point-cloud-registration

A C++ implementation for Iterative Closest Point algorithm:

• Robust ICP
• Trimmed ICP

reading point cloud:

It is only working with .ply extension using the function readPointCloud(const std::string filename) which returns an Eigen::Matrix<float, nSamples, 3> object.

ICP:

1. Create an object of icp::Icp or icp::TrIcp with the required arguments:

icp::PointCloud P, M;  
icp::Icp robusticp(P,M);  

2. run the algorithm

robusticp.run();

3. performance measurement:
   icp::Icp.getIter(): #iteration taken for until convergence
   icp::Icp.getNpo(): last #trimmedpoints considered in computations
   icp::Icp.getSts(): sum of squared distances
   icp::Icp.getMse(): mean squared error or Sts/Npo
icp::Icp.getR(): an accumulated 3X3 rotation matrix
   icp::Icp.rotationError(Eigen::Matrix3f R): computes the rotation error given a ground truth rotation
   icp::Icp.getT(): an accumulated Eigen::Vector3f translation vector

NOTE: performance results are provided here