[OSM-talk] Google Maps comparison
dair at refnum.com
Thu Oct 11 17:38:01 BST 2007
Andy Robinson wrote:
>That's really, really nice work. I was quite struck by the sheer
>number of differences. I've seen plenty in my area but as yet haven't
>looked closely enough to see just how many. Your results suggesting 9
>per square kilometre of urban area would make an astounding number
>for the whole country if it ramps linearly.
It is hard to know how to interpret them - it's quite
overpowering to see an entire town covered in push-pins, as it's
hard to convey significance.
For example, the non-existent stub opposite Lucas Grange is
"wrong" - but is not really important
<http://tinyurl.com/39bypx>. On the other hand, these two roads
have a line of trees separating them so are completely separate
for road traffic <http://tinyurl.com/3ycero>.
They're both wrong, but the latter is more likely to cause
problems (if you are routed through there, you'll have to turn back).
The approach I took was just to ask myself if the OSM map was a
better representation of reality - if it was, it got flagged.
It would be interesting to find some towns of a similar size in
different areas of the UK/different countries, and verify them
to the same degree.
After a couple of those you would start to get a feel for how
(a)typical this case is.
There are no doubt still some mistakes in the OSM map, and some
things flagged as errors are really a judgement call.
E.g., the Lucas Grange stub is actually someone's drive; these
seem to be included at random on Google Maps, even though
they're not part of the road network. You can see similar
choices with little parking stubs like this:
I always capture these, but Google/Yahoo include some but not others.
My guess is these kind of things are deliberately
included/excluded to give the map a "fingerprint", as there
doesn't seem to be any real consistency for when they're captured.
dair at refnum.com http://www.deathvalleycycle.com/
More information about the talk