Multiple classifier combination for steel surface inspection

Auteurs :  Rachid Zaghdoudi, Hamid Seridi, Adel BOUDIAF
Année : 2019
Domaine : Electronique
Type : Communication
Conférence: 2nd Conference on Informatics and Applied Mathematics, IAM 2019
Résumé en PDF :  (résumé en pdf)
Fulltext en PDF :  (.pdf)
Mots clés : machine vision, classifier combination, surface defects, support vector machine (SVM), fuzzy K-nearest neighbor (FKNN), histogram of oriented gradients (HOG), gray level co-occurrence matrix (GLCM)

Résumé : 

Vision-based steel surface inspection systems have gained increasing interest over the two past decades because they offer the possibility to meet the requirements of manufacturers in terms of time, cost and accuracy compared to traditional methods based on human vision. The main objective of this paper is to propose an efficient inspection system based on machine vision technology and multiple classifier combination to detect and classify the different types of defects in steel products. This system is based on two independent classifiers, support vector machine (SVM) and fuzzy K-nearest neighbor (FKNN). Features are extracted with the gray level co-occurrence matrix (GLCM) and Histogram of Oriented Gradients (HOG). Principal Component Analysis (PCA) is applied to these features in order to reduce the descriptor size and avoid over-fitting resulting from features redundancy. Each set of features is respectively inputted to SVM and FKNN to form four parallel classifiers. Also, seven fusion rules are applied to give the final decision. To evaluate the performance of the proposed system, a series of experiments was conducted on the NEU Surface Defects database. The results obtained demonstrate the effectiveness of the proposed approach for classifying steel surface defects.