·Home ·Table of Contents ·Computer Processing and Simulation

Weld Defect Extraction and Classification in Radiographic Testing Based Artificial Neural Networks Nacereddine Nafaâ , Drai Redouane & BENCHAALA Amar Laboratoire du Traitement du Signal et de l'Image Centre de Soudage et de Contrôle ( CSC ), Route de Dély brahim, BP 64, Chéraga, Algérie Tél/ Fax : (213) (2) 36 18 50 , E-mail : nacereddine_naf@hotmail.com Contact

Abstract

This paper contains two parts: In the first part, we show the effectiveness of using neural paradigms to detect edges in X-Ray images which are used in Non Destructive Testing. The developed classifier consisted of a multilayer feed forward network window in which the center pixel was classified using gray scale information within the window. The aim of the work in the second part, is to construct a set of weld defect descriptors in X-ray images and their recognition by the neural classifier. These descriptors are based on the geometric invariant moments which are insensitive regarding usual geometrical transformations. Once the geometric invariant features computed, a neural network classifier trained by back-propagation has to classify the defect-images in planer or volumetric defect classes.
Key words: X-Ray images, Weld defects, Artificial Neural networks, Edge detection, Invariant moments.

1.  Introduction

We intend by Nondestructive Testing (NDT) any examination of industrial materials and assemblies using methods that don't alter their structure, permitting further utilization. Each method is specific and is destined to measure certain properties or to make conspicuous certain types of defects.Then, they have to be chosen in terms of examined material and the property or the imperfection we want detect. One of the most techniques used in NDT is Radiography which is based on the transmission of X-rays or gamma-rays through an object to produce an image on radiographic film. This method is used for inspecting several types of welded assemblies such as pipe-lines, boilers, pressure vessels etc. Inspected zones may present multifarious defects such as porosity, inclusions, cracks, lack of penetration, lack of fusion etc.,repertoried by official norms of NDT [1].
This radiogram is examined by radiography interpreters whose the task consist in detect, recognize and quantify eventual defects, but the radiogram quality, the welding over-thickness, the bad contrast, the noise and the weak sizes of defects make difficult their job [2]. The defect quantification in these conditions is submitted to human judgement and subjective considerations, such as, capabilities and experience of the interpreter because, it takes time to train a film interpreter. In addition, human interpretation of weld quality based on film radiography is very subjective, inconsistent, labor intensive, and sometimes biased. It is thus desirable to develop some forms of computer-aided systems to assist the human interpreter in evaluating the quality of welded joints. This involves the digitization of film radiography and the development of algorithms to extract welds and to identify flaws in them. Therefore, we try to apply the digital image processing and artificial intelligence technologies for the defect classification [3].
One of the essential processes in computer vision consists in reduce the huge quantity of information, contained in image of objects which we have to recognize by preserving, only the most important points. Therefore, generally, the edge detection is the first stage we apply before the recognition stage [4].
In the second stage, namely, image analysis, we use geometric invariant moments for feature extraction of the defect-images. The main interest of the moments thus computed is their invariance regarding usual geometric transformations (translation, rotation, scale change) [5]. This, is very important in the case of our application, because the aim of the radiographic control automation is to identify, firstly, the different types of welding defects, in spite their emplacement, orientation or size. For example, a fissure is always identified as being a fissure, in spite its orientation or dimension.
The past decade has seen a rapid growth of interest in the computer programming techniques of artificial intelligence. One such area that has become particularly topical is that of artificial neural networks. Artificial neural networks (ANNs) are computational representations based on the biological neural architecture of the brain [6],[7]. The early work in this field dates back to 1943, although most investigation has taken place recently as computers have become more widespread and computing speeds and memory have increased, thus allowing the simulation of realistic problems. ANNs have been successfully applied to a wide range of engineering and scientific applications such as signal and image processing [8], data analysis, process control and prediction. Thus, it is expected that neural networks based techniques would perform better in image processing, pattern recognition and classification problems.
In the first part of this paper, we propose the application of Artificial Neural Networks in the edge detection of X-ray images containing defects of welding. After a description of the ANN configuration dedicated to the X-ray edge detection task, will succeed the discussion of results deriving from this application.
In the second part we propose the application of the neural networks, trained by the back-propagation, to the preliminary classification of the weld defects in X-ray images, through image analysis by geometric invariant moments. Thus, after a definition of the used features and the corresponding neural networks architecture, results will be presented and commented.

2.  Edge detection in x-ray images based Ann

2.1. Ann's configuration
Because of digital radiogram's high resolution, it is physically impossible to develop a multilayer neural network that detects all the edges on radiogram at once.
Therefore, we designed a network window classifier which classifies the central pixel of a relatively small area in the image. To extract the edges, the window must slide, pixel by pixel, over the entire image. Then, the input layer of the network corresponds to the portion of image covered by the window (3x3) [9], [10]. Therefore, the ANN has 9 neurons receiving the gray-scale values of pixels composing the square window. The output layer contains one neuron whom the state identifies the edges from the background and creates the segmented image.
There are no accurate rules for the option of the hidden layers number and the neurons number in each layer. Nevertheless, a hidden layer with 10 neurons has proven satisfying in the case of our application. The network synaptic weights are adjusted on the base of training couples set, representing examples that define possibilities of contours (rectilinear, circular, pointed, etc.). The used training rule is the back-propagation.Once, the system converges and the fixed total error is reached, the stage of training is finished and the network corresponding to the window is trained. Then, we apply it on all the image. The state of the output neuron identifies the contour. We has chosen 28 training couples representing cases of the most preponderant contours as shown in Fig. 1.

Fig 1: Some cases of contours.

Fig 2: The neural network architecture

Fig 3: Three-layer neural network used to classify welded defects


Thus, we has chosen the following architecture of the neural network
• 9 neurons in the input layer.
• A hidden layer with 10 neurons.
• An alone neuron in the output layer.

Fig. 2. illustrates the architecture of the network and the way of its application to the image.

2.2 Applications and results
The proposed approach is tested on X-ray images used in industrial radiography. We can remark that for the X-ray image without noise as shown in Fig.4.a , the network extracts the different contours of weld defects, which are put in obviousness (See Fig.4.b). These defects are represented by dark stains with lengthened form that are positioned along the welded joint. Really, in the jargon of interpreters in radiography, it concerns a lack of penetration. The image-contours emerges very well the defect boundaries. This is not the case in Fig. 6.b . That is du essentially to the fact that the defect-image in Fig.4.a is better contrasted than the one in Fig. 6.a This is why, in this case, the application of contrast enhancement is recommended before the edge detection operation.

Fig 4
Fig 6


Once, contours are extracted, they can present discontinuities. Then, we can use closing edge algorithms in order to prepare the recognition and classification stages. On the other hand, we have tested the network on noised images, as shown in Fig. 5.a . We ascertains that the network detects the contours and eliminate the noise (See Fig. 5.b). Thus, the network plays the double role : edge detection operator and noise filter. By there, we show the robustness of the proposed network regarding the noise.

Fig 5

2.3. Conclusion
The obtained results show the effectiveness of using neural paradigms to detecting edges in X-ray images which are used in Non Destructive Testing. It has therefore succeeded to deliver directly contours of welding defects present in radiograms without the application of filtering techniques, contrarily to classic edge detection operators. Indeed, the proposed neural segmentation technique has provided satisfying results on noised or variable luminance images. However, the inconvenience of this technique resides in its speed of execution that is slow enough. This is why, in this type of applications, powerful and rapid computers are recommended. Currently, our work focuses on the validation of the proposed approach by other neural models such as Kohonen and Hopfield ANNs.

3.  Preliminary classification of defects based Ann

3.1Geometric invariant moments
Moments have been used as pattern features in a number of applications, to provide invariant recognition of 2D image patterns.

The regular moments m_pq of a digital image pattern represented by f(x,y) are defined as :

(3.1)

p,q = 0,1,2,..

Hu introduced moments as image recognition features. Using non-linear combinations of normalized central moments, he derived a set of seven moments which has the desirable property of being invariant under image translation, scaling and rotation. Specifically, the central moments that have the property of translation invariance are given by :

(3.2)



The following normalized central moments are invariant under a scale change :

(3.3)

p+q = 2,3, . . .

The following moments Æ ₁, Æ ₂, . . . ,Æ ₇ , are invariant under translation, rotation and scaling :

(3.4)

We defined seven normalized values of Æ _i (i=1,2,...,7) which are applied in the input layer of neural network [11].

(3.5)

3.2. Neural network architecture
The conventional configuration for a multilayer neural network was used. This consists of an input layer of nodes with one node for each feature vector, a hidden layer, and an output layer with one node for each class. Each computational node (those in hidden and output layers) uses the sigmoid transfer function.The network weights were found by the back-propagation of errors technique.
Pattern vectors were generated by computing the invariant moments of the defects and then obtaining seven (7) values of moments (Æ'₁, Æ'₂, ... ,Æ'₇) of each defect. The resulting seven moments were the inputs to the three layer feed-forward neural network shown in Fig.3.

The two neurons in the output layer correspond to the number of pattern classes (defect classes), and the number of neurons in the middle layer was heuristically specified as 10 [12]. There are unknown rules for specifying the number of nodes in the internal layers of a neural network, so this number generally is based either on prior experience or simply chosen arbitrarily and then refined by testing.

3.3 Experimental results
We present results of combining the feature extraction technique by invariant moments and the neural network classifier described above using the welded defect X-ray images.

A. Invariant moment performance
In order to show the invariance performance of the proposed moments, we have used a set of test images representing a planer welded defect and a volumetric defect. Usual geometric transformations (rotation, scale change, the combination of the both and mirror effect) are applied on these images. Invariant moments are computed and compared between original images and their transforms. As shown in Figs and Histograms 7, the results are in reasonable agreement. The major cause of error can be attributed to the digital nature of the data.

B. Neural classifier performance
Initially, in he training process, the weights were initialized to small random values and the network was trained with the invariant moments corresponding to the planer and volumetric defects shown in Figs. 7. Each defect representation was assigned to a distinct output class. This set of couples (invariant moments, corresponding defects) are chosen as training set. This choice is justified by the fact that these defects can be considered as prototypes by referring to radiography interpreter suggestion. After the training phase was completed, the ability of the ANN to recognize defects was evaluated. The recognition phase was very simple for ANN and it consisted of only a feed-forward pass of the information presented to the input layer. The testing set is composed in the first part, by the transformed training defects by translation, rotation and scaling and in the second part by another set of defects as shown in Fig 8, and their transformed by the above transformations.

Fig 7: Training set of the ANNs and invariant moment performance a. Planer defect b. Volumetric defect


In the major classification problems, we are interested by the classification accuracy which corresponds to the outperform classified defect-images into the test set, or to the misclassification rate which corresponds to misclassified or non classified defects. In terms of classification accuracy, the simulation results are very satisfying and the two-class classification is well accomplished for the presented testing set. The results are summarized in Table 1.

Type of Testing set	nb of tested defects	nb of classified defects
Transformed Training set	9	9
Non trained set	20	19
Transformed non trained set	75	72
Table 1: Accuracy of defect classification

Fig 8: Some tested defects from the testing set of the neural network

3.4. Conclusion
In this paper, artificial neural network based approach for the classification of 2D dimensional welded defect images represented by translation, scale and rotation invariant region representation were introduced. ANN approach employing supervised learning represented by a multilayer ANN was utilized. The error back-propagation algorithm was used for the training of the multilayer ANN. Through experimentation with the defect-images for the classification problem, we show the feasibility of the proposed feature extraction and neural network paradigms, which are very promising in radiography inspection of welding joints. Presently, the use of a large bank of defects (porosity, leak of fusion, inclusions, etc.) as ANN training data is under investigation.

4. References

1. D. D. Dodge. Encyclopedia of science and Technology, McGraw-Hill. 1984.
2. B. Lavayssière & B. Georgel. Analyse multi-dimensionnelle en radiographie industrielle. Soudage et techniques connexes, Nov.Dec. 1991.
3. N. Nacereddine, R. Drai & F. Abdat. Traitement numérique des images et reconnaissance des formes par les réseaux de neurones artificiels en vue d'identification des défauts de soudures. Journées scientifiques sur le soudage et le contrôle, 6,7 Nov. 1994 Hôtel Sofitel.
4. R.C. Gonzalez & R.E. Woods. Digital Image Processing, Addison-Wesley Publishing Company. 1993.
5. M.K. Hu. Visual Pattern Recognition by Moments Invariants. IRE Trans. Info. Theory, vol. IT-8, pp. 179-187. 1962.
6. D. Rumelhart & J. McClelland. Parallel Distributed Processing, Cambridge, MA: M.I.T. Press. 1986.
7. E. D. Davalo & P. Naim. Des réseaux de neurones. Editions Eyrolles. 1990.
8. S. Ghosal & R. Mehrotra. Range Surface Characterization and Segmentation using Neural Networks. Pattern Recognition, vol. 28, no. 5, pp 711-727. 1995.
9. R. Nekovei and Y. Sun. Back-Propagation Network and its Configuration for Blood Vessel Detection in Angiograms. IEEE Trans. Neural Networks, vol. 6, no. 1, Jan. 1995.
10. N. Nacereddine, R. Drai & B.Heriouk. Artificial Neural Network and its configuration for weld defect edge detection in X-ray images. IASTED International Conference, Las Vegas, Nevada, USA, 28-31 Octobre. 1998.
11. C.H. Teh & R.T. Chin. On image analysis by the methods of moments. IEEE Trans. Pattern. Anal. Machine Intell., vol. 10, pp. 496-512. 1988.
12. N. Nacereddine. Preliminary classification of weld defects in radiograms based neural networks. IASTED International Conference, Las Vegas, Nevada, USA, 28-31 Octobre. 1998.

© AIPnD , created by NDT.net

|Home|    |Top|