Author Affiliations
College of Information Science and Technology, Northwest University, Xi′an, Shaanxi 710127, Chinashow less
Fig. 1. Diagram of local coordinate system and point projection. (a) Feature point p and its neighborhood point {pi} in local coordinate system op-xpypzp; (b) projection points on two-dimensional plane
Fig. 2. Sub-regions of regional curvature map
Fig. 3. Flow chart of proposed algorithm
Fig. 4. Iterative convergence curve
Fig. 5. Skull point cloud models to be registered
Fig. 6. Registration results of different rough registration methods. (a) PCA; (b) GA; (c) FPFH; (d) proposed method
Fig. 7. Two skulls to be registered
Fig. 8. Coarse registration results
Fig. 9. Fine registration results of ICP algorithm
Fig. 10. Fine registration results of improved ICP algorithm
Fig. 11. Initial position
Fig. 12. Coarse registration result
Fig. 13. Fine registration results of ICP algorithm
Fig. 14. Fine registration results of improved ICP algorithm
Method | Registration error /mm | Time-consuming /s |
---|
PCA | 6.254×10-1 | 20.39 | GA | 5.697×10-1 | 24.96 | FPFH | 4.638×10-1 | 16.84 | Proposed method | 3.474×10-1 | 18.25 |
|
Table 1. Comparison of registration efficiency of different rough registration methods
Algorithm | Number of iterations | Registration error /mm | Time-consuming /s |
---|
ICP | 46 | 3.652×10-2 | 47.83 | Improved ICP | 29 | 3.243×10-2 | 31.64 |
|
Table 2. Comparison of registration efficiency of skull point clouds
Algorithm | Number of iterations | Registration error /mm | Time-consuming /s |
---|
ICP | 32 | 8.495×10-3 | 21.63 | Improved ICP | 18 | 7.876×10-3 | 12.28 |
|
Table 3. Comparison of registration efficiency of bunny point clouds
Point cloud model | Number of point clouds | Algorithm | Number of iterations | Registration error /mm | Time-consuming /s |
---|
| | LO-RANSAC[26] | 48 | 5.689×10-2 | 47.86 | Skull to be registered | 210759 | Super-4PCS[27] | 35 | 4.384×10-2 | 32.75 | | | Go-ICP[28] | 53 | 4.930×10-2 | 52.59 | | | PICP[29] | 42 | 3.677×10-2 | 41.06 | Reference skull | 211234 | IRLS-ICP[30] | 37 | 3.256×10-2 | 36.63 | | | Proposed algorithm | 30 | 2.977×10-2 | 29.67 | | | LO-RANSAC[26] | 29 | 2.899×10-2 | 18.21 | Bunny to be registered | 40000 | Super-4PCS[27] | 19 | 1.556×10-2 | 13.22 | | | Go-ICP[28] | 31 | 2.017×10-2 | 20.69 | | | PICP[29] | 25 | 1.154×10-2 | 16.95 | Reference bunny | 40000 | IRLS-ICP[30] | 22 | 9.255×10-3 | 15.48 | | | Proposed algorithm | 17 | 7.688×10-3 | 11.94 |
|
Table 4. Efficiency comparison of different registration algorithms