I did some basic sanity checking in my quick script and there is a lot of points that are doubled up (i.e have the same lat/lon) which indicates that the data does form sensible/consistent boundaries.
My 'intuition' is that the shape=>boundary problem is solvable, I'll just need to put some thought in to it - actually as I write this, I think I know the approach ;-) Solving this will also help with one of the other issues that I came across, which was 'curve simplification'. There are vast numbers of redundant points in many of the boundaries where they make no difference to the shape. However doing curve simplification on closed shapes with shared boundaries results in different points being removed from the two boundaries. Small samples sounds like a good first approach - I have lots of gps tracks for the area I live in that were taken with a roof mounted aerial, so I have a reasonably high level of confidence in them cheers On Thu, Feb 5, 2009 at 4:02 PM, Darrin Smith <bel...@beldin.org> wrote: > On Thu, 5 Feb 2009 15:52:43 +1100 > Franc Carter <franc.car...@gmail.com> wrote: > > > From a 'philosophical point of view', I tend to agree that suburbs > > are made of > > a set of boundaries between adjacent areas. This was not how I did it > > in my first (very quick) attempt ;-( > > An advantage of having to sort out the legal issue means you get a bit > of time to fiddle around trying out options before you get the full > a-ok and import it ;) > > > The data is in shapefiles that define each suburb boundary > > individually, so I'll have a think about how to extract out the > > individual borders (suggestions > > welcome) > > Hmm, so there's no real surety that 2 adjacent suburbs even share the > same boundary? Perhaps then the single area option might have some > merit from a 'getting the data in there' point of view.... or we write > a convoluted script to correlate things... > > > One question about aligning them that springs to mind is 'what should > > we align' - I wonder if the accuracy of the data is better than the > > average accuracy > > of a gps or yahoo imagery. > > That's a tricky question because it might be more 'accurate' because it > might measure to an exact positional definition but is that useful or > relevant to the OSM structure whereby a boundary down the middle of the > road is more conceptually accurate.... > > Guess we have to get a small sample of the data into a city somewhere > where we have plenty of GPS as a trial run (once we have the full ok). > and see how it correlates to reality. > > GPS + Yahoo never correlate enough (at least in SA) to make it possible > for both to be relevant :) > > _______________________________________________ > Talk-au mailing list > Talk-au@openstreetmap.org > http://lists.openstreetmap.org/listinfo/talk-au > -- Franc
_______________________________________________ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
