C2C distance - Local modeling - radius (sphere) value

[Sch]

Hello,

I am beginner in CC software and I determine day by day, that it is a great, versatile and very userfriendly software. I would like to know more closely the method C2C. I use this computation process and I would like to make Local modeling. Before I compute the really values / distances, I have to decide the Radius (Sphere) value (I'd like to use this instead the kNN value, because my models contain lot of gaps and the density is low, haven't got clear density value).
The value in this case is 5.078248. How can I refine this or can I compute this with an algorithm? What is this exactly, is the default value usable or can I replace it with another calculated value?

Thanks for the answer.

[daniel]

Oh you should change it actually (it's just a random guess, based on the total bounding-box size).

The aim is to get a neighborhood radius (around each point) big enough so as to get enough points to build a reliable model, but small enough so that the shape remains simple and easy to model (more or less planar, or at least 2.5D only). As you guessed it highly depends on your reference cloud density...

[Sch]

I understand it, but the mean question is, how can I compute the value of radius exactly, which is the really result instead I estimate it? :) I will avoid 'trial and error'.

Thanks a lot for your answers.

[daniel]

No, not really sadly. You can 'eyeball' it of course, or you can also use the 'Tools > Density' tool (with the 'number of neighbors' as result) to see for a given radius value how many neighbors you get.

[Sch]

Thank you for your advice, I will try it!

[irsyadifirdaus]

Do you know the meaning of min. dist, max dist, avg dist and everything in Distance computation?

[daniel]

Those are only very rough estimates pre-computed by CloudCompare before the user can actually compute the real distances. They should basically be ignored (it's just here for very advanced user who wish to change the octree level to speed-up computations in some tricky cases ;). They are computed thanks to a distance transform on a grid and are therefore very inaccurate.

You have to compute the distances first (with the red button), and then look at the generated scalar field properties (or the histogram, or fit a statistical distribution on the distances values, etc.) to get real statistics on the true distances.

[irsyadifirdaus]

Thanks you for your information