.03 1.02 1.01 4.18
IMU-ICP 0.41 1.27 0.99 0.63 0.41 3.71

Tools

Get Citation

Copy Citation Text

Lei Geng, Chunpeng Cao, Zhitao Xiao, Fang Zhang. Multiview Point Cloud Registration Method Based on Laser Radar[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1228004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites