Making Maps from OpenStreetMap Data: Unterschied zwischen den Versionen

Version vom 17. April 2014, 14:32 Uhr

See also OSM, OpenStreetMap-Daten zu Karten aufbereiten (german)

 >> Work in progress... <<

Overview

This is a how to/knowhow page about making own web maps from OpenStreetMap data.

• If you are looking for making nice own base (or background) maps with own styling for print or web eventually within own region, and without programming skills - then this is the place to look for. Examples of personalized base maps are blind maps (maps without labels), the watercolor map from Stamen, OpenCycleMap among others.
• If you are looking for making a thematic map on top of given base maps, use Maperitive or other services (see Geovisualisierung (german)). Examples of thematic maps is Wheelmap with infoboxes.
• This is still work in progress. Our plan (see Team below) is to put everything in a single script file (tbd.).
• See also Map Making Overview at OSM Wiki.

Use Case 'Vintage Map'

As a first use case we will show how to make a Vintage Maps from OpenStreetMap data.

tbd.

Your Use Case Here?

 If you have a nice map styling (preferrably in CartoCSS) 
 based on any 'map oriented' OSM database, 
 pls. contact us (see Team below)!

Cookbook

• Software: tbd.
• Data: tbd.
• Hardware: tbd.

See workflow.

Some background: Making maps actually means in this context transforming a dataset to graphics. Before being styled (or rendered) the dataset is being preprocessed which ultimately would be a one-click process - but it's not yet since technology and user demands are changing fast.

At least, we can make life easeier when the intermediate dataset structure is has some known rules: See e.g. database schema below.

Workflow

Pls. note that this is still work in progress and thus experimental, suboptimal and at worst partially broken.

The workflow consists of followin steps:

• Step 0. Prepare software and config data.
• Step 1. Download, convert and clip OSM data. You can skip to step 2 and jump to step 3 "Style data" if you are in a hurry.
• Step 2. Enhance and preprocess SQLite dataset.
• Step 3. Style data.
• Step 4. Publish map.

As said before, our goal is to wrap step 1 and 2 into one single Python script.

Step 0. Prepare software and config data

• Install software: OGR/Python, SpatiaLite, TileMill
• Get config data, e.g. modified osmconf.ini
• Prepare clip boundary (e.g. liechtenstein_schloss_bbox.geojson)
  • Search for the desired country boundary from OpenStreetMap and copy the Relation id
  • Open http://overpass-turbo.eu/ and insert the commands below with the desired Relation id
  • Export the data as geoJSON and open as layer in TileMill

 {type=relation}
 {id=...}
 <osm-script output="json">
 <id-query type="{type}" ref="{id}"/>
 <print mode="body"/>
 <recurse type="down"/>
 <print mode="skeleton"/>
 </osm-script>

Step 1. Download, convert and clip OSM data

• Open command line shell (PBF file in, SQLite/SpatiaLite file out):

 > ogr2ogr -f "SQLite" liechtenstein.sqlite
   /vsicurl/http://download.geofabrik.de/europe/liechtenstein-latest.osm.pbf 
   -dsco SPATIALITE=YES -skipfailures -progress -overwrite 
   --config OSM_CONFIG_FILE osmconf.ini -gt 65536 
   -clipsrc liechtenstein_schloss_bbox.geojson -clipsrclayer OGRGeoJSON 
   -nlt PROMOTE_TO_MULTI

Open points:

• Convert coordinates to Mercator?

Alternatives (not used further here; this does not mean they are unusable - in contrary! - but that the are just not considered for different reasons):

• QGIS - reads PDF/OSM data with OGR (like we do here) but with default config and without preprocessing.
• spatialite_osm_map (OSM import tool from Spatialite) - no config of own tag->attribute mapping, no clipping possible?
• osm2pgsql - requires PostGIS which is cumbersome being another local server software.
• osmosis - no tag->attribute mapping?

Step 2a. Enhance SQLite dataset

• Enhance dataset by shuffling geo objects around (using Spatialite, SQLite file in/out)
• Convert Polygons to Points and add them to points as POIs.
• Eventually convert multilinestring to linestring (can TileMill/Mapnik and QGIS handle multilinestrings/multipolygons?).

Step 2b. Do some other preprocessing

• use preprocessing tools (Python, SQLite file in/out)
• line smoothing of streets
• tbd.

Sources:

• http://www.imagico.de/map/water_generalize2_en.php

Ideas:

• DB related
  • Tippfehler in Tag-Values korrigieren
  • Tags vereinheitlichen
• Geometry related:
  • Generalisierung von Wald/Gewässer/Straßen

Step 3. Style data

• Start TileMill

Goal is to make boundaries (like from buildings) look more like hand-made.

 ::outline1, ::outline2, ::outline3, ::outline4 {
   line-join: round;
   line-color:#773d00;
   line-smooth: 0.05; 
   line-width: 0.5;
 }
 ::outline1{ line-smooth: 0.04; }
 ::outline2{ line-smooth: 0.06; }
 ::outline3{ line-smooth: 0.08; }
 ::outline4{ line-smooth: 0.10; }
 

A ‘sketchy’ look to lines can be achieved by overlaying a number of line attachments with different smoothing values applied. By smoothing higher values, the lines are overlaid using the ‘multiply’ compositing operation. This make areas that overlap more appear darker.

Sources:

• CartoCSS and TileMill
• Ready made icons: Symbolkatalog

Step 4. Publish map

• On paper or on the web
• See e.g. TileMill or Geovisualisierung

Database Schema

Table names and name rules (schema):

• osm_points
• osm_lines
• osm_polygons
• osm_roads, osm_roads_gen0, ...
• osm_pois
• osm_boundaries (osm_boundaries_gen0, osm_boundaries_gen1, ...)

Table names follow the schema "<prefix>_<name>_<postfix>" e.g. for zoom levels:

• 0-10 => <name>_gen0
• 11-14 => <name>_gen1
• >= 12 => <name>

Notes

See also Diskussion:Making Maps from OpenStreetMap Data.

Team

• Prof. S. Keller
• Hao Feng Tan