[Talk-ca] Building Import

[Pierre Béland]

Bonjour Tim
Mon outil d'analyse Qualité dont les données sont publiées sur OpenDataLabRDC est basé sur PostgreSQL-Postgis.   Je suis à nettoyer / documenter le code et prévoit le publier sur github.  J'ai commencé à regarder les outils possibles, mais peu de documentation disponible. On parle par exemple de OpenCarto, mais l'info n'est plus disponible. A voir si possible à l'aide de Grass. 
https://gis.stackexchange.com/questions/15612/is-it-possible-to-simplify-orthogonal-polygons-with-opencarto-java-library
 
Pierre 
 

    Le mardi 26 mars 2019 21 h 33 min 39 s HAE, Tim Elrick <osm at elrick.de> a écrit :  
 
 I sent Daniel a sample of Montreal (Outrement) from the Open Building 
Database and Daniel's algorithm performed really well. It could reduce 
the vertices count by 13% without loosing or even improving data quality 
(as it orthogonalized the buildings). Even difficult buildings were 
treated well [1].

As OSM is mainly built on open source tools, the OSMF tries to work with 
open source tools only and the process should be reproducible (if not 
for this import, then for the next one somewhere else in the OSM 
cosmos), I suggest, we try to resemble Daniel's pre-processing in open 
source software, e.g. PostGreSQL/PostGIS. I already found the code for 
collinearity; the orthogonalization seems to be a bit trickier, but it 
should be possible to built the process in PostGIS as well, if it was 
possible to built it in FME. What do you think?

Tim

[1] https://imgur.com/a/aCKMVb7

On 2019-03-26 13:45, Jarek Piórkowski wrote:
On Tue, 26 Mar 2019 at 13:10, Begin Daniel <jfd553 at hotmail.com> wrote:
> There is actually no standard “code” available since I use FME (www.safe.com). It is a proprietary ETL application and all operations are done using “transformers” (https://www.safe.com/transformers/). I can provide you with the workbench I developed (a bunch of linked transformers) but you need a license to run it. This is why I tried to describe the operations I run on the data in the wiki.
> 
> As you did, people may send me coordinates (bounding box) of an area they know well. I’ll process the area and send the results back in OSM format. Please, be reasonable on the amount of data to process ;-)

Thanks Daniel. Let me know how it looks then!

Coming from an open-source background, the process is unusual to me,
and I have questions about scalability - will you be able to process
and provide updated data files for all of Canada then? - but if others
are comfortable with it then I won't object.

Some general thoughts regarding tooling as raised upthread:

I was initially excited to see building footprints data as they help
two quite distinct purposes:

1. they provide a mostly-automatic source of geometries for the
millions of single-family houses that wouldn't be mapped in the next
decade otherwise

2. they might provide a corrected and fairly accurate source of
geometries in heavily-built-up areas, where GPS signal is not that
reliable and it can be really difficult to get sufficiently accurate
geometries from imagery, whether because it's not sufficiently
high-resolution, two sets of imagery with conflicting offsets (Bing
and Esri are the two best sets in Toronto, and they're off by about
1-2 m on north-south axis from each other - that's not something I can
check with a consumer-grade GPS so I'm left guessing as to which is
true), or non-vertical imagery (I can count the floors on supposedly
top-down imagery in some cases).

  From what I saw, imports in the GTHA initially focused on the first
case, and I think the Tasking Manager setup was mostly sufficient for
those - where there is nothing currently on the map, or a few simple
2D geometries, a 4 sq km area can feasibly be done in under an hour.

However, as raised by others, I would really want the working squares
in Old Toronto for example to be no more than 500 m x 500 m, or no
more than 1 km x 1 km in St. Catharines. I would _love_ to have the
geometries to manually compare and adjust the 3D buildings already
existing in the area, but it will be much slower.

--Jarek

_______________________________________________
Talk-ca mailing list
Talk-ca at openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ca



_______________________________________________
Talk-ca mailing list
Talk-ca at openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ca
  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-ca/attachments/20190327/1b5e09bb/attachment.html>