CloudCompare forum

Good day,
Please, I need some clarification and guidance here.

I have two UAV LiDAR datasets, one at 46 m height and another at 30 m height. I am not sure how to check the resolution in terms of LiDAR, and to possibly resample the point cloud taken at 30 m to be comparable at least with the 46 m height dataset.

How can I do this?
Thank you

Hi @daniel , This is me trying to prompt again for a possible response from anybody available to help with my question.
Thank you.

Can you compute the (surface) 'Density' with the Tools > Other > Compute geometric features' tool. And then use the 'Edit > Cloud > Subsample' tool on the densest cloud?

See https://www.cloudcompare.org/doc/wiki/i ... c_features
See https://www.cloudcompare.org/doc/wiki/i ... CSubsample

Thank you, Daniel. I will check, but what determines what "local neighbour radius" to set?

That's something you have to determine, based on the scale of your point cloud, the density, etc.

Basically if you take a radius that is too small, you won't capture enough neighbors around each point, and your density will always be 'one point per surface unit'. And if you take a too big radius, this will take a lot of time to compute, and you will capture points so far away that it might no make a lot of sense.

You can play with different values, or maybe use a radius that correspond to an arbitrary surface (such as r = sqrt(1/pi) ~= 0.56, if you have coordinates expressed in meters, and you think a density per 1 square meter makes sense - that's just an example of course ;-) ).

Makes sense. It is clearer to me now. Thank you @Daniel.
I am however curious as to the concept of point density concerning resolution or height of flight.

If I am monitoring a plant growth (that is taking lidar images at similar heights at different growth stages in the same field), unlike pixel size in terms of images which would be the same throughout, how is this considered with regards to Lidar imaging, because if point density is is the equivalent of pixel resolution for LiDAR (i am not sure), that would be conflicting because the plant growth may contribute to the pixel density.

Could please help clarify this?

Hum, are you referring to volume density maybe? Generally, for airborne LIDAR, I believe consider surface density (at the ground level). This is rather constant (even though it's statistical, as the laser beam will never hit the same place twice between 2 flights).

Of course plants make things a little harder as they tend to reflect the laser beam from different heights, causing an error (or averaging) in the height measurement. Or even without error, they will cause high variations of heights, which may change a lot the density due to the way it is computed in CC (at least with this tool).

When looking from the top, the (2D) density should be more or less the same. If that's what you are interested in, then you might use the 'Rasterize' tool instead. If you choose a not too small raster step, then you can export the raster as a cloud with the 'population' field. This is another way to estimate the density (= population per cell divided by the cell size = step^2).

Thank you for your time, Daniel. Your responses do clarify a few things.

The main reason I ask is basically because I am trying to build a model for leaf area index from two different fields.
The ground truth was taken with a common device for LAI estimation, but then I realized that in one field the LiDAR data was taken at 46m for one date, and in another field it was taken at 30m for several dates.

I was wondering if using both fields together in a model-building process shouldn't be an issue, especially if using the height-normalized point cloud or sub-sampled data.

Yes, sorry, I was not clear, when I said it was 'rather constant', it was across the cloud assuming it was acquired at a constant altitude. Your approach is good, you should definitely resample the cloud acquired at a different altitude (if your model is sensitive to the density).

Great.
Final clarification on this topic: Use said resampling, but in cloud-compare I see subsampling, like in the link you attached earlier.

In any case, I have used the height normalized dataset because I wasn't sure or it wasn't so clear how to deal with the subsampling, especially with regards to my case of different altitudes. Can you clarify this?

Also, it'd be good if a question had been exhaustively answered, like in my case, to have a button that we can click to say answered or even a like button. I'm just thinking out loud here.