Skeletonize vectors in QGIS/Python

Question

I am trying to find a tool to create a skeleton of a large vector dataset using QGIS and its bundle. It should be directly programmatically callable through Python or commandline (QGIS, GRASS, OGR etc. are all okay).

The target is a large thin polygon with holes representing street areas and I need to obtain an approximation of the road network via skeletonization.

What I found so far:

• GRASS 7.0 has v.skeleton which should be good enough but does not seem to be in for integration with QGIS in the near future.
• CGAL has skeletonize algorithms, of course, but the Python support is very poor.
• pySkeleton does not work for me and I probably don't have time to salvage the code.

Other alternatives (openJump skeletonizer, this old ArcGIS plugin...) cannot be easily plugged into QGIS/Python.

Any other suggestions?

Answer 1

This document "Skeletonizing Polygons Using PostGIS" describes a process which utilises tools from PostGIS and GRASS, both of which might be callable. It mentions using PostGIS to create a skeleton of the vectors and cleaning it up using either PostGIS again or GRASS.

There are existing tools from the Processing Toolbox which hopefully simplifies things when calling from Python/command line such as:

• Import into PostGIS - use to import your dataset.

• PostGIS execute SQL - use to run the ST_StraightSkeleton function.

• v.clean - use to run the rmdangle tool.

Hope this helps!

Answer 2

One, somewhat round-about option might be to use SAGA's grid skeletonization module (http://www.saga-gis.org/saga_module_doc/2.1.3/imagery_segmentation_1.html), either through QGIS or using the command line saga_cmd utility. You'd have to convert your polygon to a SAGA raster first.

Otherwise, as stated by @Alpha-beta-soup, you could use v.skeleton from GRASS on the command line.

To use a commnand line tool from within Python, use the subprocess module to pass the command string to the system.