Iterative Closest Point Surface Registration
This module is a surface-to-surface global registration algorithm.
The iterative closest point algorithm (ICP) works by assuming that all
points on the source surface(s) have corresponding points on the
target surface. Each iteration consists of choosing this
correspondence using the closest point of the target surface to each
point in the source surface(s) followed by estimating the optimal
registration between these point sets. A trimmed version of ICP can
also be used by estimating the optimal registration using only a
fraction of the closest points.
Input Types
Can be applied to 3D surfaces
Input Parameters
Degrees of Freedom |
Specifies the number of degrees of freedom (dof) to allow for the registration between the point sets. The four options include: Translation (3 dof), Rigid (6 dof), Rigid and Global Scale (9 dof), and Affine (12 dof). Translation allows for a translation in each of the three coordinate directions. Rigid allows for three translations and three rotations. Global scale includes a multiplicative scale factor for each axis. Finally, affine allows freedom of all 9 parameters of the 3x3 transformation matrix in addition to the three translations. The default value is Rigid. |
|
Multi-Resolution Levels |
Specifies the number of resolution levels in the multi-resolution registration. The default value is 6. |
|
Trimmed least Sqrs Retainment |
The fraction of points to use at each iteration. Points are sorted by distance and only the specified fraction of closest points are used to calculate the optimal registration. The default value is 0.75. |
|
Error Threshold |
Specifies a threshold for the <relative change of the trimmed MSE?> as a stopping condition of the algorithm. The default is 0.0050. |
|
Max iterations |
Maximum number of iterations allowed before the process is terminated. The default value is 20. |
|
Start with PCA Alignment |
Specifies whether or not to initially align the surfaces using PCA. |
|
Example Usage
|
Example misaligned surfaces. |
|
|
Surfaces aligned after ICP registration. |
|
References
[1] D. Chetverikov, D. Svirko, D. Stepanov et al., "The trimmed iterative closest point algorithm," In Proceedings International Conference on Pattern Recognition, pp. 545-548, 2002.