MÉLIDIS : Reconnaissance de formes par modélisation mixte intrinsèque/discriminante à base de systèmes d'inférence floue hiérarchisés

Nicolas Ragot 1
1 IMADOC - Interprétation et Reconnaissance d’Images et de Documents
UR1 - Université de Rennes 1, INSA Rennes - Institut National des Sciences Appliquées - Rennes, CNRS - Centre National de la Recherche Scientifique : UMR6074
Abstract : To facilitate the integration of pattern recognition systems in real applications, we present a new approach which aim is to combine properties that are rarely fully satisfied in the same recognition system. These properties are: performances, generic architecture and learning, reliability, robustness, compactness and transparency. This last point is particularly important to make the system easier to adapt, maintain and optimize for a given application. Our approach is totally data driven and focuses on knowledge properties in a classifier. The main originality comes from the cooperation of intrinsic and discriminant knowledge that are extracted and modelized automatically. They are organized in two levels: the first one models classes with fuzzy prototypes and the second one operates discrimination on shapes that have similar intrinsic properties using fuzzy decision trees. The system is entirely formalized by fuzzy inference systems that are combined for final classification. This approach led up to the implementation of the Mélidis system that has been evaluated on several benchmarks including handwritten character recognition.
Document type :
Theses
Complete list of metadatas

Cited literature [186 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00005225
Contributor : Nicolas Ragot <>
Submitted on : Friday, March 5, 2004 - 6:01:07 PM
Last modification on : Friday, May 10, 2019 - 12:18:02 PM
Long-term archiving on : Friday, April 2, 2010 - 7:32:56 PM

Identifiers

  • HAL Id : tel-00005225, version 1

Citation

Nicolas Ragot. MÉLIDIS : Reconnaissance de formes par modélisation mixte intrinsèque/discriminante à base de systèmes d'inférence floue hiérarchisés. Interface homme-machine [cs.HC]. Université Rennes 1, 2003. Français. ⟨tel-00005225⟩

Share

Metrics

Record views

494

Files downloads

804