Ok Moritz, I understand. To my mind v.delaunay was a way to /begin/ solving the problem, given that proximity of surrounding points is already the criteria for triangulation.
Whatever the right choice, it may be interesting, on a set of data, to test both ways. In the case of a big network v.delaunay would provide a lighter net, perhaps easier to handle for v.net.salesman. Vincent Le mercredi 15 avril 2009 à 18:23 +0200, Moritz Lennert a écrit : > On 15/04/09 18:11, Vincent Bain wrote: > > Sorry if I'm beside the point, but I insist, being myself interested in > > this topic... > > > > My suggestion of a delaunay triangle net to join polygons centroids > > seems to be totally "beside the point" :-( > > I would just like to know why. Probably sth I did not catch ? > > Haven't thought this through completely, but: v.delaunay will only > create links between centroids close to each other, whereas > v.net.visibility will create links between all centroids. This leaves > v.net.salesman with a larger option of paths to choose from. As we are > not bound to any paths, this seems more appropriate to me. > > Does this sound convincing ? > > Moritz > > _______________________________________________ grass-user mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/grass-user
