OpenStreetMap The Quality Issue Oliver O’Brien
Jun 12, 2015
OpenStreetMapThe Quality Issue
Oliver O’Brien
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Contents
• Quality & Completeness– Sources of Data
• Scans of Imagery for Tracing• Imports
– Studies, Tools and Methods– Comparison– Other Considerations
• OSM in the Real World
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Quality & Completeness
• OpenStreetMap – a crowdsourced spatial database of the world.– How good is the data compared with real
life?– Is everything important on the ground in
the database?
CENTRE FOR ADVANCED SPATIAL ANALYSIS
GPS Traces
• In the beginning (~2004)– Tracks from GPS receivers
• Consumer-grade handheld devices• Prone to random and systematic error
– Particularly in urban areas (multipath)– Particularly if walking– Walk both ways down street, take the “average”?– In theory quality improves with multiple traces
• Metadata doesn’t include measured quality
– eCourier van GPS traces for London
CENTRE FOR ADVANCED SPATIAL ANALYSIS
GPS Traces
CENTRE FOR ADVANCED SPATIAL ANALYSIS
GPS Traces
CENTRE FOR ADVANCED SPATIAL ANALYSIS
GPS Traces
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Out-of-Copyright Map Tracing
• Crown Copyright expires after 50 years• Not so useful for urban areas
– Great for rural areas – less likely to change – Old, low resolution map better than nothing
• Started with Richard Fairhurst’s collection in early 2006– Quality dependent on scan resolution and
quality of rectification– Paper warps, especially over 50 years
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Out-of-Copyright Map Tracing
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Yahoo Aerial Imagery Tracing
• Agreement in late 2006 that allowed tracing– In the UK, some large
urban areas covered– Accuracy better than GPS
but not as good as newer imagery sets (e.g. Google)
• Phantom roads• Hidden barriers
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Yahoo Aerial Imagery Tracing
• Coverage of many countries– Iraq: “The whole road and rail network of
the Baghdad area has been mapped using the Yahoo! imagery. As have some parks, forested areas, etc.”
– Australia (Sydney): “Completion of the greater metropolitan railway network… several complex road interchanges not practical to map on the bike”
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Yahoo Aerial Imagery Tracing
CENTRE FOR ADVANCED SPATIAL ANALYSIS
NaPTAN
• Public transport nodes– Once-off copy of DtT database in 2009– Most useful for “less glamorous” bus stops
as railway station coverage was already complete in OpenStreetMap
– Rich metadata– Envisaged as two-way, OSM would fix poor
locations and might be fed back– Imported on a county-by-county basis
CENTRE FOR ADVANCED SPATIAL ANALYSIS
NaPTAN
• NOVAM-Viewer– Bus stops– For reconciling
NaPTAN/OSM– Graphical tool– Manage manual
verification of the locations and metadata
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Walking Papers
• Designed for large scale POI addition• The source is annotations to existing
OpenStreetMap data so ultimately dependent on the other sources
• Specially rendered to print out and add Points of Interest to
• Automatic georeferencing for tracing– If you want it
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Walking Papers
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Imports
• A common theme so far – tracing– Inherently poor quality due to user error
• Tiger (U.S.)• AND (Netherlands)• OS Open Data
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Tiger (U.S.)
• From US Government• Imported in late 2007 (data from 2005)
– Data quality low but good coverage– More difficult to create a community to fix
rather than map• Specialist tools: CloudMade has realised
potential of a cleaned-up set and has created web tools to help
– [animation]
CENTRE FOR ADVANCED SPATIAL ANALYSIS
AND (Netherlands)
• July 2007– Relatively straightforward as there was
little existing OSM mapping for the Netherlands
• From a digital mapping data company with navigation and LBS products
• First “complete” country in OSM• “15-20m” offsets in places between
Yahoo aerial imagery and the AND data
CENTRE FOR ADVANCED SPATIAL ANALYSIS
OS OpenData (U.K.)
• No bulk imports (c.f. Andy Robinson)• Community still debating
– UK coverage already too comprehensive– Possible licensing concerns– Building outlines possibly the biggest win
• Some test imports – automated feature detection from Street View rasters
• Isle of Sheppey
CENTRE FOR ADVANCED SPATIAL ANALYSIS
OS OpenData (U.K.)
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Summary of Data Sources (UK)
SourceStartDate
QualityU.K.
CoverageCompleteness Currency
Consumer GPS 2004 Very poor Most of U.K. Very poor Variable
eCourier GPS 2005 Quite poorCentral London
OK (roads only) Quite old
Out-of-Copyright Maps
2006 Quite poorAll
eventuallyReasonable Very old
Yahoo Imagery 2006 ReasonableCertain
urban areasVery good Quite old
NaPTAN 2009 GoodAll
eventuallyVery good
(bus/stops only)Fairly new
Walking Papers 2009 Variable Variable Excellent Variable
OS OpenData Meridian
2010 Quite good All (Britain)Good (certain features only)
New
OS OpenData Street View
2010 Very good All (Britain)Very good
(certain features only - no paths)
New
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Quality
• Muki Haklay (UCL CEGE)– “Haklay, M., 2010, How good is
volunteered geographical information? A comparative study of OpenStreetMap and Ordnance Survey datasets” Environment and Planning B: Planning and Design 37(4) 682 – 703• “on average within about 6 m of the position recorded
by the OS, and with approximately 80% overlap of motorway objects between the two datasets”
• Analysis in 2008 (Meridian 2)
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Quality
• Muki Haklay and Aamer Ather– Beyond good enough? Spatial Data Quality
and OpenStreetMap data• OSM vs MasterMap I.T.N., looking at road lengths• “The analysis shows that when A-roads, B-roads and
a motorway from ITN are compared to OSM data, the overlap can reach values that are over 95%. … OSM is of better quality than Meridian 2.”
• “Positional accuracy is satisfactory for many applications. Attribute accuracy is also satisfactory.”
• Spring 2009
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Quality
• “Just how good is it?”– As good as many consumer-grade online
mapping websites– Depends on the care and keenness of the
“local champion”
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Completeness
• Muki Haklay– Completeness in volunteered geographical
information – the evolution of OpenStreetMap coverage (2008-2009)• “OpenStreetMap already covers 65% of the area of
England, although when details such as street names are taken into consideration, the coverage is closer to 25%. Significantly, this 25% of England’s area covers 45% of its population.”
• Affluence bias• Covers to October 2009
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Completeness
• Muki Haklay: OSM vs Meridian 2 (red)
Left: Features only
Right: With attributes
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Studies on Completeness
• Peter Reed– DfT highway length
figures per county• “Is it complete yet?”
– Depends on what you need
– Depends on where you need it
– Improving!
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Tools for Completeness
• ITO World OS Locator
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Tools for Completeness
• “Lonely Buses” tool – 20m buffer
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Getting Completeness
• Mapping Parties– May 2006: “Mapchester”
• Chris Perkins, Andy Robinson
– January 2007: Central London• Steve Chilton• Richard Fairhurst
– May 2009: Milton Keynes • Me, Andy, Steve
– October 2009: Atlanta
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Getting Completeness
• Milton Keynes– 2 days– 25 people– [animation]
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Getting Completeness
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Getting Completeness
• London Mapping Marathon series– One hour of mapping followed by pub
• 2008: Streets• 2009: POIs• 2010: Building outlines
– Build and keep the community• Crises
– January 2010: Haiti earthquake• High quality aerial imagery
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Comparison
• Quality vs Google Maps– POIs– (Over)simplification of roads
• Completeness vs Google Maps – POIs again
• Databases vs crowdsourcing
• Coverage…– A few surprises – Pyongyang in OSM
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Forbidden City in Beijing
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Other Considerations
CENTRE FOR ADVANCED SPATIAL ANALYSIS
It Started Life as a Street Map
• It’s not Ordnance Survey MasterMap– Roads are line
features, not areal• Not yet, anyway
– OpenStreetMap is just a database• Anything can go in
it…
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Individual Trees in Bern!
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Loose Taxonomy
• Need to consider the metadata too – is it on the map, is it in the right place, is it tagged appropriately?– Community proposes and documents on
wiki but doesn’t enforce– Worldwide considerations– Metadata sometimes present for source,
always present for input/update date & user
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Loose Taxonomy
• Partial order ensured by Steve Chilton’s OSM Mapnik stylesheet– If you want to see your data appear on the
map that “most” people see, follow the guidelines
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Curves
• ...or lack of...– Ability to tag ways as curved– Still straight lines in DB
CENTRE FOR ADVANCED SPATIAL ANALYSIS
OSM in the Real World
CENTRE FOR ADVANCED SPATIAL ANALYSIS
CycleStreets
“We'll be launching a page for Local Authorities v. soon about embedded versions of the CycleStreets planner on their sites”
- CycleStreets
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Skobbler
• Commercial Sat-Nav application based on OSM data with custom additions– Motorway lanes for junction approaches
• Completeness crucial– Display Google Maps on website for
address finding (Google Local) but use OSM for routing and app 3D view
• Accuracy not so important
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Skobbler
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Skobbler
• Started in Germany– The map is more complete
• Just launched in U.K.– The map is probably good enough, though
missing a lot of turn restrictions• Users report bugs
– Skobbler fix remotely or highlight to OSM– Both communities benefit
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Flickr
• Beijing Olympics in 2008– Flickr is owned
by Yahoo
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Waze in Chile
• Recent accidental unattributed inclusion of OSM data from “unreliable source”– Fine with
attribution
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Background for Visualisation
• http://oobrien.com/vis/bikes/ (London)• http://osm.mapki.com/bikes/ (Minneapolis)
CENTRE FOR ADVANCED SPATIAL ANALYSIS
Attribution
• Various – Imagery of OpenStreetMap data which are CC-By-SA OpenStreetMap and contributors.
• Various – Imagery of Ordnance Survey OpenData Street View rasters which are Crown Copyright and Database Right 2010.
• p4 – Photograph of GPS unit by Steve Coast.• p9 – Imagery of out-of-copyright map by Richard Fairhurst• p16 – Photograph of Walking Papers in pub by Harry Wood.• p27 – Maps produced by Muki Haklay.• p28 – Map produced by Peter Reed.• p42 – Screenshot of motorway junctions in OSM by Barry Crabtree.• p43 – Photograph of ambulance by DBH NHS media services.• p50 – Photograph of Boris Bike by Pocket Lint website.
Other imagery may be copyright of the producers referred to on the same page.
CENTRE FOR ADVANCED SPATIAL ANALYSIS
More
• http://www.oliverobrien.co.uk/– My research blog
• @oobr on Twitter• [email protected]