Filling in Structure to Create Solids Given a High-Resolution 3D Point Cloud

[rw435]

Perhaps this cannot be accomplished easily with the current software, but I was wondering if the intermediate space in between points could be effectively "densified" using CloudCompare. In particular, I have a point cloud which is already over 2 million points, but it looks very grainy: . The points are locally consistent; the RGB values of points very close together are basically the same, but I just want to make the color come out to be stronger and actually describe the objects represented by the point cloud clearly.

(for reference, I'm looking to generate some segmentation results that are similar to , where objects are semantically and wholly "filled in" with one color).

[daniel]

The only tool that has a 'filling' feature in CC is the MLS filter (from the PCL library):
https://www.cloudcompare.org/doc/wiki/i ... r_(plugin) (see the last entry)

[rw435]

Hi Daniel,

My apologies for the late response, and I appreciate all your help. Maybe I'm overcomplicating my problem, but the idea is that I want to create a "solid" view of the point cloud. Perhaps I have to do some surface reconstruction for this, but I was hoping there was a simpler solution in CloudCompare. In particular, consider the following input point cloud:

grainy.PNG (429.69 KiB) Viewed 3268 times

Now when I zoom out in CloudCompare, these surfaces appear to become exactly what I want, but at a much smaller scale obviously [attachment]solid.PNG[/attachment]

Is there some way to do this so that I can get this zoomed out view, but at a much larger scale?

[daniel]

So either you increase the points size (with the +/- interactors that appear in the top-left corner of the 3D view). That's simple.

Otherwise, yes, you would need to mesh the point cloud. The most viable solution is to use the PoissonRecon plugin (https://www.cloudcompare.org/doc/wiki/i ... n_(plugin)), but you'll need normals, etc.