Road Completion Project
Towards a quality checked and complete open road network for Belgium
Ben Abelshausen
OpenStreetMap Belgium
Map
The
World!
'State of the Map' in Belgium
- Excellent basemap.
- Excellent state of transport-related data
- Good quality road network.
- Good quality transit data.
- Reasonable landuse.
- Basic address coverage
How far along are we?
How do we know?
What if things change?
Road Completion
Road Completion
Not about importing data!
Road Completion
Not about changing the way we map!
- Change detection, spot changes faster
- Completeness & quality assurances.
Road Completion
Road Completion Project
- Compare OSM <-> Other data (Open data or ML-generated)
- Mapping!
- Feedback
Loop 1
Loop 2
Step1 : Comparison
Step2 : Mapping
Step3 : Feedback
Comparison
Comparison
- Convert source data (shapefiles, wegenregister)
- ogr2ogr : reproject and convert to geojson
- nodejs : convert attributes to OSM-tags
=> this makes the comparion process independent of source.
- mapbox/tippecanoe : split into vector tiles
- Distributed comparison process:
- mapbox/tilereduce : process tiles in parallel.
GitHub: https://github.com/osmbe/road-completion
wegenregister
OSM
OSM (buffers)
comparison
diff
Comparison
(v 0.1)
Comparison
(v 0.1)
Comparison
(v 0.1)
Comparison
(v 0.1)
Comparison
(v 0.1)
Comparison
(v 0.1)
Mapping
Mapping
A lot of isolated cases: Maproulette
- allows to work on case-by-case
- keeps track of cases done
- allows mappers to report false positives
- doesn't change the mapping process
- has an API that allows us to:
- automatically generate the false-positives db
- re-run the comparison
Loop 1
Loop 2
Mid-term goals
- Verify OSM road network.
- Generate quality metrics for OSM roads in Belgium.
Do this is in the open, use open-source tools and open-data so everyone can run their own!
Why?
Trust!
Is OSM still relevant when we have ML and all that open data?
What about open data?
Is OSM still relevant with all this open-data?
Usually open-data is collected by and for the government and you are not part of that process!
What about machine learning?
"The human factor"
Commercially available data?
Response to Open Data
Use the trend to improve our own data
At the same time keep our 'chaotic structure' and openness alive
Response to ML/bigdata
Use the trend of ML-driven solutions to generate data/quality check data to improve our own data
At the same time keep our 'chaotic structure' and openness alive
open data, ML/bigdata are opportunities not 'competion'
open data, ML/bigdata
... can enable OSM to guarantee a degree of quality and completeness
open data, ML/bigdata
... can make OSM to be the best database of the physical world
We get smarter every day!
(because we are a big diverse community)
Imagine
- A costless (or very cheap) datasource.
- Quality checked and verified.
- Almost always up-to-date.
- Still community-driven!
Who can compete with this?
(almost) nobody
We
are
winning!
We need help!
(and we need sponsors!)
I am:
Ben Abelshausen
Board member of OpenStreetMap Belgium/Open Knowledge Belgium
Routing and mobility expert: