Decision Making Help during Ultrasonic Inspection of Materials

Ultrasonic testing is a type of non-destructive inspection that is widely used in industry to verify the integrity of structures. According to the advancement of technology, automatic methods and algorithms are developed to assist engineers in results interpretation.Automated procedures and the use of robots allow the inspection of larges structures (pipelines, reactor vessels…) consequently, the amount of data to analyze become more and more large and thus too long time consuming.. That is why automating results analyzes and Artificial Intelligence (AI) are mean fields of research in this domain.This article describes an AI based method to ultrasonic image classification. The objective is to recognize images relatives to defective scanned zones and images corresponding to safe zones. In the first step, image characteristics to apply as inputs to a classifier have to be determined and extracted. Exploiting the Support Vector Machine (SVM) classification models, we classified these images first by implementing the low-level features and then the high-level features.