A New Graphical Quantitative Evaluation Method of Concrete Cracks Based On Digital Image Technology and Fractal Theory
A New Graphical Quantitative Evaluation Method of Concrete Cracks Based On Digital Image Technology and Fractal Theory
A New Graphical Quantitative Evaluation Method of Concrete Cracks Based On Digital Image Technology and Fractal Theory
1 Introduction
The early age cracking of concrete is always a formidable and uncontrollable
engineering problem. The cracks will not only affect the appearance of a building, but
also weaken the strength and workability of concrete. The cracks, as a path of corrosive
ions, such as SO42-, CO32-, et al, which means exposing the inside of concrete out,
will influence the long term performance of concrete severely. Although many efforts
have been put forward to resist early age cracking of concrete, but this problem is still
one of the major concerns of the engineers during design, construction and operation.
One of the reasons that limit the investigation of early age cracking of concrete is how
to quantitatively identify the parameters of concrete cracks with efficiency. Currently,
the weighted value of concrete cracks are always defined and calculated with the length
and width of concrete cracks which are measured manually [1]. Because the evaluation
results will be subjectively affected by the measuring procedure, so this method is
*
Corresponding author: Ye Tian, Dr., Lecturer, College of Civil Engineering and Architecture,
Zhejiang University, Hangzhou, 310058, China, Tel.: +86-571-88981320.
Y. Yu, Z. Yu, and J. Zhao (Eds.): CSEEE 2011, Part I, CCIS 158, pp. 275283, 2011.
Springer-Verlag Berlin Heidelberg 2011
276
X. Jin et al.
highly in short of accuracy and efficiency. With the application of digital image
processing technology in civil engineering, the researchers established several new
methods to identify and extract the cracks based on gray threshold principle, such as
maximum entropy threshold method [2], segmentation based on edge intensity [3],
Sobel operator method [4], and maximum between-cluster variance method [5], et al.
But so far there is no universal algorithm that can isolate the complete information of
concrete cracks automatically and accurately in a complicated situation. Especially
when concrete cracks are very small, or the edges of cracks have similar gray scale with
the background, there will be a great possibility of misjudgment in extracting the cracks
from background. In addition with the denoising processing after the segment, these
methods all state a large error tolerance.
In this research, a digital image processing technology was adopted to segment and
extract concrete cracks in high accuracy with Photoshop. And the characteristic
parameters of concrete cracks, such as area, fractal dimension, length and width, were
all calculated and programmed with MATLAB. Then, the irregularity properties of
early age concrete cracks were represented using the Fractal theory.
277
the areas of crack using the Magic Wand Tool and extract the non-cracking area using
the Invert select tool in Photoshop. As shown in Fig.2, once the non-cracking area is
highlighted with white color, the segmentation of concrete cracks is accomplished. The
image resolution of the cracks is determined by the scale of pixels. In this research, the
digital camera can resolute a pixel with the length of 0.0084 mm which means the
image processing can achieve a image with the fineness of 0.0084 mm. Finally, the
binary function in MATLAB is applied to transfer the grayscale image into binary
image.
278
X. Jin et al.
(1)
where s is the area of the crack, n is the total number of pixels with the value 0 in digital
matrix of binary image.
3.3 Fractal Dimension of Concrete Cracks
In this research, the fractal dimension is calculated and programmed according to the
box dimension calculation method with MATLAB [8]. In calculating the box
dimension, the size of the box is controlled by changing the size of matrix of sub-pixels
using different size of little matrix, named as box, to cover the pixel matrix [9]. The
initial number of the boxes, M, is set as 0 in each mesh generation. If the value of every
element in a little matrix or box equals to1, then this area is background and the value of
M will not change. If the value of any element in a little matrix doesnt equal to1, then
this area is cracks and the value of M will change to M+1. Traversing all sub matrix, the
number of the boxes M is obtained with different size of little matrix covering the pixel
matrix. Then the fractal dimension D and related coefficient rd can be calculated
according to the box dimension calculation method.
13
12
11
f(x)=12.0435-1.2479x
logN
10
9
8
7
6
5
0
logb
279
( xi xi 1 )
+ ( yi yi 1 ) sb / m .
2
LI = li .
(2)
(3)
where li is the distance between two adjacent pixels l1=0, LI is the length of the upper
bounds of the crack calculated with asymptotic method.
In the same way, the length of the lower boundary of the crack can be calculated with
the same way. The larger value can be taken as the length of concrete crack.
The second method is fractal method, which is to calculate the length of cracks with
fractal theory. The specific procedure includes following steps. The left and right
ending points of concrete crack are coordinated in the digital image matrix with (xa, ya)
and (xb, yb). And the distance between the two points can be calculated with the
coordinate system. According to fractal theory, fractal dimension demonstrates the
degree of irregularity or tortuosity of concrete crack. So, the larger one of the fractal
dimensions of lower and upper boundaries can be taken as the fractal dimension of the
crack. The length of crack can be demonstrated as:
l=
( xb xa )
+ ( yb ya ) sb / m .
LII = l Db .
(4)
(5)
where L is the length of concrete crack calculated using fractal method, l is the straight
distance between the two ending points of the crack, Db is the fractal dimension of the
crack.
280
X. Jin et al.
Where
(x
xj
(y
yj
sb / m .
(6)
(7)
wmax = max( wi ).
(8)
Where wmax is the maximum width of concrete crack; wave is the average width of
concrete crack; k states the number of pixels in P matrix.
4 Experimental Results
By means of fractal theory and the image processing method, the early age cracks of
concrete slab were identified. The concrete strength grade was C30 and water/cement
ratio is 0.53. Concrete flat slab test provides a simulation of early age cracking of
concrete in restrained stated at all four sides with a steel square mould with inner size of
600mm600mm63mm. At every side of the mould, two rows of bolts were fixed by
double nuts (10100mm, 14 in total) as the restrained boundary. Concrete flat slab
was put into the curing room with the curing temperature of 303, relative
humidity of 65%5% and wind speed about 3m/s. In twenty four hours, the cracking of
concrete, as shown in Fig.5, was observed using a reading microscope.
The camera used in this research was Canon Powershot a2000IS, with image
pixels of 36482736. The reference scale was a rectangle with the size of
70mm23mm. using the proposed digital image processing technology, the
characteristic parameters of the concrete cracks were recognized. The area of the
cracks s, fractal dimension D, related coefficient rD, length of concrete cracks LI and
LII, maximum width of the cracks wmax and average width of the cracks wave are all
listed in Table 1.
281
S[mm2]
rD
LI [mm]
LII [mm]
1
2
3
4
5
6
14.275
9.251
10.910
24.262
5.291
4.443
1.4035
1.3700
1.2745
1.4193
1.3063
1.3269
0.9948
0.9946
0.9942
0.9953
0.9952
0.9945
103.821
76.440
111.003
133.152
51.823
42.609
98.589
73.667
113.786
122.649
53.960
42.108
wmax
[mm]
0.457
0.236
0.266
0.334
0.221
0.201
wave
[mm]
0.182
0.133
0.125
0.206
0.134
0.131
strue [mm2]
Ltrue [mm]
wtrue [mm]
13.975
100
0.445
9.570
75
0.225
11.265
108
0.260
24.575
130
0.345
5.120
50
0.205
4.660
44
0.195
282
X. Jin et al.
In this research, the hypothesis test is adopted to verify the new method. It is
assumed that the calculated values of area, length and width are independent with each
other and following Gaussian distributions with the Mathematical expectation of
measure values. s, LI, LII and w denote the ratios of area(s), length (LI, LII)
and maximum width wmax to their corresponding measured values. So s LI,
LII and w will follow to the normal distribution of N(1,2) in which the variance 2
are unknown. And the t test method could be employed to perform the hypotheses test
[10]
. The number of samples for each group of sLILII and w is 6. The level
of significance is assumed as 0.05. Ssam means sample variance. t represents test
statistics. H0 is the null hypothesis. H1 is the alternative hypothesis. Region of rejection
is T. The results of calculation are listed in Table 3.
Table 3. Normal population parameter test of
Inspection
ssam
H0
H1
0.0295
s=1
-0.8762
LI
0.0236
LI=1
LI1
1.8054
LII
0.0495
LII=1
LII
0.0101
0.0362
w=1
1
w1
1.9837
s, L , L
I
II and
[2.567,)
(-,-2.567][2.567,)
(-,-2.567][2.567,)
(-,-2.567][2.567,)
(-,-2.567]
Confidence
level
0.95
0.95
0.95
0.95
As demonstrated in Table 3, when the level of significance is 0.05, the hypothesis test
statistics t of s, LI, LII and w are not in its region of rejection. So s, LI,
LII and the hypothesis H0 of w can all be accepted. In comparison with the measured
value, the confidence level of the area of cracks s, the length of crack (LI LII) and the
maximum width of cracks wmax is 0.95 which means the proposed method is effective and
reliable in identifying the characteristic parameters of concrete cracks.
5 Conclusion
By comparing the characteristic parameters of concrete cracks identified with image
processing method to the measured results, it can be concluded that:
(1) The digital image extraction method proposed in this research will effectively
avoid the determination of threshold value. Then this new method shows a higher
accuracy and efficiency by eliminating the misjudgment caused by threshold value
and simplifying the denoising process, especially in case of a low digital graphic
step difference between the background and concrete cracks.
(2) The early age cracks of concrete state very clear fractal properties. It is a good way
to represent the irregularity of concrete cracks with fractal principle. And the
fractal dimensions can be taken as a characteristic parameter to describe the
tortuosity of cracks.
283
(3) In comparison with the traditional measuring method, this research put forward a
new algorithm in calculating the characteristic parameters of concrete cracks, such
as area, length and width, et al. The new method will significantly decrease the
working time and labor intensity at the measuring scene. At the same time, the new
method shows a higher accuracy in a broadened measuring range. The new method
will, undoubtedly, improve the quantitative identification of early age cracking of
concrete in accuracy, convenience, and efficiency.
Acknowledgements. The study of this paper is financially supported by National
Natural Science Foundation of China (NSFC 51008272 and 50838008) and Doctoral
Program of Higher Education (No. 20070335087).
References
1. Chen, Z.Y.: Guide to Durability and Construction of Reinforced Structures. China
Architecture & Building Press, Beijing (2005)
2. Yang, Y.J., Zhao, R.C., Wang, W.B.: The Detection of Shadow Region in Aerial Image.
Signal Processing 18(3), 228232 (2002)
3. Du, Q., Xing, J.Y., Yuan, S.C.: Study on Thresholding Segmentation for Infrared Image
Based on Edge Detection. Infrared and Laser Engineering 33(3), 7073 (2004)
4. Chen, Y., Chen, R.J., Guo, Y.H.: Graphics Programming and Image Processing with
MATLAB 6.X. XiDian University press, Xian (2002)
5. Sun, Z.L.: Image Processing with Matlab 6.x. Tsinghua University Press, Beijing (2002)
6. Ringot, E., Bascoul, A.: About the Analysis of Microcracking in Concrete. Cement and
Concrete Composites 23(2-3), 261266 (2001)
7. Ammouche, A., Breysse, D., Homain, H.: A New Image Analysis Technique for the
Quantitative Assessment of Microcracks in Cement-based Materials. Cement and Concrete
Research 30(1), 2535 (2000)
8. Xie, H.P.: Fractal Dimension-Introduction of Rock Mechanics. Science Press, Beijing
(1996)
9. Ding, B.H., Li, W.C., Wang, F.M.: Analysis of Fractal Image and Design of Fractal
Dimension Calculation Program. Journal of University of Science and Technology
Beijing 21(3), 304307 (1999)
10. Fan, D.Y., Chen, Y.H.: Probability and Mathematical Statistics. Zhejiang University Press,
Hangzhou (2003)