Using OpenStreetMap as CAD Data

Sun 27 July 2014
By foo

I've been involved with OpenStreetMap since 2012, getting my start by making a small WebGL-powered bike map of San Francisco. I'm a little hesitant to consider myself a _contributor_, however, as my involvement up to now has mostly been as a leecher.


Living in one of the best mapped areas of the world, I'm pessimistic that I have a lot of ground truth to contribute, but I am interested in developing new tools, and most of all growing the OSM contributor base. This is difficult because, unlike Wikipedia, the mediums in which data is consumed and contributed are very detached from each other.

The strongest application of OSM is display maps - these have made it into commercial use by companies like Flickr and Foursquare, and the quality is generally sufficient for the task. Consumers of slippy maps aren't terribly picky about data (I have no evidence to back this statement up.)

A different application of OSM is as a database. Driving directions, geocoding, and point-of-interest search all present OSM data to consumers programatically. WebGL vector slippy maps may fit this definition as well, if map features can be interacted with as discrete objects.

I see the OSM contributor base as broadly:

  • Slippy- and sometimes print mapmakers
  • Programmers building geo-enabled applications
  • GIS tool users, humanitarian NGOs, government agencies

I've been thinking about adapting OSM to CAD-like workflows that are common in the urban planning and architecture industries. It's someone analogous to GIS in that the ecosystem is dominated by proprietary, GUI-driven software. I've met few current OSM contributors in these fields, but have definitely noticed interest. Another nice offshoot of OSM-as-3D-data is its application in video games and 3D printing.

This is conceptually a little different than "Viewing OSM in 3D", which is well documented on the OSM wiki. Mostly because it's not simply about looking at the data but as relevant context to a 3D model. To make a standardized-testing-style analogy:

Slippy maps : geographic visualizations :: osm in 3D : CAD workflows


There's a couple of specific challenges i'm looking into:

toronto cad
  • data validation: Inaccuracy is tolerable in the slippy map world, but is it a showstopper in the 3D world? I don't expect people will be incorporating OSM into their construction diagrams. The 2D-ness of OSM can be augmented through elevation sources such as NASA SRTM.
  • Scale - Web mapmaking is coarse and deals with geography at the country and planet scale: can OSM data look good at the city and neighborhood scale? One obvious issue is projections: In OSM we're used to LatLong and Web Mercator, but for city-scale applications we may want to choose something like UTM, where we can reliably place buildings aand other features to scale. We also need to address audiences that may have less literacy with GIS concepts.
  • Lack of open source tooling around 3D - it's difficult to bootstrap an ecosystem of tools when most of your audience is using $X,XXX per head licensed software. QGIS helps 'bridge the gap' from open source tools to GIS end usersalong with standard interchange formats such as GeoJSON and Shapefiles. The 3D landscape is sparse; Blender is a popular free 3D program, with Sketchup and Rhino being affordable options compared to the incumbent AutoCAD. 3D formats are also even crazier than proprietary GIS formats: the AutoCAD DWG format is notoriously undocumented, but i've found .DXFs as pictured above to be a good starting point.