[OSM-talk] data plucked from who-knows-where?

[Andy Allan]

On Mon, Nov 3, 2008 at 2:28 PM, David Ebling <dave_ebling at yahoo.co.uk> wrote:

> On a complete tangent to this conversation... I was curious about the area
> Richard mentioned, so looked it up:
> http://www.openstreetmap.org/?lat=52.98283&lon=-1.99189&zoom=15&layers=B000FTF
>
> And was amazed that someone has obviously put a lot of work into this, and
> yet it bears very little geometric similarity to this (for comparison
> only!):
> http://www.multimap.com/maps/?qs=cheadle&countryCode=GB#map=52.98372,-2.00288|15|4&bd=useful_information&loc=GB:52.98896:-1.98721:14|cheadle|Cheadle,%20Stoke-on-Trent,%20Staffordshire,%20England,%20ST10%201
>
> I know which I am inclined to believe...

This is brilliant. It's a lot more extensive than the bit that you
linked to though - for those of you who are registered for ITOWorlds'
OSM mapper product, clicky the linky:

http://www.itoworld.com/product/osm/map?area=310:16&show=users&sort=total-

(sign in *first* :-) ). You can see that it's so extensive it's even
off the top of this area.

> I wonder how easy it will be to improve the accuracy of data such as this,
> where the topology and road naming is probably mainly correct (I imagine
> they sketched maps as they went) but the geometry is way off. When I get a
> moment I'll have to look and see whether any GPS traces already in the
> system can be used to improve this area. Maybe we could even make a
> significant improvement with Landsat images?

I've spent 5 minutes looking at it compared to NPE - it actually seems
quite reasonable topology, and looks very similar to when I've done
TIGER fixups. I'd be hesitant about using NPE for positioning, but
given how really, really inaccurate this is it might be a good start.

Cheers,
Andy