Cylinder

class Cylinder(centers, orientations, radius, height)

Bases: Parametrization

Parametrize a point cloud as a cylinder.

Parameters:

centersnp.ndarray

Center coordinates of the cylinder in X, Y, and Z.

orientationsnp.ndarray

Orientation matrix (direction vectors).

radiusfloat

Radius of the cylinder.

heightfloat

Height of the cylinder.

Methods

Cylinder.compute_normal(points)	Compute surface normals for points on the cylinder.
Cylinder.fit(positions, **kwargs)	Fit a cylinder to a point cloud.
Cylinder.points_per_sampling(sampling_density)	Computes the approximate number of random samples required to achieve a given spatial sampling_density.
Cylinder.sample(n_samples[, normal_offset, ...])	Sample points from the surface of a cylinder.

compute_normal(points)

Compute surface normals for points on the cylinder.

Parameters:

pointsnp.ndarray

Input points to compute normals for.

Returns:

normalsnp.ndarray

Computed surface normals.

classmethod fit(positions, **kwargs)

Fit a cylinder to a point cloud.

Uses PCA for the initial axis estimate, then refines center, direction, and radius via constrained optimization (SLSQP).

Parameters:

positionsnp.ndarray

Point coordinates with shape (n, 3).

Returns:

Cylinder

Fitted cylinder with estimated center, orientation, radius, and height.

points_per_sampling(sampling_density, normal_offset=None)

Computes the approximate number of random samples required to achieve a given spatial sampling_density.

Parameters:

sampling_densityfloat

Average distance between points.

normal_offsetfloat, optional

Compute number of samples on offset parametrization, instead of current.

Returns:

int

Number of required random samples.

sample(n_samples, normal_offset=0.0, mesh_init_factor=5, **kwargs)

Sample points from the surface of a cylinder.

Parameters:

n_samplesint

Number of samples to draw

normal_offsetfloat, optional

Offset points by normal_offset times their normal vector.

mesh_init_factorint, optional

Number of times the mesh should be initialized for Poisson sampling. Five appears to be a reasonable number. Higher values typically yield better sampling.

Returns:

np.ndarray

Point coordinates (n, 3).