Gemeentegrenzen uit OpenStreetMap

OpenStreetMap is de vrije wereldkaart waaraan iedereen kan bijdragen. De geodata is vrij beschikbaar volgens een Creative Commons-licentie. OpenStreetMap (OSM) bevat allang niet meer alleen straten, maar is uitgegroeid tot een veelzijdige repository van vrij beschikbare geodata. Het is alleen nog niet zo makkelijk om er uit te pakken wat je nodig hebt.
Het standaard exportformaat van OpenStreetMap is een eigen XML-formaat. Dit is met allerlei open source tools, die beschikbaar zijn via de OSM-wiki op http://wiki.openstreetmap.org of de subversion-repository op http://svn.openstreetmap.org/.
Dit artikel illustreert hoe je de actuele Nederlandse gemeentegrenzen uit de live OSM-database haalt en deze importeert in een PostGIS-database.

De database

Het startpunt voor het zoeken naar specifieke informatie in de OSM-database is de Map Features-wikipagina: http://wiki.openstreetmap.org/wiki/Map_Features. Deze pagina bevat een overzicht van alle gebruikte ‘tags’ voor objecten in de database. Gemeentegrenzen vallen onder de Administrative Boundaries: http://wiki.openstreetmap.org/wiki/Key:boundary. Hoewel de op deze pagina bijgehouden tabel met de indeling per land hier niet helemaal specifiek over is – er wordt gesproken van ‘boundaries for cities like Amsterdam but also smaller like Volendam and Lutjebroek’ – vallen de gemeentegrenzen onder admin_level=8. Op dezelfde pagina lezen we dat de modus operandi om administratieve grenzen in OSM te zetten is door gebruik te maken van ‘relations’. (OpenStreetMap kent slechts drie soorten objecten: nodes (punten), ways (lijnen) en relations (relaties tussen groepen van de andere twee types).)

Extractie

We weten nu dat we alle ‘relations’ van het type ‘admin_level=8′ willen hebben. Er zijn verschillende manieren om een dergelijke abstractie uit de live-database te maken. De ene is een actuele dump van het gewenste gebied downloaden (deze zijn beschikbaar via http://downloads.cloudmade.com/ ) en hieruit vervolgens met de command-line tool ‘osmosis’ (http://wiki.openstreetmap.org/wiki/Osmosis ) een selectie maken. Een andere manier is om gebruik te maken van de OSM Extended API (OSMXAPI, spreek uit OSM-Zappy, zie http://wiki.openstreetmap.org/wiki/Osmxapi ). De volgende URL levert dan de gemeentegrenzen op in OSM XML-formaat: www.informationfreeway.org/api/0.5/relation[admin_level=8][bbox=3.35376,50.57484,7.22095,53.51513].

Import

Het resulterende OSM-XML-bestand kun je importeren in een PostGIS-database met behulp van OSM2PGSQL: http://wiki.openstreetmap.org/wiki/Osm2pgsql.
Ervan uitgaande dat je al een spatial database hebt met de naam ‘postgis’ gaat het dan als volgt:

> osm2pgsql -H tm-sr -U postgres -W -d postgis gemeentegrenzen_081118.osm

osm2pgsql SVN version 0.55-20081118 $Rev: 10464 $

Password:
Using projection SRS 900913 (Spherical Mercator)
Setting up table: planet_osm_point
Setting up table: planet_osm_line
Setting up table: planet_osm_polygon
Setting up table: planet_osm_roads
Mid: Ram, scale=100

Reading in file: gemeentegrenzen_081118.osm
Processing: Node(110k) Way(2k) Relation(0k)
Node stats: total(110573), max(312315964)
Way stats: total(2579), max(28446793)
Relation stats: total(690), max(51805)

Writing way(0k)

Te zien is dat osm2pgsql vier tabellen aanmaakt (als deze al bestaan dan worden ze default leeggemaakt, let op dus!).
We maken ons even niet druk om spatial indexes en bekijken het resultaat:

Naschrift

Op de site van Cloudmade zijn ook ready-made shapefiles beschikbaar per land. In dit pakket zit ook een administrative shapefile, maar deze is niet goed:

Deze wat langere weg verdient dus nog steeds de voorkeur!
Overigens zijn de Nederlandse OSM-ers (waaronder ondergetekende) ook bezig met het invoegen van andere officiële en niet-officiële indelingen in de database. Denk aan COROP-gebieden, wijken en buurten, EGG-gebieden, politieregio’s, postcodegebieden en bebouwdekomgrenzen.

Note To Self: The One And Only RD Projection String

EPSG:28992, or the Dutch double stereographic RD (RijksDriehoekstelsel) projection, is quite often incompletely or just plain badly defined.

My version of MapServer for Windows (2.2.6 from september last year) states

+proj=stere +lat_0=52.15616055555555 +lon_0=5.38763888888889 +k=0.999908 +x_0=155000 +y_0=463000 +ellps=bessel +units=m +no_defs  no_defs <>

Which yields the following result when a native 28992 dataset is projected onto a Microsoft Virtual Earth (EPSG:900913 or EPSG:3785 as it is now called):

Note that the buildings layer on top of the VE aerial photos is shifted to the north, by about 100 metres.

Spatialreference.org has a slightly different take on EPSG:28992:

+proj=sterea +lat_0=52.15616055555555 +lon_0=5.38763888888889 +k=0.9999079 +x_0=155000 +y_0=463000 +ellps=bessel +units=m +no_defs

which yields an almost identical result:

These projection strings are both incomplete, because they do not take into account the datum shift that is used in the RD projection and can be approximated using the ‘towgs84′ parameter in PROJ4.

The one and only right PROJ4 projection string is

+proj=sterea +lat_0=52.15616055555555 +lon_0=5.38763888888889 +k=0.999908 +x_0=155000 +y_0=463000 +ellps=bessel +units=m +towgs84=565.2369,50.0087,465.658,-0.406857330322398,0.350732676542563,-1.8703473836068,4.0812 +no_defs <>

Links

• Explanation of the towgs84 parameter on this page
• Some discussion about the RD datum shift on the PROJ.4 mailing list
• A non-technical discourse on datum shift and coordinate systems in Dutch.
• The Dutch national survey has a website on the RD coordinate system.
• There is also a very Web 0.5 site on the RD system and NAP (Normaal Amsterdams Peil, the Dutch standard sea water level which can be observed in the Amsterdam City Hall)

The 5 minute guide to setting up GeoServer and GeoWebCache on Windows

I came across yet another tile caching implementation, GeoWebCache, through this article on the Google Open Source blog. It integrates nicely with the Geoserver OGC server, which should make it very easy to set up on a Windows box. So let’s try that.

The End Of Flickr?

Well, certainly not today, and certainly not soon, but the introduction of georeferenced photos on Google Maps this week will certainly rock the online photo communities’ boat. Sure, there are tons of websites overlaying flickr photos on top of a web map, and most are richer than what Google Maps currently offers. Take for example loc.alize.us, a flickr/Google Maps mashup that has been around for a while. It offers tag filtering, user filtering, and a very nice and clean interface. To top it off, it offers a bookmarklet that integrates georeferencing into flickr.com very nicely. I still use it, although Yahoo Maps, the mapper of choice for Flickr’s mapping needs, of course, has had adequate coverage of the Netherlands for some time now.

But still.. It’s not directly ON Google Maps, which is – at least in Western Europe at this time – the ubiquitous web map. The general public will rarely discover any layer of the geographic web beyond Google Maps and Google Earth. ‘So, if I want my photos to show up on the web, I need to be on Panoramio.’ – Panoramio being the photo sharing community that has been showing off on Google Earth for as long as I can remember, and as from now on Google Maps as well. Panoramio was acquired by Google in May, 2007.

No, I don’t expect a mass flux of flickr users towards Panoramio. The latter will see a good number of new members though, and if Google remains as picky about which photos to display within Maps – I’m still confused as to where this leaves Picasa; I guess the user base is not large enough – Panoramio might become a force to be reckoned with in the online photo community universe.

OpenStreetMap Mapping Party ‘Saendelft’

The Dutch like to live in new, modern homes with a garden front and back. This leaves the country with many a suburban jungle like the one depicted below. This also means steady jobs for surveyors with the commercial mapping companies – and many a free weekend spent mapping for a Dutch OpenStreetMap contributor.

Virtual Earth Custom Tile Layers in 3D mode — not anymore.

The Microsoft Virtual Earth API lets you add your own tile layer to your VE Map. My colleague StevenO wrote about preparing a suitable TileCache setup. This used to work in both 2D and 3D map modes. Recently, Microsoft introduced the latest version of the API, 6.1, along with a major data upgrade and a new version of the 3D control. A step forward in many respects, but the tile layers will not show up in 3D mode anymore. Let’s investigate. UPDATED 080508 11pm, see below

Importing the GML 3.2.1 namespace into .NET

There comes a time for every geo-ict professional to have his first encounter with GML. Most of the time, this is not a pretty sight. Until now, I have managed to steer clear from GML when it comes to actually incorporating it into my own software. But today, this day dawned.

Benchmarking TileCache, part 1

I have been doing some benchmarking in the wake of my TileCache installation ‘endeavor’ of last week (part 1 – part 2). In a series – well, probably two – articles, I will try to provide some insight into the performance of the TileCache – Python – Apache ensemble.

How the iPhone and iPod Touch are location aware

Being inside a Conrad store in Germany just makes you buy stuff you normally might not. I was in the Hannover branch last Sunday, on the tail of a CeBIT visit, and bought an iPod Touch. And I’m about as close to being in love as you can be with a non-human subject.The iPod Touch got a (paid) software upgrade in January, and now all of a sudden it has a Maps application, and what’s more: Maps has a location button that invokes some magic to determine your location. Without GPS. This magic is courtesy of Skyhook Wireless and uses the broadcast signals of WiFi access points to triangulate your location. It’s called WPS, WiFi Positioning System.

TileCache on Windows part 2: mod_python

The 5 minute install guide for TileCache on Windows I wrote last weekend was intended to show how easy it is to get a TileCache map server up and running. Now that it works, we should be concerned with performance. As my colleague StevenO pointed out, a CGI solution is probably not the way to go if you want to take it to production level. So I decided to try a different approach: install TileCache as a module into the local Python setup and have mod_python handle the python scripts of which TileCache is made up directly. That way, we can dispose of the CGI wrapper and thus the need to call the python interpreter for every tile request being made by the client. In theory, that should give us a huge performance gain.I’m going to look into two topics:

1. Getting TileCache to run under mod_python
2. Benchmarking TileCache and more optimazition techniques

The first topic will be covered in this post, the second in a future post.