Anda di halaman 1dari 9

A vision-based technique for damage assessment of reinforced concrete

structures
Alireza Farhidzadeh, Arvin Ebrahimkhanlou, Salvatore Salamone
*

Smart Structures Research Laboratory (SSRL), Department of Civil, Structural, and Environmental
Engineering, University at Buffalo, 212 Ketter Hall, Buffalo, NY, USA 14260



ABSTRACT

The most common damage assessment technique for concrete structures is visual inspection (VI). Condition assessed by
VI is subjective in nature, meaning it depends on the experience, knowledge, expertise, measurement accuracy, mental
attention, and judgment of the inspector carrying out the assessment. In many post-event assessments, cracks data
including width and pattern provide the most indicative information about the health or damage state of the structure.
Residual cracks are sometimes the only available data for VI. However, due to adjacent elastic members, earthquake
displacement spectrum, or re-centering systems, these measurements may lead to erroneous decisions. To overcome this
problem, this paper proposes a novel damage index based upon Fractal Dimension (FD) analysis of residual cracks as a
complementary method for VI. FD can quantify crack patterns and enhance the routine inspection procedure by
establishing a crack pattern recognition system. This algorithm was validated through an experimental study on a large
scale reinforced concrete shear wall (RCSW). The results demonstrate the novel technique as a quite accurate estimator
for damage grades and stiffness loss of the wall.


Keyword: Reinforced concrete, automated visual inspection, image processing, crack pattern quantification, fractal
analysis


1. INTRODUCTION

Civil structures such as buildings, bridges, nuclear power plants or dams include reinforced concrete (RC) structures
whose performance and function are vital to the well-being of our society. Unfortunately, many of these structures are
facing an increasing number of challenges that can jeopardize their safety and serviceability. Some of these challenges,
such as aging and maintenance, are life-cycle related and expected. Other challenges such as man-made hazards or
natural events (i.e., earthquakes, hurricanes and tsunami) are not expected and can weaken and destabilize a structure.

Therefore, an important task is to reliably assess the condition of a given structure and to ensure that safety standards are
met throughout its operational life. It is also beneficial for the owners and facility operators of a structure to know the
extent of its deterioration and to estimate its remaining life. Authors have done an extensive research on application of
state-of-the-art SHM systems for concrete structures
1-2
, however, currently, assessing the performance and safety of
these structures relies almost entirely on the routine-based visual inspection (VI). Recent collapses
3-4
due to erroneous
VI-based decisions demonstrate great need of research on this aspect of inspection. Unfortunately, condition assessed by
means of VI is subjective in nature and depends on the experience, knowledge, expertise, and judgment of the inspector
carrying out the assessment
5
. As a result there is a potential that indications of structural degradation could be missed
6
.
For instance, the complexity of crack patterns can cause inconsistency between the mapped drawings and the actual
pattern. Moreover, in case of closed cracks that can occur due to pull-back force of adjacent elastic members, re-
centering systems, and earthquake displacement spectrum, width of cracks may result in erroneous decisions. The
pattern of cracks could provide additional information about the structures health
7
.

This article presents a novel approach to enable quantitative visual inspection using automated image processing and
fractal analysis of crack patterns. The proposed algorithm will be validated through experimental studies on a full scale

*
ssalamon@buffalo.edu; phone: +1(716) 645-1523, fax: +1(716) 645-3667
Health Monitoring of Structural and Biological Systems 2014, edited by Tribikram Kundu, Proc. of SPIE Vol. 9064, 90642H
2014 SPIE CCC code: 0277-786X/14/$18 doi: 10.1117/12.2044875
Proc. of SPIE Vol. 9064 90642H-1
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms


Reinforced Concrete Shear Wall (RCSW). A preliminary study for automated crack identification from a 2D image is
investigated and then a new fractal-based damage index (DI) is introduced to provide an automatic diagnose of the
structure.


2. AUTOMATED CRACK IDENTIFICATION

Image-based algorithms for automated crack detection, such as spline fitting
8
, threshold based
9
, and edge detection
based methods
10
have received significant interest over the past few years. Despite, these novel algorithms have seen
significant developments, their reliability has been questioned. Illumination, overexposure, and existence of shadow,
blubs and dark spots on concrete surface, may hamper their reliability in terms of automatic crack detection. To
overcome current limitations pre-processing algorithms such as optimal thresholding
11-13
, and pattern recognition
methods
14-19
, have been developed. In this paper an approach based on a combination of optimal thresholding and edge
detection by Harr wavelet transform is proposed. The proposed approach consists of three steps: 1) preprocessing; 2)
edge detection, and 3) boundary tracking. Figure 1 shows a flowchart of the entire process.

2.1. Preprocessing

Preprocessing is generally the first step in all image processing algorithms. In this paper, preprocessing is done to resize
the image, remove gridlines, restore geometric distortions, and convert RGB to gray scale.

2.2. Edge detection

Edge detection is a technique in image processing that aims at identifying the pixels at which the brightness changes
20
.
As surface concrete cracks cause abrupt brightness change, edge detection algorithms is employed to detect a cracks as
an edge. In this study, Discrete Wavelet Transform with Harr mother filter
20
is used for the purpose of edge detection.
Since the majority of cracks in the tested shear wall were diagonal, the diagonal channel of discrete wavelet transform
is used. Not all edges being crack, the next two following algorithms are employed to prune the outcome of DWT. First,
an algorithm named Hystersis, divides results into three groups of strong edge, weak edge, and non-edge. Then, the
second algorithm, called non-maximal suppression, regroups weak edges into either strong edge or non-edge. Two
criteria of being in diagonal direction and at vicinity of strong edges control this regrouping decision.

2.3. Boundary tracking

Boundary tracking is the act of grouping strong edge pixels together and treating each group as a separate list of pixels
20
. Each crack is confined by two lines of boundaries. If one reduces the resolution, these two boundary lines will
converge to one line that could be either a crack, part of a crack, or noise. To classify them as crack or noise, length and
shape of these boundaries are used. For example, noise due to illumination can result in short line-like boundaries, and
noise due to concrete texture can produce curl-like boundaries. Although pattern recognition algorithms could be used to
distinguish cracks from noise, in this paper classification has been done only based on length, not shape of boundaries.

3. FRACTAL ANALYSIS

3.1. Fractal Dimension

Fractal dimension (FD) indicates the degree of complexity of a dataset, pattern, or signal extracted from phenomena such
as natural geometric objects, sound, and digital images
21
. FD can be a non-integer number varying depending on the
complexity of an object. For example, the fractal dimension for a curve will lie between 1 and 2, depending on how
much area it fills
22
. Similarly, the FD of surfaces lies between 2 and 3 depending on the roughness of surface
23
. The
complexity of two curves or two surfaces can then be easily compared, as the values of FD are not anymore restricted to
the topological dimensions of 1 and 2.

Among the several algorithms for calculating FD, the box counting is one of the most frequently used methods. In the
box counting method, the space filling properties of the curve is used as an indication of its complexity. In this
Proc. of SPIE Vol. 9064 90642H-2
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms
I n p u t I m a g e
S t e p 1 : D o w n s a m p l i n g
{
S t e p 2 : S h a r p e n i n g o f d i f f e r e n c e
o f i n r e d c h a n n e l w i t h o t h e r t w o ,
m u l t i l e v e l t h r e s h o l d i n g t o b o t h
d e t e c t a n d s u p p r e s s g r i d l i n e s
S t e p 3 : G r a y s c a l e t r a n s f o r m
S t e p 4 : H o u g h t r a n s f o r m ,
t h r e s h o l d i n g t o f i n d H o u g h
p i c k s , a n d p a r a m e t e r i z i n g g r i d
l i n e s
S t e p 5 : G e o m e t r i c t r a n s f o r m
b a s e d o n d e t e c t e d g r i d l i n e s
1
S t e p 6 : W a v e l e t t r a n s f o r m
( A p p r o x i m a t e & D i a g o n a l
c h a n n e l )
1
S t e p 7 : E d g e i m a g e
T h r e s h o l d I n g
S t e p 8 : N o n - m a x i m a l
s u p p r e s s i o n
1
S t e p 9 : H y s t e r e s i s t o f i l t e r
o u t p u t o f t h e e d g e d e t e c t o r
1
S t e p 1 0 : E d g e L i n k i n g b y
b o u n d a r y t r a c k i n g
O u t p u t : d e t e c t e d p i x e l s a s
S h e a r W a l l C r a c k '


algorithm, the area is covered with a collection of square boxes of size r, and the number of boxes which contain at least
one point of the curve is counted as N(r). Ideally, the mathematical box-counting dimension can be determined as the
ratio of log N(r) over log (1/r) when r approaches to zero as:
0
log ( )
lim
log(1/ )
r
N r
D
r


=



(1)
However, in practice, D is estimated by fitting a straight line to the log-log plot of N(r) versus 1/r over a range of box
sizes. This can be expressed as:

( ) log( ( )) log 1 N r D r C = +
(2)

where C is a constant. The slope of the least square fit line is taken as an estimator of the fractal dimension (D) of the
curve.

In this equation, range of variation of box sizes, r, has to be selected properly. First, the box size cannot be smaller than
smallest segment of cracks (i.e., in the case of photography is the size of a pixel). Otherwise, the number of counted
boxes (N) does not change with box sizes smaller than pixel size and D in equation (2) would be zero. Second, the fractal
nature of a shape in a small scale photo transits to an Euclidean (homogeneous) regime in a large scale photo. To
elaborate this matter, Figure 2 illustrates the chaotic surface cracks distribution on a concrete wall. Cracks may be
considered as a fractal object in a small scale photo (Figure 2a), or homogenous in a large scale photo (Figure 2b)
because it only focuses on one crack which resembles a line (D=1). Therefore, there should be a limit for decreasing the
size of boxes. This transition box size is called R
T
in this paper.


Figure 1. Flowchart of proposed crack detection method
Proc. of SPIE Vol. 9064 90642H-3
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms
U N
i




Figure 2. (a) Fractal object in a small scale photo (b) Euclidean object (a line) in large scale photo.


3.2. FD-based Damage Index

In order to include the proposed approach in a Structural Health Monitoring (SHM) strategy for estimating the level of
damage in RC structural elements, a new damage index (DI) is defined as follows
5
:
1
, 0 1
2
1
D D
i
DI DI
D

= <


(3)
where D
i
is the fractal dimension of the current status of visible cracks, (e.g., in the i
th
inspection); D
1
is the fractal
dimension computed during the first inspection (e.g., once the cracks become visible for the first time). DI varies
between 0 and 1 and describes the difference between the current status of crack patterns and the baseline D
1
. The
constant 2 in the denominator is the maximum value of fractal dimension for surface cracks (i.e., when cracks cover
the whole area of the concrete D tends to 2).


4. EXPERIMENTAL SETUP

The specimen was a large scale RC shear wall with a height to width ratio of 0.94, designed based on ACI 318-08.
Details about the specimen are given in Table I and the experimental setup is illustrated in Figure 3. The specimen was
subjected to a displacement controlled quasi-static reversed cyclic loading. The force-displacement hysteresis loops and
the corresponding backbone curves are illustrated in Figure 4. The specimen had a nonlinear response in LS7 and
reached its ultimate strength in LS9.

TABLE I. EXPERIMENTAL DETAILS
Height [mm] 3300
Width [mm] 3050
Thickness [mm] 200
Reinforcement Ratio [%] 0.67
Concrete compressive strength f'
c
[MPa] 24.8
Reinforcing bars yield strength [MPa] 464
Reinforcing bars ultimate strength [MPa] 708


(a)
(b)
1:267 mm
1:33 mm
Proc. of SPIE Vol. 9064 90642H-4
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms


F



5.1. Visual I

This section
grades (DGs)
the IAEA gu
maximum wi
above 1 mm
cracks) are co
development
crack widths

Figure 6 show
deformation
indicated by

Figure 7 illus
be seen that,
assessment b
cracks width
after an earth

5.2. Automa

Figure 8 dem
shadow, and
intended prop

5.3. Fractal

The box-coun
superimposed
counted. The
spacing, r. Th
versus 1/r. Th

Figure 3. Experim
Inspection / Cr
presents the re
). Considering
uidebook on no
idth is less than
as grade III. In
onsidered as m
ts in each load
are indicated i
ws the maximu
of each load st
horizontal and
strates cracks w
based on resid
based on residu
are used curre
hquake).
atic crack dete
monstrates resu
rough concret
posed, which i
Dimension Re
nting algorithm
d over each sch
e method was r
herefore, the e
his procedure i
mental setup
1

5
rack Width M
esults of visual
that specimen
n-destructive t
n 0.2 mm, the d
n general, non-
moderate damag
step are illustr
in each picture
um and modal
tep. Crack wid
d vertical dashe
width measured
dual cracks wid
ual cracks width
ently by engine
ection
lt of applicatio
e texture. Alth
s calculation o
esults
m was used to c
hematic picture
repeated with a
stimated D wa
is illustrated in

5. EXPERIM
Measurement
observations a
s used in this s
testing of concr
damage is clas
-structural or ha
ge or grade II,
rated in Figure
e with a square
crack width m
th thresholds b
ed lines, respec
d once loading
dth DG II initia
h would provid
eers to determin
on of the propo
ough there are
of a rough estim
compute a frac
e (Figure 9a). N
a denser and de
as the slope of l
n Figure 9 at LS
F
o
r
c
e
[
k
N
]
Figure 4. F
MENTAL R
and crack width
study are typica
rete structures
ssified as grade
airline cracks a
and critical da
5. The maximu
dot () and wi
measured on the
based on the IA
ctively.
g was ceased an
ates in LS7 wh
de very inaccur
ne the type of r
sed method on
still some und
mate of general
ctal dimension
Next the numb
enser grid to de
least square fit
S5.
-60 -4
-1500
-1000
-500
0
500
1000
1500
F
o
r
c
e

[
k
N
]

1st
2n
Force-displacem
RESULTS
h measuremen
ally being used
10
was used to
e I, between 0.2
are treated as g
amage or wide
um and modal
ith a circle (),
e surface of the
AEA guidelines
nd cracks were
ereas DG III in
rate results. It s
repair required
n a shear wall im
detected cracks
l location of cr
(D) for each im
ber of boxes tha
efine the numb
line from R
S
to
40 -20 0
Displacemen
t cycles
nd cycles
ent response
1

ts along with i
d for nuclear sa
define the DG
2 mm and 1 mm
grade I, structur
cracks as grade
(i.e., appearin
, respectively.
e wall when it w
s and DG boun
closed (i.e., re
nitiates in LS8.
should be men
d for a damaged
mage containin
s, the approach
acks.
mage. A grid o
at intersect cra
er of boxes as
o R
T
in the log
20 40 60
nt [mm]

dentified dama
afety related str
Gs. In particular
m as grade II, a
ral cracks (or m
e III. Residual
g most frequen
was at the peak
ndaries are also
esidual cracks)
. Therefore, co
ntioned that res
d component (e
ng lots of blob
is doing well i
of square boxes
cks, N(r), was
a function of th
-log plot of N(
80

age
ructures,
r, if the
and
medium
cracks
ntly)
k
o
. It can
ondition
idual
e.g.,
s,
in its
s was
he grid
(r)
Proc. of SPIE Vol. 9064 90642H-5
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms
I ,




Figure 5. Cracks pattern
5




Figure 6. Width of cracks at peak displacement vs. load
steps and damage grades
5



Figure 7. Width of residual cracks vs. load steps and
damage grades
5



(a)

(b)

(c)
Figure 8. Final result of preliminary crack detection (a) initial image, (b) edge detection, (c) boundary tracking

The fractal dimension was calculated for each load step. The FD variation trend versus load steps and its similarity with
the outcome of another experiment on a shorter shear wall
5
, led to introducing a phenomenological behavior of fractal
dimension versus load, illustrated in Figure 10. It exhibits an ascending trend when small lateral deformations are


0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
0 1 2 3 4 5 6 7 8 9 10
C
r
a
c
k

w
i
d
t
h

a
t

p
e
a
k

d
e
f
o
r
m
a
t
i
o
n

[
m
m
]
Load step
Maximum width
Modal width
I II III
SW1
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1 2 3 4 5 6 7 8 9 10
R
e
s
i
d
u
a
l

c
r
a
c
k

w
i
d
t
h

[
m
m
]
Load step
Maximum width
Modal width
SW1
I II III
SW1
Proc. of SPIE Vol. 9064 90642H-6
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms


imposed. The walls are purely elastic in this range. Then, this trend follows a plateau path where cracks are starting to
open until the onset of plasticity. Thereafter, the trend reports a sudden jump as indicator of significant damage.


(a)

(b)

(c)

Figure 9. Steps of FD calculation on residual cracks for LS5 as an example; (a) coarse mesh, (b) fine mesh, (c) estimation of FD
5



Figure 10. Variation of fractal dimension versus increasing load steps


5.4. Damage Index Results

In this section the results of the proposed damage index (DI) are presented and compared with the relative stiffness
loss (RSL) defined as:
1
1
K
i
RSL
K
=

(4)
where K
i
is the lateral secant stiffness of the wall at the i
th
load step (i.e., i=1,2,10) and K
1
is the initial stiffness of the
wall. Figure 11 shows the results of RSL and DI for both walls. Interestingly, DI can generally track the relative stiffness
loss. Therefore, DI can be exploited to infer the remnant lateral stiffness of the wall. In this work two threshold levels,
j
,
are proposed in order to discriminate damage grades (i.e., j=1 for DG II and j=2 for DG III). In particular, a
1
=0.4 is
defined for DG II and
2
=0.7 for the onset of DGIII
5
. Using these thresholds for the FD, the damage classification in
comparison to the measurement of residual cracks and cracks at load peak is depicted in Figure 12. DI could
significantly remedy the erroneous DGII at LS6 relying on width of residual cracks although it could not improve the
misclassification of LS7 as DGII. Another study by authors on a shorter RCSW showed that even DGIII could be
modified using this novel DI
5
.

R
S
R
T
-2 -1.5 -1
1
1.5
2
2.5
log(1/r)
l
o
g
(
N
(
r
)
)
FD=1.63
F
r
a
c
t
a
l

d
i
m
e
n
s
i
o
n
Load
I II III
Significant
macro-crack
formations
Elastic
response
Yielding
Ultimate
strength
Proc. of SPIE Vol. 9064 90642H-7
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms



Figure 11. Comparison of DI and relative stiffness loss
5



Figure 12. Damage grade classification by different
methods.

6. CONCLUSIONS

This article presented a procedure to automate visual inspection of reinforce concrete structures. The proposed algorithm
was verified through experimental studies on a large RC shear wall. The proposed damage index based on fractal
dimension of visible residual cracks could acceptably estimate damage grades better than crack width measurement.
Furthermore, it could accurately estimate the stiffness loss of the specimen by relying only on the residual cracks pattern.
This methodology is of utmost importance since instead of relying on a subjective decision about the safety operation of
the structure, it could develop an advance and automatic visual inspection by improving damage grades estimation and
stiffness loss evaluation. Further studies can incorporate this methodology with automated crack detection in concrete
structures and provide a complete systematic visual SHM system.

ACKNOWLEDGMENTS

The authors acknowledge National Science Foundation (NSF) for providing the financial support under Grant No.
CMMI-1333506. The experiments presented herein could not have been completed without contributions from the staff
of the Structural Engineering and Earthquake Simulation Laboratory (SEESL) of the State University of New York at
Buffalo. The authors acknowledge the advice and help provided by Prof. Andrew Whittaker as well as the technical staff
at the NEES Equipment Site at the University at Buffalo.

REFERENCES

[1] Farhidzadeh, A., Salamone, S., and Singla, P., A Probabilistic Approach for Damage Identification and Crack
Classification in Reinforced Concrete Structures, J. Intell. Matls. Sys. & Struct., 24 (14), 1722-1735 (2013).
[2] Farhidzadeh, A., Salamone, S., Luna, B., and Whittaker, A., Acoustic Emission Monitoring of a Reinforced
Concrete Shear Wall by b-value based Outlier Analysis, J. Struct Health Monit: an Int J. 12(1):3-13 (2013).
[3] DBH, Christchurch CBD Buildings 22 February 2011 Aftershock Stage 1 Expert Panel Report, New Zealand
Department of Building and Housing, New Zealand (2011).
[4] RAIU, Malahide Viaduct Collapse on the Dublin to Belfast Line, on the 21st August 2009, Dublin, Ireland, (2010).
[5] Farhidzadeh, A., Dehghan-Niri, E., Mustafa, A., Salamone, S., and Whittaker, A., Damage Assessment of
Reinforced Concrete Structures Using Fractal Analysis of Residual Crack Patterns, Exp Mech, Volume 53 (9),
1607-1619 (2013).
[6] IAEA, Guidebook on non-destructive testing of concrete structures, Vienna, Austria (2002).
[7] Vanniamparambil, P. A., Bolhassani, M., Carmi, R., Khan, F., Bartoli, I., Moon, F. L., Hamid, A., and Kontsos, A.,
A data fusion approach for progressive damage quantification in reinforced concrete masonry walls, Smart Mater
Struct, 23(1) 1-11 (2014).
[8] Dare, P., Hanley, H., Fraser, C., An operational application of automatic feature extraction: the measurement of
cracks in concrete structures. Photogramm Rec, 17(99): 453464 (2002).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
R
e
l
a
t
i
v
e

s
t
i
f
f
n
e
s
s

l
o
s
s

(
R
S
L
)
D
a
m
a
g
e

i
n
d
e
x

(
D
I
)
Drift ratio [%]
DI
RSL
SW1
0
2
4
6
8
10
12
0 1 2 3 4 5 6 7 8 9 10
DI
Residual crack width
Grade I
Grade II
Grade III
M
e
t
h
o
d
1 2 3 4 5 6 7 8 9 10
Load step
SW1
Crack width at peak deformation
Proc. of SPIE Vol. 9064 90642H-8
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms


[9] Ito, A., Aoki, Y., Hashimoto, S., Accurate extraction and measurement of fine cracks from concrete block surface
image, IECON 02, 3, 2202,2207, (2002).
[10] Abdel-Qader I, Analysis of edge-detection techniques for crack identification in bridges, J. Comput. Civ. Eng.,
17(4), 255263 (2003).
[11] Fujita, Y., Mitani, Y., Hamamoto, Y., A Method for Crack Detection on a Concrete Structure, 18th Int. Conf.
Pattern Recognit. IEEE, 1821 (2006).
[12] Fujita Y, Hamamoto, Y., A robust automatic crack detection method from noisy concrete surfaces, Mach. Vis.
Appl., 22 (2), 245-254 (2011);
[13] Hutchinson, T., Chen, Z., Improved image analysis for evaluating concrete damage, J. Comput. Civ. Eng., 20 (3),
210-216, (2006).
[14] Kim, K., Sim, K., Ahn, S., Recognition of concrete surface cracks using the ART1-based RBF network, Adv.
Neural Networks, 3972, 669-675 (2006).
[15] Yamaguchi, T., Nakamura, S., Image-Based Crack Detection for Real Concrete Surfaces, IEEJ T Electr Electr,
3(1), 128135 (2008).
[16] Yamaguchi, T., Hashimoto, S., Fast crack detection method for large-size concrete surface images using
percolation-based image processing, Mach. Vis. Appl. 21(5), 797-809 (2010).
[17] Chen, Z., Hutchinson, T.C., Image-Based Framework for Concrete Surface Crack Monitoring and Quantification,
Adv. Civ. Eng. 1-18 (2010).
[18] Jahanshahi, M.R., Masri, S., Padgett, C., Sukhatme, G.S., An innovative methodology for detection and
quantification of cracks through incorporation of depth perception, Mach. Vis. Appl. 24, 227-41 (2013).
[19] Jahanshahi, M.R., Masri, S.F. Adaptive vision-based crack detection using 3D scene reconstruction for condition
assessment of structures, Autom. Constr. 22, 56776 (2012).
[20] Sonka, M., Hlavak, V., and Boyle, R., Image Processing, Analysis, and Machine , 3rd ed., Thomson Engineering
(2007)
[21] Mandelbrot, B.B., Fractals and Chaos: The Mandelbrot Set and Beyond, New York, USA: Springer-Verlag New
York Inc. (2004).
[22] Moustafa, A., Dehghan Niri, E., Farhidzadeh, A., and Salamone. S. Corrosion monitoring of post-tensioned
concrete structures using fractal analysis of guided ultrasonic waves, Struct Control Health Monit, DOI:
10.1002/stc.1586 (2013).
[23] Dubuc, B., Quiniou, J.F., Roques-Carmes, C., Tricot, C. and Zucker, S.W. Evaluating the fractal dimension of
profiles, Phys Rev A, 39(3),1500-1512 (1989).
Proc. of SPIE Vol. 9064 90642H-9
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/12/2014 Terms of Use: http://spiedl.org/terms

Anda mungkin juga menyukai