Project

General

Profile

SimpleTree plugin and not 0,0,0 centered input clouds

Added by Hackenberg Jan over 3 years ago

Hi Everyone

I received a bug report which I think is purely caused by clouds not centered around the coordinate origin. Normally those are georeferenced clouds. Those are the most troublesomes. I know also that a little shift in the z-axis can also cause (less impactful) errors. I tried to place already a lot warnings in my documentation, this topic is another one.

If you work with such data, here is what I suggest to do:

The whole plugin is 99% based on point cloud library. PCL uses float instead of doubles for floating point numbers which is a known problem. Not only the voxelgridfilter will do problems.

The solution:
You bring your plot which you want to process to the screen. You select (left click) the cloud and click in upper right on Item information. You will get the center x,y,z.
You write down the center coordinates and use the step Points→Transform→Points translation. You just enter -x,-y,-z and have shifted your cloud to the origin.

Cheers
Jan


Replies (2)

RE: SimpleTree plugin and not 0,0,0 centered input clouds - Added by Piboule Alexandre over 3 years ago

Hello,

There is also the step named “Center point clouds” from toolkit plugin, in Points/Transform menu.
This step directly allow do apply a translation to a cloud bringing its center to (0,0,0), and giving the inverse transformation matrix.

4 modes are available (for the new center):
- Bounding box center
- Bounding box center, but with minimum height of the cloud for Z coordinate
- Points centroid
- Points centroid, but with minimum height of the cloud for Z value

Best regards.
Alexandre Piboule.

RE: SimpleTree plugin and not 0,0,0 centered input clouds - Added by Hackenberg Jan over 3 years ago

Yes now I remember. When I told Alexandre I would like to write that step he told me about the existence of his. I cannot check tonight, cause I cut my remote connection:

My QSM exporter can take that inverse transformation matrix as an optinoal input. It will backtransform the QMSs to orginal coordinates. I hope I really wrote this extension and its not only
on my mental todo list. If I still have to implement it, you can do for now a manual backtransformation in R.

Cheers
Jan

    (1-2/2)