Analyse et classification des signatures des véhicules provenant de capteurs magnétiques pour le développement des algorithmes « Intelligents » de gestion du trafic

Abstract : Road traffic is at the heart of concerns for society due to issues of spatial development, mobility, the fight for better road safety or, more recently, environmentally friendly considerations. Observation and knowledge of travel patterns can partly help to answer these concerns. The development of a way to measure individual journeys can be achieved using vehicle tracking. To be able to anonymously track vehicles, magnetic sensors are chosen rather than the main traffic sensors. After a preliminary study of the physical properties of both the inductive loop and magnetometer, three steps in the monitoring process (detection, pre-processing and re-identification) are developed. Firstly, a state machine is provided to improve vehicle detection using a magnetometer. Then, two new pre-processing steps are available. The first concerns the use of a novel method of blind deconvolution for the "inductive loop" sensor. The second concerns the selection of characterizing variables by principal component analysis. Subsequently, the SVM method is adapted for the re-identification of vehicles. A unanimous voting process on either fuzzy logic, a Bayesian approach or similarity measurement is offered and compared in relation to the use of a decision threshold. A new independent predictor of traffic modelling is available to evaluate this reidentification. Finally, all the suggestions are evaluated during different experiments with the goal of obtaining individual travel time measurements or estimates of the origin – destination matrix.
Complete list of metadatas

Cited literature [93 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/tel-01133516
Contributor : Sandrine Charlier <>
Submitted on : Thursday, March 19, 2015 - 3:18:59 PM
Last modification on : Friday, November 16, 2018 - 1:29:17 AM
Long-term archiving on : Monday, April 17, 2017 - 6:39:37 PM

Licence


Public Domain

Identifiers

  • HAL Id : tel-01133516, version 1

Citation

David Guilbert. Analyse et classification des signatures des véhicules provenant de capteurs magnétiques pour le développement des algorithmes « Intelligents » de gestion du trafic. Traitement du signal et de l'image [eess.SP]. UNIVERSITE DE NANTES, 2015. Français. ⟨tel-01133516⟩

Share

Metrics

Record views

2289

Files downloads

2987