Thinning ground survey

[AndrewPerth]

Hey guys,

I had a look around but wasn't able to find a feature of plugin that does this (I very easily could have missed it) but is there anything that can complete topographic survey thinning similar to this process:

"Points are removed if they are less than 10m from the nearest adjacent point and can be interpolated to within 15cm in elevation by the surrounding points. No points are removed near changes of slope. This preserves most of the detail in the original survey grid."
Taken from: https://www.photosat.ca/gis/thinned-grid/

kind regards,

Andrew

[daniel]

Removing points closer than 10m from the others can be achieved with the subsampling tool (with 'SPATIAL' mode, and 10 m. as the main parameter).

Making the radius of exclusion change depending on the slope is a little bit harder to do, but can be achieved by computing a scalar field (e.g. the curvature, the dip angle also maybe) and then using the subsampling 'use actif SF' mode so as to make the radius change based on the scalar value associated to each point: http://www.cloudcompare.org/doc/wiki/in ... CSubsample

[AndrewPerth]

Hi Daniel,

cheers for that. That is definitely an option, my only concern is how to do a QA/QC check to see whether the resulting point cloud at least reasonably represents the overall shape of the terrain with not too much of a volumetric error. I'll give it a try and see what I can come up with.

thanks again for your assistance!

Andrew

[AndrewPerth]

Hi Daniel,

I had a crack at it and had difficulty achieving what i was attempting to generate.

I computed the curvature and then had a look at what areas were what curvature, checking to see whether distinct man made features or sharp features were either above or bellow a set curvature. However when you then go to do the sub sampling it is a relationship between the maximum curvature value and the minimum curvature value and two set sub-sampling distances.

What would be incredibly useful would be to set the curvature value, e.g. no sub-sampling on any points above a certain value of curvature. Or simply have it as a binary, sub sample when the curvature is equal to or less than a cutoff value and leave any points above that curvature as is. This is really pronounced in the data i'm looking at as the histogram of the scalar field looks like this:

topo curvature scalar.PNG (7.32 KiB) Viewed 16908 times

kind regards,

Andrew

[daniel]

For this you can use the 'Edit > SF > Filter by value' tool to split the cloud. Then use the subsampling tool only on one part. And then merge the result with the other part of the cloud.

And of course you may want to use another scalar field than the curvature.

[toxicmag]

.... using the subsampling 'use actif SF' mode so as to make the radius change based on the scalar value associated to each point: http://www.cloudcompare.org/doc/wiki/in ... CSubsample

Daniel could you check this again, please?

I still experience a bug when putting the higher distance value to the lower SF value and the lower distance to the higher SF. It works the other way around.

Tried with 11alpha, beta and latest stable.

[daniel]

Can you post a snapshot of the susampling tool dialog? (in the failing case). I'm not sure to understand how you set it up.

[toxicmag]

Hi Daniel,

here are 5 screnshots with the following:

1) orig. Pointcloud (gravel pit) wit metric values for coordinates
2) SF by "surface variation" (radius 0.5) with very nice results of the "breaklines"
3) SF clipped a little responding to the values
4) subsampling with 0 and 1 just to show it works depending on SF values
5) subsampling fails when higher distance values corresponds to lower SF and lower distance to higher SF...

I see this in current CC stable, alpha and beta.
Maybe some internal issue with my laptop. I will check at a collegues PC right away.

Update: It's the same experience at my colleagues computer. If subsampling is done like in #5 the scalar field seems to be skipped. All resulting points are equally spread over the area.

[toxicmag]

Dear group,

some of you - and also me - reported about some difficulties to thin the cloud by curvatures.
One of the posts is this one here and from time to time i think about wether there were some changes already.

Am i wrong and the process still is working for all of you out there?
Or can it not be done in CloudCompare anymore like in former times and we have to close this question here?

Thanks for your feedback and all the best to you out there.


Alex

[daniel]

Normally it should work, but it depends on your scalar field and how you set the value (which can be tricky)