Publication type: Conference paper
Type of review: Peer review (publication)
Title: Map-generation algorithm using low-frequency vehicle position data
Authors: Steiner, Albert
Leonhardt, Axel
Proceedings: TRB 90th Annual Meeting Compendium of Papers
Page(s): 1
Pages to: 17
Conference details: 90th Annual Meeting of the Transportation Research Board, Washington, USA, 23-27 January 2011
Issue Date: 2011
Publisher / Ed. Institution: Transportation Research Board
Publisher / Ed. Institution: Washington
Language: English
Subjects: TRB; Map generation; Traffic; Floating car data
Subject (DDC): 370: Education
Abstract: Providing accurate map data is essential for a variety of applications in traffic engineering, e.g. navigation systems. To overcome costs and effort associated with classical map generation methods (e.g. surveying or specialized vehicles) a method for automatic map generation based on position data stemming from vehicle fleets that are already equipped with GPS and participate in day-to-day traffic is presented. The method is designed in particular for urban road networks, where building work is carried out frequently. In this paper we make use of taxi fleet data for map generation. The data used were recorded at relatively low frequencies (15 to 90 seconds), such that a trivial connection of subsequent positions is not possible. The problem is solved by processing the data through a series of algorithmic steps. After data cleaning and transforming raw data by a Transverse Mercator projection to a plane, density regions based on geographic position and estimated headings are calculated. The density images are filtered by a Gaussian kernel and a watershed transform is applied. From the resulting image a skeleton is derived and intersections are detected by dedicated rules. The resulting directed graph is compared with aerial images from Google Earth and an existing OpenStreetMap digital road network. Using data from vehicles participating in day-to-day traffic allows not only to derive static network information (nodes / links), but - depending on the vehicle fleet used - to annotate the resulting graph with attributes directly deduced from data, like typical link speed distributions or turning fractions at intersections.
Fulltext version: Published version
License (according to publishing contract): Licence according to publishing contract
Departement: School of Engineering
Organisational Unit: Institute of Data Analysis and Process Design (IDP)
Appears in collections:Publikationen School of Engineering

Files in This Item:
There are no files associated with this item.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.