[Talk-ca] OSM Geobase import: giving a try
Frank Steggink
steggink at steggink.org
Tue Feb 17 02:31:32 GMT 2009
Hi Richard, Sam, Steve,
Thanks for your feedback. I'll continue using PostGIS (good for my own skills as
well), and look at the other suggestions you've made. I'll also leave the NIDs
on the data which I will eventually upload. They don't hurt, although there is
the theoretical possibility that due to subsequent edits they get shifted
(splitting and joining).
I'll also investigate PostGIS buffering a bit more. 10 to 20 meters should be
enough. In cases of roads which have been shifted 100 meters, they need better
investigation anyways. In case they turn out to be drawn from low-res imagery,
they can better be replaced by the Geobase data. Hopefully the buffering won't
cause data to be removed accidentally. For areas where the OSM density is quite
small (as in my test areas) it is no big deal to remove duplicate Geobase roads
manually. Better to be safe than sorry. How would RoadMatcher deal with a case
where two roads are aligned, but one of them has been shifted?
With regard of this I also think a manual step should always take place. (Maybe
even to evaluate discarded data.) We must only make sure that it covers those
parts which can't be done automatically, so the required amount of time spent of
it is as small as possible. Our brains can still make judgments better than the
computer. I don't think anyone has very sophisticated algorithms which eliminate
human input in the process ;)
I hope to really start with the Geobase import soon. I came across the Global
Administrative Areas (GADM) database, which is part of the BioGeo project
mentioned at the potential datasources. I've looked more closely at the data,
and exchanged some ideas about the import of this data as administrative
boundaries with the guys on #osm-nl. (Sorry, sometimes it's necessary to chat in
my mother language ;)) I also inquired after use of the data by OSM, because of
their copyright (CC-BY-NC-SA 3.0), and because they aggregated data from various
other sources.
Regards,
Frank
More information about the Talk-ca
mailing list