392

ISSN 1310-3946

NDT days 2017/ 2017

Year / XXV Number/ 1 (216) June/ 2017

COMPUTED RADIOGRAPHY VISION IN WELD TESTING CR-VISION-WT

Nafaa Nacereddine, e-mail: n.nacereddine@crti.dz, Research Center in Industrial Technologies, CRTI, Algeria Nadia Mhamda, e-mail: n.mhamda@crti.dz, Research Center in Industrial Technologies, CRTI, Algeria

Aicha Baya Goumeidane, e-mail: a.goumeidane@crti.dz, Research Center in Industrial Technologies, CRTI, Algeria

Abstract: With an aim of gathering some works carried out during several years within the Pattern Recognition team of Signal Processing and Imagery Division in our research center, we have conceived software which we have called Computed Radiography Vision in Weld Testing (CRVision-WT). This software is dedicated to the image processing and analysis in industrial radiography for the purposes of detecting, locating, quantifying and identifying possible discontinuities and defects present in a welded joint. The ultimate processing step in this software is related to decision-making about the acceptance or rejection of the discontinuity in question according to international standards such as API 1104 and ASME Sect. V. For a given processing stage, the software offers sometimes several routines in order to seek, in an interactive way, the more appropriate ones for the current processed radiographic image. Keywords: radiographic testing, weld defect, image processing and analysis, software. 1. Introduction

The interpretation of possible weld discontinuities in industrial radiography is ensured by human interpreters. Consequently, it is submitted to subjective considerations such as the aptitude and the experiment of the interpreter, in addition of the poor quality of radiographic images, the weak size of defect, etc., due essentially to the exposure conditions. These considerations make sometimes the weld quality interpretation inconsistent, labor intensive and biased. It is thus opportune to develop computer-aided techniques [1-14] to help the interpreter in evaluating the quality of the welded joints. We present in the Fig. 1 a general configuration of a radiographic testing system used in the inspection of the welded joint.

Fig. 1 General configuration of a welded joint inspection system using radiography

Synoptic scheme of the proposed software is displayed in Fig. 2.

Fig. 2 Global structure of the proposed software

In the software proposed in this paper, a toolbar provide shortcuts to some processing operations (see Fig. 3).

Fig. 3 The software: Computed Radiography Vision in Weld Testing, CRVision-WT

2. Image operations

In the first rubric, we present to the user various operations to handle the image to be processed. For example, load image, select the region of interest, display and save the resulted images (Figs. 4 and 5).

mailto:n.nacereddine@crti.dzmailto:n.mhamda@crti.dzmailto:a.goumeidane@crti.dz

393

Fig. 4 Handle the image to process and the resulted images

Fig. 5 Select the region of interest

3. Preprocessing This rubric in Fig. 6 proposes several preprocessing

algorithms to enhance the quality of the image to be processed. We can find:

Noise removal using the filters (Fig. 7) Contrast enhancement by image stretching, (Fig. 8) Contrast enhancement by mathematical morphology,

statistical and adaptive enhancement (Fig. 9).

Fig. 6 Preprocessing and image improvement

Fig. 7 Example of median filter

Fig. 8 Image stretching

Fig. 9 Example of statistical enhancement

4. Segmentation

The segmentation is the main step of image processing. From this operation, we try to delimitate the region of the object representing the discontinuity or the defct in the welded joint (Fig. 10). This software offers several segmentation methods

Binarization: Histogram 1D : Otsu, Kapur, Kittler et Tsai, Histogram 2D : local , joint and relative entropies, Adaptive locally : Niblack and Sauvola (Fig. 11),

Active contours: polygonal statistical (Fig. 12) et polygonal statistical with automatic ROI,

Finite mixture models: Gaussian (Fig. 13) and generalized Gaussian distributions.

394

Fig. 10 Image segmentation toolbar

Fig. 11 Example of binarization with Sauvolas method

Fig. 12 Example of active contour-based segmentation

Fig. 13 Example of segmentation by a Gaussian mixture

model

5. Post-processing The procedures in this rubric aim to improve the results of

segmentation in order to better fit the objet representing the weld defect. In post-processing (Fig. 14), we have:

Median filter Morphological operators, dilation, erosion, opening,

closing, artefact removal, etc. (Fig. 15)

Fig. 14 Segmentation without post-processing

Fig. 15 Applying of morphological operators 6. Description or feature extraction

In this rubric (Fig. 16), we compute the geometric

descriptor (Fig. 17) and the generic Fourier descriptor (GFD) in order to describe the defect in terms of shape (elongation, rectangularity, symmetry, etc. These descriptors constitute the inputs of the weld defect classification stage.

395

Fig. 16 Feature extraction toolbar

Fig. 17 Example of image description using the shape

geometric descriptor 7. Classification

We have chosen four types of weld defects for this software: Crack (Cr), Lack of penetration (LP), Porosity (Po) and Solid inclusion (SI). In the rubric classification (Fig. 18, 19), our purpose is to classify the defect indication present the radiographic image, if it exists, into one of the cited defects using several classification and clustering techniques such as finite mixture model, support vector machine, artificial neural network and Bayesian networks.

Fig. 18 Classification of the segmented image

Fig. 19 Example of classification for the discontinuity in Fig. 17

8. Diagnosis/Standards

In this rubric, we apply the international standards dealing with radiographic testing in order to accept or reject a weld (Fig. 20). These international standards are based on criteria computed according to the dimensions of the extracted defect indications. Here, API 1104 standard is used in the proposed software (Fig. 21).

The software application on the other standards (ASME Sect. V, API 5L, etc.) are under construction.

Fig. 20 Decision making according to international standards

396

Fig. 21 Example of applying of the standard API 1104 9. Conclusion

The software CRVision-WT is an image processing tool dedicated to nondestructive testing of weld by radiography. From the input radiographic film image of a weld, the software apply incrementally all the processing stages involved in a machine vision system, namely, preprocessing, segmentation, feature extraction, classification, decision-making based on norms and standards governing NDT by radiography.

10. References 1. Nacereddine N., Weld defect detection and recognition in industrial radiography based image analysis and neuronal classifiers, Fourth International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurized Components, London, UK, 6-8 Dec. 2004. 2. Nacereddine N., M. Tridi, Computer-aided shape analysis and classification of weld defects in industrial radiography based invariant attributes and neural networks, Fourth IEEE International Symposium on Image and Signal Processing and Analysis, Zagreb, Croatia, 15-17 Sept. 2005. 3. Nacereddine N., L. Hamami, D. Ziou, Thresholding Techniques and their Performance Evaluation for Weld Defect Detection in Radiographic Testing, Machine Graphics & Vision, ISSN: 1230-0535, vol. 15, no 3/4, pp. 557-566, 2006. 4. Nacereddine N., L. Hamami, D. Ziou, M. Tridi, Statistical tools for weld defect evaluation in radiographic testing, 9th European Conference in Non Destructive testing, ECNDT, Berlin, 25-29 Sept. 2006. 5. Mekhalfa F., N. Nacereddine, A.B. Goumeidane, Unsupervised Algorithm for Radiographic image Segmentation based on the Gaussian Mixture Model, IEEE Conf. on Computer as a tool EUROCON 2007, Varsovie, Pologne, 9-12 Sept. 2007. 6. Goumeidane A.B., M. Khamadja, N. Nacereddine, F. Mekhalfa, Statistical Deformable Model Based Weld Defect Contour Estimation in Radiographic Inspection, IEEE International Conference on Computational Intelligence for Modelling, Control and Automation, CIMCA08, Vienna, Austria, pp. 420-425, Dec. 1012, 2008. 7. Nacereddine N., S. Tabbone, D. Ziou, L. Hamami, Lalgorithme EM et le Modle de Mlanges de Gaussiennes Gnralises pour la Segmentation dimages. Application au contrle des joints souds par radiographie, Traitement et Analyse de l'Information : Mthodes et Applications, TAIMA09, Hammamet, Tunisie, pp. 217-222, May 2009. 8. Goumeidane A.B., M. Khamadja, N. Nacereddine, Bayesian Pressure Snake for Weld Defect Detection, in Proc. de Advanced Concepts for Intelligent Vision Systems ACIVS09, Sep.28Oct.2, 2009, Bordeaux, France. In Lecture Notes in Computer Science LNCS 5807, pp. 309319, 2009. 9. Nacereddine N., S. Tabbone, D. Ziou, L. Hamami, Asymmetric Generalized Gaussian Mixture Models and EM algorithm for Image Segmentation, in Proc. of 20th Int. Conf. on Pattern Recognition ICPR10, pp. 4557-4560, Istanbul, 23-26 Aug. 2010. 10. Nacereddine N., D. Ziou, Invariant shape features and Relevance Feedback for Weld Defect Image Retrieval, VIth International Workshop NDT in Progress, Prague, 10-12 Oct. 2011. 11. Nacereddine N., D. Ziou, L. Hamami Fusion-based Shape Descriptor for Weld Defect Radiographic I