[Talk-us] Baltimore Building Outlines Import

Matthew Petroff openstreetmap at mpetroff.net
Sun May 5 07:10:57 UTC 2013

Thank you everyone for the suggestions.

I had originally tried using QGIS's simplify function, but it simplified things
far too much, even with the minimum tolerance, and I had given up on
simplification. I tried ogr2ogr's simplify function due to the suggestion, and
it worked much better. This reduced the file from about 3.5 million nodes to
2 million nodes.

I had originally assigned addresses using QGIS's "join attributes by location"
function, set to take the attributes of the first located feature using the
building outlines and the parcel map. This lead to errors due to slight
misalignments of the building outlines and the parcel map, as it would often
use information from adjacent parcels if there was slight overlap. I have since
redone address assignments using PostGIS to assign the address of the parcel
that intersects with the building's centroid. This should result in correct
address information in virtually all cases, as it is quite rare for a building
to be on two or more parcels. Multiple addresses on large buildings are not
being handled; the only address data available from the city is the parcel map,
so buildings are being assigned that address. In regards to labeling garages,
it's much easier to do so. I suppose one could only label the largest building
on each parcel, but that would probably lead to buildings that should be labeled
not being labeled.

I used ogr2osm for conversion and tagging this time around; it made things much
easier. I am planning on tagging both the individual buildings and the
changesets with source information. I have removed the "addr:state,"
"addr:country," and "addr:inclusion" tags.

When run through the JOSM validator, there are no errors, but there are ~47k
warning, mostly for "crossed buildings" and "building inside building," with a
few "intersection between multipolygon ways," "self-intersecting ways," and
"overlapping ways." From the cases I looked at, these are the result of slight
overlap between adjacent row houses, and I don't think they are an issue.

I have not yet decided how I will upload the data. I'll probably test a few
things on the dev server first.

In regards to licensing, I think the confusion results from multiple departments
adding data to OpenBaltimore. I will get in contact with the city again and try
to get a more formal public domain dedication for the data.

Here is an updated copy of the buildings:

I left the data in one file this time after I found that JOSM still worked with
the large amount of data, although it was a bit sluggish.


On Sat, May 4, 2013 at 12:03 AM, Matthew Petroff
<openstreetmap at mpetroff.net> wrote:
> Hello,
> The City of Baltimore provides a large amount of public domain GIS data through
> their data portal [1]. Included are city wide building footprints [2], which I
> would like to import into OpenStreetMap. Other users have already started
> importing this data in a less automated way. I confirmed with the city GIS
> office that the data is indeed public domain.
> Using QGIS, I assigned approximate street addresses to each building using a
> parcel map [3] and used the field calculator to clean up the labels and remove
> abbreviations. In addition, I removed all data that intersected with existing
> buildings to preserve existing work. I then separated the data into smaller
> chunks and converted the Shapefiles to OSM with Merkaartor. After simplifying
> Merkaartor's output using osmconvert's "--drop-author" switch, I tagged the
> data with sed, before finally using JOSM to remove duplicate vertices and empty
> tags. My only qualm with the data is that some buildings have more nodes than
> they need, but I'm not sure what can be done about it besides manually reviewing
> and simplifying all 200k+ outlines.
> I've currently tagged the data with:
> building=yes
> addr:city=Baltimore
> addr:state=MD
> addr:country=US
> addr:inclusion=estimate
> source=Baltimore City GIS
> addr:housenumber=[Based on parcel data]
> addr:street=[Based on parcel data]
> I've uploaded the files here for feedback:
> http://db.tt/P9qh1hIm
> Thoughts, suggestions, questions?
> -Matthew Petroff
> [1] https://data.baltimorecity.gov/
> [2] https://data.baltimorecity.gov/Geographic/Building-Footprint-Shape/deus-s85f
> [3] https://data.baltimorecity.gov/Geographic/Parcels-Shape/jk3c-vrfy

More information about the Talk-us mailing list