reconstructing movement traces throug a hybrid map matching algorithm
DESCRIPTION
Kevin Baker, Pascal Brackman, Philippe De Maeyer, Rik Van de Walle University Ghent, Belgium; RouteYou, Belgium Topic: “Reconstructing movement traces through a hybrid map-matching algorithm”TRANSCRIPT
AGILE 2013 – Leuven, May 14-17, 2013
Reconstructing movement traces through a hybrid map-matching
algorithm
Understanding Urban Cycling: A Data Challenge
Kevin Baker
AGILE 2013 – Leuven, May 14-17, 2013
Research Framework
• Branding slogan: “Plan your nicest route”
– Specific information about the road infrastructure and surroundings focused on his application
AGILE 2013 – Leuven, May 14-17, 2013
Research Framework
• PhD research:
– “Mapping Linear Landscapes - Geosemantic methods for information extraction, validation and enrichment using dynamic geodata”
– intelligent aggregation and combination of novel geographic information from a dynamic community
AGILE 2013 – Leuven, May 14-17, 2013
Algorithm
AGILE 2013 – Leuven, May 14-17, 2013
Algorithm
AGILE 2013 – Leuven, May 14-17, 2013
Algorithm
AGILE 2013 – Leuven, May 14-17, 2013
Algorithm
• Software: FME / Python/OSRM Routing Engine
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Hybrid map-matching algorithm • Geographical: Point Search Algorithm
• Semantic: Attribute matching
• Topological: Shortest Path Routing Engine
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Vector Database:
– TomTom
• Preprocessing steps twofold: Routable dataset Point cloud
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Raw dataset
– Create plausible trips per PERSONID
• Time between registrations < 5 minutes
• Remove outliers/error: – HDOP < 5
– Distance between
registrations (<2500 m)
– outliers/error in Lat/Lon
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Raw dataset
– Analyse and filter trips on time passed and meanspeed
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Raw dataset
– Resulting trips (3302)
• PersonID
• TripID
• Starttime
• Endtime
• Meanspeed
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Point Search Algorithm
– Detect unambigious points along a trace:
• Dual carriage way
• Parallel roads
• Bearing difference
• Analyse closest
candidates
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Point Search Algorithm
– Variable parameters in function of quality:
• Dynamic search distance (B)
• Search interval (ΔA)
• Allowed bearing difference
• Number of closest candidates to analyse
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Similarity measure
– A is routed segment
– B is original segment of trace
A
B
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Similarity measures (Quality dependend threshold)
– Frèchet distance
– Relative/Absolute length difference
– Area between segments
– Turning function
AGILE 2013 – Leuven, May 14-17, 2013
Workflow
• Hybrid map-matching algorithm
AGILE 2013 – Leuven, May 14-17, 2013
Result: overview
Good similarity
Bad similarity
AGILE 2013 – Leuven, May 14-17, 2013
Result: overview
Good similarity Bad similarity
AGILE 2013 – Leuven, May 14-17, 2013
Result: comparison
• Individual trips
Created Trip TRIPID: 3034 Meanspeed: 4.88 Person : 207 Traveltime: 0:36:36
Additional trail TRIPID: 207-961 Person : 207 Traveltime: 0:16:00
AGILE 2013 – Leuven, May 14-17, 2013
Result: comparison
• Individual trips
Created Trip TRIPID: 1328 Meanspeed: 6.9955 Person : 146 Traveltime: 0:49:25
Additional trail TRIPID: 146-964 Person : 146 Traveltime: 0:15:00
AGILE 2013 – Leuven, May 14-17, 2013
Result: comparison
• Heatmap of ‘additional cycle trips’ (Raster) Max
Min
Remark: Pixelsize of 4m
AGILE 2013 – Leuven, May 14-17, 2013
Result: comparison
• Vector dataset with continuous color on count Max
Min
AGILE 2013 – Leuven, May 14-17, 2013
Result: comparison
AGILE 2013 – Leuven, May 14-17, 2013
Discussion and Research Outlook
• Customize routable data to specific activity – Exclude specific highwayclasses?
– Resistance on edges (fix vs variable)?
• Extensive post processing of delta (red lines) – How far can the automatic integration go?
– Crowdsourcing/Outsourcing?
• Fine-tuning the Point Search Algorithm
• Extend similarity measure and thresholds
• Resulting trip parameters
AGILE 2013 – Leuven, May 14-17, 2013
Q&A