Fusion multi-sources pour l'interprétation d'un environnement routier

Abstract : Exceeding speed limits is a major cause of road accidents, which could be reduced by the use of robust detection of speed limits that may continuously inform the driver of the proper speed limitation. The work presented in this document relate to the achievement of such a system based on a visual detection of speed limit signs. To make the system robust, it is necessary to merge the results of these detections with information from other sensors to interpret the results of the visual detection. For this aim, two algorithms were developed. First, a specific geographic information system was developed in order to expand the electronic horizon of the vehicle. The fusion process in place addressing these various sources of information is based on model-based rules to overcome the problems inherent to the probabilistic fusion process that can sometimes lead to uncertain situations putting the whole system in global fault. These works are the fruit of collaboration with an automotive supplier and the prototype has been validated experimentally on the road and in real conditions. A ground truth tool has been specially developed to quantify the results. The system shows excellent results with high detection and classification rates for speed limit signs recognition and complex situations analysis.
Document type :
Theses
Complete list of metadatas

Cited literature [128 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00005997
Contributor : Ecole Mines Paristech <>
Submitted on : Wednesday, December 15, 2010 - 4:40:04 PM
Last modification on : Monday, November 12, 2018 - 10:56:59 AM
Long-term archiving on : Friday, September 10, 2010 - 3:06:14 PM

Identifiers

  • HAL Id : pastel-00005997, version 1

Citation

Alexandre Bargeton. Fusion multi-sources pour l'interprétation d'un environnement routier. Mathématiques [math]. École Nationale Supérieure des Mines de Paris, 2009. Français. ⟨pastel-00005997⟩

Share

Metrics

Record views

712

Files downloads

2701