pcl::IterativeClosestPoint -> Correspondence rejector not realy implemented

[hdiethelm]

In registration/include/pcl/registration/impl/icp.hpp: line 145
// We should be doing something like this
// for (size_t i = 0; i < correspondence_rejectors_.size (); ++i)
// {
// correspondence_rejectors_[i]->setTargetCloud (target_);
// if (target_has_normals_)
// correspondence_rejectors_[i]->setTargetNormals (target_);
// }
and line 177:
// We should be doing something like this
// correspondence_rejectors_[i]->setInputSource (input_transformed);
// if (source_has_normals_)
// correspondence_rejectors_[i]->setInputNormals (input_transformed);
I think code should be uncommented for any CorrespondenceRejector to work. But just uncommenting doesn't help, because the setXXX methods are not defined in class CorrespondenceRejector.

[sdmiller]

Sorry for such a late response. I'm the one who commented those, and indeed, this is an issue which should be fixed in the long run. It would require adding some generic methods to the CorrespondenceRejector class, and switching all current subclasses to follow it.

The problem is that not everything operates on normals, or even on a cloud. Since everything is templated, this gets a bit gross. It may be smarter to have:

bool CorrespondenceRejector::requiresNormals ();
bool CorrespondenceRejector::requiresCloud ();
void CorrespondenceRejector::set[Input/Target]Cloud (pcl::PCLPointCloud2::Ptr cloud)
void CorrespondenceRejector::set[Input/Target]Normals (pcl::PCLPointCloud2::Ptr normals)

And use the [to/from]PCLPointCloud2 () methods to convert between the blob and templated types. I think it's time to do this, but I'm not sure I have the cycles.

[sdmiller]

Unless someone else has noticed missing points, I think this was solved with #419 .