[Talk-ca] Merging osm + geobase status

[Jean-Sébastien Moreau]

It seems to be a very good start! I will try to test it as soon as possible.

Feel free to modify and update the algorithm part I've created in the geobase
import wiki page so that everyone can understand the logic of your script.

http://wiki.openstreetmap.org/wiki/GeoBase_Import#AutoMatch_Process

> I figured that the initial step to do a comparison between the OSM and
> geobase datasets would be to split the OSM data along all junctions (so
> that it's representing geometry in roughly the same way as the geobase
> one), then see how much deviation there is between the endpoints of
> nodes in both sets (by trying to find the way which corresponds most
> closely with the geobase one). So that's what I did. You can find the
> script which does the comparison between two OpenStreetMap files
> attached below.

More advanced matching logic could be built using spatial operations. You might
want to take a look at GEOS. It's a c++ geometry engine but it has a python
binding using swig.

http://trac.osgeo.org/geos/

Could luck with your experiments!

Jean-Sebastien

Quoting William Lachance <wrlach at gmail.com>:

> Hey all,
>
> I haven't had as much time to work on this as I thought I would, but I
> thought I'd give a quick update on the status of this project anyway,
> since I have made some progress in the few hours I've spent on it.
>
> I figured that the initial step to do a comparison between the OSM and
> geobase datasets would be to split the OSM data along all junctions (so
> that it's representing geometry in roughly the same way as the geobase
> one), then see how much deviation there is between the endpoints of
> nodes in both sets (by trying to find the way which corresponds most
> closely with the geobase one). So that's what I did. You can find the
> script which does the comparison between two OpenStreetMap files
> attached below.
>
> The results so far have been pretty interesting. I've been doing a small
> test with my neighbourhood in Halifax (the north end) which has very
> good crowdsourced coverage. It looks like there's a match between the
> geobase ways and the osm ones in at least a majority of cases, which
> indicates to me that I'm at least doing something right. :)
>
> I've become convinced that actually sorting out which bits of the data
> should be kept is basically an AI-complete problem. That is, some human
> intervention will be required to figure out the best strategy for
> merging the two data sets. I believe I should be able to create a fairly
> decent UI for this using the PyGTK and PyCairo libraries.
>
> I do understand that this might result in inconsistent results depending
> on the quality of the GeoBase data and the human assisting with the
> import process, but the nice thing about OpenStreetMap data is that we
> can take an iterative approach. We can revisit the same data many times
> using the script as it improves in quality.
>
> My first project will be to create a user assisted way of merging the
> geobase tags with the OSM ones (e.g. using the geobase name of the
> street if it's better, getting the lane count) in the case where there's
> a match between the data.
>
> After that, I'm going to experiment with creating a UI to assist with
> grafting in new geobase information into the map, as we discussed on the
> list earlier.
>
> While I'm working on this, I'll be sending sample results to the list,
> so people can comment on my strategy and approach.
> --
> William Lachance <wrlach at gmail.com>
>