[Talk-us] Massachusetts addresses/tiger addresses

[Greg Troxel]

Jason Remillard <remillard.jason at gmail.com> writes:

> I am not sure if anybody is actually working on importing the tiger address
> ranges or not. Just in case somebody is ... The Massachusetts mapping
> department (MassGIS) just this week released data for all of the buildings
> in the state!
>
> http://www.mass.gov/anf/research-and-tech/it-serv-and-support/application-serv/office-of-geographic-information-massgis/datalayers/structures.html

That's great to hear they have released this.  I've heard that it was in
progress for a while, and have been cogitating on how to use the data.

> MassGIS also has very good parcel data with addresses. I think we have at
> this point everything we need to do an address import into each building,
> rather than using address interpolation. For MA, using the state data
> instead of the tiger data would probably be better.

I agree that the state data is likely far higher quality than TIGER,
based on what I've seen of the data so far (buildings near Boston,
parcel data).

I think that getting from here to there, where there means

  all building footprints added and carefully merged with the many
  hand-drawn buildings

  some sort of tying of address data from parcels to buildings that are
  within parcels

is going to be a long process, and we shouldn't expect that it will be
easy, especially given the need to be very careful about the social
consequences of imports (both the actual consequences and the
perceptions within the world OSM community :-).  But, with a bunch of
people working together I think it's reasonable to achieve in 2013.


My take on first steps is to get code that translates the state data to
osm format, and to be able to separate it into per-town chunks (if it
doesn't already come that way), and maybe smaller chunks, and to review
that.  Then we can see what's next.  Probably something that takes a
$town-state-buildings.osm file and diffs it agains the osm database and
produces

  A) a subset that doesn't overlay any existing buildings
  B) a subset that is an exact match to existing builds
  C) a subset that have a very good match to existing buildings
  D) a subset that is not A, B or C

Presumably A can be spot-checked and added.  B is completed work
(somehow).  C is probably not worth worrying about, but compulsive types
can look at the buildings and imagery and adjust if warranted.   D will
be hopefully not too many, and deserve manual one-building-at-a-time
attention.


For addressing, I wonder about:

  For each parcel, finding the point of tangency to the road matching
  the parcel name, and putting in a point 5m off the road midway along
  the tangency.   This has a good chance of being either close enough
  that if you are at the point the address location is obvious or close
  to the driveway.  This could be done without buildings.

  For each parcel, see how many buildings are in in.  If exactly 1 and
  there are no address tags in the parcel, put the address tags on the
  building, guessing that it's probably right.


Greg

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20121207/380296e4/attachment.pgp>