A Step Towards Automatic Defect Pattern Analysis and

Evaluation in Industrial Radiography using Digital

Image Processing
Hemanth Jagannathan, Narayana Bhaskar, P. Sriraman C.N., Vijay N.A.
Department of Electrical & Electronics Engineering, Sri Venkateswara College of
Engineering, Sriperumbudur, Tamilnadu, India.
B. Venkatraman, P. Kalyanasundaram, Baldev Raj
Indira Gandhi Centre for Atomic Research, Kalpakkam-603 102
Contact

Abstract

Industrial radiography is a traditional technique for the identification and

evaluation of flaws, or defects, such as cracks, porosity and foreign
inclusions found in casting and machined parts. In the past 40 years
radiography has become one of the major techniques of industrial non-
destructive testing. Although this technique has been well developed,
improving both the inspection process and cycle time, it does suffer from
several drawbacks.

Digital image processing techniques allow the interpretation of the image

to be automated, avoiding the presence of human operators making the
inspection system more reliable, reproducible and faster. Moreover the
high-level image processing methods can even replace the expert's
knowledge.

This paper describes our attempt to develop and implement neoteric

algorithms for the purpose of automatic defect detection in radiographic
images by digital image processing. The various defects in radiographic
images are identified by means of various image-processing algorithms
suitable for defect detection. These are well established algorithms adapted
for use with radiographic images which include those for improving the
quality of the radiographic image, such as reducing image noise and
increasing contrast, and algorithms for analysis of image contents, such as
locating edges or regions and segmentation from the background image.
The standards and type of defects is used to train a neural network. After
preprocessing, the radiographic image is fed to the neural network. The
result is a summary of the type of defect, its extent, size and other
necessary details need for analysis. These defects are therefore
 automatically detected and evaluated with reference to standard
specifications.

Problem Specification

Non-destructive testing (NDT) is widely used in engineering industry to

detect and evaluate all type of defects in materials. Among other NDT
techniques, radiographic testing is widely used for identification and
evaluation of flaws, or defects, such as cracks, porosity and foreign
inclusions found in casting and machined parts.

The conventional radiographic system suffers from several drawbacks.

These include high work-in-progress and high operational costs.
Additionally, due to the nature of image formation and the resultant image
quality, the radiographic image presents many problems to the human
inspector, which makes interpretation of image content very difficult and
detection of defects inconsistent.

In the recent years, radiographic imaging is becoming a very important

area of interest for engineering industry. The radiographic images are
contaminated with noise and are also blurred. In order to improve the
image for observation and accurate analysis, various digital image-
processing techniques can be applied. The restoration of radiographic
images is possible through the use of various image processing methods
such as convolution, noise removal, edge detection filters and
morphological operations such as dilation, erosion, outlining. Thus there is
a need for the development of a Comprehensive Software Package for
Digital Image Processing of radiographic images. Further, using this
comprehensive package, it is planned to carry out detailed investigations
on a variety of defects, to ultimately develop procedures for automated
detection and evaluation of defects in imaged regions.

Conventional Radiographic System

The figure 1.0 gives a pictorial representation of a conventional digital

radiographic system. An X-ray source is used to obtain the radiographic
image of the test object. The image intensifier improves the density and
resolution of the image and the image is then passed to the Image
acquisition and processing system. This system has in-built image
processing routines. The image got and processed is stored on video tapes
or taken as print outs. This system suffers from a number of drawbacks,
namely
 1. High Initial Cost
2. Archiving of images is time consuming and not easy
3. The system is bulky and not portable
4. Upgradation is difficult

Fig 1: Conventional Radiographic System

Proposed Radiographic System

The figure 2.0 gives a pictorial representation of our proposed

radiographic system. Here the Image Acquisition and Processing System is
replaced by a Frame Grabber Card and a Personal Computer loaded with
our Image Processing software. The Frame Grabber Card is used to
interface the Personal Computer to the X-ray machine. The images after
processing are archived or stored on the Computer's hard disk or any other
external media.

Fig 2: Proposed Radiographic System

This system has the following advantages

1. Cut Down in Principle & Operating Costs

2. Archiving of Images is easy and possible
3. Improved operation due to increased speed of detection
4. Portability
 5. Can be easily upgraded in the future
6. Has compatibility with scanner and any other devices

Development of Methodology for Automation

Automated detection and evaluation of defects is gaining importance. Thus

it is intended to develop a methodology to process and evaluate
automatically the radiographic images produced.

Advances in real time radiography, or 'radioscopy', are serving to reduce

the limits on the speed, flexibility and performance of inspection cycles.
The counter result of these advances is that conventional manual
inspection can often no longer cope with the inspection rates and
intensities required. To address this problem various work has been
directed at automating the inspection task by a large number researchers
worldwide. The most difficult problem in the inspection cycle is the
accurate detection of defects in a given radioscopic image, and it is to this
area which the majority of research has been applied, though with limited
success. The general failings of the majority of published techniques can
be attributed to four key areas:

1. Unacceptable false alarm rates due to component structure and

noise,
2. Inability to detect defects of all orientations and types,
3. Inability to detect defects across different applications, and
4. Non realistic computation times.

For these reasons, work in fully automated on-line radiographic inspection

is being investigated for a number of industrial NDT applications. The
methodology developed for these tasks are artificial neural network
(ANN's) and Multi channel filtering schemes.

Once potential defective areas in an image have been highlighted by these

operators, artificial intelligence techniques, including both Expert Systems
and ANN's, can be successfully applied to provide full defect classification
and overall product quality interpretation.

Artificial Neural Networks for Defect Detection

The majority of methods proposed for automatic defect detection in NDT

x-ray images generally tend to create problems due to the various other
artifacts which typically appear in an image: notably a high noise content
 and component structures (edges), which serve to mimic the local
properties of a defect, and hence may be detected as false alarms. False
positive reactions can not only complicate subsequent processing which
must remove them, but may also, and perhaps more significantly, serve to
mask actual defect areas.

False negative reactions (i.e. the overlooking of a defect area) are also
potentially prevalent with many approaches. This is because the sharp
intensity changes that characterize a defect are often blurred by the x-ray
image process, and the transition between background and defect becomes
so slight as to be of similar magnitude to noise spikes.

The problem of finding an ideal defect detection operator is to produce a

very sensitive operator which is not prone to false alarms due to noise or
component structure. This is a very difficult task since the local area
properties of, for example, a weld edge very closely resemble those of a
true defect. However, inspired by the fact that artificial neural networks (or
ANNs) are, theoretically at least, able to simulate any arbitrary mapping,
we are working on employing such structures in an effort to solve the
problem of defect detection without false alarms.

Automated detection and evaluation of defects is gaining importance. Thus

it is intended to develop a methodology to process and evaluate
automatically the radiographic images produced.

Multi-channel filtering for defect detection

Defects are characterized in a radioscopic picture by local changes in x-ray

collection density which translates to a two dimensional transition in gray
level intensity. Image integration is useful in order to reduce noise whilst
the image is digitized. A set of particular convolution kernels can be used
to enhance defective areas within an image. Some filters are particularly
effective at detecting blob-like defects (e.g. porosity, inclusions) whilst
others are more useful for detecting longitudinal type defects such as
cracks.

Once an image has been filtered, one of two post processing operators is
applied in order to segment suspect defect pixels from the image
background. These are: -

1. A neural network trained to detect areas of 'roughness' in an image

 2. Calculation of the 2nd moment (variance) of a local area of the
image.

The neural network method is more sensitive and hence is likely to locate
very subtle areas but has a high false alarm rate. The variance operator is
cruder, but is also faster (speed increase of x5) and is also more 'tunable'
so it is applicable to most image types. Once suspect defect pixels have
been located, some blob analysis is carried out to eliminate any areas that
are either too small to warrant further investigation, or are likely to be part
of the component geometry (i.e. the edges of the weld structure, etc.).

Defect areas are then grouped and geometrically analyzed to give an

indication of their length, area, and position. It is envisaged that this
information will be used in future work to give a classification of each
defect, which is located.

The proposed methods reduce the operator-induced errors and ensure

reliable and repeatable results during inspection.

Conclusion

The automation of defect pattern analysis in industrial radiography using

Image Processing can overcome the difficulties faced in the conventional
model. Moreover the principal and running cost are considerably lower.

This technique of defect pattern recognition and evaluation lends to

improvement by incorporating Artificial Intelligence for intelligent
analysis and report generation.

In conclusion we have presented a model for Automated Defect Pattern

Analysis which has the advantage of low cost, ease of manipulation,
portability and real time testing.

References

1. A Kehoe and GA Parker Image Processing for Industrial Radiographic

Inspection, British Journal of NDT 1990 April Vol. 32 no. 4 p 183-189
2. V Kaftandjian, y M Zhu, G Piex and D Babot Automatic recognition of
defects inside aluminum ingots by X-ray imaging, INSIGHT 1996
September Vol. 38 no. 9 p 618-625
3. X-Ray Real-Time Imaging (Radioscopy) for Weld Inspection : IIW
Summary Report British Journal of NDT 1990 January Vol. 32 no. 1 p 9-
14
4. W Thale, R Clausen and T Just ( Translated by T Kowol) Digital image
processing for the interpretation of radiographs used for condition
monitoring, INSIGHT 1996 September Vol. 38 no. 9 p 632-635
5. Nockeman C., Heidt H., Thomsen N., Reliability in NDT : ROC study of
radiographic weld inspections, NDT&E International, 1991, Vol. 24, no. 5,
p.235-245.
6. Kehoe A., Parker G.A., An intelligent knowledge based approach for the
automated inspection of castings, NDT&E International, 1992, Vol. 25,
no. 1, p.23-36.
7. Sera J., Image analysis and mathematical morphology, New-York :
Academic Press, 1984, vol.I, 610 p.
8. William K Pratt Digital Image Processing
9. Anil K Jain Fundamentals of Digital Image Processing
10.Rafael C Gonzalez, Richard E Woods Digital Image Processing
11.Maher A Sid-Ahmed Image Processing- Theory, Algorithms, and
Architectures
12.Adrian Low Introductory Computer Vision and Image Processing
13.Edward R. Dougherty, Charles R. Giardina Matrix Structured Image
Processing
14.William B. Green Digital Image Processing
15.Dwayne Phillips Image Processing in C
16.Kenneth R. Castleman, DIGITAL IMAGE PROCESSING, Prentice-Hall
(Signal Processing Series), 1979.
17.Hedengren.k.h., McCary.R.O. and Young.J.D., Rev. of Prog. In
Quantitative NDE, Eds. D.O. Thompson and D.E. Chimenti, Plenum
Press, New York, Vol.7A, 1988, pp.354-372.
18.Copley. David. C., Rev. of Prog. In Quantitative NDE, Eds. D.O.
Thompson and D.E. Chimenti, Plenum Press, New York, Vol. 2B, 1993,
pp.1517-1550.
19.Benoist.B, Attoui.P, La.R, Lengelle.R, Gaillard.P, Reuchet.J, Rev. of Prog.
In Quantitative NDE, Eds. D.O. Thompson and D.E. Chimenti, Plenum
Press, New York, Vol. 17, 1998, pp.1954-1973.
20.Steblev.Yu.L, skorobogatov.E.G, Russian Journal of NDT, Vol.33, 1997,
pp.230-240.
21.Young-Won Song and Udpa. Satish S., Rev. of Prog. in Quantitative NDE,
Eds. D.O. Thompson and D.E. Chimenti, Plenum Press, New York,
Vol.16, 1997, pp. 720-735.
 22.Hedengren.K.H., McCary.R.O. and Young.J.D., Rev. of Prog. in
Quantitative NDE, Eds. D.O. Thompson and D.E. Chimenti, Plenum
Press, New York, Vol. 8A, 1989, pp.975-983.
23.Bernd Jahne, Digital Image Processing Concepts, Algorithms, and
Scientific Applications, Second Edition, John Wiley and Sons, Inc., 1991.
24.Robert J. Schalkoff, Digital Image Processing and Computer Vision, John
Wiley and Sons, Inc., 1989.
25.G.W.Awcock and R.Thomas, Applied Image Processing, McGraw Hill
International Editions, 1996.

https://www.ndt.net/article/wcndt00/papers/idn673/idn673.htm