R help - Apr 2009 - error in trmesh (alphahull package)

Hello R community,

I have cross-posted with r-sig-geo as this issue could fall under either 
interest group I believe. 

I just came accross the alphahull package and am very pleased I may not 
need to use CGAL anymore for this purpose.   However, I am having a 
problem computing alpha shapes with my point data, and it seems to have 
to do with the spatial configuration of my points (which form long, 
skinny, contiguous clusters).  If I use rnorm to generate the same 
sample size with the same mean and SD, I can compute the alpha shape 
without trouble (which is why I think it might have to do with the 
spatial arrangement of my points).  See below:

# first here's a quick sample of my data (UTM coordinates)
 > xcoords[1:5]
[1] 670080.2 670080.2 670080.2 670080.2 670080.2
 > ycoords[1:5]
[1] 5005501 5005499 5005498 5005497 5005495
 > xcoords[1:5]
#try the ashape routine with error
 > alpha.shape<-ashape(xcoords,ycoords,15)
Error in tri.mesh(X) : error in trmesh

#get statistics to generate a similar dataset to test against
 > length(xcoords)
[1] 26257
 > length(ycoords)
[1] 26257
 > mean(xcoords)
[1] 670462.4
 > mean(ycoords)
[1] 5005382
 > sd(xcoords)
[1] 149.3114
 > sd(ycoords)
[1] 181.5950

#generate the test data
 > xtest<-rnorm(26257,670462.4,149.3)
 > ytest<-rnorm(26257,5005382,181.60)

# try ashape routine with success
 > alpha.shape<-ashape(xtest,ytest,15)
 > class(alpha.shape)
[1] "ashape"

Thanks for any insight into this!

Murray

ps I am able to compute the alpha shapes for this same dataset without 
problem using CGAL but I find it a pain to work with....

Hello,
The problem is that your first three points CANNOT be collinear. This is 
a feature of trmesh in tripack. So: either you "unsort" your data 
(keeping the indices you can sort them again) or alter (e.g. jitter) the 
coordinates very slightly. Either solution will do the work.
Cheers,
Jos? M. Blanco


Murray Richardson escribi?:
> Hello R community,
>
> I have cross-posted with r-sig-geo as this issue could fall under 
> either interest group I believe.
> I just came accross the alphahull package and am very pleased I may 
> not need to use CGAL anymore for this purpose.   However, I am having 
> a problem computing alpha shapes with my point data, and it seems to 
> have to do with the spatial configuration of my points (which form 
> long, skinny, contiguous clusters).  If I use rnorm to generate the 
> same sample size with the same mean and SD, I can compute the alpha 
> shape without trouble (which is why I think it might have to do with 
> the spatial arrangement of my points).  See below:
>
> # first here's a quick sample of my data (UTM coordinates)
> > xcoords[1:5]
> [1] 670080.2 670080.2 670080.2 670080.2 670080.2
> > ycoords[1:5]
> [1] 5005501 5005499 5005498 5005497 5005495
> > xcoords[1:5]
> #try the ashape routine with error
> > alpha.shape<-ashape(xcoords,ycoords,15)
> Error in tri.mesh(X) : error in trmesh
>
> #get statistics to generate a similar dataset to test against
> > length(xcoords)
> [1] 26257
> > length(ycoords)
> [1] 26257
> > mean(xcoords)
> [1] 670462.4
> > mean(ycoords)
> [1] 5005382
> > sd(xcoords)
> [1] 149.3114
> > sd(ycoords)
> [1] 181.5950
>
> #generate the test data
> > xtest<-rnorm(26257,670462.4,149.3)
> > ytest<-rnorm(26257,5005382,181.60)
>
> # try ashape routine with success
> > alpha.shape<-ashape(xtest,ytest,15)
> > class(alpha.shape)
> [1] "ashape"
>
> Thanks for any insight into this!
>
> Murray
>
> ps I am able to compute the alpha shapes for this same dataset without 
> problem using CGAL but I find it a pain to work with....
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

-- 
---------------------------------------
Jos? M. Blanco Moreno

Dept. de Biologia Vegetal (Bot?nica)
Universitat de Barcelona
Av. Diagonal 645 
08028 Barcelona SPAIN
---------------------------------------

phone: (+34) 934 039 863
fax: (+34) 934 112 842
e-mail: jmblanco at ub.edu