Unsupervised weld defect classification in radiographic images using multivariate generalized Gaussian mixture model with exact computation of mean and shape parameters Nafaa Nacereddine a, *, Aicha Baya Goumeidane a , Djemel Ziou b a Research Center in Industrial Technologies CRTI, P.O.Box 64, Chéraga 16014, Algiers, Algeria b DMI, Université de Sherbrooke, Québec, QC J1K 2R1, Canada A R T I C L E I N F O Article history: Received 2 October 2018 Received in revised form 23 January 2019 Accepted 11 February 2019 Available online xxx Keywords: Mixture model Multivariate GGD Radiography Weld defect Classification A B S T R A C T In industry, the welding inspection is considered as a mandatory stage in the process of quality assurance/ quality control. This inspection should satisfy the requirements of the standards and codes governing the manufacturing process in order to prevent unfair harm to the industrial plant in construction. For this purpose, in this paper, a software specially conceived for computer-aided diagnosis in weld radiographic testing is presented, where a succession of operations of preprocessing, image segmentation, feature extraction and finally defects classification is carried out on radiographic images. The last operation which is the main contribution in this paper consists in an unsupervised classifier based on a finite mixture model using the multivariate generalized Gaussian distribution (MGGD). This classifier is newly applied on a dataset of weld defect radiographic images. The parameters of the nonzero-mean MGGD- based mixture model are estimated using the Expectation-Maximization algorithm where, exact computations of mean and shape parameters are originally provided. The weld defect database represent four weld defect types (crack, lack of penetration, porosity and solid inclusion) which are indexed by a shape geometric descriptor composed of geometric measures. An outstanding performance of the proposed mixture model, compared to the one using the multivariate Gaussian distribution, is shown, where the classification rate is improved by 3.2% for the whole database, to reach more than 96%. The efficiency of the proposed classifier is mainly due to the flexible fitting of the input data, thanks to the MGGD shape parameter. © 2019 Elsevier B.V. All rights reserved. 1. Introduction In any welding process the quality of the weld is a function of the interaction of a large number of variables. That is why, it is difficult to completely control the welding process because, even if the best care is taken during selecting the materials, joint design or welding procedure, many of the welds will contain some imperfections referred to as discontinuities. The latter must imperatively be detected and identified to prevent any damage to the industrial plant in question. Obviously, any manufacturer aims to prevent as far as possible the production of defective welds. However, the difficulty and the cost of consistently producing welds without any defect is such that we must accept the fact that the welded structures produced at an economic cost may contain a proportion of weld defects. Therefore, we need to know the significance of the different weld defects in terms of weld performance so that we can define a safe and realistic tolerance limit for defects [1]. One way to perform weld inspection is to impart Nondestructive Testing (NDT) procedures (1) immediately after the fabrication to make sure the welded joint is defect-free and (2) during the service life of welded components to ensure that no unacceptable defects are present and grow [2]. Mainly used in the petroleum, petrochemical, nuclear and power generation industries, the radiographic testing (RT), being a part of NDT, has played an important role in the inspection of welds. This technique uses penetrating and ionizing radiations such as X-rays or gamma rays to detect, in a welded joint, internal discontinuities such as, porosity, crack, lack of penetration, lack of fusion, solid inclusion, undercut, etc. [3]. The obtained radiograms are then manually examined by a radiography expert in order to detect the presence of possible flaws that could occur after the welding operation. Afterward, the expert proceeds to their quantification and their identification to decide about their acceptance or rejection according to the requirements of the weld RT standards and codes. To this end, because of the high developed * Corresponding author. E-mail addresses: n.nacereddine@crti.dz (N. Nacereddine), a.goumeidane@crti.dz (A.B. Goumeidane), djemel.ziou@usherbrooke.ca (D. Ziou). https://doi.org/10.1016/j.compind.2019.02.010 0166-3615/© 2019 Elsevier B.V. All rights reserved. Computers in Industry 108 (2019) 132–149 Contents lists available at ScienceDirect Computers in Industry journal homepage: www.elsevier.com/locat e/compind