Bayesian Networks-Based Defects ClassesDiscrimination in Weld Radiographic Images

Auteurs :  Aicha Baya Goumeidane, Abdessalem Bouzaieni, Nafaa Nacereddine, Antoine Tabbone
Année : 2015
Domaine : Electronique
Type : Article de journal
Revue : Lecture Notes in Computer Science (LNCS)
Résumé en PDF :  (résumé en pdf)
Fulltext en PDF :  (.pdf)
Mots clés : Bayesian networks, weld defects, Geometric descriptors, radiography

Résumé : 

Bayesian (also called Belief) Networks (BN) is a powerfulknowledge representation and reasoning mechanism. Based on probabilitytheory involving a graphical structure and random variables, BN iswidely used for classification tasks and in this paper, BN is used as aclass discrimination tool for a set of weld defects radiographic imagesusing suitable attributes based on invariant geometric descriptors. Testsare performed on a database of few hundred elements where the resultsare outstanding and very promising, since they outperform those givenby powerful SVM classifiers.