Hi Saikrishna, Just a quick look.
- The street number tag is addr:housenumber, not addr:houseNumber. OSM tags don't use CamelCase - didn't the source data also have zip codes? - If you are going to dump an address, you should write it out into a separate file that will be reviewed by hand. These are the interesting bits.. - I am not sure the check that looks for addresses too close to a street makes sense. - I don't see a check in the code to look for addresses that don't match any street name. Thanks Jason. . On Sat, Jan 11, 2014 at 6:58 PM, Saikrishna Arcot <[email protected]> wrote: > Another update: > > I've added checks into the application to make sure that addresses are > between 5 and 100 meters from the street centerline, and that each address > is at least 4 meters from another address. The source code located here has > been updated. Note that the code with the distance checks is in the > geos-library branch, not the master branch. I've updated the file here to > show a sample output. > > Saikrishna Arcot > > On 01/09/2014 11:54 AM, Saikrishna Arcot wrote: > > (In case my previous message doesn't go through) > > For the record, I took a quick look at the converted data (after processing > and all) in JOSM, and the address data seems fairly accurate (at least in > midtown Atlanta). There are a few points that are half-a-house over and some > building that have two points, but mostly all points seem to be located > correctly. > > A slightly bigger issue I'm seeing is POIs and buildings that have an > addr:housenumber, but no addr:street, which makes it difficult to determine > what street the building is on and to detect that as an existing address. In > the Atlanta region, there are a few that can probably be manually handled. > Currently, I'm ignoring anything that has an addr:housenumber but no > addr:street, but this will cause some duplicate data. > > As a status update, I can generate a sample OsmChange file after running it > through my application, but it does very few checks to see if the address is > located at a reasonable location. I've uploaded a sample file for review so > far here. Note that addr:city is missing. > > I'll get in touch with the Fulton County GIS to see if they can allow the > building footprint data to be used in OSM. However, I don't see how the tax > parcel data can help in OSM. > > Saikrishna Arcot > > On 01/09/2014 11:49 AM, Saikrishna Arcot wrote: > > For the record, I took a quick look at the converted data (after processing > and all) in JOSM, and the address data seems fairly accurate (at least in > midtown Atlanta). There are a few points that are half-a-house over and some > building that have two points, but mostly all points seem to be located > correctly. > > A slightly bigger issue I'm seeing is POIs and buildings that have an > addr:housenumber, but no addr:street, which makes it difficult to determine > what street the building is on and to detect that as an existing address. In > the Atlanta region, there are a few that can probably be manually handled. > Currently, I'm ignoring anything that has an addr:housenumber but no > addr:street, but this will cause some duplicate data. > > As a status update, I can generate a sample OsmChange file after running it > through my application, but it does very few checks to see if the address is > located at a reasonable location. I've attached a sample file for review so > far. Note that addr:city is missing. > > I'll get in touch with the Fulton County GIS to see if they can allow the > building footprint data to be used in OSM. However, I don't see how the tax > parcel data can help in OSM. > > Saikrishna Arcot > > On 01/09/2014 11:08 AM, Serge Wroclawski wrote: > > > On Thu, Jan 9, 2014 at 10:59 AM, Clifford Snow <[email protected]> > wrote: >> >> >> I agree with Martijn. Addresses add real value to OSM. With an address >> node, routing works better, adding POI information is easier, and it aids >> disaster recovery work. I'll admit, building outlines look nice and alone >> they aid mappers adding POIs and also can be used in disaster recovery work. >> So each adds value. >> > >> >> About the disaster recovery work. I had a conversation with a CERT member >> last week. He was frustrated using address interpolation to plan for >> response efforts. He would love to see address nodes added. >> > > Let's not confuse addresses as raw nodes with no addresses. > > That's a bit like me saying a healthy meal is better than junk food, and the > response being that junk food is better than no food. It's not what I'm > saying. > > The issues with address points are: > > 1. They're often in the wrong places. You just don't know it until you have > building footprints to compare it to. > > 2. There's not a whole lot of data about addresses being updated > > 3. (as was brought up in the NYC building point discussion) There's no code > that I'm aware of that correctly parses address points inside building > polyogons in such a way that non-addressed POIs get the address attributes. > > 4. If geolocation is the issue (which it seems to be), then we could feed > that data to the geolocator directly, rather than placing it in OSM. > > - Serge > > > _______________________________________________ > Imports mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/imports > > > > > > _______________________________________________ > Imports mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/imports > _______________________________________________ Imports mailing list [email protected] https://lists.openstreetmap.org/listinfo/imports
