DataSetFilters.align

DataSetFilters.align(

target: DataSet | _vtk.vtkDataSet,

max_landmarks: int = 100,

max_mean_distance: float = 1e-05,

max_iterations: int = 500,

check_mean_distance: bool = True,

start_by_matching_centroids: bool = True,

return_matrix: bool = False,

)

Align a dataset to another.

Uses the iterative closest point algorithm to align the points of the two meshes. See the VTK class vtkIterativeClosestPointTransform.

Parameters:

targetpyvista.DataSet

The target dataset to align to.

max_landmarksint, default: 100

The maximum number of landmarks.

max_mean_distancefloat, default: 1e-5

The maximum mean distance for convergence.

max_iterationsint, default: 500

The maximum number of iterations.

check_mean_distancebool, default: True

Whether to check the mean distance for convergence.

start_by_matching_centroidsbool, default: True

Whether to start the alignment by matching centroids. Default is True.

return_matrixbool, default: False

Return the transform matrix as well as the aligned mesh.

Returns:

alignedpyvista.DataSet

The dataset aligned to the target mesh.

matrixnumpy.ndarray

Transform matrix to transform the input dataset to the target dataset.

See also

align_xyz

Align a dataset to the x-y-z axes.

Examples

Download Python source code | Download Jupyter notebook

Create a cylinder, translate it, and use iterative closest point to align mesh to its original position.

>>> import pyvista as pv
>>> import numpy as np
>>> source = pv.Cylinder(resolution=30).triangulate().subdivide(1)
>>> transformed = source.rotate_y(20).translate([-0.75, -0.5, 0.5])
>>> aligned = transformed.align(source)
>>> _, closest_points = aligned.find_closest_cell(
...     source.points, return_closest_point=True
... )
>>> dist = np.linalg.norm(source.points - closest_points, axis=1)

Visualize the source, transformed, and aligned meshes.

>>> pl = pv.Plotter(shape=(1, 2))
>>> _ = pl.add_text('Before Alignment')
>>> _ = pl.add_mesh(source, style='wireframe', opacity=0.5, line_width=2)
>>> _ = pl.add_mesh(transformed)
>>> pl.subplot(0, 1)
>>> _ = pl.add_text('After Alignment')
>>> _ = pl.add_mesh(source, style='wireframe', opacity=0.5, line_width=2)
>>> _ = pl.add_mesh(
...     aligned,
...     scalars=dist,
...     scalar_bar_args={
...         'title': 'Distance to Source',
...         'fmt': '%.1E',
...     },
... )
>>> pl.show()

Static Scene

Interactive Scene

Show that the mean distance between the source and the target is nearly zero.

>>> np.abs(dist).mean()
9.997635192915073e-05

See Register a Surface with ICP for more examples using this filter.