[OSM-talk-be] wegenregister now open data

joost schouppe joost.schouppe at gmail.com
Fri Jul 1 16:53:23 UTC 2016


Hi,

As you might already be aware off, the Wegenregister is now open data [1].
Basically, it is what you get when you mash NGI data together with GRB for
geometry and CRAB for streetnames. Strangely enough, that means that the
federal NGI data is now open if it is in Flanders (or Brussels).

Even though it is very much still a work in progress, the data quality
looks rather good. As is tradition, I made a simple Mapbox map [2] where
you can visually compare OSM and Wegenregister. I still haven't found a way
to make OSM roads appear at low zoom levels, so zoom on in for a realistic
comparison. Blue OSM on red Wegenregister, so clear red means we're missing
something, clear blue means Wegenregister is missing something. Yellow
extra lines mean these are planned roads.
Special thanks to Mikel at Mapbox, as I would normally not have been able
to pull this of with my free account.

Here's some things where it could be useful for us:

- "slow roads": these are much more complete in the NGI data than in OSM.
However, as these change all the time, it is probably best only to use them
combined with local knowledge, aerial imagery and/or gps tracks (especially
the Strava layers are becoming rather useful, available too for the
JOSM-illiterate here [3]). I was surprised how many paths were still
missing in my own neighbourhood. Even in the larger parks of Antwerp, their
network is much more complete.
- planned roads: the idea is that roads will be in Wegenregister even
before they exist. A lot of those are missing in OSM.
- private roads: the NGI seem to have gone where we can't survey
- improving geometry

It is probably not the best place to get your streetnames from, as the flow
is from CRAB to GRB to Wegenregister, where each of these steps still
involves manual labour. In the case that was on Radio 2 recently [4],
corrections were made in CRAB as early as January 2015, but were not
reflected in GRB roads as late as December 2015. So get your CRAB road
segments for that, for example through the Lara application.

The first thing to do is of course surveying. To make things easier, I am
planning to make a dataset of Wegenregister roads which have a very high
probability of being missing in OSM. That dataset should then be small
enough to load in your Osmand or JOSM without any splitting necessary.

I actually have the general technique to do that already available at work,
so this shouldn't be a real issue (though it is made in FME, so we would
need to re-build the model in Postgres to build something more permanent).
In fact, I'll be doing an analysis for my employer, the city of Antwerp to
see where OSM can be used to improve Wegenregister. This, to me, is a clear
indication of government in Flanders finally waking up to OSM as a
datasource -though most people still cannot help but giggle when talking
about it :)
In fact, it looks like AGIV is interested in talking with us about ways in
which we can work together. Contact me if you want to be involved.

It should be clear that any competition between Wegenregister and OSM is a
bit absurd. Wegenregister has no ambition of being a routable dataset. It
is merely the minimum data that the government should know about roads: who
operates them, what's their surfacing, what kind of road is it. I do
believe the idea is to make it easy to link other data to this dataset,
with road segments having kind-of-stable unique IDs.

That means that a direct "quality" comparison between OSM and Wegenregister
really doesn't make much sense. That doesn't mean it hasn't been done. In
fact, as some of you might remember, there is a thesis on this subject. In
the course of their investigation, they also did a little vandalism, as
escada discovered [5]. It will not surprise you that many of us have had
serious issues with this research. As it was mentioned in several
presentations at the latest AIV Trefdag (where the entire government
related GIS world comes together), we can't really ignore it. Because of
that, I have asked for the thesis to be made public, so we can at least
read it and comment on it. And all relevant parties have accepted that, so
that's good news. You can now download the thesis [6].

Is a response in order? I think it probably is. I already wrote a comment
before (in private), Glenn had some pointed remarks too. I don't know if we
should respond with just an analysis of what they did and why it's wrong.
If you think that's useful, let's start a Hackpad!
I would like to respond with research of our own. Basically, looking at OSM
as it was now and year by year earlier, how have we evolved in terms of
network completeness.

Well that's a wall of text! Sorry about that.

1:
https://download.agiv.be/Producten/Detail?id=1429&title=Wegenregister_03_06_2016

2:
https://api.mapbox.com/styles/v1/joostschouppe/cipy8xm3a006bdnncjuo4jc5g.html?title=true&access_token=pk.eyJ1Ijoiam9vc3RzY2hvdXBwZSIsImEiOiJjaWh2djF1c2owMmJrdDNtMWV2c2Rld3QwIn0.9zXJJWZ4rOcspyFIdEC3Rw#15.8642209654081/51.2262361907535/4.4063452275206885/0
3: http://strava.github.io/iD/
4: https://lists.openstreetmap.org/pipermail/talk-be/2016-June/008786.html
5: https://www.openstreetmap.org/user/escada/diary/35244
6: https://iiw.kuleuven.be/denayer/thesis-openstreetmaps-wegenregister (sic)
-- 
Joost @
Openstreetmap <http://www.openstreetmap.org/user/joost%20schouppe/> |
Twitter <https://twitter.com/joostjakob> | LinkedIn
<https://www.linkedin.com/pub/joost-schouppe/48/939/603> | Meetup
<http://www.meetup.com/OpenStreetMap-Belgium/members/97979802/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-be/attachments/20160701/33f1ed6e/attachment.htm>


More information about the Talk-be mailing list