smoothing and adjusting set of point cloud

[Jaume]

Hi,

I would advise you to certain issue. I need to make an inventory of road using point clouds aquired by my drone. To processing my data I have to remove noisy scattered data points and get smoothed set of points in order to prepare proper point cloud to further process. As the result I need to reconstruct surface that is the best adjusted and approximated to the real road surface.

I have read about Moving Least Squares (MLS) surface reconstruction. It looks great but I have a doubt it distorts the model of the surface. Always when I use this method the final surface is slightly below the whole point cloud. But I need to adjust my surface to  agglomeration of points (exclude noisy points).

I am attaching a image of example points to submit what I mean.

Thank you for any suggestion in this case.

Best wishes,

Jakub

[stihl]

What I'd do if I were you is to build a DEM based on the Dense point cloudand and export that.
Then create a Mesh from the Dense point cloud and run 5 passes of mesh smoothing (Tools > Mesh > Smooth Mesh). Then create a DEM from the Mesh and export that as well.

Then import that latest file into any GIS and crop the file so only the road remains (create an poly outline around the road borders and export cropped to outline)
Then import the first DEM you made and overlay the cropped road over it so you've got a clear DEM but with a smoothed road. If the result isn't good enough, try smoothing the mesh with more passes.

[Jaume]

Thank you for your idea. It works pretty good. I have compared MLS and Photoscan method with raw point cloud from UAV. When I used 5 passes the smoothed surface is more and less among the point cloud.

But I am wondering how CloudCompare handles with this issue. 

[stihl]

Glad to hear this worked for you. Photoscan isn't the best with asfalt roads by default.

What do you mean about CloudCompare? How it would handle the smoothing?

[Jaume]

I think it is possible to smoothing a set of points in CloudCompare. I saw several options to modify, clean, process point cloud but I have no idea what the best option for this issue is.

I can not allow to overly modification of original point cloud - As the result of my project is computation of earthwork volume between a real and design road based on guidelines.

I posted in CC forum about assitance in CC procedure to gain sufficient accuracy.

[stephan]

Cloudcompare uses algorithms from PCL ( http://pointclouds.org/ ) to clean noisy point clouds etc... The results can be excellent and it would be really nice to have the option to use these algorithms in photoscan to automatically clean the dense point cloud for example.