normals_Hough

Normal Estimation in Unstructured Point Clouds with Hough transform

Paper

Please acknowledge our reference paper :

"Fast and robust normal estimation for point clouds with sharp features" by Alexandre Boulch and Renaud Marlet, Symposium on Geometry Processing 2012 (SGP 2012) and Computer Graphics Forum

Code

The code for normal estimation is C++ hearder only. Three version are proposed, previous version are located in archives/cgal/ and archives/pcl/ relies on CGAL and PCL libriaries. The current version Normals.h relies on Eigen and nanoflann (we provide them in the third_party_includes folder).

Parameters

• neighborhood_size (default 200) the neighborhood size for computing the normals.
• n_planes (default 700) the number of random sample to be picked to estimate the distribution.
• n_rot (default 5) the number of random rotations of the accumulator.
• n_phi (default 15) the discretization of the sphere accumulator.
• tol_angle_rad (default 0.79) the maximal angle used for normal cluster (final normal decision).
• k_density (defautl 5) the neighborhood size for density computation.

Examples and tests

Python

Using the library with Python requires building the wrapper. All python code has been built and tested under Unbuntu 18.04 and Anaconda.

Building the wrapper

pip install -ve /path/to/notmals_Hough/

Python test

From the root directory:

python example.py

The script creates a point cloud on a cube and estimate the normals. It produces a .xyz containing both points and normals, it can be displayed using Meshlab or CloudCompare.

C++

In the folder test_cpp, the executable can be generated using CMake.

Usage

Hough_Exec [options] -i input_file.xyz -o output_file.xyz

Author webpage

Alexandre Boulch

License

The code is released under a GPLv3 license. For commercial purposes contact the authors. The detailed licence is here.

Previous versions

Previous version of the code are located in folders archives/cgal/ and archives/pcl/