Skip to content

Latest commit

 

History

History

ICP-Registration

README.md

An implementation of Iterative Closest Point for point cloud registration.

An application that allows to align two point clouds of the same object in an interactive way, using ICP [1]. Implementation of the Iterative Closest Point algorithm, and the Point-To-Plane [2] variant.

Usage

Dependencies

Libraries needed :

  • LibIGL . Put the libigl source code folder two levels above this repository.
  • ANN (Approximate Nearest Neighbor). Put the ANN source code folder one level above this repository, or change the include_directories and link_directories of the CMake to the corresponding paths.

Building and running the code

Build using CMake and the CMakeLists.txt provided. Run the ICP_registration_viewer target. Change the line string filenames[num_data_total] with the right filenames of your point clouds, in main.cpp.

Using the interface

The libigl GUI menu has several options you can manipulate :

  • Change the "ICP mode" slider in order to perfom point-to-point ICP (0) or point-to-plane ICP (1).
  • Click the "ICP step" button to perform one registration step of the algorithm.
  • Click the "ICP registration" button to perform the algorithm until convergence.
  • Change the values of "Max num iterations" and "Error threshold" to define the convergence stopping conditions.
  • You can roughly pre-align the point clouds by manipulating the "Rotation slider" and "Translation slider".

References

[1] Paul J. Besl et N.D. McKay. A Method for Registration of 3-D Shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence.

[2] LOW, K.-L. 2004. Linear least-squares optimization for point-to- plane icp surface registration. Tech. rep., Chapel Hill, University of North Carolina.