Automation in Construction 86 (2018) 240–255

Contents lists available at ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Evaluation of pavement surface drainage using an automated image T

acquisition and processing system
⁎
Behrouz Mataeia, Fereidoon Moghadas Nejada, , Mohsen Zahedib, Hamzeh Zakeria
a
Department of Civil and Environment Engineering, Amirkabir University of Technology, No. 424, Hafez Ave., Tehran, Iran
b
Department of Civil Engineering, Razi University, Kermanshah, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: Network level evaluation of pavement surface drainage plays a crucial role in the improvement of pavement
Pavement surface drainage safety and reducing accident rates. Hydroplaning, as the main considered cause of accidents in wet weather
Automated system conditions, is a consequence of the low quality draining in the pavement surface. Since no automated system
Image processing currently exists for the pavement drainage evaluation, this work was conducted to present a new system to assess
the quality of the surface drainage process. To this end, an innovative device is presented to simulate the sa-
turation condition of the pavement surface and acquire photos from the drainage process of the pavement
surface after saturation. Next, an image processing method was applied to produce proper indices for drainage
quality assessment. The preprocessing and enhancement of images was performed using shearlet transform. The
rate of surface drainage progress was evaluated by three indices extracted from the images. Finally, pavements
were classified into three categories according to the indices extracted for their surface drainage. The validation
of the proposed method by the confusion matrix shows the high performance of the system in simulation and
assessment of surface drainage of the road pavements.

1. Introduction Many studies demonstrated the relationship between wet pavement

1.1. Background
One of the most important parameters influencing road safety is the
climatic (wet or dry) conditions of the pavement surface [1–8]. Many
researchers proposed a relationship between accidents and weather
conditions [9–11]. In areas with long intervals between precipitations,
after a dry period, the number of accidents increases during the first
precipitation [12]. In wet conditions, the layer of water covering the
pavement acts as a lubricant and reduces the contact between the tires
and the pavement surface [13]. Therefore, the friction decreases and
the pavement surface exhibits a lower friction than the dry-pavement
surface. In addition to this lubricating effect of water at high speeds,
lack of drainage facility in the presence of certain depths of water film
may result in hydroplaning. Hydroplaning is a phenomenon, which
occurs when water film is developed between the tires of the vehicle
and the pavement surface. This phenomenon results in the reduced
traction and disables the vehicle from responding to actions such as
steering, braking, or accelerating [14]. Accordingly, hydroplaning is
considered as the main cause of accidents in wet weather conditions
[13].
crashes and characteristics of pavement surface texture [15–18]. Due to
the importance of pavement texture characteristics in the safety of
roads, the pavement surface is required to be monitored continuously.
Two important surface characteristics are microtexture and macro-
texture. Macrotexture refers to the coarse-scale texture irregularities of
the pavement surface that affects the hysteresis component of the
friction. These irregularities are associated with the void area between
aggregate particles. The magnitude of this component will depend on
the size, shape, and distribution of coarse aggregates used in pavement
construction, the nominal maximum size of aggregates and the parti-
cular construction methods used in the implementation of the pavement
surface layer [4,19].
Microtexture refers to fine-scale texture irregularities in the surface
of the aggregate particles that are measured at the micron scale of
harshness and are known to be mainly a function of aggregate particle
mineralogy. Stone particle smoothness or harshness depends on these
irregularities. The magnitude of microtexture depends on the initial
roughness of the aggregates surface and the resistance of the aggregates
against the polishing action of traffic and environmental factors. The
adhesion component of the friction is influenced by microtexture [19].
Wet pavement friction is influenced by both macrotexture and

⁎
Corresponding author.
E-mail addresses: behrouz.mataei@aut.ac.ir (B. Mataei), moghadas@aut.ac.ir (F. Moghadas Nejad), zahedi@razi.ac.ir (M. Zahedi), h-zakeri@aut.ac.ir (H. Zakeri).

https://doi.org/10.1016/j.autcon.2017.11.010
Received 21 November 2016; Received in revised form 3 July 2017; Accepted 11 November 2017
0926-5805/ © 2017 Elsevier B.V. All rights reserved.
B. Mataei et al.
microtexture of the pavement surface [20]. Macrotexture that is an
overall asphalt mixture characteristic reduces dynamic hydroplaning
development by providing channels for removal of water [21]. Micro-
texture that is primarily an aggregate surface characteristic reduces the
viscous hydroplaning development by draining the viscous water film
between tire and pavement [22].
The drainage process is controlled by the shape of microasperities
[23]. Savkoor showed that amplitude and number of microasperities on
pavement surface controls the drainage of the water film between tire
and pavement [24].
Ong et al. proved that using materials with better microtexture re-
duces the chance of hydroplaning [22]. Their study showed that in the
pavements comprised of coarse aggregates with high microtexture in
the range of 0.2 mm to 0.5 mm, hydroplaning occurs at a 20% higher
speed. Horne also reported that hydroplaning could be delayed in pa-
vements with a good microtexture [25]. Pelloli found that the amount
of microtexture is determinant on the relationship between friction
coefficient and the water depth accumulated on the surface [26].
Implementing corrective actions in hazardous areas can reduce the
rate of these accidents. Mataei et al. [27] reviewed various methods
applied for measuring pavement surface texture. According to their
work, there are a number of tests, procedures, and devices available for
evaluation of the pavement surface texture. These methods mainly
consist of field tests, high-speed and laboratory tests, and low-speed
approaches. The locked wheel test is a method for measuring friction
force at 100% slip situation. In addition, the sideway force test is an-
other technique for friction force measurement through a wheel ro-
tating with a 20° yaw angle along the motion direction [4,5,15]. The
fixed slip test is conducted to measure the friction force of the con-
stantly slipping wheel, whereas the variable slip test measures the
friction force at any desired slip [4,15,28]. Methods applied for mea-
suring pavement surface texture are divided into two groups of contact
and non-contact methods. In the contact methods domain, two la-
boratory testers including British Pendulum Tester (BPT) and Dynamic
Friction Tester (DFT) are typically applied. Through these tests, the
pavement surface friction is measured by determining the loss in kinetic
energy of a sliding pendulum or rotating disc when in touch with the
roadway surface and converted to a frictional force. Measuring friction
at various speeds gives the DFT device the ability to measure the speed
dependency of the pavement friction [1,4,15].
The OutFlow Meter Test (OFT) and Sand Patch Test (SPT) are the
most popular methods that measure the macrotexture of the pavement
surface with contact. In OFT, a transparent vertical cylinder with a
rubber ring under it, is placed on the pavement surface. Then, water is
allowed to flow into the pavement, and the required time for passing a
determined volume of water in the transparent vertical cylinder is re-
corded. Also referred to as outflow time, this parameter, which is at-
tributed to pavement macrotexture, indicates the ability of the pave-
ment surface to drain water and shows how fast water depletes from the
surface [2–4,28–31].
In addition, the sand patch test is a volumetric method for macro-
texture measurement, through which a known volume of a homo-
geneous material (sand, glass beads, or grease) is spread on the pave-
ment surface and the resulting area is measured. Eventually, by
dividing the initial volume by the area, Mean Texture Depth (MTD) is
calculated via this test [2,4,15,28]. In the non-contact methods, the
circular track meter (CT Meter) and vehicle-mounted laser devices are
the most common [2,28,29].
The CT Meter is a laser-based device for measuring the mean profile
depth (MPD) of a pavement at a static location. This device can be used
both in the laboratory and field. However, the CT Meter cannot mea-
sure the MPD of a pavement at highway speeds [3,15,29,31]. Therefore,
vehicle-mounted laser devices were developed to measure the macro-
texture by vehicle movement at the wheel pass line and without dis-
rupting traffic flow [2,4,29]. Recently, the image-processing techniques
are increasingly applied to develop a non-contact method for pavement
241
Automation in Construction 86 (2018) 240–255
Table 1
Methods for measuring pavement surface characteristics [27].
General classification Methods
Field measurement locked wheel test
sideway force test
Fixed slip test
Variable slip test
Portable and laboratory testers Contact British Pendulum Tester (BPT)
Dynamic Friction Tester (DFT)
OutFlow Meter Test (OFT)
Sand Patch Test (SPT)
Non-contact Laser-based
Image based
texture assessment [32,33].
Table 1 summarizes the classification of the mentioned methods
applied for measuring pavement surface characteristics. It is note-
worthy that none of these methods are capable of evaluating pavement
surface drainage.
The methods summarized in Table 1 have many advantages which
present them as traditional texture measurement methods. However,
deficiencies in some of these methods like BPT, SPT, DFT, and OFT
make them unreliable. Being time-consuming, requiring traffic control
and expert operators, unrepeatable results, requiring high memory for
data storage and expensive equipment are among these drawbacks.
Therefore, researchers, trying to cover these shortcomings, have re-
cently presented new methods for pavement texture measurement.
De Leon Izeppi et al. presented a new low-cost system for pavement
surface texture measurement using a stereo vision camera [34]. Also,
Elunai et al. proposed a method based on image processing to estimate
the road surface texture coarseness distribution from its edge profiles
[35]. Wang et al. introduced a low-cost portable laser device that cal-
culates the profile of the pavement surface texture and records the
mean profile depth (MPD) by scanning the road surface [29]. Else-
where, Elunai et al. suggested using autocorrelation and Wavelet
Transform (WT) as two new image processing methods for measuring
texture depth [31]. Zelelew et al. recommended a WT approach for
analysis and interpretation of the macrotexture data obtained using a
circular track meter (CT Meter) device [36]. Pidwerbesky et al. sug-
gested replacing the sand circle test with a safe and reliable approach
by presenting an image processing based method using Fast Fourier
transform (FFT) [37]. Finally, Cigada et al. proposed a laser-based
method using two identical industrial laser triangulation displacement
transducers [38]. These cases are examples of recent achievements of
researchers in presenting new methods and procedures for the evalua-
tion of pavement surface texture.
The aggregate imaging system (AIMS) introduced by Masad et al. is
one of the recent methods for measuring the aggregate texture directly
using a microscope and a digital image processing technique [39]. This
method is an important development in texture measurement methods
as it allows measuring physical characteristics of the aggregate. This
advanced technology consists of an automated video system that di-
rectly analyzes texture, angularity, and the shape of aggregates. Victor
et al. compared the results obtained by conventional tests and the ag-
gregate image measurement system (AIMS) for measuring texture
characteristics of aggregates and hot mix asphalt (HMA)s [40]. The
results showed that, in relation to the microtexture of field asphalt
samples, AIMS gives a good correlation with the results from the British
pendulum.
In addition to surface texture, other factors affect the surface drai-
nage such as rutting, pits, cross and longitudinal gradients, surface signs
and colors, type of materials and bitumen, etc. This study is focused on
surface texture.
B. Mataei et al.
Fig. 1. The structure of the proposed system.
1.2. Problem statement
Although many researchers tried to present methods and devices for
a better evaluation of pavement surface features, surface drainage as an
important subject in road safety has not been studied yet. The standard
procedure for assessment of pavement surface features is to perform it
on saturated pavements. Hydroplaning, which is considered the main
cause of accidents in wet weather conditions, leads to the loss of friction
and thus disables the vehicle from responding to actions such as
steering, braking, and accelerating. The occurrence of this phenomenon
has a great relationship with the quality of pavement surface drainage.
Although the assessment of the drainage of the pavement surface is of
highest significance, no automated system currently exists for the
drainage evaluation. This study presents an automated system based on
image processing that allows simulating surface drainage of the pave-
ments and presents indices for surface drainage evaluation. The pro-
posed system consists of an innovative device and a software sum-
marized in Fig. 1.
1.3. Objective of the research
The objective of this paper is to propose a device to prepare in-
formation from pavement drainage by saturating the surface and image
acquisition from the drainage process, and then use image processing
techniques to process and analyze the pavement images and present a
new method for evaluation of pavement surface drainage. Also the ef-
ficacy of texture characteristics in surface drainage is evaluated.
The rest of this paper is organized as follows: Section 2 describes
theoretical information. Section 3 presents the proposed method in
detail. Finally, Section 4 presents the discussion and conclusion.
2. Theoretical information
2.1. Shearlet transform
During the last decade, a new generation of multiscale systems has
emerged. This system combines the power of the classical multi-
resolution analysis with the ability to process directional information
with a high efficiency. Some of the most notable examples of such
systems are the curvelets [41], the contourlets [42], and the shearlets
[43]. Unlike classical wavelets, the elements of such systems form a
pyramid of well-localized waveforms ranging not only across various
242
Automation in Construction 86 (2018) 240–255
scales and locations but also across various orientations and highly
anisotropic shapes. Due to their richer structure, these more sophisti-
cated multiscale systems are able to overcome the poor directional
sensitivity of traditional multiscale systems. Thus, they have been used
to derive several state-of-the-art algorithms in image and signal pro-
cessing [44,45].
Shearlets, in particular, offer a unique combination of remarkable
features in that they have a simple and well-understood mathematical
structure derived from the theory of affine systems [43,46]; they pro-
vide optimally sparse representations, in a precise sense, for a large
class of images and other multidimensional data where wavelets are
suboptimal [47,48]; and the directionality in these systems is controlled
by shear matrices rather than rotations. This last property, in particular,
enables a unified framework for a continuous and discrete setting since
shear transformations preserve the rectangular lattice which is an ad-
vantage in deriving reliable digital implementations [49,50].
Shearlets are mathematically defined as follows:
For all functions fϵL2(R2), the shearlet transform can be written as:
SHψ f (a, s, t ) = < f , ψa, s, t > (1)
Where ψast is the affine function which is defined as follows:
−3
ψa, s, t = a 4 ψ (Ss−1Aa−1 (x − t ))
a 0 ⎞
where Aa = ⎛
⎝0 a ⎠
, Ss =
1 s
0 1 ( )
and ψϵL2(R2).a represent image scale
parameter, s represents shear direction, and t represents translation.
The function ψ should meet the following conditions:
 (ε ) = ψ  (ε1, ε2) = ψ (ε1) ψ
 ε2 ε
1- ψ 1 2 ( )
1

ψ = Fourier transform of the function ψ,
ε = (ε1 , ε2 )ϵ R2 , ε2 ≠ 0.
2- ψ1ϵL2(R) is a continuous wavelet, satisfying Calderon condition
∞
∫ |ψ (aε )|2 ada2α = 1 and supp ψ1 ⊂ ⎡⎣−2,− 12 ⎤⎦ ∪ ⎡⎣ 12 , 1⎤⎦.
0
3- ψ2ϵL2(R),‖ψ2‖L2 =1 and supp ψ2 ⊂ [−1, 1].
The shearlet transform is a compactly supported function in the
frequency domain and has a fast decay in the spatial domain. This
transform has directional sensitivity, by which the number of orienta-
tions is doubled at each finer scale. Shearlets are obtained by transla-
tion on the lattice Z2 for any fixed scale and orientation. The shearlet
transform is found especially effective for edge discontinuities and
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 2. The tilings of the frequency plane made by the shearlets (n = 2).

Table 2
List of different multiscale geometric analysis methods in the field of asphalt pavement distress detection, classification, and evaluation.

N Methods Related researches Advantages Shortcomings

1 Wavelet [55–61] Regular in shape; Not sparse for 2-D signals, which leads to a low accuracy for
Compact support; crack representation;
Efficient for objects with point singularities; No smoothness along the edge;
Decomposes an image into subimages; Fail to represent objects with anisotropic elements;
Decomposes an image according to scale and Non-geometrical; and
translation; and Limited capability in dealing with directional information due
Ability to deliver highly sparse approximations to their isotropic support and limited directional sensitivity.
2 Curvelet [58,62–64] High directional sensitivity; Curvelet transform is not directly constructed in the discrete
High anisotropy; domain and implementation involves a low efficiency;
Availability of fast transform algorithms; and Non-adaptive methods; and
Tree-structured filter bank Pseudo-Gibbs artifacts
3 Contourlet [62] The combination of a multi-scale and a Less clear directional features;
directional filter bank in the discrete domain Non-adaptive methods; and
Large computation
4 Ridgelet [58,59,62,64] High directional sensitivity Pseudo-Gibbs artifacts
Highly anisotropic;
Availability of fast transform algorithms
5 Bandelet [62] Adaptive method; Regular with compact support;
Time-frequency localization; The algorithm of searching the best Bandelet basis is very
Directionality; complicated
Anisotropy; and
Strict sampling
6 Shearlet [53,54,60,62,64] Less pseudo-Gibbs artifacts;
High anisotropy;
Adaptive method;
time-frequency localization;
Directional filter bank;
Different decomposition scales;
Optimal approximation; and
Unique combination of mathematical rigidness
7 Beyond (Directionlet, Brushlet, [64]
Tetrolet, Wedgelet, Grouplet)

other singularity identifications [51,52]. Because of this property, In this investigation, a discrete shearlet transform is utilized to
shearlet transform was found useful for edge detection and anomaly process the pavement surface images. This method is based on a
detection [53,54]. Laplacian pyramid combined with appropriate shearing filter. Fig. 2

243
B. Mataei et al.
Fig. 3. Hardware of the proposed system.
shows the schematic representation of the shearlet transform. As shown
in Fig. 2, this transform decomposes the image into two subbands by
the Laplacian pyramid, namely a low-pass and a high-pass subband.
The high-pass subband means the difference between the original
image and the low-band pass image. Then, the high-pass subband is
imported into appropriate shearing filters to complete multidirectional
decomposition. Similarly, the same process is also executed on the low-
pass subband to obtain a multi-scale decomposition.
However, to the best of our knowledge, no work has been done on
evaluation of drainage in asphalt pavement surface using shearlet
transform (Table 2).
3. Proposed method
3.1. Hardware
A hardware set was manufactured to obtain images from the pa-
vement surface drainage. This system is based on a digital camera that
records the pavement surface texture and drainage process. A dynamic
244
Automation in Construction 86 (2018) 240–255
frame, with all imaging components and saturating tools that are placed
inside, was developed in this work. The proposed device consists of
several parts which are explained in the following.
3.1.1. Imaging equipment
A digital camera is embedded in a device that is able to capture
images and videos with desired quality and resolution from the pave-
ment surface and surface drainage process in adjustable time intervals.
3.1.2. Saturation equipment
This system consists of reservoir, pump, pipes and waterjets that are
designed to simulate saturation conditions. Many calculations were
done to adjust the spray intensity, angle of spraying, and installation
height of waterjets. Finally, two waterjets were installed on two sides of
the inner section of the device to prepare the best condition for sa-
turation.
3.1.3. Lighting system
To prepare a constant lighting condition without efficacy of the
B. Mataei et al.
surrounding light, a system consisting of LED lamps was designed to
provide a stable condition for imaging without any reflection and ne-
gative effects in the recorded images. Using this system, image acqui-
sition can be performed at all ambient light conditions.
3.1.4. Temperature recording system
A system with four thermometers is designed to record the tem-
perature of the pavement surface during the images acquisition and
transfer them to the computer to monitor how temperature influences
the surface drainage of pavements.
3.1.5. Movement equipment
The proposed device is able to move in any direction automatically
with a radio controller to facilitate the movement of the device. Two
motors supply the energy for the device to move with the command of a
radio controller in a directional path and turnings.
It is noteworthy that all energy required for different parts of the
proposed device is supplied by a battery embedded on the device,
which makes it totally portable.
Fig. 3 illustrates the hardware of the proposed system and the
procedure of data acquisition.
Fig. 4 presents a schematic illustration of different parts of the
proposed device.
3.2. Database (image base)
To develop the required database, the drainage process of 40 pa-
vements was captured using the assembled hardware. The purpose of
collecting this database is to evaluate the proposed device and method
and assess the efficacy of pavement surface texture in the surface
drainage while surface texture is the determinant parameter. For this
purpose, the database collection is based on these conditions:
- the pavements surface is without any failure such as rutting, frac-
tion, pit, etc.
- the cross and longitudinal gradients are below 1%.
- the surface is without any color or signs.
- the materials and bitumen are homogenous.
245
Automation in Construction 86 (2018) 240–255
Fig. 4. A schematic illustration of the proposed
device: (1) camera; (2) thermometer; (3) waterjet;
(4) LED lamps; (5) height adjustment tools; (6)
USB cable; (7) wheels; and (8) pavement surface.
To perform the data acquisition process, the pavements were in-
itially saturated by the saturation tools placed in the device, and then
the imaging process started immediately after saturation by 2-second
intervals until the end of the drainage process. To facilitate the analysis
process of the images taken by the device, a special white liquid was
used for saturation to create a contrast with the aggregates color, pro-
vide a better visualization for water drainage, and extruding of the
surface aggregates from the liquid. Fig. 5 presents an example of images
taken from the surface drainage of a pavement using the proposed
device.
3.3. Proposed algorithm
This study presents a new method, based on the applied hardware
and software to determine the pavement surface drainage indices from
extracted digital images. Fig. 6 presents a flowchart of the pavement
surface drainage index system.
The algorithm of the proposed system is presented in Fig. 7. The first
section is preprocessing and preparing images, and then analyzing these
images to evaluate the surface drainage of pavement and finally clas-
sification of pavements based on their surface drainage quality.
3.3.1. Preprocessing
There are numerous types of noise and artifacts in imaging methods
that degrade the image quality. Such degradations can have a sig-
nificant impact on the image quality and, as a result, affect human
interpretation and accuracy of computer-assisted methods.
Additionally, due to the poor quality of images, feature extraction,
analysis, recognition, and quantitative measurements become difficult
and unreliable. Denoising and compressing are two actions to pre-
process and enhance the image quality. Denoising is the removal of
noise from images while retaining the important signal features as
much as possible. Moreover, image compression means reducing the
volume of data needed for representing a given image. Computer
images are extremely data-intensive and hence require a large size of
memory for storage. The compression techniques reduce the size of a
file in order to facilitate the efficient transfer of their storage. In this
regard, it is of high importance to compress images without losing their
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 5. Images taken from the surface drainage of a pavement using the proposed device with 2-s intervals.

valuable features for future analysis.

In this step, five wavelet, ridgelet, curvelet, shearlet, and contourlet
transforms were utilized to denoise and compress pavement surface
images for the next step. The best quality after denoising and com-
pressing was for the images preprocessed by shearlet transform
(Table 3). Thus, this transform was used to improve pavement surface
images. Fig. 8 shows a pavement image denoised and compressed by
shearlet transform.

3.3.2. Morphological feature extraction

Fig. 6. A flowchart of the Pavement Surface Drainage index system.
3.3.2.1. Granulometry. In this step, using proper thresholding, the
improved images were transformed in the binary mode. Then, the
granulometry of the aggregate protruding from the surface of the water
was determined. Granulometry estimates the intensity surface area
distribution of aggregates as a function of size. Granulometry likens
image objects to stones whose sizes can be determined by sifting them
through screens of increasing mesh size and collecting what remains
after each pass. Image objects are sifted by opening the image with a
structuring element of increasing size and counting the remaining
intensity surface area (summation of pixel values in the image) after
each opening. Fig. 9 shows the shearlet coefficients extracted from the
pavement surface.
The features of pavement surface texture and edge of surface

246
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 7. The algorithm of the proposed system.

Table 3
The global comparison of the performance of transforms.

Transform Filter Time (s) Psnr (db) SNR (db) MSE MAE UQI SSIM

Wavelet dmey 2.36 25.5104 16.4626 6.2025e + 004 232.2252 0.9708 0.9995
Ridgelet Db1 1.7 101.2772 92.2320 6.1974e + 004 232.1057 1 1
Curvelet dmey 120 34.5235 25.4757 6.2025e + 004 232.2252 0.9952 0.9999
Shearlet pyrexc 6.3 309.9993 300.9446 6.2028e + 004 232.2581 1 1
Contourlet Db1 13.7 297.4252 288.3705 6.2028e + 004 232.2581 1 1

Fig. 8. Pavement image denoised and compressed by

shearlet transform: (a) the original image and (b) the en-
hanced image.

aggregates were extracted and detected by shearlet transform. Fig. 10
shows edge detection and feature extraction of the pavement surface by
shearlet transform. This figure illustrates the high efficiency of shearlet
transform in the detection of edges and surface texture features.
Extracting granulometry of the whole pavement surface images
taken from saturation time until completion of drainage may present
the difference of the protruded aggregates of the pavement surface from
the water during the drainage and trends of water discharge process
from the pavement surface over time. Fig. 11 presents a granulometry
diagram of the pavement surface aggregate protruding from the water
over time (2-second intervals) for a given pavement. The lowest curve is
related to the granulometry diagram of the pavement surface in the
saturation time (0 s) and the other curves from bottom to top refer to
the granulometry diagram over time.

247
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 9. Pavement surface and shearlet coefficients.

Fig. 10. Pavement surface edge detection and shearlet coefficients extraction.

The granulometry diagram provides valuable information about
surface drainage. As shown in Fig. 11, the granulometry curves have
raised which means that the area of protruded aggregates is increasing
with time. The rate of this increase is higher in the first seconds but
decreases with the passage of a certain amount of time. This trend
implies that water is drained from the pavement surface after a while
and drainage continues with a low speed. Clearly, the section with
higher drainage speed (in this diagram about 12-16th seconds) is more
important in terms of pavement drainage evaluation. After passing this
section, the occurrence probability of hydroplaning is drastically re-
duced. Thus, it can be concluded that the lower the time of this section,
the faster the drainage of water.
3.3.2.2. Pavement surface drainage. To develop an automated method
for pavement surface drainage assessment, three general rules were
applied:
1- The volume of surface water decreases by time.
2- The number of protruding aggregates increases by time, but after a
while, due to the decreasing water head, a number of aggregates will
join together and form a new object with a larger area. Thus, the
number of objects will decrease at some points.
3- The area of objects will increase by time. In some pavements, the
presence of obstacles on the path of water drainage causes water
returning and sectional decreasing of the objects area.

248
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 11. Granulometry diagram of the pavement surface aggregates protruding from the water.

Fig. 12. A schematic illustration for the three proposed rules.

Fig. 12 presents a schematic explanation for these three rules. These indices can be calculated as below:
Fig. 13 also presents a 3D presentation of water level changes and n
∑i = 0 (Ni )
number and area of protruded objects over time. the rate of protruded objects numbers = QN = ,
According to the rules listed, three indices were presented to surface t
drainage evaluation: n
∑i = 0 (A i)
the rate of protruded objects area = QA = ,
1- The number of protruded objects over time; t
2- area of protruded objects over time; and the rate of protruded objects surfaces scattering = QP
3- Surface scattering of protruded objects over time. n n
(∑i = 0 (Ni )/ ∑i = 0 (Ai ) )
= .
t

249
B. Mataei et al. Automation in Construction 86 (2018) 240–255

where i is image number, t is imaging time interval, Ni is the number of

n
protruded objects, Ai is area of protruded objects, and ∑i = 0 is summa-
tion vector for time with i steps.
Table 4 presents these indices for a pavement surface 30 s after
saturation.
As can be seen in Table 4, the growth in the number of protruded
objects occurs with a high rate in the first seconds but, it decreases after
a while. This behavior also can be seen in the rate of the objects area.
The higher rate of growth of number and area in the first seconds and
their declination in the following substantiates the higher drainage
speed in the first seconds and decreasing it after a few seconds. The row
of the rate of objects number included positive and negative numbers.
Negative numbers mean the decrease in the number of protruding ob-
jects. The cause of this phenomenon is due to the drop in water level,
where some small objects join together and generate one object with a
bigger area. Therefore, the number of objects reduced while the rate of
protruded objects area is still increasing. The rate of protruded objects
surface scattering has an almost constant trend during the drainage
process. Clearly, the higher the number of objects than the area of
objects, the bigger the rate of protruded objects surface scattering.
As mentioned earlier, the appearance of negative numbers in the
rate of objects number means that a large part of water is drained and
small, protruded objects are substituted with bigger objects. Thus, it can
be concluded that whatever causes the occurrence of hydroplaning,
diminishes after this period and the pavement surface is placed in a safe
mode.

3.3.2.3. Pavement surface drainage diagram. As previously mentioned,

the granulometry diagram of protrusion changes over time. The rate of
this change is variable: in the first seconds after saturation, it is higher
but it decreases after a while due to the reducing drainage speed.
Extraction of the difference of the granulometry diagrams over time for
a given pavement presents the rate of water discharge or surface
drainage rate of the pavement. Fig. 14 presents surface drainage rate
diagrams for 6 pavement surfaces.
As shown in Fig. 14, the diagrams have higher values in the first
seconds implying that the drainage has a higher speed at this time. In
the following, the values decrease and drainage continues with a lower
speed. Every surface drainage diagram has a minimum point which is of
great importance; because after this point the behavior of the surface
drainage changes. Studying the characteristics of these diagrams in two
sections before and after the minimum point and measuring statistical
parameters including mean, variance, minimum, and maximum of
these sections shows that:

1- The first section has a higher change in growth rate of objects than
the second section and drainage speed is higher while in the second
section drainage has a constant rate and drainage is almost stopped.
2- The time of minimum point occurrence is an important issue. The
lower the time of minimum point, the higher the speed of the
drainage.
3- In the minimum point of each drainage rate pavement surface, water
is partially available on the pavement surface. Thus, water is
drained and the risk of hydroplaning is minimized.
4- Some drainage diagrams of pavement surfaces have a negative
Fig. 13. 3D presentation of water level fluctuations and number and area of protruded minimum point. In such pavements, drained water returns due to
objects over time. the existence of obstacles in the discharging route. This returned

Table 4
The indices calculated for a pavement surface for 30 s after saturation.

Time(s) 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30

QN 49.5 132 44.5 57.5 − 5.5 − 3.5 − 13.5 34 − 19 − 30 −4 − 16.5 11.5 −29 −4
QA 1469 2682 2768 769 1974 1639 1217 290 955 502 680 514 163 664 185
QP ∗10−4 169 246 80 373 − 14 − 11 − 55 586 − 99 − 298 − 29 − 161 353 −218 − 108

250
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Fig. 14. A drainage rate diagram of 6 pavement surfaces.

Fig. 15. A drainage rate diagram of pavement surfaces with coarse and fine texture.

water is placed in the drainage process once more and increases the occurs in lower time and it can be concluded that drainage speed in
drainage time. pavements with fine texture is higher that drainage speed in coarse
5- The drainage speed has a relationship with the surface texture of the texture pavements (Fig. 15).
pavement. In the pavements with coarse texture, the minimum point

251
B. Mataei et al. Automation in Construction 86 (2018) 240–255

Table 5 To use the system results and make it work on full scale, the data
The variation range of indices for three class of pavements. (images) of surface drainage of the intended pavement is obtained by
the presented device. These images are analyzed by the presented
Type of pavement surface drainage QN QA QP
quality method and the characteristics of surface drainage of the pavement will
be examined. Finally, three indices related to pavement surface drai-
Good > 25 > 1000 0.015 < , < 0.02 nage are calculated and the drainage quality is placed in the range
Normal > 21 > 900 0.013 < , < 0.018
presented for three classes of pavements. Thus, the drainage quality of
Bad < 21 < 900 < 0.013
the pavement is characterized.
The numbers are related to mean of calculated parameters over time for every pavement.
3.3.4. The method validation
Table 6 In this section, the performance of the presented system is evaluated
Confusion matrix. using the confusion matrix. This matrix provides the basis for evalu-
ating the performance of any classifier through its four components,
System result
i.e., True Positive (TP), False Negative (FN), False Positive (FP), and
Reference images Good Normal Bad True Negative (TN). Among others, classification accuracy is the major
Good 12 2 0 parameter to judge the efficiency of a classifier. Classification accuracy
Normal 3 11 3 of a classifier on a given data set refers to the percentage of test set
Bad 0 2 17
tuples that is correctly classified by the classifier. It reflects how well
the classifier recognizes tuples of various classes. This measure is de-
Table 7 fined by:
Confusion table for “good” class.
TP + TN
accuracy (ACC) =
TP = 12 FN = 2 P+N
A number of measures of system performance can be defined based
FP = 3 TN = 28
on the confusion matrix. Some common measures are a sensitivity or
true positive rate (TPR), specificity (SPC), True Negative Rate (TNR),
Table 8 and precision or positive predictive value (PPV).
Confusion table for “normal” class. Sensitivity is a measure of the ability of a prediction model to select
instances of a certain class from a data set, defined as:
TP = 11 FN = 6
TP
FP = 4 TN = 29 Sensitivity =
TP + FN
Specificity is a measure of the ability of a prediction model to select
Table 9 instances of other classes in their own classes. This parameter is defined
Confusion table for “bad” class. as:
TP = 17 FN = 2 TN
Specifity =
FP + TN
FP = 3 TN = 23
Precision is a measure of the accuracy by which a specific class is
predicted. It is defined by:
Table 10
Performance evaluation of the presented system. TP
precision =
TP + FP
Type of pavement surface Accuracy Precision Sensitivity Specificity
drainage quality A confusion matrix has two dimensions; one dimension is indexed
by an actual class of an object and the other is indexed by the class that
Good 88.9% 80% 85.7% 90.3%
the classifier predicts. Table 6 presents the confusion matrix for this
Normal 80% 73.3% 64.7% 87.8%
Bad 88.9% 85% 89% 88.5%
work. For this evaluation, reference images were used as a comparison
system 86% 79.4% 80% 88.9% with the method results. In Table 7 the confusion table for "Good" class
and in Table 8 and Table 9 the confusion tables for respectively
"Normal" and "Bad" classes are presented.
Based on these facts, it is possible to assess the surface drainage As shown in Table 10, the presented system showed an 86% accuracy,
quality of different pavements. 79.4% precision, 80% sensitivity, and 88.9% specificity. Thus, it can be
concluded that the system presents appropriate indices and has a high per-
3.3.3. Final surface drainage evaluation formance in simulation and assessment of surface drainage of pavements.
In the previous sections, three parameters (e.g., the growth rate of
protruded objects number, the growth rate of protruded objects area, 3.4. User interface
and the growth rate of protruded objects surface scattering) were in-
troduced as indices for evaluation of pavement surface drainage. To produce a good user interface, a graphical user interface (GUI)
These parameters were calculated for surface drainage of 40 pave- was designed (Fig. 16). This GUI is user-friendly and bears all necessary
ments. The drainage process of pavements is reviewed by experts in the information. The operation starts by loading the pavement image (“load
pavement surface and pavements were classified into three classes: pavement” command). Then, the system analyzes the series of images
good, normal, and bad. taken from the drainage process and the surface drainage diagram is
The variation range of three indices for every class was calculated displayed as the “analyze” command. Finally, three indices calculated
by assessing the indices for different pavements according to their for the current pavement are shown and the drainage situation is
surface drainage situation. Table 5 presents the variation range of in- evaluated and classified into one of three good, normal, or bad classes
dices for three classes of pavements. by the “pavement drainage situation” command.

252
B. Mataei et al.
Fig. 16. The graphical user interface of the presented automated system.
4. Conclusions
In this study, an automated system was developed based on image
processing techniques to evaluate the pavement surface drainage. The
results indicate the acceptable performance of the proposed system.
Briefly, the shearlet transform was found as an effective tool for
denoising and compressing the pavement images taken by the proposed
hardware. The changes in the granulometry of the protruded aggregates
over time with 2-s intervals were recorded and the granulometry dia-
gram of every section was plotted. Moreover, the diagram of pavement
surface drainage rate was obtained by the differences of the granulo-
metry diagram. Analysis of these diagrams provides valuable informa-
tion about the drainage pattern of pavements. Finally, three indices
were presented for evaluation of surface drainage based on the number,
area, and surface scattering of the protruded aggregates from the water
over time. Evaluating the performance of the proposed system with
reference images using the confusion matrix presented a considerable
86% accuracy for the proposed system.
In this work, considerable time and effort were allocated to attain
reliable results which cannot be easily affected by environmental fac-
tors. Although the image acquisition process needs short-time traffic
controlling (because drainage needs time to accomplish), the analysis
process of the proposed system was very fast (less than a minute). This
short period allows giving immediate responses to a given index during
real-time applications.
The advantage of this system is its fast processing, operation ease,
and user-friendly function. Besides, it is able to predict the occurrence
probability of the hydroplaning phenomenon, which is an important
253
Automation in Construction 86 (2018) 240–255
issue in road safety assessment. The results of the proposed system can
be implemented in the network-level and project-level pavement
management systems for road safety enhancement.
5. Future work
Despite all advantages mentioned for the proposed method, it is a
few steps taken toward the application of new methods and technolo-
gies for assessing pavement surface drainage. There are many methods
and devices for evaluation of pavement texture characteristics, such as
skid resistance and roughness. The majority of these methods are based
on the dry situation of the pavement surface. Also, other methods which
work on wet pavements cannot simulate the drainage process of the
pavement from saturation to completion of drainage. By now, there is
no method and device proposed for evaluation of pavement surface
drainage. Therefore, the proposed method is the first system for eval-
uating surface drainage of pavement. There are many options to de-
velop the present method to achieve more efficient results. The future
work will focus on using new technologies to prepare better data for
processing and analysis like using scanning laser, Kinect camera, and
other equipment to provide more accurate visualization and analysis of
the surface in three dimensions (3D), or in the processing step, finding
better filters for image enhancement using transforms, extracting more
appropriate indices for drainage evaluation, etc.
References
[1] Ibrahim M. Asi, Evaluating skid resistance of different asphalt concrete mixes,
Build. Environ. 42 (1) (2007) 325–329, http://dx.doi.org/10.1016/j.buildenv.
B. Mataei et al.
2005.08.020.
[2] Gerardo Flintsch, et al., Pavement surface macrotexture measurement and appli-
cations, Transp. Res. Rec. 1860 (2003) 168–177, http://dx.doi.org/10.3141/
1860-19.
[3] Brian Prowell, Douglas Hanson, Evaluation of circular texture meter for measuring
surface texture of pavements, Transp. Res. Rec. 2005 (1929) 88–96, http://dx.doi.
org/10.3141/1929-11.
[4] Arash Rezaei, et al., Predicting asphalt mixture skid resistance by aggregate char-
acteristics and gradation, Transp. Res. Rec. 2104 (2009) 24–33, http://dx.doi.org/
10.3141/2104-03.
[5] Mayora, José M. Pardillo, Rafael Jurado Piña, An assessment of the skid resistance
effect on traffic safety under wet-pavement conditions, Accid. Anal. Prev. 41 (4)
(2009) 881–886, http://dx.doi.org/10.1016/j.aap.2009.05.004.
[6] Soon Mi Oh, et al., Evaluation of Traffic and Environment Effects on Skid Resistance
in California, Safe Transportation Research & Education Center, 2010 https://
EconPapers.repec.org/RePEc:cdl:itsrrp:qt01p1d923.
[7] Julio Alberto Roa, Evaluation of International Friction Index and High-friction
Surfaces, Diss. Virginia Tech, 2008 https://vtechworks.lib.vt.edu/handle/10919/
36277.
[8] Murat Ergun, Sukriye Iyinam, A. Faik Iyinam, Prediction of road surface friction
coefficient using only macro-and microtexture measurements, J. Transp. Eng. 131
(4) (2005) 311–319, http://dx.doi.org/10.1061/(ASCE)0733-947X(2005)
131:4(311).
[9] Tom Brijs, Dimitris Karlis, Geert Wets, Studying the effect of weather conditions on
daily crash counts using a discrete time-series model, Accid. Anal. Prev. 40 (3)
(2008) 1180–1190, http://dx.doi.org/10.1016/j.aap.2008.01.001.
[10] Anna K. Andersson, Chapman Lee, The impact of climate change on winter road
maintenance and traffic accidents in West Midlands, UK, Accid. Anal. Prev. 43 (1)
(2011) 284–289, http://dx.doi.org/10.1016/j.aap.2010.08.025.
[11] Chengcheng Xu, Wei Wang, Liu Pan, Identifying crash-prone traffic conditions
under different weather on freeways, J. Saf. Res. 46 (2013) 135–144, http://dx.doi.
org/10.1016/j.jsr.2013.04.007.
[12] W.R. Tyfour, Tire skid resistance on contaminated wet pavements, Jordan J. Mech.
Indust. Eng. 3 (2) (2009) 119–124 (ISSN:1995-6665).
[13] Gerardo W. Flintsch, Yingjian Luo, I.L. Al-Qadi, Analysis of the Effect of Pavement
Temperature on the Frictional Properties of Flexible Pavement Surfaces, 84th
Transportation Research Board Annual Meeting Vol. 21 (2005) https://trid.trb.org/
view/1157120.
[14] Burak Sengoz, Ali Topal, Serhan Tanyel, Comparison of pavement surface texture
determination by sand patch test and 3D laser scanning, Period. Polytech., Civ. Eng.
56 (1) (2012) 73, http://dx.doi.org/10.3311/pp.ci.2012-1.08.
[15] J.W. Hall, et al., Guide for Pavement Friction, Transportation Research Board of the
National Academies, Washington DC, USA, 2009 http://www.trb.org/Publications/
Blurbs/161756.aspx.
[16] James C. Wambold, John J. Henry, Rudolph R. Hegmon, Skid Resistance of Wet-
weather Accident Sites, The Tire Pavement Interface, ASTM International, 1986
https://www.astm.org/DIGITAL_LIBRARY/STP/PAGES/STP20001S.htm.
[17] P.G. Roe, David C. Webster, G. West, The Relation Between the Surface Texture of
Roads and Accidents, (1991) https://trid.trb.org/view.aspx?id=1174885.
[18] M.M. Miller, H.D. Johnson, Effect of Resistance to Skidding on Accidents: Surface
Dressing an Elevated Section of M4 Motorway. No. R&D Rpt, (1973) https://trid.
trb.org/view.aspx?id=137522.
[19] David A. Noyce, et al., Incorporating Road Safety into Pavement Management:
Maximizing Asphalt Pavement Surface Friction for Road Safety Improvements,
Draft Literature Review and State Surveys, Midwest Regional University
Transportation Center (UMTRI), Madison, Wisconsin, 2005 http://www.wistrans.
org/mrutc/files/Lit-rev-and-info.pdf . (10.1.1.130.4067).
[20] Hartwig W. Kummer, Wolfgang Eberhard Meyer, Tentative Skid-resistance
Requirements for Main Rural Highways, NCHRP Report 37, (1967) https://trid.trb.
org/view.aspx?id=104847.
[21] B.M. Gallaway, et al., Tentative Pavement and Geometric Design Criteria for
Minimizing Hydroplaning, No. FHWA-RD-75-11 Final Rpt. (1975) https://trid.trb.
org/view.aspx?id=62125.
[22] G. Ong, T. Fwa, Jing Guo, Modeling hydroplaning and effects of pavement micro-
texture, Transp. Res. Rec. 1905 (2005) 166–176, http://dx.doi.org/10.3141/
1905-19.
[23] Minh-Tan Do, Hassan Zahouani, Roberto Vargiolu, Angular parameter for char-
acterizing road surface microtexture, Transp. Res. Rec. 2000 (1723) 66–72, http://
dx.doi.org/10.3141/1723-09.
[24] A.R. Savkoor, Paper viii (iii) tribology of tyre traction on dry and wet roads, Tribol.
Ser. 18 (1991) 213–228, http://dx.doi.org/10.1016/S0167-8922(08)70137-8.
[25] Walter B. Horne, Status of runway slipperiness research, Transp. Res. Rec. 624
(1977) 95–121 https://trid.trb.org/view.aspx?id=59344.
[26] Rita Pelloli, Road surface characteristics and hydroplaning, Transp. Res. Rec. 624
(1977) 27–32 https://trid.trb.org/view.aspx?id=59336.
[27] Behrouz Mataei, et al., Pavement friction and skid resistance measurement
methods: a literature review, Open J. Civ. Eng. 6 (04) (2016) 537, http://dx.doi.
org/10.4236/ojce.2016.64046.
[28] James C. Wambold, J.J. Henry, International Piarc Experiment to Compare and
Harmonize Texture and Skid Resistance Measurement, Nordic Road and Transport
Research, (1994) https://trid.trb.org/view.aspx?id=551495.
[29] Weifeng Wang, et al., Design and verification of a laser based device for pavement
macrotexture measurement, Transp. Res. C 19 (4) (2011) 682–694, http://dx.doi.
org/10.1016/j.trc.2010.12.001.
[30] Eyad Masad, et al., Field Evaluation of Asphalt Mixture Skid Resistance and Its
Relationship to Aggregate Characteristics, No. 0-5627-3, Texas Transportation
254
Automation in Construction 86 (2018) 240–255
Institute, 2010 https://texashistory.unt.edu/ark:/67531/metapth303684/m2/1/
high_res_d/txcs-0092.pdf.
[31] Ronald Elunai, Vinod Chandran, Prosper Mabukwa, Digital Image Processing
Techniques for Pavement Macro-texture Analysis, Proceedings of the 24th ARRB
Conference: Building on 50 Years of Road Transport Research, ARRB Group Ltd.,
2010No. 0572-1 https://eprints.qut.edu.au/40770/.
[32] Fereidoon Moghadas Nejad, Nasrin Karimi, H. Zakeri, Automatic image acquisition
with knowledge-based approach for multi-directional determination of skid re-
sistance of pavements, Autom. Constr. 71 (2016) 414–429, http://dx.doi.org/10.
1016/j.autcon.2016.08.003.
[33] H. Zakeri, Fereidoon Moghadas Nejad, Ahmad Fahimifar, Rahbin: a quadcopter
unmanned aerial vehicle based on a systematic image processing approach toward
an automated asphalt pavement inspection, Autom. Constr. 72 (2016) 211–235,
http://dx.doi.org/10.1016/j.autcon.2016.09.002.
[34] Izeppi de Leon, Edgar David, et al., Area-based Macrotexture Measurements: Stereo
Vision Approach, Transportation Research Board 88th Annual Meeting. No. 09-
3067, (2009) https://trid.trb.org/view.aspx?id=882081.
[35] Ronald Elunai, Vinod Chandran, Edith Gallagher, Asphalt concrete surfaces mac-
rotexture determination from still images, IEEE Trans. Intell. Transp. Syst. 12 (3)
(2011) 857–869, http://dx.doi.org/10.1109/TITS.2011.2116784.
[36] H.M. Zelelew, A.T. Papagiannakis, E.D. de León Izeppi, Pavement macro-texture
analysis using wavelets, Int. J. Pavement Eng. 14 (8) (2013) 725–735, http://dx.
doi.org/10.1080/10298436.2012.705004.
[37] B. Pidwerbesky, et al., Road surface texture measurement using digital image
processing and information theory, Land Transport New Zealand Research Report,
290 2006, pp. 65–79 ISBN 0-478-28701-X.
[38] Alfredo Cigada, et al., Laser-triangulation device for in-line measurement of road
texture at medium and high speed, Mech. Syst. Signal Process. 24 (7) (2010)
2225–2234, http://dx.doi.org/10.1016/j.ymssp.2010.05.002.
[39] E. Masad, et al., Test Methods for Characterizing Aggregate Shape, Texture and
Angularity, Technical Report, National Cooperative Highway Research
Program–NCHRP, 2005 https://trid.trb.org/view.aspx?id=810799.
[40] Victor M.C. Araujo, Iuri S. Bessa, Verônica T.F. Castelo Branco, Measuring skid
resistance of hot mix asphalt using the aggregate image measurement system
(AIMS), Constr. Build. Mater. 98 (2015) 476–481, http://dx.doi.org/10.1016/j.
conbuildmat.2015.08.117.
[41] Emmanuel J. Candès, David L. Donoho, New tight frames of curvelets and optimal
representations of objects with piecewise C2 singularities, Commun. Pure Appl.
Math. 57 (2) (2004) 219–266, http://dx.doi.org/10.1002/cpa.10116.
[42] Da Cunha, L. Arthur, Jianping Zhou, N.Do. Minh, The nonsubsampled contourlet
transform: theory, design, and applications, IEEE Trans. Image Process. 15 (10)
(2006) 3089–3101, http://dx.doi.org/10.1109/TIP.2006.877507.
[43] Demetrio Labate, et al., Sparse multidimensional representation using shearlets,
Adv. Opt. Photon. (2005), http://dx.doi.org/10.1117/12.613494 (International
Society for Optics and Photonics, 2005).
[44] G. Kutyniok, Shearlets: Multiscale Analysis for Multivariate Data, Springer Science
& Business Media, https://doi.org/10.1007/978-0-8176-8316-0, (2012).
[45] Jean-Luc Starck, Fionn Murtagh, Jalal M. Fadili, Sparse Image and Signal
Processing: Wavelets, Curvelets, Morphological Diversity, Cambridge University
Press, 2010 https://dl.acm.org/citation.cfm?id=1830428.
[46] Kanghui Guo, Demetrio Labate, The construction of smooth Parseval frames of
shearlets, Modelling of Natural Phenomena 8 (1) (2013) 82–105, http://dx.doi.org/
10.1051/mmnp/20138106.
[47] Kanghui Guo, Demetrio Labate, Optimally sparse multidimensional representation
using shearlets, SIAM J. Math. Anal. 39 (1) (2007) 298–318, http://dx.doi.org/10.
1137/060649781.
[48] Kanghui Guo, Demetrio Labate, Optimally sparse representations of 3D data with
C^2 surface singularities using Parseval frames of shearlets, SIAM J. Math. Anal. 44
(2) (2012) 851–886, http://dx.doi.org/10.1137/100813397.
[49] Glenn Easley, Demetrio Labate, Wang-Q Lim, Sparse directional image re-
presentations using the discrete shearlet transform, Appl. Comput. Harmon. Anal.
25 (1) (2008) 25–46, http://dx.doi.org/10.1016/j.acha.2007.09.003.
[50] Gitta Kutyniok, Morteza Shahram, Xiaosheng Zhuang, Shearlab: a rational design of
a digital parabolic scaling algorithm, SIAM J. Imag. Sci. 5 (4) (2012) 1291–1332,
http://dx.doi.org/10.1137/110854497.
[51] Kanghui Guo, Demetrio Labate, Characterization and analysis of edges using the
continuous shearlet transform, SIAM J. Imag. Sci. 2 (3) (2009) 959–986, http://dx.
doi.org/10.1137/080741537.
[52] Kanghui Guo, Demetrio Labate, Characterization of piecewise-smooth surfaces
using the 3D continuous shearlet transform, J. Fourier Anal. Appl. 18 (3) (2012)
488–516, http://dx.doi.org/10.1007/s00041-011-9209-y.
[53] Sheng Yi, et al., A shearlet approach to edge analysis and detection, IEEE Trans.
Image Process. 18 (5) (2009) 929–941, http://dx.doi.org/10.1109/TIP.2009.
2013082.
[54] Xavier Gibert, et al., Discrete shearlet transform on GPU with applications in
anomaly detection and denoising, EURASIP J. Adv. Signal Proc. 2014 (1) (2014) 64,
http://dx.doi.org/10.1186/1687-6180-2014-64.
[55] Jian Zhou, Peisen S. Huang, Fu-Pen Chiang, Wavelet-based pavement distress de-
tection and evaluation, Opt. Eng. 45 (2) (2006) 027007, http://dx.doi.org/10.
1117/1.2172917.
[56] H. Zakeri, Fereidoon Moghadas Nejad, Ahmad Fahimifar, Image based techniques
for crack detection, classification and quantification in asphalt pavement: a review,
Arch. Comput. Meth. Eng. (2016) 1–43, http://dx.doi.org/10.1007/s11831-016-
9194-z.
[57] S. Arivazhagan, L. Ganesan, T.G. Subash Kumar, Texture classification using rid-
gelet transform, Pattern Recogn. Lett. 27 (16) (2006) 1875–1883, http://dx.doi.
B. Mataei et al.
org/10.1016/j.patrec.2006.04.013.
[58] Jalal Fadili, Jean-Luc Starck, Curvelets and ridgelets, Encyclopedia of Complexity
and Systems Science, Springer, New York, 2009, pp. 1718–1738, , http://dx.doi.
org/10.1007/978-0-387-30440-3_111.
[59] Daqi Zhang, et al., Automatic ridgelet image enhancement algorithm for road crack
image based on fuzzy entropy and fuzzy divergence, Opt. Lasers Eng. 47 (11) (2009)
1216–1225, http://dx.doi.org/10.1016/j.optlaseng.2009.05.014.
[60] James M. Murphy, Jacqueline Le Moigne, David J. Harding, Automatic image re-
gistration of multimodal remotely sensed data with global shearlet features, IEEE
Trans. Geosci. Remote Sens. 54 (3) (2016) 1685–1704, http://dx.doi.org/10.1109/
TGRS.2015.2487457.
255
Automation in Construction 86 (2018) 240–255
[61] Gitta Kutyniok, Wang-Q Lim, Rafael Reisenhofer, Shearlab 3D: Faithful digital
shearlet transforms based on compactly supported shearlets, ACM Trans. Math.
Softw. 42 (1) (2016) 5, http://dx.doi.org/10.1145/2740960.
[62] Ke Xu, Shunhua Liu, Yonghao Ai, Application of shearlet transform to classification
of surface defects for metals, Image Vis. Comput. 35 (2015) 23–30, http://dx.doi.
org/10.1016/j.imavis.2015.01.001.
[63] Wang-Q Lim, Nonseparable shearlet transform, IEEE Trans. Image Process. 22 (5)
(2013) 2056–2065, http://dx.doi.org/10.1109/TIP.2013.2244223.
[64] Laurent Jacques, et al., A panorama on multiscale geometric representations, in-
tertwining spatial, directional and frequency selectivity, Signal Process. 91 (12)
(2011) 2699–2730, http://dx.doi.org/10.1016/j.sigpro.2011.04.025.