PCL's registration API has methods to set the search method for source and target points. However, the only supported method appears to be a kd tree. Since I have organized point clouds, I would like to employ PCL's OrganizedNeighbor search method in the correspondence estimation.
As far as I can tell, the registration classes use from the search method, or the CorrespondenceEstimation containing it these functions
nearestKSearch()-> available for all search methodssetInputCloud()-> available for all search methodssetPointRepresentation()-> only kd Trees have this, for converting points do k-dimensional vectors
So only the point representation stuff seems to be specific to kd trees. When looking at implementing a custom CorrespondenceEstimation subclass, it appears to not be supported to use other search methods.
Is there a way to accomplish this?