[Imports] Building Footprint Import for Canada: Regina, SK

David MacNeil david_macneil at hotmail.com
Sat Jul 9 18:47:29 UTC 2016


I neglected to engage prior to import but figured I should at least post my methodology and source here after getting called on it by a member of the DWG.

Footprint data obtained from http://open.regina.ca/ where City of Regina provides a variety of datasets where they explicitly state the data may be used publicly.  Footprints are based on a 2014 aerial survey of the city.  The data was available as an ESRI shapefile.  The data contained footprint shapes only with no additional meta data.  The area covered by the import can be found here: http://www.openstreetmap.org/#map=12/50.4391/-104.6194

Prior to the import:
- individual address nodes were present for the majority of buildings in the city from a previous import (this address data would be transferred from nodes to shapes where reasonable to do so)
- most large buildings had previously been manually drawn and contained the address information as part of the shape, rather than as a node.  effectively making the scope of my import individual houses for the most part.
- another user had started performing a building footprint import covering several large areas, then ceased several months ago

Processing:
The shape file contained approximately 95,000 shapes.  It was too large for JOSM to easily handle even when allocating more memory, so I split the shape file into 12 sections using major roads / natural features as boundaries.  Some of these sections were still too large and were further split in half.

The general workflow for each section was:
- open building footprint file
- visually confirm footprints against satellite imagery
- download data for the area in a separate layer
- delete any footprint from the footprint file which corresponded to a building already present in the OSM database
- delete any footprints outside the area of interest 
- apply the filter (-addr | building) to the OSM layer to isolate address tags existing as individual nodes, and copy them to a new layer
- on the address node layer re-arrange the nodes so they are encompassed by the footprint to which they belong (to allow a script to transfer the tags from the nodes to the shapes)
- delete nodes from address layer for any building that contained more than one node (duplexes mostly).  Those buildings would be uploaded as a shape only and the existing address nodes on the OSM server left unmodified
- save a copy of the modified footprint file, and the layer containing the re-arranged address nodes
- combine addresses and footprints using the script: merge-building-addrs.py (https://gist.github.com/balrog-kun/4241509)
  > script only transfers the address tags if a single node falls within the shape,  if multiple nodes are present they are all copied to the output file and none transferred to the object
- review the merged address / footprint file for address nodes, indicating either multiple addresses for a footprint or bad alignment prior to running the script
- manually copy / paste individual address tags to footprints where alignment was the problem, and delete nodes where the footprint contains more than one address as they would remain unmodified on the OSM server.
- use filter to isolate footprints with an address, then visually determine what type of building= tag to assign.  If I was certain it was a residence I assigned building=house,  same for building=apartments (though most of those were already on the server and fell outside the scope of my work), I made limited use of building=yes preferring to be as specific as possible.
- use filter to isolate footprints without an address.  This consisted mostly of duplex housing with multiple addresses, detached garages and utility structures.  Again I made an effort to be as specific as possible in the tagging.
- switch back to the downloaded OSM layer and filter it to display only address nodes
- manually delete any node that corresponds to a footprint that will contain the addressing information
- particular care taken to leave nodes belonging to buildings with multiple addresses, or areas where no footprint was present such as houses constructed after the imagery was taken.
- copied and pasted processed footprints from the working layer to the OSM layer
- pressed upload, addressed quality assurance messages from JOSM if any
- reviewed area after upload
- checked area on KeepRight about a week after uploading, something I do on a regular basis anyway regardless of whether I have been editing or not.

Any feedback would be appreciated.  I hope that this import can be allowed to remain despite the breach of protocol.


More information about the Imports mailing list