C++ implementation for Oriented Point Sampling for Plane Detection in Unorganized Point Clouds by Bo Sun and Philippos Mordohai

Link to their paper

The algorithm detects all planes in an unorganized point cloud. Normals are computed only for a small fraction of all the points in the point cloud (usually between 0.3% and 3%). This allows for single point hypotheses (unlike most approaches that use three unoriented points) without being computationally intensive.

Build

Dependencies are PCL 1.10 (point cloud processing and visualization) and Boost library

mkdir build && cd build
cmake ..
make -j6

Usage

./orientedPointSampling <inputFilename> input_filename is a .pcd file containing a point cloud

ToDO

• [x] Implement merging process
• [ ] Implement ground removal