5

How to extract street graph or network from OpenStreetMap ?

ethicnology
  • 474
  • 5
  • 13

1 Answers1

5

Solutions:

There are many solutions to achieve this goal, I listed some of them below.

- Overpass-api

Overpass-api & overpass-turbo let you use overpass query language to collect ways and nodes of type highway for a city :

[out:xml]; area[name = "Pantin"]; (way(area)[highway]; ); (._;>;); out;

Pantin is a city in France overpass-turbo

- Geofabrik & Osmium

Geofabrik allows you to download various datasets from continents to cities.

Next, extract nodes and ways of type highway using Osmium tag-filters:

osmium tags-filter map.osm w/highway -o highways-ways.osm

NOTE: osmium tags-filter also works with .pbf files

- Ophois

Ophois is a CLI tool written in Rust, i created to:

  • Download a map from overpass-api
  • Process data from Overpass or Geofabrik to extract the street graph
  • Simplify the extracted graph with detailed heuristics
  • Discretize the extracted or simplified graph to a distance in meter

I also created a simple tool to display the generated graph on a Leaflet map to check the simplification process, cartographe.
Cartographe let you check the node id and the distance of links in meters using haversine formula.

Extracted

Pantin extracted from overpass-api using Ophois

Simplified

Pantin simplified using Ophois

Simplified and Discretized

Pantin simplified and discretized using Ophois

NOTE: Simplified and discretized with 10 meters parameter

- OSMnx

OSMnx: Python for street networks. Retrieve, model, analyze, and visualize street networks and other spatial data from OpenStreetMap.
Pantin-osmnx

NOTE: Pantin using OSMnx

ethicnology
  • 474
  • 5
  • 13
  • 1
    AWESOME! I want to add other three options, 1. [osm2pgrouting](https://github.com/pgRouting/osm2pgrouting) 2. [osm2po](http://osm2po.de/) 3. [osm4routing2](https://github.com/Tristramg/osm4routing2) – kangkang Feb 27 '23 at 19:17
  • I am also wondering if you could use any of these tools for big graph extraction, eg, for the whole planet. – kangkang Feb 27 '23 at 19:23
  • @kangkang When I composed [ophois](https://github.com/ethicnology/ophois), I designed it to process streams of data so it can perform big graphs. You can download datasets of big regions [here](http://download.geofabrik.de/). For the whole earth you should concatenate and maybe pre-process these datasets – ethicnology Feb 28 '23 at 08:22