Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz, Shahid Jabbar , and Miguel Liebe

16
On Constructing a Base Map for Collaborative Map Generation and its Application in Urban Mobility Planning Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz, Shahid Jabbar , and Miguel Liebe Chair for Programming Systems, University of Dortmund, Germany

description

On Constructing a Base Map for Collaborative Map Generation and its Application in Urban Mobility Planning. Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz, Shahid Jabbar , and Miguel Liebe Chair for Programming Systems, University of Dortmund, Germany. Motivation. Problem: - PowerPoint PPT Presentation

Transcript of Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz, Shahid Jabbar , and Miguel Liebe

Page 1: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

On Constructing a Base Map for Collaborative Map Generation and its Application in Urban Mobility

Planning

Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz, Shahid Jabbar, and Miguel LiebeChair for Programming Systems, University of Dortmund, Germany

Page 2: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

2

Motivation

Problem: Computer assisted urban mobility planning

requires good vector maps. Good vector maps are not always

available, especially for many third world countries.

Solution: Web 2.0 Collaborative map generation

GPS-Tracks, Wikimapia, Open Street Map, etc.

Page 3: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

3

Challenges and Solutions

Combining the GPS traces collected by people. Through Computational Geometry algorithms

[Shahid Jabbar, Master’s Thesis, University of Freiburg, Germany, 2003]

[Edelkamp, Jabbar, Willhalm, ITSC 2003] [Edelkamp, Jabbar, Willhalm, IEEE Transactions on ITS vol. 6 no. 1

(2005)] AI clustering methods to combine these traces in order to

infer road geometry [Brüntrup, Edelkamp, Jabbar, Scholz, ITSC’05]

A reliable integration of traces require a good base map that can act as the template.

This paper discusses our approach to generate such a vector base map. Borrows several techniques from Digital Image Processing

and Computational Geometry. Extracts calibrated road topology from raster maps. Integrated with SUMO (by German Aerospace Agency, DLR)

Page 4: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

4

Raster Maps

•Can be collected easily from city authorities or through scanning paper maps.

•A 2D arrangement of pixels.

•Raster Maps from Dortmund, Germany.

•Collected from the City authority of Dortmund.

Page 5: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

5

Extraction of Road Surfaces

Streets’ extraction by color values.

Problem: Railway tracks and street names are also black!

Page 6: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

6

Erosion

Street names and railway tracks are eliminated.

3x3 Mask

Page 7: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

7

Dilatation

Street lines might become distorted by erosion Made thicker again.

Small holes due to street names are filled

Page 8: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

8

Other Filters

Morphological Opening and Closing

Gap closing Fragment Elimination Smoothen

ing of contours

Page 9: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

9

Road Skeleton Computation•Extraction of the center lines of the thick surfaces.

Skeleton of a Pixel Map: A set of thin curves denoting the centerlines of the black surfaces.

Medial Axis Transformation

Page 10: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

10

Graph Construction

Sweep-line paradigm: process pixels in columns

For each crossing, start a traversal in all possible directions!

Need a hash table to avoid duplicate work

Page 11: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

11

Graph Simplification

Several thousands of nodes are generated!

Not all are required or – more precisely – “interesting”.

Employ a similar algorithm as Douglas-Peucker simplification.

Co-linearity test?

133

122

111

yx

yx

yx

d

(x1,y1)

(x2,y2)

(x3,y3)

If d = 0, (x2,y2) can be deleted!

ε (epsilon) as the accurac

y paramet

er

Page 12: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

12

SUMO – Simulation for Urban Mobility (by DLR) A start-of-the-art tool for traffic simulation Used during FIFA-06 and Catholic Youth day,

along with a Zeppelin to give real-time guidance to the traffic authority.

SUMO

Nodes + Edges in

XML

Routes

Simulation ResultsRaster to

Vector Transformation

Raster

Maps

Page 13: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

13

Integration with SUMO

Page 14: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

14

Summary

Urban mobility planning require a good vector map. Collaborative map generation needs a base map to

correct the inaccuracies that can be added by people.

Raster maps are inexpensive and widely available. Good quality maps can be obtained from the city

authority. We propose:

Extract a vector map from a raster map. Digital Image Processing techniques can be helpful.

Integrated with SUMO – a state-of-the-art tool for traffic simulation.

Page 15: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Constructing a Base Map for Collaborative Map Generation

Drodzynski, Edelkamp, Gaubatz, Jabbar & Liebe

15

Future extensions

Better image processing for Bridges – 3D.

Integration with lane information. Traffic Signals etc.

Special Thanks to: Daniel Krajzewicz at German Aerospace Agency (DLR)

Page 16: Maik Drodzynski, Stefan Edelkamp, Andreas Gaubatz,  Shahid Jabbar , and Miguel Liebe

Thank You!Questions ?