Routing-based map matching for extracting routes from GPS trajectories

Authors:
Terry Griffin;Yan Huang;Shawn Seals
Affiliations:
Midwestern State University, Wichita Falls, Texas;University of North Texas, Denton, TX;Midwestern State University, Wichita Falls, Texas
Venue:
Proceedings of the 2nd International Conference on Computing for Geospatial Research & Applications
Year:
2011

Citing 5
Cited 2

C4.5: programs for machine learning

C4.5: programs for machine learning
Matching planar maps

SODA '03 Proceedings of the fourteenth annual ACM-SIAM symposium on Discrete algorithms
Mapping Hacks

Mapping Hacks
Correction to "A Formal Basis for the Heuristic Determination of Minimum Cost Paths"

ACM SIGART Bulletin
From GPS traces to a routable road map

Proceedings of the 17th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems

Fast Viterbi map matching with tunable weight functions

Proceedings of the 20th International Conference on Advances in Geographic Information Systems
Map matching: comparison of approaches using sparse and noisy data

Proceedings of the 21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper introduces a novel offline map matching approach. We develop a routing-based map matching approach for standardizing identified routes in a collected set of GPS trajectories. Our approach first identifies key waypoints in a user's GPS trajectory using a modified Peucker curve reduction algorithm. Subsequently, it sends the key waypoints to a black-box driving directions service which returns a route utilizing each of the key waypoints. The returned route is a standardized representation of the original GPS trajectory constructed using the minimum necessary set of points. A filter-and-refine approach is used to identify the incorrect portion of the returned route and a refine step is carried out by eliminating the waypoints what leads to the incorrect matching. Experiments results showed that the proposed approach works well for a data-set of 10 volunteers each collecting data an average of 34.3 days.