CGAL users discussion list

["Scriven, David" <>]

They are related - the first problem was  a step in the construction of the multi-clusters - (assigning a number of clusters to a convex hull - your solution worked, thank you). Now I'm trying to find an efficient way to to get the nearest neighbour distance between these clusters. I've reduced the problem to finding nearby clusters to the the cluster of interest and then measuring the nearest neighbour distances between the each cluster of the group and the cluster of interest - I need to find which is the nearest cluster as well. What I do works, but I can see where it might fail.

I probably should rephrase it - I need an efficient and robust way to find the neighbouring clusters of a cluster. The rest I do already.

David Scriven, Ph.D. e-mail : Dept Cellular & Physiological Sci. ph: 604-822-7812 2350 Health Sciences Mall, fax: 604-822-2316 University of British Columbia, Vancouver, BC. V6T 1Z3, Canada

---

From: <> on behalf of Sebastien Loriot <>
Sent: 07 July 2022 01:07:58
To:
Subject: Re: [cgal-discuss] Calculating minimimum edge to edge distance between large clusters of points

 

[CAUTION: Non-UBC Email]

how is it different to the other problem you posted and for which I
gave you an answer?

Best,

Sebastien.

On 7/7/22 03:03, "Scriven, David" ( via cgal-discuss
Mailing List) wrote:
> I have a problem in which there are 2D planes on which there are
> thousands of clusters of points (x,y coordinates) - the size of the
> clusters varies from many tens to many thousands of points and the
> shapes are variable. For every cluster I have to find the nearest
> neighbour edge-to-edge distance. My solution so far has been to identify
> the nearby clusters through their centroids and then do a nearest
> neighbour search between the  cluster of interest and each cluster that
> was  identified as nearby. I have to  be able to identify the nearest
> cluster so I can examine at its properties, so I cannot pool the data.
>
>
> I can see many ways that this approach could fail - if one doesn't
> search far enough or if the centroids are very far from the edge of the
> cluster (imagine two very elongated clusters, one above the other  that
> have their tips very close but their centroids very far apart).  This
> sort of scenario becomes more likely when studying the relationship
> between multi-clusters (clusters of clusters, where the formation is
> defined by some parameter).
>
>
> Can someone suggest a rapid, robust and reliable method that could solve
> this problem?
>
>
> I was wondering if there was a way to use a Delaunay triangulation on
> all the points and then, in some way, pick out individual clusters and
> look at the connections from the edge of that cluster, but it's not
> obvious to me how to make this work.
>
>
> David
>
>
> David Scriven, Ph.D.                 e-mail :  <>Dept Cellular & Physiological Sci. ph: 604-822-7812 <tel:604-822-7812>
> 2350 Health Sciences Mall, fax: 604-822-2316 <tel:604-822-2316>
> University of British Columbia, Vancouver, BC. V6T 1Z3, Canada
>
>
> --
> You are currently subscribed to cgal-discuss.
> To unsubscribe or access the archives, go to
> https://sympa.inria.fr/sympa/info/cgal-discuss
>

--
You are currently subscribed to cgal-discuss.
To unsubscribe or access the archives, go to
https://sympa.inria.fr/sympa/info/cgal-discuss