On 06/09/2013 02:47 PM, Serge Wroclawski wrote:
Hi Brian,
I think we talked this weekend, but if not, and you're here in SF,
I WISH! I was at SOTM US 2012 and took inspiration from that meeting to
continue on with this. Esp. Steve Coast asking people to put in address
data. Been ticking away at it off and on since then.
Does your building data data have the building type? building=shed or
building=greenhouse are valid tags.
SELECT COUNT(gid),layer FROM buildings GROUP BY layer ORDER BY layer;
count | layer
-------+-------------------------
29003 | BUILDING
93 | BUILDING INTERIOR
160 | CONCRETE
184 | DECK
545 | GREENHOUSE
42 | MISCELLANEOUS STRUCTURE
1678 | OBSCURED LINE
92 | POOL
116 | RUINS
85 | UNDER CONSTRUCTION
For now I immediately eliminate CONCRETE, DECK, POOL, RUINS and
"BUILDING INTERIOR" (which is the donut hole in buildings with atriums,
since the BUILDING polygon has a hole it's redundant)
That leaves BUILDING and which I will tag as "building=yes". Greenhouse
is also unambiguous so it will get "building=greenhouse". OBSCURED LINE
simply refers to some visual obstruction in the air photo required
interpolating an edge so it will get "building=yes" too.
Should I tag MISCELLANEOUS STRUCTURE and UNDER CONSTRUCTIONS as
"building=yes" and be done with it? MISC does not tell me anything and
UNDER CONSTRUCTION says it will be a building... could be a greenhouse
but I doubt it. I think those are probably all permitted building projects.
I suppose I could leave the small structures in. Initially I was trying
to identify houses as "building=residential" based on bedrooms > 0 from
the taxlots data but I realized I could not differentiate between a
house and a barn or large garage anyway.
There are a few reasons that parcels are not useful in OSM. Here are a few:
Thanks, I had a pretty good idea. I added a version of your comments to
http://wiki.openstreetmap.org/wiki/Import/Past_Problems#Not_all_data_should_be_imported
Now, let me ask you a few questions:
1. Can you point us to this data? Is it available online?
I started this page some time ago. It is for Corvallis but comes up if
you do a search for Benton County too --
http://wiki.openstreetmap.org/wiki/Corvallis,_Oregon
Most of the county data are online here:
http://gis.co.benton.or.us/GISDataDownload
One of the questions for me, for example, is the data license.
2. Are these buildings addressed?
I have taxlots and I have already added address data from taxlots to the
buildings in PostGIS for exampe
situs building
-----------------------------------------------------+------------
428 NW 26TH ST CORVALLIS, OR 97330-5423 | yes
4500 SW CAMPUS WAY CORVALLIS, OR 97333 |
900 SE CENTERPOINTE DR CORVALLIS, OR 97333 | no
3516 SE SHORELINE DR CORVALLIS, OR 97333 | no
3519 SE SHORELINE DR CORVALLIS, OR 97333 | greenhouse
2860 NW POLK AVE CORVALLIS, OR 97330 | no
I intend to conflate the addresses as the streets are already done.
TO me it seems a little bit odd to tag a 100 sq ft shed with an address
pulled from the taxlot data.
I tend to think the address should go on a point feature placed on the
centroid of the taxlot and then manually adjusted as needed.
I have a 'houses' point layer containing situs data too but it looks
like it covers only about 10% of the actual structures and all outside
city limits.
3. What existing data exists?
You mentioned that there are buildings on the local university. It's
quite common for a given town to have *some* editing, so your process
needs to take that into account.
Do you have a sense of how much existing data there is in your area?
There are roughly 80,000 structures in the buildings shapefile, which I
have reduced to 40,000 by clipping to Benton county and eliminating
small structures and as noted above things like "ruins" and "concrete"
(I don't have permission for Lane or Polk counties yet and probably
never will get Linn County (he's not amenable to data sharing even
though he is a promoter of open source))
I have not tried to count how many buildings are already in
OpenStreetMap in Benton county but I'd estimate it's in the low 100's,
all of them either on the campus or downtown. Manually checking them
should be easy. This could be an activity at our next local meetup.
If you look at the Corvallis page there are more comments on available
data there. I will update it when I get past this stage and move on.
I also would encourage you to work with local community for a hybrid
import, of the type that's being done in Seattle.
I just checked their page
http://wiki.openstreetmap.org/wiki/Seattle_Import
I can handle most of this with python and sql scripts, it's what I do
for fun as I am forced to do C# programming right now in my day job. :-)
Then we can get some volunteers to look at the results.
I think a good first step would be to show the data source, and also
the .osm file your process would create.
Okay. My first thought (12-Oct-2012) was that I can slam the building
footprints in because there is so little existing data in this area.
Then I realized I had to tag them accurately... :-)
I promised to show how to do data import at my next local meeting. I
will probably choose something simple for demo / community engagement as
a starting point, such as city parks. I have a shapefile that we can use
to double check existing hand drawn data and to add any parks that have
not already been drawn in by my friend MapJunkie.
- Serge
Cheers and thanks
Brian
_______________________________________________
Imports-us mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/imports-us
