flaw detection in radiographic weld images using morphological approach
TRANSCRIPT
Flaw detection in radiographic weld images using morphological approach
Alaknandaa, R.S. Ananda,*, Pradeep Kumarb
aDepartment of Electrical Engineering, Indian Institute of Technology, Roorkee, Roorkee 247667, IndiabDepartment of Mechanical and Industrial Engineering, Indian Institute of Technology, Roorkee, Roorkee 247667, India
Received 23 February 2005; revised 13 May 2005; accepted 18 May 2005
Available online 26 July 2005
Abstract
It is necessary to detect suspected defect regions in the radiographic weld images to find the type of flaw and its causative factors. This
requires processing of radiographic images by a suitable approach. This paper presents an approach to process these radiographic weld
images of the weld specimens considering morphological aspects of the image. The proposed approach first determines the flaw boundaries
by applying the Canny operator after choosing an appropriate threshold value. The boundaries are then fixed using a morphological image
processing approach i.e. dilating few similar boundaries and eroding some irrelevant boundaries decided on the basis of pixel characteristics.
The flaws detected by this approach are categorized according to their properties.
q 2005 Published by Elsevier Ltd.
Keywords: Radiographic images; Flaw detection; Edge detection; Dilation; Erosion
1. Introduction
Some times, weldments are not properly joined together
during the welding process due to different reasons, which
produce flaws of different types in weldments. These flaws
are seen in the form of gaps, cracks or pores and cavities.
Common weld defects include—lack of fusion, lack of
penetration or excess penetration, gas cavities, slag
inclusions, cracks, undercuts, lamellar tearing, shrinking
cavities etc. The lack of fusion results from too little heat
input and/or too rapid traverse movement of the welding
torch (gas or electric). Lack of penetration or excess
penetration arises from a too high heat input and/or too
slow movements of the welding torch (gas or electric). Gas
cavities occur when gases are trapped in the solidifying
weld metal. During solidification gases are trapped in the
welded zone resulting in porosity. Slag inclusion is of
different types. It can be of any shape in any direction—
slag lines (elongated cavities containing slag or other
foreign matter), wearing faults, faults from bad chipping,
faults at electrode change, faults at junction of seams.
0963-8695/$ - see front matter q 2005 Published by Elsevier Ltd.
doi:10.1016/j.ndteint.2005.05.005
* Corresponding author. Tel.: C91 1332285590; fax: C91 1332273560.
E-mail address: [email protected] (R.S. Anand).
Cracks are often caused by sulphur and phosphorus and are
more likely to occur in higher carbon steels. They are of
two types: longitudinal cracks and transverse cracks.
Longitudinal cracks normally appear in straight lines
along the centerline of the weld bead while transverse
cracks are straight lines perpendicular to centerline and
occasionally appear. Undercut is due to the reduced
thickness of one (or both) of the sheets at the toe of
weld, due to inaccurate settings/procedure. Lamellar
tearing is mainly a problem with low quality steels. It
occurs in plates that have low ductility with laminar
segregation. Shrinkage cavities are due to thermal
shrinkage or due to a combination of steam accompanying
phase change and their shrinkage cavities [1,2].
Radiography using X-rays is one of the NDT techniques
which are used for imaging the weldments for detecting and
locating the defects in them. In an X-ray image of
weldments, variation in intensity is caused by inhomogen-
eity of the weldments. Light patches or lines seen on the
films correspond to different types of discontinuities due to
lack of penetration, cracks, cavities, slag inclusion, under-
cuts and shrinkage cavities. The thickness of material in
welding process, weld type, weld position and radiographic
methods plays an important role in image quality. The flaws
can also be detected during the welding process on hot
welds, which further helps in quality control as well as in
cost reduction.
NDT&E International 39 (2006) 29–33
www.elsevier.com/locate/ndteint
Fig. 1. Original X-Ray figure
Alaknanda et al. / NDT&E International 39 (2006) 29–3330
In classical methods of analyzing the images, human
experts make the quality control of welded joints by
identifying defect characteristics in NDT images and the
results depend on quality of images, previous experiences
of human experts and their mental fitness. The classical
process is slow and results are varying from expert to
expert and even the same person can give different
results for the same image depending on his mental
state [3].
Image processing plays an important role to detect flaws
in weldments. With the help of image processing a semi-
automated system can be developed which is more reliable
as compared to the classical methods and also helpful
in taking decision with the help of a set of manipulating
tools [4].
2. Morphological image processing
The flaw information in the weld images is obtained by
segmenting the weld image using a boundary-based
approach. Segmentation of an image using boundary-
based approach involves following approaches:
Fig. 2. Binary gradient mask of the original image.
2.1. Edge detection
The edges in the image are formed where there are
abrupt changes in the intensity of the pixels. There are
different types of edge detectors, which can be used for
detecting the edges [5,6]. The Canny edge detector [7] is a
very important tool for local preprocessing to locate
changes in intensity function, having very good signal to
noise ratio. By using the Canny operator, the main
criterions are the detection of all important edges, the
distance between the actual and the located position of the
edge should be minimal and the minimum number of
multiple response to a single edge. In order to reduce noise
in Canny operation, first of all the image f(x) is smoothened
by convolution with a Gaussian function of scale s as
given below in Eqs. (1) and (2).
Gðx; sÞ Z eKðx2=2s2Þ (1)
Fðx;sÞ Z f ðxÞ!Gðx; sÞ (2)
where s is the standard deviation and changes according to
images. F(x,s) represent a surface on the (x,s) plane.
In an image, an edge point is defined as a point whose
strength is maximal in the direction of the gradient. For
each pixel, the direction n of the local edge for the
Canny-operation is estimated by Eq. (3) and expressed as
below
n ZVðG!f Þ
jVðG!f Þj(3)
where G is a 2D Gaussian function and f is an image
function.
The location of an edge (also known as non-maximal
suppression) is obtained with the help of Eq. (4).
v2
vn2G!f Z 0 (4)
The location is tracked along the top of obtained ridges.
All pixels that are not actually on the ridge top are set to
zero. This procedure of tracking the location of edges has
been set to obtain a thin line at the output. Further, the
spurious edges are eliminated by thresholding with
hysterisis. For thresholding, the magnitude of edge is
computed by calculating Eq. (5) as expressed below.
jGn !f j Z jVðG!f Þj (5)
Then two thresholds T1 and T2, satisfying T1!T2, are
used to threshold ridge pixels. Ridges having a value greater
than T2 are said to be ‘strong’ and other ridge values are
considered as ‘weak’. The image obtained after applying the
Canny operation on the image of Fig. 1 is shown in Fig. 2.
The image so obtained has contours as is seen in the image.
It is the binary gradient of the original image.
2.2. Morphological processing
The output obtained in the form of contours has
discontinuities due to crack edges. To obtain a closed
contour morphological image processing is done on the
image. The image is dilated first and then eroding is
performed [8,9]. In this process, high contrast lines having
gaps in the binary gradient mask image disappear using
appropriate structuring elements in dilation. High contrast
lines do not delineate the outline of the object of interest
very well. As compared to the original image, gaps can be
seen in the lines surrounding the object in the gradient mask
image. These linear gaps will disappear if the image after
performing the Canny operation is dilated using linear
structuring elements. Dilation is an operation that ‘grows’ or
Fig. 3. Dilated gradient mask of the binary image. Fig. 5. Superimposition of the segmented image on the original image.
Alaknanda et al. / NDT&E International 39 (2006) 29–33 31
‘thickens’ objects in the binary images. Mathematically
speaking, dilation of A by B denoted as A4B, is defined as
Eq. (6)
A4B Z fz=ðBÞzhAsFÞ (6)
where F is the empty set, B is the structuring element and A
is the object to be dilated.
Dilation is used to fill small holes and narrow gulfs in
objects. The binary gradient mask image is dilated using the
vertical structuring element followed by the horizontal
structuring element. The dilated gradient mask of the binary
image is shown in Fig. 3.
After dilation of an output image of the edge detection
operation, smoothening of the object is done in order to
make the segmented object look natural. The object is
smoothened by combining dilation with erosion to preserve
the original size. Erosion of A by B is the set of all points z
such that B, translated by z, is contained in A and
represented by Eq. (7)
AQB Z fz=ðBÞz 4Ag (7)
The object is smoothened by eroding the image twice
using appropriate structuring element (e.g. line, disk or
diamond). When the dilation is followed by erosion it is
known as morphological closing, which is mathematically
denoted by Eq. (8)
A$B Z ðA4BÞQB (8)
The output of the morphological function applied on the
dilated image is shown in Fig. 4. The superimposition of the
segmented image on the original image is shown in Fig. 5.
In this image, an incomplete penetration type of weld defect
is seen.
Fig. 4. Segmented images.
3. Results and discussions
It is very useful in industries to find out the flaws
even during process. The discussed technique can be
used to find out the position of the flaw in the weldment,
which will further help in finding out the type and the
cause due to which flaws are produced. This can be
useful in reducing the flaws by controlling the relevant
weld parameters.
The algorithm has been tested on different industrial
radiographic images depicting the various types of
welding defects choosing appropriate threshold value
for the Canny operation to find the edges. Different
contours of weld defects are extracted which are
smoothened by using morphological functions to detect
the flaws in an image. On digitized image of Fig. 6a
discontinuities are detected using the Canny edge
detection. The obtained edge profile is presented in
binary format using an appropriate threshold value.
Morphological operations are applied to dilate and to
smoothen the obtained binary image, which gives outer
border of segmented objects of an image after super-
imposition on the original image. The image obtained
after superimposition is shown in Fig. 6(b). Fig. 6(b)
shows a slag line defect present in the weld. The above
algorithm has been applied on four such radiographic
images having different type of welding flaws. Both, the
original and processed images are shown in each case.
Segmented images of respective original images
Figs. 7(a)–10(a) have been shown in Figs. 7(b)–10(b).
Fig. 6. (a) Original image of a slag line defect. (b) Segmented image of the
slag line defect.
Fig. 7. (a) Original image of a lack of fusion type defect. (b) Segmented
image of a lack of fusion type defect.
Fig. 8. (a) Original image of slag inclusions defects. (b) Segmented image
of slag inclusions defects.
Fig. 10. (a) Original image of an incomplete penetration defect. (b)
Segmented image of an incomplete penetration defect.
Table 1
Category of defects in segmented radiographic images
Segmented image appearance Type of flaw
Thin line along the edge of weld. Line may be wavy
and diffuse depending upon the orientation of defect
with respect to the X-ray beam
Lack of fusion
Edge line in the middle of weld Incomplete
penetration
Line more or less interrupted, parallel to the edges of
weld
Slag line
Irregular contour of any shape and size Slag inclusion
Fine line, straight or wandering in direction Cracks
Line some times broad and diffused along the edge
of weld
Undercuts
Rounded contours having dark shadows Porosity (gas
cavities)
Alaknanda et al. / NDT&E International 39 (2006) 29–3332
These radiographic images can be categorized on the
basis of information available in them. Table 1 summar-
izes this information.
4. Conclusions
The morphological approach has been implemented here
for the segmentation of weld defects in radiographic weld
images. The segmented image can be used to characterize
the flaws in weldments. Flaws characterized in segmented
images can be categorized in different types like lack of
Fig. 9. (a) Original image of gas cavity type defects. (b) Segmented image
of gas cavity type defects.
fusion, incomplete penetration, slag line, slag inclusion,
cracks, undercuts, porosity and wormholes. Once the
images are categorized the causes due to which flaws are
generated can be analyzed and can be taken care off in
future during the welding process.
Acknowledgements
The authors are thankful to the Head, Department of
Electrical Engineering, Indian Institute of Technology,
Roorkee for providing the required computational and
experimental facilities. The financial assistance provided by
the Council of Scientific and Industrial Research, New
Delhi, India is gratefully acknowledged.
References
[1] Vamos G, Lovanyi I, Nagy A, Kiss B. Flaw detection in metallic fusion
welds on X-ray images using Bayesian networks. Proceedings of first
Hungarian conference; 2002. p. 118–23.
[2] Grieve DJ. Welding defects. www.tech.plym.ac.uk/sme/strc201/wde-
fects.htm; 2003.
Alaknanda et al. / NDT&E International 39 (2006) 29–33 33
[3] Nockeman C, Heidt H, Tomsen N. Reliability in NDT: ROC study of
radiographic weld inspection. NDT&E Int 1991;24(5):235–45.
[4] Kaftandjian V, Joly A, Odievre T, Courbiere M, Hantrais C. Automatic
detection and characterisation of aluminium weld defects: comparison
between radiography, radioscopy and human interpretation. Seventh
ECNDT 1998;3(10):1–7.
[5] Laggoune H, Gouton SP. Dimensional analysis of the welding zone
Proceedings of 22nd international conferences on information
technology interface 2000. p. 719–24.
[6] Daillant G, Micollet D, Paindavoine M. Defect in a weld: a complete
radiographic processing line. Industrial electronics, control, and
instrumentation Proceedings of the 1996 IEEE IECON 22nd
international conference; 1996. p. 719–24.
[7] Canny J. A computational approach to edge detection. IEEE Trans
Pattern Anal Mach Intell 1986;8(6).
[8] Gonzalez RC, Woods RE. Digital image processing. 2nd ed; 2002.
[9] Sonka M, Hlavac V, Boyle R. Image processing, analysis, and machine
vision. 2nd ed; 2003.