Road Map Inference from GPS Traces: A Segmentation and Grouping Framework

Date
2015-09-25
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
A road network is one of the most fundamental data of geospatial information. In order to update road maps promptly and consistently, map inference is proposed to automatically generate roads' geometric positions and topological connections from Global Positioning System (GPS) traces. Most of the existing methods are designed to deal with low-noise, densely sampled and uniformly distributed GPS traces. In this research, we propose a novel point clouds segmentation and grouping framework to infer high-quality road maps from high-noise and sparsely sampled GPS traces. First, we segment the points of GPS traces into clusters to represent nearly straight roads. Second, we group the adjacent clusters according to their spatial proximities. Finally, we generate centerlines from the clusters and refine the intersections to form road networks. Experimental results show that our methods are robust to noises and sampling rates. The generated road maps have better geometric accuracy compare to the existing methods.
Description
Keywords
Physical Geography, Computer Science
Citation
Qiu, J. (2015). Road Map Inference from GPS Traces: A Segmentation and Grouping Framework (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/27672