[Imports] [Talk-ca] [Import] Ottawa Buildings & Addresses [Statistics Canada project]
james2432 at gmail.com
Thu Oct 20 03:17:09 UTC 2016
Steve, the only replacement, is if ottawa's geometry is better. There are a
lot of buildings in Ottawa that are CanVec 6.0 which is ancient! Not to
mention pretty crumby/blocky.
As for manually mapped buildings like stittsville(AJ Ashton's buildings)
just re-align it as they offset to bing(2-10meters off) in the most cases
Osmose is a tool we can use to further validate that our data on the map,
it's not ment to be an after thought. JOSM errors should be cleared before
uploading(we don't need to burdon these problems on new mappers: see canvec
errors). Yes I do believe if there's a river going through a house(there
isnt by the way) it should be corrected. This will also lead to a map with
less errors on it. Expected errors that might occur are self intersecting
ways(there's often a small tail pointing in wards to the building then
coming back on the same node) I'm not sure if this is like a trap street
for the data(to avoid unauthorised copying) as it is present once or twice
in about 70% of the tiles.
Of the tiles I've faced with my 2 other collegues, there were only 2-3 that
may need surveying not hundreds. The major issue with addresses is
placement in the backyard and not directly on the building (big back yards
or parks might have 2-3 nodes spread over the parcel, but is still accurate.
There are no more tiles in OSM that would be considered a success as
currently woodpecker(Frederik Ramm) is reverting the all of them.
The people that will be dealing with the import are expert level mappers.
One of them works in the GIS field another sells satelite imagery. Myself,
I am on the programming side of things and have been mapping for many
years(have dealt with importing canvec data and clearing errors).
Johnwhelan(expert mapper and frequent contributor to HOT tasks) has also
offered to help with the validation process.
As the documentation states, this import will be handled by the OSMCanada
team and a few expert mappers if they wish to join. We are not picking a
day one mapper and expecting him to know what has to be done.
Once the data has been imported building footprints and addresses with
coincide together(no backyard addresses) and will provide us with more
accurate routing/geolocalization power
On Oct 19, 2016 9:59 PM, "Steve Singer" <steve at ssinger.info> wrote:
> On Wed, 19 Oct 2016, James wrote:
> Seems like a good enough time like any other to talk about the import of
>> Ottawa buildings and addresses
>> into OpenStreetMap.
>> Documentation is available here:
> A few comments
> * Replacing Building Geometries - I don't like the policy of replacing
> existing building geometries with imported ones. I feel that large scale
> replacement of hand traced work with automated/imported work will
> discourage future mappers. Also sometimes which is 'better' is a matter of
> opinion and if a few mappers are replacing hundreds or thousands of
> buildings in a short period of time it is hard to have discussion on an
> individual building basis. I don't have issues with a mapper manually
> changing or replacing the geometries but the bulk non-traced nature
> concerns me. I have heard numerous complaints over the years from mappers
> who traced lakes manually only to see them replaced by canvec imported
> * Notes - How many 'What is the address of this house?' notes do we think
> this is going to generate? Adding a few dozen notes in the Ottawa area
> isn't an issue but if this process adds hundreds of notes that require
> manual survey then I'm not sure that is better than just skipping the
> address(but this could be discussed)
> I think the import plan needs to talk more about what the expected
> 'quality' is. I am not talking about the quality of the data but the
> quality of the merging
> For example:
> * Do we expect that before uploading a tile all JOSM validator errors get
> cleared on the objects being uploaded?
> * If an existing object such as a road, alley or stream runs through the
> geometry of the building then will the pre-upload validation process
> require that this be fixed
> Part of the problem with the Canvec imports is that quality is very much a
> factor of who imported a tile in a particular area. We never really came
> up with minimum standards for the validation and repair requirements. I
> think a checklist of specific types of problems that we expect will need
> correcting/validating by the person uploading the tile would help
> The plan also talks about using post upload validation tools like Osmose.
> Is the proposed workflow something along the lines of
> 1) Grab a .osm tile with buildings
> 2) Work in JOSM to address any issues the importer finds by eyeballing
> things or using the validator
> 3) Upload
> 4) Pick another tile repeat
> 5) Once there are no more tiles left check Osmose and other tools
> I think it would be better to fix all the issues with the footprints in
> the first tile before moving onto a second tile.
> Are there examples of tasking manager coordinated building imports that
> the community considers a success? If so how does this plan compare to what
> has worked well. (This isn't specifically directed at James)
> I haven't yet looked at samples of the .OSM files
>> Discussion with local mappers has happened in person for multiple
>> months(community buy in is very high in
>> these meetings and we all saw benefit in including this data) :
>> Once positive discussion of a period of 2 weeks has been met, we would
>> like to start the import of said
>> Thank you.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Imports