Acta Optica Sinica, Volume. 40, Issue 6, 0615001(2020)
Skull Point Cloud Registration Algorithm Based on Hierarchical Optimization Strategy
Figures & Tables(13)
![Comparison of coarse registration results. (a) Registration results without optimization of k-means algorithm; (b) registration results optimized by proposed algorithm; (c) rough registration results of Ref. [28] method; (d) rough registration results of Ref. [37] method](/Images/icon/loading.gif){kind=icon}
Fig. 4. Comparison of coarse registration results. (a) Registration results without optimization of k-means algorithm; (b) registration results optimized by proposed algorithm; (c) rough registration results of Ref. [28] method; (d) rough registration results of Ref. [37] method
Fig. 5. Comparison of fine registration results. (a) Registration results of classical ICP algorithm; (b) registration results with k-d tree optimization; (c) registration results with geometric feature constraint optimization; (d) registration results of the improved ICP algorithm
Table 1. Efficiency comparison of different rough registration methods
View tableTable 1. Efficiency comparison of different rough registration methods
Algorithm Registration error /mm Time- consuming /s Without k-means optimization 5.259×10-1 18.322 Ref. [28] method 6.126×10-1 24.235 Ref. [37] method 5.387×10-1 37.448 Proposed method 3.432×10-1 22.839
Table 2. Efficiency comparison of different fine registration methods
View tableTable 2. Efficiency comparison of different fine registration methods
Algorithm Number of iterations Registration error /mm Time-consuming /s Classical ICP method 49 4.568×10-2 44.688 Adding k-d tree optimization 32 4.287×10-2 21.557 ICP method with geometric feature constraints 45 3.062×10-2 37.951 Improved ICP method 31 2.934×10-2 28.124
Table 3. Efficiency comparison of different registration methods
View tableTable 3. Efficiency comparison of different registration methods
Algorithm Matching rate /% Registration error /mm Time- consuming /s RANSAC[ 38 ]61.3 8.022×10-2 65.493 LO-RANSAC[ 39 ]72.6 5.556×10-2 57.739 4PCS[ 40 ]78.8 4.760×10-2 54.875 Super-4PCS[ 41 ]84.1 3.011×10-2 45.618 ICP[ 7 ]76.6 4.826×10-2 61.942 Go-ICP[ 42 ]76.2 4.986×10-2 61.557 GA-ICP[ 43 ]75.5 5.159×10-2 57.334 ICP-CP[ 14 ]84.8 2.994×10-2 49.561 IRLS-ICP[ 44 ]82.7 3.558×10-2 52.188 Proposed method 89.9 2.355×10-2 42.632
Table 4. Efficiency comparison of different registration methods
View tableTable 4. Efficiency comparison of different registration methods
Algorithm Matching rate /% Registration error /mm Time- consuming /s RANSAC[ 38 ]54.5 9.113×10-2 31.224 LO-RANSAC[ 39 ]63.3 7.739×10-2 57.739 4PCS[ 40 ]79.6 4.609×10-2 24.682 Super-4PCS[ 41 ]80.7 4.288×10-2 22.899 ICP[ 7 ]79.1 4.977×10-2 26.861 Go-ICP[ 42 ]78.6 5.096×10-2 25.933 GA-ICP[ 43 ]77.8 5.355×10-2 24.129 ICP-CP[ 14 ]81.1 4.071×10-2 22.567 IRLS-ICP[ 44 ]80.4 4.525×10-2 23.337 Proposed method 81.6 3.749×10-2 21.735
Tools
Get Citation
Copy Citation Text
Wen Yang, Mingquan Zhou, Xiangkui Zhang, Guohua Geng, Xiaoning Liu, Yangyang Liu. Skull Point Cloud Registration Algorithm Based on Hierarchical Optimization Strategy[J]. Acta Optica Sinica, 2020, 40(6): 0615001
Paper Information
Category: Machine Vision
Received: Oct. 20, 2019
Accepted: Nov. 30, 2019
Published Online: Mar. 6, 2020
The Author Email: Zhou Mingquan (nwuzmq@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20