CloudCompare forum

Hi, I'm working with aerial lidar data.
I've subsampled by space to get to a manageable file size, but what I would like to do now is further optimise the point cloud by reducing the density of points in areas where the ground surface is relatively smooth, and retain high density in areas where the ground is complex.
I'm sure this is a common task, but can't figure it out.
Any clues?
Thanks
Mick

Ah, not that common (or at least, it's not that simple ;).

The only way to do that is to use the 'Subsample' tool with a scalar field (see https://www.cloudcompare.org/doc/wiki/i ... 8option.29). And you would typically use the curvature or another measurement to drive the subsampling variable spacing.

OK thanks, that looks like it might work for me.

The attached shows an example where this is working in principle: I have created an SF based on curvature and then subsampled by space inversely proportional to curvature. The settings I've used are 1m radius for the curvature local neighbourhood radius, which resulted in curvature levels from 0.0001 to 0.25, to which I applied the sub-sampling distance 1m to 0.1m respectively.

In the attached example, the relatively smooth blue areas have sub-sampled point spacing (white dots) at around 1m as expected, however the red areas should ideally have a higher density than shown (approaching 0.1m spacing), but I can't seem to achieve this. The original cloud is on a 0.1m grid.

Any suggestions?
thanks
Mick

It's probably because the repartition of the curvature values are not very 'linear' (while this tools make the density vary in a linear way between 2 scalar values). So the only idea I have is to use the SF arithmetic tool with the log function for instance to change this?

I think that's getting there - I was unaware of the arithmetic transforms. Modifying the initial curvature radius and playing with a combination of logs and other functions get me somewhat closer to a linear distribution and a better subsampling distribution.
Thanks Daniel.