I have looked for how to implement Generalized ICP(GICP) with PCL. I found a sample code in test/registration/test_registration.cpp in PCL repository on github. The sample code uses GICP as following. Could you tell me following procedure is correct one to use GICP with PCL?
The fucntion "align" is a function of IterativeClosestPoint class. It means that "align" does not consider point-to-plane that AV Segal et al refer in their paper. I'm wondering if this is a correct procedure to use GICP with PCL. In addition, I don't know why PCL does not provide us sample codes which use estimateRigidTransformationBFGS the metod of GeneralizedIterativeClosestPoint class.
GeneralizedIterativeClosestPoint<PointT, PointT> reg_guess;
reg_guess.setInputSource (src);
reg_guess.setInputTarget (transformed_tgt);
reg_guess.setMaximumIterations (50);
reg_guess.setTransformationEpsilon (1e-8);
reg_guess.align (output, transform.matrix ());