Drost PPF

PPF (point pair feature) is a point registration method proposed by Drost. et al. in 2010

Model Globally, Match Locally: Efficient and Robust 3D Object Recognition

Thanks for the great effort of pcl(point cloud lib), the method has been implemented. You can find the usage of PPF in their document: here

3D is here: Point Cloud Library (PCL)

Although the algorithm in the PCL works well and has been well managed, I tried to reproduce the method and figure out the some details about the article. That is the reason for establishing this repo. 😃

Dependence

• Cmake 3.5
• PCL 1.8
• Boost (Although boost has been in integrated in C++ standard(std::), unfortunately, I still use many func in the boost)
• OpenMP
• Eigen3

Usage

Compile the code

    mkdir build && cd build && cmake .. && make

Run the code

    ./Drost-PPF /path_to_src_point_cloud /path_to_tgt_point_cloud

• The .pcd point cloud file is needed
• arg1 is the model and arg2 is scene point cloud

Warning

• This is not a complete project and I just finished the key code and check the correctness.
• If you wanna use Drost PPF to establish some test for research, PCL lib is recommended to use.
• I did not write the Config file , so some param need to be adjusted from the code and don't forget recompile.
  • Point Cloud downsample leaf size : ./Drost_PPF.cpp -> line 123.
  • OpenMP process number can be adjusted in the file ./PPFRegistration.cpp -> line 132.
  • Many params are set in the file ./Drost_PPF.cpp's "test" function.