Optik 127 (2016) 1030–1033

Contents lists available at ScienceDirect

Optik
journal homepage: www.elsevier.de/ijleo

Detection crack in image using Otsu method and multiple filtering in

image processing techniques
Ahmed Mahgoub Ahmed Talab a,b,∗ , Zhangcan Huang a , Fan Xi a , Liu HaiMing a
a
School of Sciences, Wuhan University of Technology, Wuhan 430070, China
b
College of Engineering, Elimam ELMahdi University, Kosti, Sudan

a r t ic l e in f o a b s t r a c t

Article history: Cracks in concrete structures can indicate major structural problems and can harm the appearance of
Received 21 October 2014 monolithic construction. This article presents a new approach in image processing for detecting cracks
Accepted 9 September 2015 in images of concrete structures. This method involves three steps: First; change the image to a gray
and using edge of the image and use Sobel’s method and development of an image filter using Sobel’s
Keywords: filter for detecting cracks. Second; using a suitable threshold in a binary image and classifies all pixels
Crack detection
two categorizations background and foreground, and gets the region area after that use filter area and
Otsu method
changes the area if less than the specific number to back. Third; after using Sobel’s filtering to elimination
Sobel filter
of residual noise, and detecting major cracks using Otsu method. This paper describes a method for
detection crack patterns in cement use image processing techniques. The advantage of this method is
clearly and accurate detection of cracks in images. Experimental work shows that our method is improved
relatively to the other widely used techniques.
© 2015 Published by Elsevier GmbH.

1. Introduction for quantitative micro-structural investigation of concrete and let

the crack pattern in these materials to be measured [3]. In this arti-
Cracks are one of the most serious defects in concrete struc- cle, we use experimental program to investigation many objectives.
tures because when they are developed, they tend to reduce the Use experimental program to find the specimen, and use this image
effective loading area which is leads to increase of stress and sub- of a specimen to find a new approach to detection crack in image
sequently failure of the concrete [1]. Although cracks originated processing.
at the sections surface, it is difficult to detect visually if the crack
width small [2]. The hydrated cement deterioration takes place due
2. Experimental program
to many factors; the combination of these factors returns differ-
ent configuration of cracks in concrete. Each mode of cracks will
2.1. Equipment
yield diverse damage and failure modes in concrete structures [3].
Fresh concrete subjected to crack during tropical or cold conditions
For Uniaxial tensile test using Universal Testing Machine
[4–8], especially if it is insufficiently cured. The trend of concrete
(Instron 5882), the Loading speed ratio was 0.4 mm/min which was
to cracking due to temperature emanating from weather condition
controlled under the displacement control mode.
or even the hydration process has been known since the beginning
of concrete technology [3]. The most widely applied techniques
for examining cracks in concrete are scanning electron microscopy 2.2. Test method
and optical fluorescent microscopy [9–13]. Two tools should be
adopted to characterize cracks: one for observation and one for Data collection, including capacity and strain. The test method
quantitative analysis. Currently developments in microscopy and referred to the Japanese standard. The strain data was measured by
image processing and analysis techniques render powerful tools the 10 mm LVDT which was fixed on both sides of the sample. And
the sampling frequency was 1 Hz. The sample and the LVDT were
shown as in Fig. 1.
∗ Corresponding author at: School of Sciences, Wuhan University of Technology, The prepare and product is located in a laboratory at Wuhan
Wuhan 430070, China. Tel.: +86 13720263554/24918015860. University of Technology. The main steps of producing the crack
E-mail address: [email protected] (A.M.A. Talab). in this specimen as follows; firstly is to put the specimen in

http://dx.doi.org/10.1016/j.ijleo.2015.09.147
0030-4026/© 2015 Published by Elsevier GmbH.
 A.M.A. Talab et al. / Optik 127 (2016) 1030–1033 1031

First step:

(1) Use the image (I) RGB, and change the image to gray.
(2) Use the edge of the image.
(3) Use the filter of the Sobel.

Second step:

(4) Using a suitable threshold (T) in a binary image that classi-

fies all pixels in two classifications background and foreground.
According to this equation.

!
1 im > T
I= (1)
0 otherwise

(5) Use the function (bwlabed).

(6) Use the function (region props) to get the region area, and get
the filter area.
(7) Change the small area than 30 to back.

Third step:
Fig. 1. The specimen in this paper.
(8) Use the Otsu method [14]. The main idea in proposed method
Universal Testing Machine (Instron 5882), and second step is to to detection crack in this paper to depend of Otsu method.
apply automatic loading down as the same time reading the mag-
nitude of applied loading in the programmer; then the next step
3.1. Otsu algorithm analysis
is performed if the cracks start; by locating the digital camera as
shown in Fig. 1.
The Otsu method is the basic idea let the threshold image
into two groups. The target number of pixels representing the
3. The proposed method
ratio of the image ω0 , the average gray level "0 , the Background
image represent the ratio of the number of pixels ω1 , the aver-
The object of this paper is propose new approach at binarization
age gray level "1 . The total average gray level of the image
method in image processing for detection crack in image by opti-
is: " = ω0 (t)"0 (t) + ω1 (t)"1 (t). From the minimum value to the
cal fluorescent microscopy, as well as use combination method to
maximum gradation value gray traversal t, When making value
detection crack in this image. The proposed algorithm consists of
g = ω0 ("0 − ")2 + ω1 ("1 − ")2 Segmentation is the optimal maxi-
three steps (Fig. 2):
mum threshold. On the Otsu method can be understood as follows:
the formula is actually between-class variance, threshold seg-
mentation of the target and the background of two parts of the
whole image, and the target value "0 , probability ω0 , background
values "1 , probability ω1 , total mean ", according to the defini-
tion of variance that was the style. Gray distribution of variance
is a measure of homogeneity. Suppose f(x, y) to N × M image
(i, j) gray value at the point of, the gray level ", let’s assume
f(i, j) value [0, m − 1]. Suppose p(k) the gray value k frequency,
then:
1 "
p(k) = 1 (2)
MN
f (i,j)=k

Suppose t is a gray value threshold segmentation of target and

background are as follows:
"
ω0 (t) = p(i) (3)
0≤i≤t

Target part points:

"
N0 (t) = MN p(i) (4)
0≤i≤t

Background section ratio:

"
ω1 (t) = p(i) (5)
Fig. 2. Flow chart of the proposed method. t<i≤m− 1
1032 A.M.A. Talab et al. / Optik 127 (2016) 1030–1033

The right-hand side brackets actually inter-class variance, the

threshold value g segmented object and the background of two
parts of the whole image, The target "0 (t), probability ω1 (t), back-
ground values "1 (t), probability ω0 (t), total mean " [14].

4. Experimental and discussion

In this section, we show experimental result of the proposed

method. The proposed algorithm was implemented in MATLAB
(R2012 b) and tested in windows 8 with Intel (R) core (TM) i7-
3632QM CPU@ 2.20 GHz with a memory of 4GB.
To assess the characteristics of the crack network, the (Fig. 3.) use
proposed method. This result in Fig. 3(b) to explain the proposed
Fig. 3. The typical images of a cement specimen.
method is a good approach to detection crack in image concrete. The
advantage of this method is clearly, accurate and no noise to detec-
tion cracks in the image. The (Fig. 3C) use proposed method and
Background section points: Skeleton (function of thin in Matlab package in image processing)
" to reduce the cracks to single pixel lines [3].
N1 (t) = MN P(i) (6)
In order to make comparisons, we implemented three binariza-
t<i≤m− 1 tion methods, all method are global threshold methods are Otsu
Target mean: method [14], Kittler met method [15] and the proposed method.
The extracted crack shown in Fig. 4 by the proposed method is
" ip(i)
"0 (t) = (7) clearer and well separated from the surrounding objects when
ω0 (t) compared with the other methods. Moreover, the resulted image
0≤i≤t
has less noise than the other approaches.
Background mean:
" The crack in Otsu method has noise while the resulted crack of
(8)"1 (t) = ip(i) the Kittler method is not well separated and has thicker lines and
ω1 (t)
t<i≤m− 1
has the highest noise in contrast with the other methods.
Total mean:(9)" = ω0 (t)"0 (t) + ω1 (t)"1 (t)
Otsu method for image that the best threshold(10)g = 5. Conclusion
Arg Max [ω0 (t)("0 (t) − ")2 + ω1 (t)("1 (t) − ")2 ]
0≤t≤m− 1
This paper proposes a new approach in image processing
method for detecting cracks in image. In this method, we use more
than one filter like Sobel and filter of Area and change the small
area to background, and detecting major cracks using Otsu method.
The experimental results of the proposed method in this paper
successfully extract background and foreground image. The advan-
tage of this method is clearly and accurate detection of cracks in
images.

Acknowledgments

The authors would like to thank all students in the laboratory

in the Mechanical Department of Wuhan University of Technol-
ogy to assist in the preparation of the specimen in this paper, and
gratefully appreciate Mr. Xiong Jiang for his assistance within the
research.

References

[1] T. Nishikawa, J. Yoshida, T. Sugiyama, Y. Fujino, Concrete crack detection by

multiple sequential image filtering, Comput.-Aided Civ. Infrastruct. Eng. 27
(2012) 29–47.
[2] P. Broberg, Surface crack detection in welds using thermography, NDT&E Int.
57 (2013) 69–73.
[3] A. Litorowicz, Identification and quantification of cracks in concrete by optical
fluorescent microscopy, Cem. Concr. Res. 36 (2006) 1508–1515.
[4] C. Qi, J. Weiss, J. Olek, Characterization of plastic shrinkage cracking in fiber rein-
forced concrete using image analysis and a modified Weibull function, Mater.
Struct. 36 (2003) 386–395.
[5] K. Wang, S.P. Shah, P. Phuaksuk, Plastic shrinkage cracking in concrete materi-
als? Influence of fly ash and fibers, ACI Mater. J. 98 (2001).
[6] Y. Ma, M. Tan, K. Wu, Effect of different geometric polypropylene fibers
on plastic shrinkage cracking of cement mortars, Mater. Struct. 35 (2002)
165–169.
[7] A. Brandt, V. Li, I. Marshall, Application of uv image analysis for eval-
uation of thermal cracking in concrete, Brittle Matrix Compos. 7 (2003)
Fig. 4. Experiment results for global binarization methods. 101.
 A.M.A. Talab et al. / Optik 127 (2016) 1030–1033 1033

[8] S. Lykke, E. Skotting, U. Kjaer, Prediction and control of early-age cracking: [12] P. Soroushian, M. Elzafraney, A. Nossoni, Specimen preparation and image
experiences from the Oresund Tunnel, Concr. Int. 22 (2000). processing and analysis techniques for automated quantification of concrete
[9] H.C. Gran, Fluorescent liquid replacement technique. A means of crack detec- microcracks and voids, Cem. Concr. Res. 33 (2003) 1949–1962.
tion and water: binder ratio determination in high strength concretes, Cem. [13] J.R.A. Ammouche, D. Breysse, J. Marchand, Image analysis for the automated
Concr. Res. 25 (1995) 1063–1074. study of microcrack in concrete, Cem. Concr. Compos. (2001) 267–278.
[10] M. Elzafraney, P. Soroushian, Assessment of microcrack development in con- [14] N. Otsu, A threshold selection method from gray-level histograms, IEEE Trans.
crete materials of different strengths, Mater. Struct. 37 (2004) 724–731. Syst. Man Cybern. SMC-9 (1) (1979) 62–66.
[11] K.M. Nemati, P. Stroeven, Stereological analysis of micromechanical behavior [15] J. Kittler, J. Illingworth, Minimum error thresholding, Pattern Recognit. 19
of concrete, Mater. Struct. 34 (2001) 486–494. (1986) 41–47.