Skeletron
Skeletron generalizes collections of lines to a specific spherical mercator zoom level and pixel precision, using a polygon buffer and voronoi diagram as described in a 1996 paper by Alnoor Ladak and Roberto B. Martinez, "Automated Derivation of High Accuracy Road Centrelines Thiessen Polygons Technique" (http://proceedings.esri.com/library/userconf/proc96/TO400/PAP370/P370.HTM).
Required dependencies:
- qhull binary (http://www.qhull.org)
- shapely 1.2+ (http://pypi.python.org/pypi/Shapely)
- pyproj (http://code.google.com/p/pyproj)
- networkx 1.5+ (http://networkx.lanl.gov)
- StreetNames 0.1+ (https://github.com/nvkelso/map-label-style-manual/tree/master/tools/street_names)
You'd typically use it via one of the provided utility scripts, currently just these two:
skeletron-osm-streets.py
Accepts OpenStreetMap XML input and generates GeoJSON output for streets using the "name" and "highway" tags to group collections of ways.
skeletron-osm-route-rels.py
Accepts OpenStreetMap XML input and generates GeoJSON output for routes using the "network", "ref" and "modifier" tags to group relations. More on route relations: http://wiki.openstreetmap.org/wiki/Relation:route
The Name
The first two implementations of Skeletron used the "straight skeleton" of a polygon to find a generalized center, and ultimately didn't work very well.
The straight skeleton: http://twak.blogspot.com/2009/01/that-straight-skeleton-again.html
How it's useful for maps: http://aci.ign.fr/Leicester/paper/Haunert-v2-ICAWorkshop.pdf