PATTERN RECOGNITION IN ULTRASONIC IMAGERY USING THE HOUGH TRANSFORM T.Merazi Meksen R.Drai F.Sellidj Laboratoire de traitement du signal et d’imagerie Centre de recherche en Soudage et Contrôle Algerie

In non destructive testing of materials, the ultrasonic imagery is a field in full rise. Indeed , more than one convivial representation of the results , the operations of detection , localization and sizing of the defects can be carried out automatically by the analysis of synthesized images. The problem breaks up in general into a phase of preprocessing to only limit the quantity of information to useful and a processing phase in order to characterize the defect. In our article, we will briefly describe the the ultrasonic images formation said C-SCAN and TOFD, before seeing how a high pass filtering makes it possible to reduce the image c-scan to the only points of the defect edges. This preprocessing reduces considerably calculations for using the Hough transform in patterns recognition of defects in the case of C-SCAN image , and the localization of cracks in the case of the TOFD image.