2026-07-22T22:13:55Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:9339 2024-03-06T10:44:05Z ddc:004 open_access

Detection and Localization of Traffic Signals with GPS Floating Car Data and Random Forest Méneroux, Yann Kanasugi, Hiroshi Saint Pierre, Guillaume Le Guilcher, Arnaud Mustière, Sébastien Shibasaki, Ryosuke Kato, Yugo Map Inference Machine Learning GPS Traces Traffic Signal As Floating Car Data are becoming increasingly available, in recent years many research works focused on leveraging them to infer road map geometry, topology and attributes. In this paper, we present an algorithm, relying on supervised learning to detect and localize traffic signals based on the spatial distribution of vehicle stop points. Our main contribution is to provide a single framework to address both problems. The proposed method has been experimented with a one-month dataset of real-world GPS traces, collected on the road network of Mitaka (Japan). The results show that this method provides accurate results in terms of localization and performs advantageously compared to the OpenStreetMap database in exhaustivity. Among many potential applications, the output predictions may be used as a prior map and/or combined with other sources of data to guide autonomous vehicles. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Yann Méneroux and Hiroshi Kanasugi and Guillaume Saint Pierre and Arnaud Le Guilcher and Sébastien Mustière and Ryosuke Shibasaki and Yugo Kato 2018 Is Part Of LIPIcs, Volume 114, 10th International Conference on Geographic Information Science (GIScience 2018) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/LIPIcs.GISCIENCE.2018.11 urn:nbn:de:0030-drops-93397 https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.GISCIENCE.2018.11 eng https://creativecommons.org/licenses/by/3.0/legalcode