[postgis-users] identifying similar geometries

[Charles Galpin]

Thanks for that.  Unfortunately it is also subject to the same problem of me not knowing how the old links were split. If the old link was split into a very small piece at the start, and then one or more larger pieces following it, the larger pieces show a better match for the old link.  In general this would be a good way to get the biggest part that matches, but since I know these are replacements for the old links I'd rather match the first portion of the link.

I did some tests comparing the hausdorff distance between the first 10% of the links, and this works better, but that is also dependent on link lengths. I suppose if I calculated the actual length of the 10% and mapped that back to the appropriate/equivalent sized portion of the new link it would be more accurate, assuming 10% of the old is not bigger than the first new segment.

But this should come in handy if I take the approach of starting at the start point and keep adding links to the new link until the hausdorff distance decreases to zero to be able to tell I have the right links. Although at that point I think I can just check the end points :)

Thanks,
charles

On Aug 11, 2011, at 3:16 AM, Sandro Santilli wrote:

> On Wed, Aug 10, 2011 at 05:26:06PM -0400, Charles Galpin wrote:
> 
>> Does anyone know of a way to say "this geometry looks [a lot] like this one"?
> 
> ST_HausdorffDistance 
> 
> --strk;