Hi,

On Mon, Feb 24, 2014 at 04:46:25PM -0500, Saikrishna Arcot wrote:
> I'm dropping the city field from the import because the city in the import 
> data doesn't accurately correspond to the actual city. As it is, the cities 
> in OSM most likely have well-defined borders in this county. As for the zip 
> codes, the zip code for an address will be included if it is located inside 
> the zip code area defined in the ZCTA; otherwise, the zip code will be 
> excluded.

>From previous discussions with people in the US I understand that
the administrative boundaries of a city do not necessarily line up
with the postal cities. I wonder if that might be the reason why you
see mismatches. In that case, it might be well worth including addr:city
in the import, so that OSM gets better postal addresses.

There is a similar issue with zip codes. The areas in ZCTA are only
approximated because they use some aggregations[1] and US zip codes 
are in general difficult to describe as areas. Your data might actually
be more precise. Do you have any means to point check the data?

Sarah

[1] http://en.wikipedia.org/wiki/ZCTA


> There are two expected errors to result from this import: If a building has 
> multiple addresses (which are all located outside of the building), then the 
> building may have one of those addresses instead of not having a single 
> address. In addition, a building that has only one address might not have the 
> address node merged in; this is because the address node is too far away.
> 
> I'm currently dividing Fulton County into grids somewhat based on census 
> block groups so that there is at most 4000 nodes and ways in each upload. 
> Some sections will only have 2000 nodes and ways, while a few sections may go 
> above the 4000 limit. The import will start in Atlanta and spread outwards. 
> Once this is done, I'll notify mappers in the region who have previously 
> edited/added addresses and building about the import so that they are aware 
> about this.
> 
> -- 
> Saikrishna Arcot
> On Wednesday, February 05, 2014 10:00:53 PM you wrote:
> > Upon first glance at a sample in the suburbs, I saw that many addresses 
> > that could be imported weren't because the address points and the 
> > buildings were too far away. Therefore, I bumped up the maximum 
> > distance between building and address to 10 meters (approximately 32 
> > feet). Most addresses are (supposedly) located in Roswell, GA, along 
> > with some addresses having a city of Atlanta and Alpharetta.
> > 
> > Sample files for suburbs:
> > Suburbs.osc: 
> > https://drive.google.com/file/d/0B30vrP6AZTFydFdLcDhjcDNSRFU/edit?usp=sharing
> > Suburbs_errors.osc: 
> > https://drive.google.com/file/d/0B30vrP6AZTFyaFR0Sm5CaVVBakE/edit?usp=sharing
> > Suburbs.log: 
> > https://drive.google.com/file/d/0B30vrP6AZTFybHBycWVaWjhoUzg/edit?usp=sharing
> > 
> > Saikrishna Arcot
> > 
> > On Wed 05 Feb 2014 03:47:49 PM EST, Saikrishna Arcot wrote:
> > > The conversion program now also reads in the relations for the
> > > building data. Also, for the address-merging, if there is exactly one
> > > address inside the building, then that address is merged into the
> > > building way. For any buildings that don't have addresses after this
> > > process, if there is an address node within 5 meters of the building,
> > > that address is used for the building.
> > >
> > > Zipcode data is also being read in. Currently, if the zipcode for an
> > > address has a zipcode area that was read in, but is not in the zipcode
> > > area, then the address's zipcode is excluded from the output file.
> > >
> > > Also, existing buildings are read in (ways only). If a building in
> > > this dataset intersects an existing building, then the building is not
> > > added.
> > >
> > > Updated sample files:
> > > Output.osc:
> > > https://drive.google.com/file/d/0B30vrP6AZTFyYldqT0hJaVVCdGc/edit?usp=sharing
> > > Output_errors.osc:
> > > https://drive.google.com/file/d/0B30vrP6AZTFyRjdVSWdIb1E3Tm8/edit?usp=sharing
> > > Output.log:
> > > https://drive.google.com/file/d/0B30vrP6AZTFyOGJEeTkwanFscVU/edit?usp=sharing
> > >
> > > I'll be updating the wiki-page soon and adding sample files for an
> > > area outside of Atlanta.
> > > Saikrishna Arcot
> > > On 01/29/2014 09:15 AM, Saikrishna Arcot wrote:
> > >> Updates:
> > >>
> > >> The program now reads in the building data (no relations yet) from
> > >> the dataset and merges the address point data into the building if
> > >> the point is inside the building. It also removes some unnecessary
> > >> nodes from the building.
> > >>
> > >> I'll start working on reading in the zipcode data and adjusting the
> > >> zipcodes for addresses.
> > >>
> > >> I've updated the wiki page
> > >> <https://wiki.openstreetmap.org/wiki/Address_Data_Import_for_Fulton_County,_Georgia>
> > >> to include data about the building dataset.
> > >>
> > >> Saikrishna Arcot
> > >>
> > >> On Sat 18 Jan 2014 10:07:12 PM EST, Saikrishna Arcot wrote:
> > >>>
> > >>> Carl,
> > >>>
> > >>> I managed to grab hold of the ZCTA5 dataset through the HTTP interface
> > >>> (FTP, for some reason, doesn't like me).
> > >>>
> > >>> I do recall seeing the years on the buildings. A few of the overlapping
> > >>> buildings had different years, and so that made it easy to determine
> > >>> which one to use, but many have the same year. This dataset will take
> > >>> some work.
> > >>>
> > >>> Saikrishna Arcot
> > >>>
> > >>> On Sat 18 Jan 2014 06:50:03 PM EST, Carl Anderson wrote:
> > >>>>
> > >>>> Saikrishna,
> > >>>>
> > >>>> From personal knowledge I can tell you that the Fulton County
> > >>>> buildings data layer contains data of several lineages, time frames,
> > >>>> and data qualities. There are flags indicating the known time of each
> > >>>> outline but the flags should be use with caution.
> > >>>>
> > >>>> The vast majority of data is from a 1988 planimetric data capture from
> > >>>> ortho-stereo pairs. Those have an NSSDA-part3 E95* positional
> > >>>> accuracy of 2.5 feet. Some of the buildings captured no longer exist.
> > >>>> Some data is derived from other remote sensing data. Some data is
> > >>>> from "heads up" data capture methods. Finally some data is derived
> > >>>> from "Tax Assessment measured drawings" depicted as turtle graphics
> > >>>> that were repositioned into the correct space in State Plane
> > >>>> Coordinate system.
> > >>>>
> > >>>> Be careful.
> > >>>>
> > >>>> C.
> > >>>>
> > >>>> * NSSDA-psrt3
> > >>>> http://www.fgdc.gov/standards/projects/FGDC-standards-projects/accuracy/part3/chapter3
> > >>>>
> > >>>>
> > >>>> On Sat, Jan 18, 2014 at 2:48 PM, Saikrishna Arcot
> > >>>> <[email protected]> wrote:
> > >>>>>
> > >>>>> A few updates:
> > >>>>>
> > >>>>> Since I can't get the ZCTA5 file, I'll just have the existing city
> > >>>>> name
> > >>>>> converted to title case, and add in an exception for DeKalb, since
> > >>>>> that
> > >>>>> seems to be the only one with mixed case.
> > >>>>>
> > >>>>> Also, I looked at the building data from Fulton County GIS, and it
> > >>>>> looks
> > >>>>> fairly good. There is at least one main issues there: Some
> > >>>>> buildings have
> > >>>>> overlapping buildings; in other words, the same building has been
> > >>>>> drawn two
> > >>>>> or three times. The other building ways are generally shifted
> > >>>>> slightly, and
> > >>>>> have more/less detail than the others.
> > >>>>>
> > >>>>> If we are to import these buildings, I can add in a check to see
> > >>>>> if the
> > >>>>> building being imported intersects with or is contained in an existing
> > >>>>> building. I added the source dataset and converted OSM files
> > >>>>> (using the
> > >>>>> OpenData plugin in JOSM and a splitter script to break it into smaller
> > >>>>> files) on the wiki page. The area that each file covers is listed
> > >>>>> in the
> > >>>>> areas.list file.
> > >>>>>
> > >>>>>
> > >>>>> On Fri 17 Jan 2014 01:25:48 PM EST, Bryce Nesbitt wrote:
> > >>>>>
> > >>>>>
> > >>>>> The issue of 'source attribution' is one that comes up over and over
> > >>>>> again.
> > >>>>> Having a good standard answer to that would be nice. I think OSM
> > >>>>> should tag the contributing source
> > >>>>> in a way that would allow any region to be selected, and show a list
> > >>>>> of contributors, be they OSM
> > >>>>> users, background map providers, or data providers.
> > >>>>>
> > >>>>>
> > >>>>> _______________________________________________
> > >>>>> Imports mailing list
> > >>>>> [email protected]
> > >>>>> https://lists.openstreetmap.org/listinfo/imports
> > >>>>>
> > >>>>> --
> > >>>>> Saikrishna Arcot
> > >>>>>
> > >>>>>
> > >>>>> _______________________________________________
> > >>>>> Imports mailing list
> > >>>>> [email protected]
> > >>>>> https://lists.openstreetmap.org/listinfo/imports
> > >>>>>
> > >>>>> --
> > >>>>> Saikrishna Arcot
> > >



> _______________________________________________
> Imports mailing list
> [email protected]
> https://lists.openstreetmap.org/listinfo/imports


_______________________________________________
Imports mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/imports

Reply via email to