C# "equalize" random points

Replies to This Discussion

Permalink Reply by Thomas on September 19, 2014 at 6:55pm

Hi!

Script 1:

If I understand you correctly you have a list of points and you want to select the points where the shortest distance to all other points is within a given interval.

A very simple attempt:

The result might depend on the order of the points (first point -> higher priority).

Attachments:

• filtergh.png, 134 KB
• filter.gh, 11 KB

Permalink Reply by peter fotiadis on September 20, 2014 at 12:35am

Hi Thomas,

There's something in the core logic that stops adding correct points (see attached : only a "region" of points is added, whilst others although suitable are excluded).

Attachments:

• projectpointsonmesh4.gh, 338 KB

Permalink Reply by peter fotiadis on September 20, 2014 at 2:20am

Additionally this projection does quite weird things  :

by weird I mean this  point to projected (on mesh) point oddity:

PS change tol (projection tolerance) and/or MaxDist :  spot extra weird results as well

Attachments:

• projectpointsonmesh5.gh, 340 KB

Permalink Reply by Thomas on September 20, 2014 at 3:43am

Hmm... the algorithm works correctly (some null testing etc. is missing...) but maybe I misunderstood the problem.

The algorithm starts with the first point in the set (green) and searches all untested points within the given interval (cyan):

It removes the points that are to close to each other (gray) and adds the rest to the final set:

It repeats the hole thing with the last added points:

After 7 iteration steps there are no untested points (pink) that matches the criterion:

All of them are too far away or too close to already added points.

The algorithm gives you different results depending on the order of the points (I don't think you can avoid this). Maybe you can filter more points with a different order.

Cheers!

Permalink Reply by peter fotiadis on September 20, 2014 at 4:29am

Hmm... so the solution could be finding  "regions/sets" of points ("clusters" if I may use the term) that satisfy the min/max criterion: meaning re-setting the green (start) point from the pink set until all pinks are tested.

 

Permalink Reply by Thomas on September 20, 2014 at 5:12am

You can do that but that means that you drop the max criterion (and in some cases also the min criterion...). Then the final set will contain points where the distance to the nearest neighbor is greater than the given max value.

Instead of filtering points from a random set you maybe can create your own point generator that takes regard of the min-max criterion.

Permalink Reply by Thomas on September 20, 2014 at 9:36am

1. Start with a given point

2. Populate the ring around the point with some random points:

3. Remove points that are to close to others

4. Repeat the whole thing with the new points

Attachments:

• populateMinMax.gh, 11 KB

Permalink Reply by peter fotiadis on September 20, 2014 at 11:13am

Well...that's really cool (I like the filter thing : better safe than sorry, he he)

And since the case is resolved (grazie mille : if you plan vacations in Greece, a ballistic Ducati 1199 is waiting for you) I was thinking the following :

Since my initial stupid approach actually creates sets of (potentially "isolated" ) points ... what about storing them in branches (per iteration) and then testing branch x against the rest about if there's any point in any other branch == with a point of the branch x (if yes addrange, remove and create a sum of the 2 sets etc etc). That way we could create "islands" (sets) of OK points potentially isolated (violating the initial requirement - but just for the shake of it).

Of course this is pointless ... but because of that it's worth doing it.

best, Peter

Permalink Reply by peter fotiadis on September 21, 2014 at 4:46am

OK, the last remaining thing:

In order to make this query to work :

we need a faceC of type curve, right?

so the final question is : given the fact that the BrepFace shown has holes (inner loops) how we can do it? (test shown turns GH script red by attempting to declare a curve as union (of Array of curves)).

Since I never design things with Rhino, the following "terminology" may be totally wrong:  we need a difference between BrepFace Outer Loop and Inner Loops - doe's such "curve" type exist (a region ???) - and if it is does can accept the .Contains(point)?

Permalink Reply by Thomas on September 21, 2014 at 7:56am

You're welcome!

Maybe you can change the type of "bounds" to Brep. Then you can use the yellow projection as "bounds" and check if a point lies on the Brep with something like this:

.Where(p => bounds.ClosestPoint(p).DistanceTo(p) < 1e-12)

And yes: With some improvements you can use the algorithm to create "islands" of valid points.

Permalink Reply by peter fotiadis on September 21, 2014 at 9:22am

Thomas

bounds.ClosestPoint(p, out u, out v) is bool

Permalink Reply by peter fotiadis on September 21, 2014 at 10:28am

Arrived in the known jelly fish mind state > close computer > start the Ducati.

Attachments:

• projectpointsonmesh5.gh, 380 KB
• projectpointsonmesh5.3dm, 334 KB

• ‹ Previous
• 1
• 2
• Next ›
• Page