Skip to content

jakarto3d/jakteristics

Repository files navigation

Jakteristics

Jakteristics is a python package to compute point cloud geometric features.

A geometric feature is a description of the geometric shape around a point based on its neighborhood. For example, a point located on a wall will have a high planarity.

The features used in this package are described in the paper Contour detection in unstructured 3D point clouds. They are computed based on the eigenvalues and eigenvectors:

  • Eigenvalue sum
  • Omnivariance
  • Eigenentropy
  • Anisotropy
  • Planarity
  • Linearity
  • PCA1
  • PCA2
  • Surface Variation
  • Sphericity
  • Verticality
  • Nx, Ny, Nz (The normal vector)

It's inspired from a similar tool in CloudCompare.

It's implemented in cython using the BLAS and LAPACK scipy wrappers. It can use multiple cpus, and the performance is quite good (at least twice as fast as CloudCompare).

Installation

python -m pip install jakteristics

Usage

Refer to the documentation for more details.

From python

from jakteristics import compute_features

features = compute_features(xyz, search_radius=0.15, feature_names=['planarity', 'linearity'])

CLI

Once the package is installed, you can use the jakteristics command:

jakteristics input/las/file.las output/file.las --search-radius 0.15 --num-threads 4

Run tests

python -m pip install -r requirements-dev.txt
python setup.py pytest