I've noticed that tf2::doTransform from tf2_sensor_msgs only transforms the XYZ fields, and leaves normals and viewpoints untransformed. I can either send a PR adding support for this (if the channels are present), or there should be a warning that the pointcloud was transformed but not its normals and viewpoints. What solution does sound better?
There could also be a new helper function called transformChannel or similar, that would take a pointcloud, channel prefix and transform, and apply the transform to all points in the given channel (appending x, y and z to the channel prefix).
I've noticed that tf2::doTransform from tf2_sensor_msgs only transforms the XYZ fields, and leaves normals and viewpoints untransformed. I can either send a PR adding support for this (if the channels are present), or there should be a warning that the pointcloud was transformed but not its normals and viewpoints. What solution does sound better?
There could also be a new helper function called
transformChannel
or similar, that would take a pointcloud, channel prefix and transform, and apply the transform to all points in the given channel (appendingx
,y
andz
to the channel prefix).