Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

https://doi.org/10.1007/s40996-023-01240-5

REVIEW PAPER

Airport Pavement Distress Analysis

Hamid Noori1 · Raju Sarkar1

Received: 27 June 2023 / Accepted: 15 September 2023 / Published online: 26 October 2023
© The Author(s), under exclusive licence to Shiraz University 2023

Abstract
Airport pavement evaluation and surface condition assessment is the core topic of researchers today. When performing
surveys for airport pavement surface conditions, it is desirable but not feasible to check all types of distresses that appeared
in the pavement area of an airport. The Airport’s main operation areas are airside, consisting of a Runway, Taxiway, and
Apron used for the takeoff, landing, and taxing, including Parking of the aircraft. The primary objective of evaluating pave-
ment conditions is to focus on the current and future pavement conditions. The pavement conditions are assessed based on
Surface distresses, which are collected manually or automatically and are primarily human-dependent, time-consuming,
insufficient/or inefficient, and expensive. To this extent, the collection of literature in this review article focuses on identifying
related research topics to assist in describing distress types, severity levels, and densities, including the procedure followed
by researchers to analyze the pavement condition. The overall systematic literature review demonstrated the number of
published articles from various Journals concentrating on the evaluation of pavement structures using traditional methods
and provided an alternative solution relevant to this research area for identifying distress measurement and detection using
Artificial intelligence. Finally, summarize the paper with a qualitative overview, and comparisons of different models based
on quantitative, and qualitative with reflection due to limitations and restrictions and the possibility of expansion for future
research work on the subject area for a better understanding and fast decision-making in airport pavement evaluation program.

Keywords Airside · Surface distresses · Pavement conditions · Distress measurement · Decision-making · Pavement
evaluation program

1 Introduction which are functional and structural. For the safe operation
of the aircraft that travels at high speed, it is needed to deter-
The performance of the Pavement is measured by the pave- mine the condition of the pavement, therefore, it is necessary
ment condition Index which indicates the present condition to measure both functional and structural-related induced
of the pavement in terms of structural integrity and surface distresses that are appearing on the surface of the pavement
operational conditions (Shahin 2005). Pavement condition structures. A good understanding of the types, quantities,
evaluation provides information to make more cost-effective severity, and locations of distresses, plus the causes of those
and reliable decisions regarding the management of pave- distresses, is essential in selecting a cost-effective rehabili-
ment networks. (Majidifard et al. 2020). In particular, air- tation method. Consequently, a pavement condition survey
port pavements deteriorate more rapidly as heavy loads are must first be conducted. However, this can only ascertain
imposed by aircraft traveling at high speeds (Hachiya et al. the actual distress present during the survey. To evaluate the
2013). Pavement distressed information is usually collected progression of the distresses, regular surveys are indispensa-
manually or automatically. In fact, distresses are two types ble (Mulry et al. 2015). The latest developments in computer
applications and technology have made it possible to ease
the detection and measurement of pavement distress auto-
* Hamid Noori matically (Coenen and Golroo 2017). Mohan and Poobal
[email protected]
(2018) studied the analysis of crack detection using Image
Raju Sarkar processing from India. A similar review article was produced
[email protected]
by Mathavan et al. (2015) for pavement distress detection
1
Department of Civil Engineering, Delhi Technological and measurements using 3D imaging technology. Pavement
University, New Delhi 110042, India

13
Vol.:(0123456789)
1172 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

distress segmentation methods were developed by (Tsai et al. proposed by Wesolowski and Iwanowski (2020) for cement
2010). The authors used and tested actual pavement images concrete airport pavements to scope air operation safety
with varying lighting conditions, shadows, and crack posi- and air transport participants’ lives. H. Wang et al. (2016)
tions to differentiate their performance. Image segmenta- assessed Life cycle of airport pavement design alternatives
tion is the process of extracting the object of interest (dis- for energy and Environmental impacts. Shah et al. (2004)
tresses) from the background of an image. Thresholding and developed a pavement performance evaluation model at
edge detection are two principal segmentation approaches. Canadian airfield pavement management.
Shahnazari et al. (2012) developed an alternative method of
measuring the Pavement condition index with the help of
an optimization technique based on ANN and genetic algo- 2 Research Objectives
rithms. Statistical models and mathematical procedures such
as regression technique is also been implemented to predict Airport authorities are investing a considerable amount in
the pavement condition by determining the remaining ser- airport infrastructure to enhance safe operation and move-
vice life RSL based on the PCI method by the researcher in ments for users.
Indonesia (Setywan et al. 2015). Various methods of deter- The following objectives are considered to be studied in
mining the pavement condition index (PCI) are listed in liter- this research article.
ature and operational practice based mainly on the inventory
of pavement surface deterioration. The standard methods 1. Identify the present procedure for the evaluation of Air-
used worldwide are the ASTM D5340 and ASTM D6433, port pavement distresses.
published in the American standards described by (Shahin 2. Investigate the distress levels and classify different cat-
2005). The others include the PAVER procedure used world- egories of the distresses appearing in the surface section
wide. Shah et al. (2013) from India developed overall pave- of the network, particularly the Runway, Taxiway, and
ment condition index for the Urban-road network, including Aprons area.
pavement, distress roughness, structure, and skid parameter. 3. Identify and describe artificial intelligence procedures
Mubaraki (2014), Identified pavement distresses and evalu- including machine learning, deep learning, and image
ated pavement conditions using the PCI procedure. Chen analysis methods for detection and analysis of distress
et al. (2012), developed a method for calculating the PCI on pavement structure.
index based on mathematical models, including weighting 4. Compare the image analysis and traditional distress
factors for specific types of pavement deterioration. Sharaf analysis methods.
et al. (1987) applied the ANOVA method to predict Remain- 5. An overall review of airport pavement distress analysis
ing Service Life based on the PCI indicator. The author con- and techniques followed by different organizations/insti-
cluded that if the proper range of PCI is selected, a higher tutions and researchers for implementing the pavement
possibility of determining Pavement maintenance and cost management systems.
scenario is achievable. Majidifard et al. (2020) applied Deep
Machine Learning to develop a new Asphalt PCI. Sue et al.
(2002) established mathematical prediction models for the 3 Pavement Distresses Analysis Methods
deterioration of rigid airfield pavement in South Korea. A
similar approach was proposed by (Camarena et al. 2018) to Airport pavements can generally deteriorate rapidly due
predict Airport pavement management numerically in Peru. to aircraft load and climatic effects. Thus, according to the
Liu et al. (2022) developed a new set of indices and thresh- severity of the distress, appropriate measures need to be
olds for preventive maintenance decision-making for Airport taken to maintain pavement performance (Havhiya et al.
Runways with composite pavement structures. Performance 2013). The first phase of the evaluation process is the iden-
models for Oklahoma Airfield Pavement were studied by tification of pavement distress and recognizing the types
(Yuan et al. 2003), and family modeling methods based on appearing on the Surface of the pavement structure; next, it
PCI were used to forecast the deterioration rate. Greene et al. is necessary to measure the level of defects and their depth.
(2004) assessed airfield pavement condition using pavement Recognizing distress is a critical phase of the evaluation
condition index, SCI characteristics, and FOD potential. Guo process because no specific severity-related instructions
et al. (2021) used a comparative Analysis of Airfield pave- exist for an approach that could influence the classification
ment life. The author found the relationship of PCI, SCI, variation (Ragnoli et al. 2018). ASTM D6433-16, which
and FOD in evaluating the service life of Airfield Pavement provides distress criteria identification and classification
based on PAVEAIR. Arhin et al. (2015) studied the pre- for both flexible and rigid pavement, is a well-known and
diction of pavement condition index using the IRI method diffused reference.
in a Dense Urban Area. An APCI evaluation method was

13
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1173

3.1 Pavement Condition Index m/km, mm/m) is given. Arhin et al. (2015) predicted the
pavement condition index using the IRI method in a Dense
The United States of America primarily applies this method Urban Area. Sandra and Sarkar (2013), Developed a model
to evaluate airport pavement surface conditions, and ASTM for estimating IRI from pavement Distresses. Loprencipe
D5340 standardized PCI in 1988. PCI system performed and Zoccali (2019) used to compare methods for evaluation
by the visual condition survey identifying the severity and of pavement roughness.
quantity of the defect (Shahin 2005). The PCI process is
a simple and convenient method to measure and evaluate 3.3 Pavement Rehabilitation Index (PRI)
pavement conditions in terms of surface distress and provide
a solution for maintenance and rehabilitation (Karim et al. The pavement rehabilitation index was established to inves-
2016). Different types and categories of distress for flexible tigate the pavement surface and structure of the Pavement
Pavement and Rigid Pavement are included in this method, in Japan to help in decision-making for rehabilitation and
as illustrated in Tables 1 and 2. maintenance requirements. The Pavement Surface condi-
The PCI technique is not a straightforward method used tions are quantified by the index of PRI, which includes
to calculate the structure bearing capacity or evenness or a general sense of the pavement’s existing state and the
surface roughness (Shahin and Becker 1984); however, it rehabilitation necessity. Equations to calculate the PRI for
is an impartial system used for assessing a paved network asphalt pavements and concrete pavements are listed as fol-
section maintenance and rehabilitation needs concerning an lows (Eqs. (3) and (4), respectively) (Hachiya et al. 2013).
entire pavement system (Hajj et al. 2011; Tarefder and Rah- (a) Asphalt Pavement
man 2016). Specifically, PCI is a number from zero to 100
(Shahin and Becker 1984). The lowest value indicates poor PRI = 10 − 0.45CR − 0.0511RD − 0.655SV (3)
condition, and the highest value denotes the perfect state of
where CR: cracking ratio (%), RD: rutting (mm), SV: rough-
the pavement.
ness (Standard Deviation of longitudinal elevation (mm).
PCIi = PCImax − Max corrected Deduct value (CDV) (1) (b) Concrete Pavement

where, PCIi= individual condition index based on measured PRI = 10 − 0.29CR − 0.296JC − 0.535SV (4)
condition, PCImax = value for a perfect condition with no where CR: Cracking ratio (Cm/m2), JC: ratio of joint fail-
measured defects, deduct = deduct value assigned to distress ures (%), SV: faulting (Maximum, mm).
type, severity, and extent from the result using the Eq. (1).

3.4 Structural Ability Evaluation

3.2 International Roughness Index (IRI)
The structural evaluation aims to assess the pavement’s
The IRI method is used to measure the roughness of the strength and maximum allowable traffic loading by the
pavement surface. It is directly related to the ride quality and pavement and forecast the pavement’s future service life. In
computed based on the surface elevation Shahin and Becker the late 1950s, ICAO provided the LCN (Load Classifica-
(1984). It is measured by using high-performance equipment tion Number) method to determine load bearing capacity of
such as a profiler, profilometer, and profile graph. The IRI pavements (Hachiya et al. 2013). This method changed into
method detects road profiles in real-time and the acquired the ACN/PCN method in the 1980s (ICAO 2006). Under
data are analyzed in terms of frequency-weighted vertical this technique, airport pavements are considered fit for use
acceleration ( awz ) according to ISO 2631. The evaluation by aircraft if the ACN of the plane in question is equal to
is performed by model, calculating the simulated suspen- or smaller than the load-carrying capacity of the relevant
sion motion on profile and dividing the sum by the distance pavements, expressed as PCN. ACN is an indexed number
traveled according to the Eq. (2): expressing the impact of aircraft on the pavement. At the
same time, PCN does the same for load bearing capac-
1∕l
ity of pavement. In other words, they indicate the aircraft
l∫
1
IRI = |ż − ż |dt
|s u| (2) load that can be tolerated without restricting traffic volume
0
(ICAO 2016). The method of calculating ACN depends on
where L is the length of the profile in km, v is the simulated the type of pavement, i.e., whether it is made of concrete or
speed equal to 80 km/h, żs is the time derivative of vertical asphalt; for asphalt pavement, it is calculated from the pave-
displacement of the sprung mass in meters and żu is the time ment thickness obtained using the CBR method concerning
derivative of vertical displacement of the unsparing mass aircraft mass, traffic volume, and subgrade strength. Based
in meters. In this way, an index having slope units (e.g., on Westergaard’s central loading stresses, it is calculated

13
 Table 1  Types of distress used to calculate PCI for flexible (Asphalt) Pavement (Shahin 1989)
1174

No Type Description Severity level classification Repair method

1 Alligator or fatigue cracking Repeated traffic loading Probability of Spalling development in the Patching, overlay, reconstruction
In parallel patterns and the interconnected crack area

13
cracks look like alligator skin
2 Longitudinal and transverse cracking Zigzag Pattern running parallel to the traffic Spalling or Spalling-like secondary cracking Crack Sealing, Partial-depth Patching
direction
Perpendicular to the pavement centerline
Occurrence is usually independent of traffic
loading
3 Joint reflection cracking A crack reflecting a crack or joint in a concrete A fine crack with the onset of spalling or Crack sealing, Repair, reconstruction
slab with an asphalt overlay spalling-like secondary Cracking
4 Patching and utility cut patching Considered a defect regardless of its current Slight spalling development along the patch Crack sealing, full-depth repatching, Patch
condition boundary distress repair
5 Raveling and weathering The result from a loss of asphalt binder and A dislodged coarse aggregate count of 10–30 Surface treatment
free/thaw cycle-induced dislodging of aggre- per 1 m2 area
gate particles from the pavement surface
6 Rutting Rutting occurs along wheel paths because of Average rutting depth of 13–25 mm Partial/Full-depth patch repair, overlay
compression or lateral movement of aggregate
under traffic load
7 Swelling Caused by freezing-induced 13 mm ≤ base to peak height < 25 mm Reconstruction
Swelling of subgrade materials or soil swelling Readily detectable by visual inspection
8 Block cracking Caused by asphalt shrinkage and daily tem- Little or no spalling development along Crack sealing, reconstruction of the affected area,
perature change cycle, independent of traffic unsealed cracks (> 6 mm) overlay
loading
9 Depression Depression is a specific portion of a pavement 1. Runway and rapid-exit Taxiway: Patching, reconstruction
of lower-than-normal elevation due to subsoil 13 mm ≤ Max depth < 25 mm
settlement or improper construction 2. Taxiway and aprons: 25 mm ≤ Max
depth < 51 mm
10 Polished aggregate Polished aggregate occurs when river gravel is No specific severity-related instructions are
Surface treatment or overly
used as aggregate or the aggregate Surface is provided
worn by repeated traffic loading
Too small a portion of the aggregate extends
above the pavement surface, or too small
several surface particles
11 Bleeding An excessive amount of asphalt cement, tars, or No specific severity-related instructions are No need for Repair, surface treatment, or overlay,
low air-void content is in the mix provided depending on the extent of distress
12 Corrugation Regular wavy ripples formed along the pave- 1. Runway and rapid-exit Taxiway: 6 mm < max Reconstruction
ment to the traffic direction depression < 13 mm
It may occur as a result of the combination of 2. Taxiway and aprons 13 mm < max depres-
an unstable asphalt mix and traffic loads sion < 25
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190
 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1175

following the Portland Cement Association (PCA) method

for concrete pavements. There are two ways of calculating
PCN- one based on experience in aircraft operations and
the other on pavement structure evaluation. The pavement
No repair or patching on the condition

No repair or patching on the condition

No repair or patching on the condition

evaluation method is divided into two types; one is based on
a back calculation of the pavement structure design method,
and the other involves measuring pavement response when
the load is applied. In the former, the thickness of each pave-
ment layer and subgrade, material properties, traffic volume,
and other details are first determined then the design method
Repair method

is drawn backward. With the latter, the number of passé to

failure is calculated from deflection based on non-destruc-
tive tests. This produces the allowable load of the pavement,
which is taken as the PCN.
No specific severity-related instructions exist.
Only the presence of distress documented

No specific severity-related instructions are

No specific severity-related instructions are

4 Artificial Intelligence (AI) Methods

Artificial intelligence (AI) methods are based on assump-

tions, pre-defined algorithms, and consequences. This
Severity level classification

method is appropriate when it comes to the point where a

direct mathematical relationship between cause and effect
cannot be determined. For distress identification and quan-
tification, AI procedure has been used very long and the
application area is extensive. Figure 1 shows the models and
provided

provided

approaches that are applicable to apply. Machine learning

and image processing method are the two most useful pro-
cedures that have reached great success by choosing image
analysis methodology for the best outcome of implementing
or the bond between the Surface and the layer
to contamination by aircraft fuel, oil, or other

Occur when a loa-strength surface mix is used,

through sliding in the traffic direction under
Darkened as a result of burning or carboniza-

The Surface becomes smooth or worsens due

Crescent or half-moon-shaped cracks formed

tion of the bituminous binder by aircraft jet

various computer vision tasks by means of higher accuracy

and precision. These techniques have been widely used for
the detection and measurement of different types of distress
and defects appearing on the surface of pavement structures.
beneath the Surface is poor

4.1 Traditional Machine Learning Algorithms

Distresses can be assessed using traditional Machine Learn-

wheel loading

ing algorithms in qualitative and quantitative methods. Ran-

Description

dom Forest (Li et al. 2019a), Boosting (Cord and Chambon

liquids

2012) and decision tree, Artificial Neural Network based

blast

Analysis (Vyas et al. 2021a, b; Liu et al. 2020, Meignen et al.

1997), and Entropy-based fuzzy SWOT decision-making
and other methodologies are included in this section. The
most advantages of machine learning in pavement condition
evaluation are their robust learning algorithms which lead
to the ability to extract rules and features (i.e., Distress of
Pavements) from pavement datasets (Sholevar et al. 2022).
15 Slippage cracking

Using ML has shown a better performance result in terms of

13 Jet blast erosion
Table 1  (continued)

accuracy and computational time compared to conventional

14 Oil spillage

image processing techniques because they effectively learn

so many features and rules essential in detecting pavement
No Type

distress due to their complicated patterns. Table 3 shows the

13
 Table 2  Types of distress used to calculate PCI for rigid (Concrete) pavement (Shahin 1989)
1176

No Type Description Severity level classification Repair method

1 Linear cracking Accumulated fatigue caused by repeated traffic Non-filled cracks with average crack width > 3 mm Crack filling
loading

13
Shrinkage stress due to changes in temperature and Filled cracks in Poor condition
moisture
2 Spalling The breakdown of a slab corner or Joint Broken pieces likely to become loose (slight FOD Partial-depth Patching
risk)
Secondary cracks cause spalling and slight spalling
around the cracks with FOD risk
3 Patching Post-patching deterioration is attributable to an A partial secondary cracking condition requiring Repatching
inadequate mix design for the patch material, FOD preventive treatment
construction error, or insufficient bounding
between the original and new pavement areas
4 Durability crack (D-crack) Caused by environmental factors such as freeze/ Spalling or loss of pieces in the cracks partly Crack sealing, joint reconstruction
thaw cycle covering the slab
5 Scaling map crack/ crazing, Caused by freeze/thaw cycling and involves the Progression of fine cracks on a small portion Full-depth Repair, Slab replacement
shrinkage cracking breakdown of the slab surface (< 5%) of the slab surface that may result in
Poor curing of concrete occurs only on the slab scaling
surface
6 Corner cracking A Triangle-shaped crack that penetrates the entire Low-severity cracks with slight spalling or second- Crack Sealing, Repair slab replacement
slab thickness at the intersection of longitudinal ary Cracking (a slight FOD risk)
and transverse joints
7 Intersecting cracking Cracking a slab into four or more pieces is caused The slab has broken into six or more pieces with Crack sealing, Full-depth Repair, Slab replacement
by overloading or insufficient support low-severity cracks (> 85%) with little spalling
The highest severity level of this distress is consid- The slab has broken into 4 or 5 pieces with
ered a shattered slab M-severity cracks (> 15%) and No H-severity
cracks that may lead to FOD
8 Joint seal damage The accumulation of foreign objects causes joint or Overall fair condition throughout the sample unit Crack sealing, Full-depth repair, Slab replacement
moisture infiltration into joints Preventive joint resealing required
Characterized by the hardening, loss of bond, or Joint sealants are still in place, but a noticeable gap
stripping of joint sealant material (< 3 mm) that allows water infiltration
9 Alkali silica reactivity (ASR) It refers to the formation of gel-type material in Cracks on a small portion of the slabs (≤ 5%) have Slab replacement
concrete when the alkalis in cement react with a slight FOD risk
the silica in aggregate
This gel formation includes the slab to swell Noticeable trace of ASR at the slab Surface
because of its high moisture absorption capacity
10 Settlement or faulting A difference in elevation across a joint or crack Runway/taxiway: 6 mm ≤ elevation difference < 13 Diamond grinding
due to curling or loss of support mm
Aprons: 13 mm ≤ elevation difference < 23 mm
11 Pumping Pumping is the ejection of water that infiltrates The definition of separated pumping severities is Sub-sealing and patching
through joints or cracks under traffic loads not provided
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190
 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1177

details of detection using traditional ML models, adopted by

Sholevar et al (2022).
Before emerging deep learning algorithms, traditional
machine learning methods such as support vector machines

Partial/Full-depth repair, reconstruction

and decision-tree procedures were extensively employed
in image analysis and pattern recognition (Sholevar et al.
2022). The advantages of these models are that they are rela-
tively less dependent on a high volume of learning data. For
the past two decades, many researchers have been develop-
No need for repairs

ing pavement distress detection and recognition algorithms

along with improved AI and laser lighting, however, fully
Repair method

automated pavement distress detection and classification in

a real-time environment has remained a challenge. Deploy-
ment of ML for automated pavement distress recognition is
not novel anymore, but using the Deep learning method is an
alternative to fulfill the needs of distress identification, and
considered. The distress severity levels are not

classification, and to quantify the distresses based on object

Aprons 25 mm ≤ elevation difference ≤ 51 mm
It occurs when a small piece of pavement becomes Only an extensive distribution of pop-outs is

Runway/Taxiway: 13 mm ≤ elevation differ-

detection and segmentation. (Majidifard et al. 2020). Image

segmentation is the crucial first step of an automated pave-
ment inspection system. The most promising approaches
for automated distress analysis, leverage image processing
and computer vision algorithms to detect edges of differ-
Severity level classification

ent types of distress and severities of the cracks in pave-

ment images (Majifard et al. 2020). Nhat-Duc et al. (2018),
applied the AI method for Pothole Detection of Asphalt
ence ≤ 25 mm

pavement using Least Square SVM and ANN technique

with steerable Filter-based feature extraction. The author
defined

found that both SVM and ANN method has the ability to
detect pothole distress with an accuracy level of classifi-
cation of more than 85%. Zhao et al. (2021) introduced a
materials, such as stones, into the joints or cracks
loose due to the breakdown of the pavement sur-
face by aggregate swelling or freeze/thaw action
It occurs at transverse joints or cracks that do not

detection method for asphalt pavement segregation, based

because of the infiltration of incompressible
have enough room to expand in hot weather

on the traditional algorithm “LBP” local binary pattern and

gray-level- co-occurrence matrix in combination with SVM.
The model results obtained show an accuracy of diagno-
sis of 94%. Sulistyaningrum et al. (2021), implemented a
multi-class Support Vector Machine (SVM) to detect alliga-
tors, Potholes, and Cracks. The authors used a cell size of
16 X 16 pixels during the Histogram of Oriented Gradients
feature extraction process with an accuracy of 78.85%. Cui
et al. (2015) proposed a method of detection of distress by
Description

using random decision forests. They have introduced color

gradient features at multiple scales commonly used in con-
tour detection in order to represent cracks with more details.
Oliveira and Correia (2013) developed an automatic road
crack detection model using CrackIT. Shi et al. (2016) pur-
posed CrackForest, a road crack detection framework based
on random structured forests. Li et al. (2019b) studied the
performance of different models in the detection of Asphalt
Table 2  (continued)

pavement such as potholes and cracks based on UAV mul-

tispectral imagery by using machine learning algorithms
12 Pop-outs

13 Blowup

such as SVM, ANN, and random forest. Similar research

No Type

was conducted by Li et al. (2019a), for identifying Asphalt

pavement distress using UAV LiDAR point cloud data and

13
1178 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

Fig. 1  Image analysis method

Classifica on

Deep Learning Object Detec on

Seman c
Segmenta on
Machine Learning

Image Analysis
SVM

Decision Tree and

Tradi onal ML Rondom Forest
Algorithms
Boos ng
Edge Detec on
ANN
Image Processing Thresolds

Manual Filters

random forest classification. Issa et al. (2021) developed of any arbitrary size by parsing the image block by block
classical and Neural Networks method for modeling PCI and clarifying each block into target classes, image clas-
using cascade architecture. Gogoi and Dutta (2020) from sification models consist of CNNs, with easy-to-understand
India studied the maintenance prioritization of interlocking architecture to extract features or patterns from images.
concrete block pavement by applying fuzzy logic. These CNN models are the base for other learning tech-
niques such as object detection and image segmentation.
4.2 Deep Learning Algorithms Typically, these models have a high speed during the train-
ing and testing phases. The output results from the image
The advantage of DL models over more traditional ML algo- classification method do not directly indicate the distress
rithms and other IP methods is the data-driven end-to-end location, severity, and density. It is only giving an idea that
learning process that eliminates the need for handcraft fea- indicated whether or not distresses are present on the pave-
tures engineering leading to higher accuracy, more acces- ment. (Sholevar et al. 2022).
sible and faster operation, and more generalized (Sholevar
et al. 2022). The recent progress in DL has directed sub- 4.2.1.1 Binary Classification/Multi‑class Classification Classi-
stantial improvements in the ability to analyze the stream of fications models have been developed to recognize particular
videos and images at unprecedented accuracy (Majidifard distress types which exist or do not on the surface of the pave-
et al. 2020). Due to limitations that traditional Pavement ment structure. Using the Binary Classification CNN method
condition indices and techniques have, researchers have been is applied to classify images acquired from smartphones that
deploying different research and proposed new approaches indicate a crack or no crack (Zhang et al. 2016a, b). CNN-
to analyzing pavement condition indices and optimizing the CDM model was developed for the detection of Cracking by
scope of accuracy and precision levels. (Nhar-Duc et al. 2018) using Edge detection algorithms such
as Sobel and Canny and illustrated 12% good improvements
4.2.1 Image Classification compared to the previous version. (Nguyen et al. 2018) devel-
oped a CNN network with 5 layers on public binary datasets
The image classification method was the first machine (i.e., CrackIT and JTIRC), with application and extraction of
vision problem solved by using ANN. This method basi- 40,000 patches from road images and shows a significant out-
cally assigns an image to pre-defined classes if the image put detection with an accuracy level of 99.92 percent. Some
includes a certain amount of target class pixels. This method researchers used pre-trained models such as VGG-16 architec-
applies the sliding window technique to analyze images ture on the ImageNet database and they fine-tuned the model

13
 Table 3  Details of distress detection using traditional ML models (Sholevar et al. 2022)
Authors Model Distress types Task Dataset # of Inputs Input size Evaluation matrix % Hardware Framework
Precision Recall FI

Li et al. (2019a) MFCD C SE CDF 118 – 89.9 89.47 88 Core i5 4200U, MATLAB
APR 33 65.88 79.62 69.8 4GB
Ai et al. (2018) PGM, SVM, C SE CFD 118 – 90.7 84.6 87 C: 3.7 GHz, R: OpenCV
WDO 32GB
Li et al. (2019b) SMFB, IP ANN, C SE Private 200 2048 × 4096 88.38 93.15 90.6 C: 3GHz, R: 32 MATLAB
SVM 88.38 GB
92
91
Ahmadi et al. KNN, Tree LC, TC, DG, BC CL Private/ iPhone6 400 1334 × 750 88 MATLAB
(2022) Ensemble 85
89
Cao et al. (2021) LSSVC, SVM rutting CL Private 400 32 × 32 99 98 98 C:3.7GHz, –
Accuracy = 98 R:16GB
Pan et al. (2017) ANN, RF P, C CL Private, UAV 2000 – Accuracy = 98 C:2.6GHz, G: –
captured Quadro
Accuracy = 98 M1000M,
R:16GB
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

Oliveira et al. Crack IT C SE public – 75 × 75 92 95 93 C: 2.6GHz, - MATLAB

(2013) R:4GB
Moussa and Hus- SVM C SE private 87 – 97
sain (2011)
Lin and Liu SVM P CL private 80 64 × 64 97 C: 2.13 GHz, –
(2010) R:4GB
Pan et al. (2017) KNN, SVM, C, P CL private 126 feature extraction 98 –
ANN, RF 98
98
98

C—CPU; G—GPU; R—RAM; LC—longitudinal Crack; TC—transversal crack; DC—diagonal crack; BC—block crack; C—crack; P—pothole; SE—segmentation; CL—classification;
WDO—weighted dilation operation; SMFB—steerable matched filter bank; IP—image processing technique; LSSVS—least square support vector classification; RDG—random decision forests

13
1179
1180 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

on some particular cracking datasets. (Gopalakrishnan et al. models apply the classification task on each pixel and fine
2017; Simonyan and Zisserman 2015; Deng et al. 2009). This pre-defined object at the pixel level so they can be used to
pre-trained database with the model tunning such as VGG16, determine pavement distress severity and density. Two com-
ResNet34, and ResNet101, indicates an achieving result to mon architectures for segmentation models are the encoder-
a deeper ResNet with a superior network (Riid et al. 2019). decoder and fully convolutional networks (FCN). In crack
Wang and Hu (2017) used grid-based pavement crack analysis segmentation the severity of the distresses is detectable
and also applied the CNN method for the detection of distress. although there is no ability to classify them into different
The author used the CNN approach in order to sketch and classes or groups. U-net and crack U-net are the most popu-
build diagrams for cracks. (Eisenbach et al. 2017) performed lar encoder model used for the segmentation of defects in
to compare four models including (RCD-net, ASINVOS-Mod imaging. U-Net is the most popular encoder-decoder model
with RCD-net, and Crack-pot Net) to classify six types of developed by (Ronneberger et al. 2015) for medical pur-
pavement distresses. The authors used the GAPs V2 model poses. Huyan et al. (2020); Jenkins et al. (2018) provided
to assess the result with ResNet output. It was shown a very the application of a model using U-Net based for cracks
significant improvement compared to ResNet Architecture segmentation. Lau et al. (2020) applied and used pre-trained
instead of VGG-type networks. Shah and Deshmukh (2019) ResNet-34, for U-Net-based crack detection considering the
employed a pre-trained (ResNet-50) model and fine-tuned it pixel level on the image. Majidifard et al. (2020) proposed
to classify some distresses such as (Pothole and Bumps) using using U-net and YOLO back-to-back to segment the out-
the CNN method. Fan et al. (2018) proposed an automatic put bounding boxes of YOLO to quantify cracks severity
pavement detection model based on structured prediction level, the author used this procedure and developed a new
based on multi-label classification CNN for crack prediction. method called PASER. Zhang et al. (2017) proposed Crack-
Net and later a modified version called Crack-Net-R. Chen
4.2.2 Object Detection et al. (2020b, a) developed PCSN with pre-trained VGG16
as its encoder. Yang et al. (2018) applied FCN for crack seg-
The object detection model can identify and recognize an mentation in which a model was trained on 800 images with
object’s location in an image (Sholevar et al. 2022). R-CNN, diverse crack widths. Although the results were inferior to
You Look Only Once “YOLO” is both a practical and use- Crack Tree, the training time significantly decreased (from
ful deep learning model which belongs to ML groups that nine hours to less than 1 hour). Song and Wang (2021) used
categorize distress. The classification process is easier than a “multiscale dilated convolutional module,” which helps
segmentations (Majidifard et al. 2020). Anand et al. (2018). the network learn complex features. The author purposed the
Introduced a real-time crack and pothole detection model. Crack-Seg model which helps to detect a crack. Yang et al.
Majidifard et al. (2020) used YOLO v2, to detect nine types (2020) introduced feature pyramid and hierarchical boosting
of distress from 7500 images extracted with Python-based network (FPHBN) to detect pavement cracks by integrat-
coding from the database. The author used U-net, a pre- ing low-level and high-level features using a hierarchical
trained model to quantify the distresses. Majidifard et al. pyramid architecture. Liu et al. (2021) proposed a modified
(2020) applied both YOLO v3 and ResNet models for Faster R-CNN by adding a parallel segmentation branch to
object detection and data extraction from the street view perform pixel-vise classification on the bounding boxes. The
images dataset. The author used to extract and locate eight model was trained to detect Cracks, Potholes, and Patches.
types of distress in this procedure. Ghosh and Smadi (2021)
employed both YOLO v3 and Faster R-CNN to detect vari- 4.3 Crack Detection Based on Image Processing
ous distress types. Song and Wang (2021) deployed Faster
R-CNN to detect cracks, potholes, dot surfaces, and oil Yi Yang et al. (2014) have proposed crack detection algo-
bleeding. The authors trained 20 models with the same rithms using digital IP processing, similarly, Crack detec-
structure and different configurations to select the best tion based on IP was critically assessed by researchers from
model. Hu et al. (2021) developed a series of YOLO with India. A review article was produced by Zakeri et al. (2017)
four different initial weights provided by YOLO v5, and all Image-based technique for crack detection, classification and
models showed above 85% accuracy. quantification in asphalt pavement. They showed a general
architecture for crack detection based on IP. The steps in the
4.2.3 Semantic Segmentation IP technique are as follows: (1) initially collect the image
of the structure which will be subjected to the crack detec-
These models can express the object size along with its tion process using the camera (2) pre-processed (3) Image
details of location and quantity. The segmentation pro- processing (4) crack detection and (5) parameter estimation
cess provides the measurement of distress severity and (Mohan and Poobal 2018; Lins and Givigi 2016). Adhikari
density while others only detect and classify them. These et al. (2014) developed a model to represent numerically the

13
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1181

defect measurements. Also, an automatic procedure for crack and limitations. Researchers are exploring new ideas for
detection was developed by (Zou et al. 2012; Salman et al. easy interpretation and analysis to use for pavement manage-
2013; Shan et al. 2016; Yamaguchi and Hashimoto 2010; ment system. The literature review analyzes and determines
Talab et al. 2016; Yang et al. 2015; Oliverira and Correia the previous work done by the researchers to identify the
2012; Lins and Givigi 2016; Li et al. 2014; Lee et al. 2009; most critical condition in their scope of work considering the
Jahanshahi and Masri 2012; Gunkel et al. 2012; Fujita and functional and structural classification performance activity
Hamamoto 2011; Glud et al. 2016; Shan et al. 2016; Kim in particular for airports pavement.
and Ryy 2014; Zhang et al. 2016b). Sinha et al. 2006; inves- This systematic literature review represents the follow-
tigated the cracks by using a two-step statistical approach ing main topic, which is associated with the subjects listed
to filter design for crack detection. A 2D image of concrete as follow:
types cracks using an edge detection model was proposed by
Nguyen et al. (2014). The challenges related to the IP meth- 1. Automatic detection of distresses and measurement.
ods are that these procedures mostly rely on image quality 2. Development of prediction model for pavement condi-
and image configurations. Camera resolutions and installa- tion index using AI.
tion of the system in a static position play an important role 3. Technologies and applications used for inspection and
to collect the images for processing with higher accuracy of survey (Manual & Automatic).
measurement. Table 4 provides details for different image 4. Distress data collection and assessment using image
processing models that are applicable to detect and assess analysis procedure.
distress. 5. Types of distress that are under focus while using tradi-
tional and IA methods.

5 Model Comparisons Analysis For this purpose, the Google Scholar search database
was used to identify the most relevant papers. The search
Table 5 provides the model comparisons based on traditional options contain the article’s title, abstract, and keywords
and AI methods. It includes quantitative and qualitative from scientific journals, conferences, and other sources of
approaches including artificial intelligence such as machine various publications, such as book chapters. The database
learning, deep learning, and image processing method that includes the methodology and finding segments that define
enable us to distinguish the different models based on data the work written by the particular research authors, which
collection, advantages, and limitations. For pavement evalu- guide the systematic review. These selection systems based
ation, there are different indices that are utilized to charac- on different categories of research studies were applied to
terize pavement conditions. The pavement condition index distinguish the studies conducted by various organizations
(PCI), structural condition index (SCI), international rough- and published by reputed journals in specific areas of state
ness index (IRI), pavement rehabilitation index (PRI), and of the art such as subject, methodology, variables/factors,
foreign object debris (FOD) are quantitative-based pave- and techniques employed.
ment conditions indices, while present serviceability index,
present serviceability rating and pavement surface evalua-
tion and rating can be grouped as qualitative based. The AI 7 Qualitative Overview
mode is a separate procedure that has both quantitative- and
qualitative-based techniques advantages. The pavement evaluation and distress identification for the
Pavement Management System require considerable work
and effort. The time, money, and resources used for all these
6 The Review Procedure performances, such as (Manpower and equipment) includ-
ing programs and technologies for support, are essential
This review paper outlines the work involved in identifying to regulate the functional and structural condition of the
different types of distress induced by traffic and environmen- pavement structure and work data. From the papers studied
tal aspects. This review article describes the methodology in this review, it is concluded that most of researchers are
applied for the assessment of distresses in pavement net- considering to apply innovative idea and implement new
works. The traditional pavement condition evaluation meth- technologies such as AI for comparing with traditional pro-
ods are based on a quantitative approach; however, Artificial cedures to improve the accuracy level. Although using new
intelligence such as machine learning, Deep learning and technologies and techniques requires Intelligent Technol-
imager Analysis are as discussed in details are following ogy system and smart application tools and equipment’s that
qualitative and quantitative means. Table 5 provides with the are more expensive to apply, however, safety measures and
comparison of the models with describing the advantages speed of the operation are reducing comparatively by using

13
 1182

13
Table 4  Different IP methodology for distress detection and assessment
Authors Methods of IP Application /area Remarks

Martin et al. (2016); Broberg (2013); Brooks Infrared (IR) based image processing Image rectification with the extraction of The IR method additionally provides geomet-
et al. 2015; Pei et al. 2015; Xu et al. (2014); isotherms ric characterizations and orientation of the
Guo et al. (2013); de Moura et al. (2021) cracks. brooks et al. used the laser-excited
thermography method. xu et al. proposed
super-pixel to detect cracks. Guo et al. use
ultrasonic IR thermography
Dhital et al. (2012); Pascale et al. (2015); Ultrasonic-based image processing Uses ultrasonic laser scanning excitation and UFT and WUPI algorithms were used to extract
Shirahata et al. (2014), Wolf et al. (2015); piezoelectric air-coupled sensing technique damaged features. 96% accuracy level. Pas-
Iliopoulos et al. (2015); cale et al. used this method to determine the
depth of cracks by 20 mm
Sun et al. (2012); Mathavan et al. (2015), Laser-based image processing Detection of cracks in the structure The capability of measuring defects using 3d
Nazaryan et al. (2013); Merazi-Meksen et al. space
(2014)
Anwer et al. (2014); Electroluminescence (EL) Optoelectrical The filtering algorithms used in this procedure
showed an accuracy level of 90% from the
results
Pereira et al. (2015) UAV camera image Unmanned aerial vehicle is used for automa- Used based on filtering algorithms and also
tion purposes provide with particle filter that relates the
probability of an image to characterized and
segment distresses based on pixel intensity
Vidal et al. (2016); Chen et al. (2012); Oliveri Scanning electron microscopy, probe image, Different Techniques based on IP processing Edge detection, thresholding, and filtering
et al. (2015); Agarwal et al. (2015); Hei- microwave image, radar image, sensor
deklang et al. (2015); Xue-jun et al. (2013); image, video image,
Salari and Bao (2011)
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190
 Table 5  Traditional model comparisons with artificial intelligence procedures for distress analysis
Mode Model Method Data collection type Advantages Limitations

Quantitative Pavement condition index “PCI” PCI (Manual/Automatic) Considering both flexible and rigid pavement Provide functional evaluation of the pavement network
All types of distresses are taken into account according to (Table 1, Surface layer is considered only
Table2) Dependent on the Visual observations and Data survey
Applicable to Road and Airfield Pavements Costly
The distress types, severity, and density have to be measured manually Time-consuming
Accuracy level of data collection is less due to human error Human- dependent
Structural condition index “SCI” SCI (Manual) Considering both flexible and rigid pavement Provide Structural Evaluation of the pavement network
Evaluate the bearing capacity of the Base, sub-base, and subgrade Dependent on the Visual observations and Lab data results
layer of the structure Costly due to Laboratory and on-site Testing /sampling
Applicable to Road and Airfield Pavements Time-Consuming
The CBR value for Flexible Pavement and K- Value for Rigid type Human- Dependent
pavement has to be determined Machinery and Lab Testing is a principal part of the
FWD data or DCP test results are more appropriate when determining evaluation activity
the back data calculation is required DT&NDT Data needed for Models development which is
Accuracy level of data collection is higher than PCI expensive and time-consuming
International roughness index “IRI” IRI (Manual) Provide with Network profile (roughness) Using the IRI method, the correct assessment of the
VOC can be obtained by Using IRI effect on aircraft operations of long wavelengths is not
Using the IRI procedure for the prediction of pavement condition possible
could eliminate the subjectivity of determining the PCI value IRI method locates roughness zone but it is not able to
Can be measured by automation using a road profiler including the adequately assess the effect on aircraft response
profile graph Time-consuming
Human-dependent
Advance Technologies are expensive
Pavement rehabilitation index “PRI” PRI (Manual/Auto) Investigate the pavement surface and structure Applicable to Japan region
Separate models are available for both Flexible and Rigid type pave- Used for Airport Pavement
ment
Models provide an index for the following Pavements:
Cracking Ratio, Roughness, and Rutting (Flexible) Pavement
Cracking Ratio, Ration of Joint Failure and SV(Faulting) (Rigid)
Pavement
Foreign object debris “FOD” FOD (Manual or Automatic) Provide excellent results output while using AI method for FOD Huge potential risk to aircraft during operations at Airside
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

detection at Airfield
Accuracy level of FOD detection is much higher compared to Manual
detection
Qualitative Present serviceability index “PSI” PSI (Manual) NIL NIL
Present serviceability rating “PSR” PSR (Manual)
Pavement surface evaluation & rating PASER (Manual)

13
1183
 Table 5  (continued)
1184

Mode Model Method Data collection type Advantages Limitations

Artificial intel- Traditional ML algorithm Regression (Manual or Automatic) Have been widely used in various Image processing and Image analy- Training and testing data set on private datasets has
ligence & image analysis sis methods for distress detection made it difficult to compare models using ML group
analysis Less dependent on a high volume of learning data algorithms
Ability to combine in all other methodology

13
Logistic
regression Provide Statistical result output

Decision tree
(LTTP)
Random forest
(RF)
SVM, LS-
SVM
Hybrid ML
ANN
boosting
Deep learn- Classifica- Binary CNN, Automatic/ IP Base Computer-Based Vision A pre-defined Database is required to train model
ing tion classifi- Cost-effective Limited types of distress are considered for analysis
method cation Consistent Decision-making Detect & Classify Distress appearing on the surface of the
Multi- Accurate, and precise results obtainable pavement structure without providing the severity level
class More accessible and Faster Operations and density measures
classifi- More Generalized Advanced Hardware System needed
cation Provide an end-to-end learning Process that eliminates handcraft Higher Resolutions Camera or Technology is essential
features while collecting data Output results are not details in terms of distress
High Speed at training and testing phases Dependency on Training Data
Data preparation can be easily handled
Object detection R-CNN Automatic/ IP base Models can recognize the location of an object along with the types Complex architecture
method You Look of distress Time-consuming (training)
Only Once Simple, and has the feature of an extractor
“YOLO” Has the ability to find and recognize distress in a large Image
Preparation of the dataset is simple for training models
SSD
Semantic segmentation Crack U-Net / Automatic/ IP base Provide the quantity of distress along with the location and presence Low Speed in training and testing
method Crack-Net- Able to measure distress severity and density Effortful data preparation
R/ Crack Provide best result performance
Tree
U-net Based
Algorithms
FBHPN
Crack-Seg
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190
 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1185

traditional method. It was found that most of researcher are

Performance deteriorates when the image has shadow and

These methods will not perform well when the image set
The camera image-based detection, the low resolution of
applying ML and DL for better consequences of the model
predations, yet the introducing all the distresses categories
are not included in the proposed models. Until recently few
works have been done on the measurements and analysis of
various types of distress to address, specifically the issue
the image is a factor in mislead

is diverse and heterogeneous of fine-grained classification and assessment of distress

Computation time is very long

More expensive technology pavement. The adoption of recent advances in AI has histri-
onically enhanced to strength of distress classification and
segmentations of the object (distress), however, despite the
lighting defects

great achievements of using ML and DL including Image

Limitations

Analysis, existing of noise, environmental affect factors and

loading configurations of different types of vehicles that uses
the pavements Areas have not been considered. For Image
analysis the existence of poor lighting and shadow including
low contrast has made it very challenging for the researchers
to develop such models that capable of detecting and assess-
ing distress in such environment.

8 Summarizations of Papers Dealing

with MODEL Categories, Pavement Types
& Performance Analysis Using Traditional
Method

The following pile charts represent a summary of the

research article conducted by different authors considering
High accuracy

the Pavement Types such as Flexible, Rigid, and Composite

Advantages

High safety

Realistic

type structures for their studies to determine the Pavement

condition. It shows that 39% of the data demonstrated the
Rigid type, and 35% of overall highlighted Flexible type
pavement illustrated in Fig. 2. The Performance Analysis
Data collection type

shown in Fig. 3 indicated that 50% of total statistics repre-

sent Functional Performance Analysis, followed by 19% with
IP based

Structural and 31% for Structural and Functional criteria.

The studied model categories are Empirical, Mechanistic,
and Empirical-mechanistic models that have been verified
shown in Fig. 4. It shows that 54% of the studies are based
Method

on the Empirical-mechanistic model category, followed by

Edge detection

Manual filters

33% and 13% of Empirical and Mechanistic models, respec-

Thresholds

tively. For all these research topics, the first essential data
set is to conduct a survey and collect distress data required
for assessing the pavement structure. It is noticeable that a
large number of variables may be analyzed in the decision-
making of the maintenance programs, which may affect the
objectives proposed by researchers. Different countries fol-
Image pro-

low change policies to determine and apply a cost-effective

cessing
Model

maintenance performance to run the operation without delay.

Table 5  (continued)

To this extent most of the researchers that have been used the
AI tools have not discussed about the factors that have been
considered in their model while assessing the distresses. A
lot of studies have been focusing on detection of Crack type
Mode

13
1186 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

PAVEMENT TYPES PERFORMANCE ANALYSIS

19%
26% 32%
39%

35% 49%

Structural & Funconal Funconal Structural

Rigid Flexible Composit
Fig. 3  Performance analysis models
Fig. 2  Pavement structure types

distress including Potholes and rutting on the surface of

pavement structures. Mechani
scs
13% Empirical
-
9 Conclusion Empirical Mechani
33% scs
The assessment conducted in this research article is intended 54%
to summarize and guide researchers on pavement evalua-
tion. To define specific terms and follow the right path, it
is necessary to implement the rule, assign the most reliable
tools, and describe variables very specifically to calculate
the distresses. Different categories of distress are listed and
Fig. 4  Studied models categories
categorized for both flexible and rigid Pavement with an
explanation of the type, severity level, and measurement
procedure that have been followed by many researchers by considering all the types of distresses and development of
applying traditional method. The different methods applied performance response model by different types of pavement
in various countries and practices for conducting surveys at categories using Image Analysis methods or any AI based
network and project levels are dissimilar. As a result of the techniques such as Machine learning or deep learning that
review, we can outline the following points that are present would be helpful to all users.
in the bibliometric analysis: (1) The subject of the study
Acknowledgments The authors gratefully acknowledge and thank the
area that most researchers are trying to explore in particular
Service Center of Port Engineering (SCOPE) and, in particular, Yoshi-
using Machine Learning, Deep Learning method and Image taka Hachiya for providing technical support and material for the study.
Analysis for detection and measurement of Distresses; (2)
Development of new prediction models based on AI for Declarations
assessment and analysis; (3) Comparisons of Traditional
Conflict of interest The authors declare that they have no known com-
Method and Artificial Intelligence method, describing the
peting financial interests or personal relationships that could have ap-
advantages and limitations; (4) Finally, it was concluded peared to influence the work reported in this paper.
with a qualitative overview summarizing the outcomes for
predicting modeling and methodologies that are applicable Ethical Approval This article does not contain any studies with human
participants or animals performed by any of the authors.
to be applied in pavement evaluation and performance analy-
sis modeling using Traditional method. Within this context, Informed Consent For this type of study, formal consent is not
it is suggested that future studies of the pavement distress required.
analysis procedure for detection and measurement modeling

13
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1187

References De Moura IR, dos Santos Silva FJ, Costa LHG, Neto ED, Viana HRG
(2021) Airport pavement evaluation systems for maintenance
strategies development: a systematic literature review. Int J
Adhikari RS, Moselhil O, Bagchi A (2014) Image-based retrieval of
Pavement Res Technol 14(6):676–687. https://​doi.​org/​10.​1007/​
concrete crack properties for bridge inspection. Autom Constr
s42947-​020-​0255-1
39:180–194
Deng J, Dong W, Socher R, Li L-J, Li K, Fei-Fei L (2009) ImageNet:
Agarwal S, Singh D (2015) An adaptive statistical approach for non-
a large-scale hierarchical image database. In: 2009 IEEE confer-
destructive underline crack detection of ceramic tiles using
ence on computer vision and pattern recognition pp 248–255,
millimeter wave imaging radar for industrial application. IEEE
https://​doi.​org/​10.​1109/​CVPR.​2009.​52068​48
Sens J 15(12):7036–7044
Dhital D, Lee JR (2012) A fully non-contact ultrasonic propagation
Ahmadi A, Khalesi S, Golroo A (2022) An integrated machine learn-
imaging system for closed surface crack evaluation. Exp Mech
ing model for automatic road crack detection and classification
52(8):1111–1122
in urban areas. Int. J. Pavement Eng 23(10):3536–3552. https://​
Eisenbach M, Stricker R, Seichter D, Amende K, Debes K, Sesselmann
doi.​org/​10.​1080/​10298​436.​2021.​19058​08
M, Ebersbach D, Stoeckert U, Gross HM (2017) How to get
Ai D, Jiang G, Siew Kei L, Li C (2018) Automatic pixel-level pavement
pavement distress detection ready for deep learning? A system-
crack detection using information of multi-scale neighborhoods.
atic approach. In: proceedings of the international joint confer-
IEEE Access 6:24452–24463. https://​doi.​org/​10.​1109/​ACCESS.​
ence on neural networks, pp 2039–2047. https://d​ oi.o​ rg/1​ 0.1​ 109/​
2018.​28293​47
IJCNN.​2017.​79661​01, 2017-May
Anand S, Gupta S, Darbari V, Kohli S (2018) Crack-pot: autono-
Fan Z, Wu Y, Lu J, Li W (2018) Automatic pavement crack detection
mous road crack and pothole detection. In: 2018 digital image
based on structured prediction with the convolutional neural,
computing: techniques and applications (DICTA), IEEE pp
Network 1–9. http:// arxiv.org/abs/1802.02208
1–6.https://​doi.​org/​10.​1109/​DICTA.​2018.​86158​19
Fujita Y, Hamamoto Y (2011) A robust automatic crack detec-
Anwar SA, Abdullah MohdZaid (2014) Micro-crack detection of
tion method from noisy concrete surfaces. Mach vis Appl
multicrystalline solar cells featuring an improved anisotropic
22(2):245–254
diffusion filter and image segmentation technique. EURASIP
Ghosh RR, Smadi O (2021) Automated detection and classification
J Image Video Process 2014:1–7
of pavement distresses using 3D pavement surface images and
Arhin SA, Williams LN, Ribbiso A, Anderson MF (2015) Predicting
deep learning. Transp Res Record 9:1359–1374. https://​doi.​org/​
pavement condition index using international roughness index
10.​1177/​03611​98121​10074​81
in a dense urban area. J Civ Eng Res 5(1):10–17. https://​doi.​
Glud JA, Dulieu-Barton JM, Thomsen OT, Overgaard LCT (2016)
org/​10.​5923/j.​jce.​20150​501.​02
Automated counting of off-axis tunneling cracks using digital
Broberg P (2013) Surface crack detection in welds using thermog-
image processing. Compos Sci Technol 125:80–89
raphy. NDT E Int. 57:69–73
Gogoi R, Dutta B (2020) Maintenance prioritization of interlock-
Brooks WSM, Lamb DA, Irvine SJC (2015) IR reflectance imaging
ing concrete block pavement using fuzzy logic. Int J Pave-
for crystalline Si solar cell crack detection. IEEE J Photovolt
ment Res Technol 13(2):168–175. https://​d oi.​o rg/​1 0.​1 007/​
5(5):1271–1275
s42947-​019-​0098-9
Camarena KA, Eng B, Flores L (2018) Proposal of numerical model
Gopalakrishnan K, Khaitan SK, Choudhary A, Agrawal A (2017)
for airport pavement management purposes. https://​doi.​org/​10.​
Deep convolutional neural networks with transfer learning for
18687/​LACCE​I2018.1.​1.​417
computer vision-based data-driven pavement distress detection.
Cao M-T, Chang K-T, Nguyen N-M, Tran V-D, Tran X-L, Hoang N-D
Constr Build Mater 157:322–330. https://​doi.​org/​10.​1016/j.​
(2021) Image processing-based automatic detection of asphalt
conbu​ildmat.​2017.​09.​110
pavement rutting using a novel metaheuristic optimized machine
Greene J, Shahin MY, Alexander DR (2004) Airfield pavement condi-
learning approach. Soft Comput 25:12839–12855. https://​doi.​
tion assessment. Transp Res Rec 1889(1):63–70
org/​10.​1007/​s00500-​021-​06086-5
Gunkel C, Stepper A, Mu¨ller AC, Mu¨ller CH (2012) Micro crack
Chen X, Michaels JE, Lee SJ, Michaels TE (2012) Load-differential
detection with Dijkstra’s shortest path algorithm. Mach vis Appl
imaging for detection and localization of fatigue cracks using
23(3):589–601
Lamb waves. NDT E Int 51:142–149
Guo X, Vavilov V (2013) Crack detection in aluminum parts by using
Chen H, Yao M, Gu Q (2020a) Pothole detection using location-
ultrasound-excited infrared thermography. Infrared Phys Technol
aware convolutional neural networks. Int J Mach Learn Cybern
61:149–156
11:899–911. https://​doi.​org/​10.​1007/​s13042-​020-​01078-7
Guo L, Wang H, Gagnon J (2021) Comparison analysis of airfield pave-
Chen T, Cai Z, Zhao X, Chen C, Liang X, Zou T, Wang P (2020b)
ment life estimated from different pavement condition indexes. J
Pavement crack detection and recognition using the architec-
Trans Eng Part B: Pavements 147(2):04021002. https://​doi.​org/​
ture of segNet. J Ind Inf Integr. https://​doi.​org/​10.​1016/j.​jii.​
10.​1061/​JPEODX.​00002​54
2020.​100144
Hachiya Y, Watanabe T, Kitaochi K (2013) Some issues on surface
Coenen TB, Golroo A (2017) A review on automated pavement dis-
distress of airport pavements in Japan. Service Center of Port
tress detection methods. Cogent Eng 4(1):1374822. https://​doi.​
Engineering (SCOPE), Tokyo, Japan
org/​10.​1080/​23311​916.​2017.​13748​22
Hajj E, Loria L, Sebbaaly P, Borroel C, Leiva P (2011) Optimum time
Cord A, Chambon S (2012) Automatic Road defect detection by
for application of slurry seal to asphalt concrete pavements,
textural pattern recognition based on AdaBoost. Comput Aided
Paper No 11–4071, United States
Civ Infrastruct Eng 27:244–259. https://​d oi.​o rg/​1 0.​1 111/j.​
Heideklang R, Shokouhi P (2015) Multi-sensor imagefusion at sig-
1467-​8667.​2011.​00736.x
nal level for improved near-surface crack detection. NDT E Int
Cui L, Qi Z, Chen Z, Meng F, Shi Y (2015) Pavement distress detec-
71:16–22
tion using random decision forests. In: Lecture notes in com-
Hu GX, Hu BL, Yang Z, Huang L, Li P (2021) Pavement crack detec-
puter science (Including Subseries Lecture Notes in Artificial
tion method based on deep learning models. Wirel Commun
Intelligence and Lecture Notes in Bioinformatics) pp 95–102
Mob Comput. https://​doi.​org/​10.​1155/​2021/​55735​90
https://​doi.​org/​10.​1007/​978-3-​319-​24474-7_​14
Huyan J, Li W, Tighe S, Xu Z, Zhai J (2020) CrackU-net: a novel
deep convolutional neural network for pixel-wise pavement crack

13
1188 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

detection. Struct Control Health Monit 27:1–19. https://​doi.​org/​ Pavements 148(1):04022001. https://​doi.​org/​10.​1061/​JPEODX.​
10.​1002/​stc.​2551 00003​42
ICAO (International Civil Aviation Organization) (2006) Aerodrome Loprencipe G, Zoccali P (2019) Comparison of methods for evaluating
design manual, 3rd Edition, Document (9157) airport pavement roughness. Int J Pavement Eng 20(7):782–791.
ICAO (International Civil Aviation Organization) (2016) Annex 14, https://​doi.​org/​10.​1080/​10298​436.​2017.​13455​54
Aerodromes-Volume 1, Aerodrome design and operations, Sev- Majidifard H, Adu-Gyamfi Y, Buttlar WG (2020) Deep machine learn-
enth Edition, July 2016 ing approach to develop a new asphalt pavement condition index.
Iliopoulos S, Aggelis DG, Pyl L, Vantomme J, Van Marcke P, Cop- Constr Build Mater. https://​doi.​org/​10.​1016/j.​conbu​ildmat.​2020.​
pens E, Areias L (2015) Detection and evaluation of cracks in 118513
the concrete buffer of the Belgian Nuclear Waste container using Mathavan S, Kamal K, Rahman M (2015) A review of three-dimen-
combined NDT techniques. Constr Build Mater 78:369–378 sional imaging technologies for pavement distress detection and
Issa A, Sammaneh H, Abaza K (2021) Modeling pavement condition measurements. IEEE Trans Intell Transp Syst 16(5):2353–2362.
index using cascade architecture: classical and neural network https://​doi.​org/​10.​1109/​TITS.​2015.​24286​55
methods, Iran. J Sci Technol Trans Civ Eng. https://​doi.​org/​10.​ Meignen D, Bernadet M, Briand H (1997) One application of neural
1007/​s40996-​021-​00678-9 networks for detection of defects using video data bases: identi-
Jahanshahi MR, Masri SF (2012) Adaptive vision-based crack detec- fication of road distresses, database and expert systems applica-
tion using 3D scene reconstruction for condition assessment of tions. In: 8th international conference, proceedings pp 459–464
structures. Autom Constr 22:567–576 https://​doi.​org/​10.​1109/​DEXA.​1997.​617332
Jenkins MD, Carr TA, Iglesias MI, Buggy T, Morison G (2018) A deep Merazi-Meksen T, Boudraa M, Boudraa B (2014) Mathematical mor-
convolutional neural network for semantic pixel-wise segmenta- phology for TOFD image analysis and automatic crack detection.
tion of road and pavement surface cracks. In: European signal Ultrasonics 54(6):1642–1648
processing conference pp 2120–2124. https://​doi.​org/​10.​23919/​ Mohan A, Poobal S (2018) Crack detection using image processing: a
EUSIP​CO.​2018.​85532​80 critical review and analysis. Alex Eng J 57(2):787–798. https://​
Karim FM, Rubasi KAH, Saleh AA (2016) The road pavement condi- doi.​org/​10.​1016/j.​aej.​2017.​01.​020
tion index (PCI) evaluation and maintenance: a case study of Moussa G, Hussain K (2011) A new technique for automatic detection
Yemen. Organiz Technol Manag Constr: Int J 8(1):1446–1455 and parameters estimation of pavement crack. In: 4th interna-
Kim T, Ryu SK (2014) Review and analysis of pothole detection meth- tional multi-conference on engineering and technological innova-
ods. J Emerg Trends Comput Inf Sci 5(8):603–608 tion, pp 11–16. https://​doi.​org/​10.​13140/2.​1.​3191.​2001
Lau SLHH, Chong EKPP, Yang X, Wang X (2020) Automated pave- Mubaraki M (2014) Identification of pavement distress types and pave-
ment crack segmentation using U-net-based convolutional neu- ment condition evaluation based on network-level inspection for
ral network. IEEE Access 8:114892–114899. https://​doi.​org/​10.​ the Jazan city road network. J Eng Res [TJER] 11(1):44–54
1109/​ACCESS.​2020.​30036​38 Mulry B, Jordan M, O'Brien DA (2015) Automated pavement condition
Lee SG, Mao Y, Gokhaleb AM, Harris J, Horstemeyer MF (2009) assessment using laser crack measurement system (LCMS) on
Application of digital image processing for automatic detection airfield pavements in Ireland. In: 9th International conference on
and characterization of cracked constituent particles/inclusions in managing pavement assets
wrought aluminum alloys. Mater Charact 60(9):964–970 Nazaryan N, Campana C, Moslehpour S, Shetty D (2013) Application
Li X, Jiang H, Yin G (2014) Detection of surface crack defects on fer- of a He–Ne infrared laser source for detection of geometrical
rite magnetic tile. NDT E Int 62:6–13 dimensions of cracks and scratches on finished surfaces of met-
Li Z, Cheng C, Kwan MP, Tong X, Tian S (2019a) Identifying asphalt als. Opt Lasers Eng 51(12):1360–1367
pavement distress using UAV LiDAR point cloud data and ran- Nguyen H-N, Kam T-Y, Cheng P-Y (2014) An automatic approach for
dom forest classification. Int Soc Photogramm Remote Sens accurate edge detection of concrete crack utilizing 2D geometric
(ISPRS) Int J Geo-Inform. https://​doi.​org/​10.​3390/​ijgi8​010039 features of crack. J Signal Process Syst 77(3):221–240. https://​
Li B, Wang KCP, Zhang A, Fei Y, Sollazzo G (2019b) Automatic doi.​org/​10.​1007/​s11265-​013-​0813-8
segmentation and enhancement of pavement cracks based on 3D Nguyen NTH, Le TH, Perry S, Nguyen TT (2018) Pavement crack
pavement images. J Adv Transp 2019:1813763. https://​doi.​org/​ detection using convolutional neural network. In: Proceedings of
10.​1155/​2019/​18137​63 the ninth international symposium on information and commu-
Lin J, Liu Y (2010) Potholes detection based on SVM in the pavement nication technology pp 251–256 https://​doi.​org/​10.​1145/​32879​
distress image. In: Proceedings of the 9th international sym- 21.​32879​49
posium on distributed computing and applications to business, Nhat-Duc H, Nguyen QL, Tran VD (2018) Automatic recognition of
engineering and science, pp 544–547. https://​doi.​org/​10.​1109/​ asphalt pavement cracks using metaheuristic optimized edge
DCABES.​2010.​115 detection algorithms and convolution neural network. Autom
Lins RG, Givigi SN (2016) Automatic crack detection and meas- Constr 94:203–213. https://d​ oi.o​ rg/1​ 0.1​ 016/j.a​ utcon.2​ 018.0​ 7.0​ 08
urement based on image analysis. IEEE Trans Instrum Meas Oliveira H, Correia PL (2012) Automatic Road crack detection and
65(3):583–590 characterization. IEEE Trans Intell Transp Syst 14(1):155–168
Liu G, Niu F, Wu Z (2020) Life-cycle performance prediction for rigid Oliveira H, Correia PL (2013) Automatic Road crack detection and
runway pavement using artificial neural network. Int J Pavement characterization. IEEE Trans Intell Transp Syst 14:155–168.
Eng 21(14):1806–1814. https://d​ oi.o​ rg/1​ 0.1​ 080/1​ 02984​ 36.2​ 019.​ https://​doi.​org/​10.​1109/​TITS.​2012.​22086​30
15679​22 Oliveri G, Ding P-P, Poli L (2015) 3-D crack detection in anisotropic
Liu C, Li J, Gao J, Gao Z, Chen Z (2021) Combination of pixel-wise layered media through a sparseness regularized solver. IEEE
and region-based deep learning for pavement inspection and Antennas Wirel Propag Lett 14:1031–1034
segmentation. Int J Pavement Eng. https://​d oi.​o rg/​1 0.​1 080/​ Pan Y, Zhang X, Sun M, Zhao Q (2017) Object-based and super-
10298​436.​2021.​18777​04 vised detection of potholes and cracks from the pavement
Liu S, Ling J, Wu Z, Li P, Zhang J (2022) A new set of indexes images acquired by UAV. Int Arch Photogramm Remote
and thresholds of preventive maintenance decision making Sens Spat Inf Sci. https:// ​ d oi. ​ o rg/ ​ 1 0. ​ 5 194/ ​ i sprs- ​ a rchi​
for airport runway composite pavement. J Transp Eng Part B: ves-​XLII-4-​W4-​209-​2017

13
Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190 1189

Pascale G, Lolli A (2015) Crack assessment in marble sculptures using Sharaf EA, Reichelt E, Shahin MY, Sinha KC (1987) Development
ultrasonic measurements: laboratory tests and application on the of a methodology to estimate pavement maintenance and repair
statue of David by Michelangelo. J Cultural Herit 16(6):813–821 costs for different ranges of Pavement Condition Index. Transp
Pei C, Qiu J, Liu H, Chen Z (2015) Simulation of surface cracks meas- Res Rec 1123:30–39
urement in first walls by laser spot array thermography. Fusion Shi Y, Cui L, Qi Z, Meng F, Chen Z (2016) Automatic road crack
Eng Des. https://​doi.​org/​10.​1016/j.​fusen​gdes.​2015.​11.​055 detection using random structured forests. IEEE Trans Intell
Pereira FC, Pereira CE (2015) Embedded image processing systems Transp Syst 17:3434–3445. https://​doi.​org/​10.​1109/​TITS.​2016.​
for automatic recognition of cracks using UAVs. IFAC-Paper- 25522​48
sOnLine 48(10):16–21 Shirahata H, Miki C, Yamaguchi R, Kinoshita K, Yaginuma Y (2014)
Qin Zou Yu, Cao QL, Mao Q, Wang S (2012) CrackTree: automatic Fatigue crack detection by the use of ultrasonic echo height
crack detection from pavement images. Pattern Recogn Lett change with crack tip opening. Weld World 58(5):681–690
33(3):227–238 Sholevar N, Golroo A, Esfahani SR (2022) Machine learning tech-
Ragnoli A, De Blasiis MR, Di Benedetto A (2018) Pavement distress niques for pavement condition evaluation. Autom Constr
detection methods: a review. Infrastructures 3(4):58. https://​doi.​ 136:104190. https://​doi.​org/​10.​1016/​jautc​on.​2022.​104190
org/​10.​3390/​infra​struc​tures​30400​58 Simonyan K, Zisserman A (2015) Very deep convolutional networks
Riid A, L˜ouk R, Pihlak R, Tepljakov A, Vassiljeva K (2019) Pave- for large-scale image recognition. In: 3rd international confer-
ment distress detection with deep learning using the ortho-frames ence on learning representations, (ICLR) 2015 —conference
acquired by a mobile mapping system. Appl Sci (switzerland). track proceedings pp 1–14. https://​arxiv.​org/ abs/1409.1556
https://​doi.​org/​10.​3390/​app92​24829 Sinha SK, Fieguth PW (2006) Automated detection of cracks in buried
Rodrı´guez-Martı´na M, Lagu¨ elaa S, Gonza´ lez-Aguileraa D, Mar- concrete pipe images. Autom Const 15(1):58–72
tinezb J (2016) Thermographic test for the geometric characteri- Song L, Wang X (2021) Faster region convolutional neural network for
zation of cracks in welding using IR image rectification. Autom automated pavement distress detection. Road Mater Pavement
Constr 61:58–65 Des 22:23–41. https://​doi.​org/​10.​1080/​14680​629.​2019.​16149​69
Ronneberger O, Fischer P, Brox T (2015) U-net: convolutional net- Suh YC, Park DY, Jeong KY (2002) Development of deterioration
works for biomedical image segmentation. In: Lecture notes in prediction models for rigid airfield pavements. Transp Res Rec
computer science (Including Subseries Lecture Notes in Artifi- 1788(1):132–137
cial Intelligence and Lecture Notes in Bioinformatics) 9351 pp Sulistyaningrum DR, Putri SAM, Setiyono B, Ahyudanari E, Oranova
234–241. https://​doi.​org/​10.​1007/​978-3-​319-​24574-4_​28 D (2021) Classification of damaged road types using multiclass
Salari E, Bao G (2011) Pavement distress detection and severity support vector machine (SVM). J Phys Conf Ser. https://​doi.​org/​
analysis. In: image processing: machine vision applications IV, 10.​1088/​1742-​6596/​1821/1/​012048
78770C. https://​doi.​org/​10.​1117/​12.​876724 Sun X, Huang J, Liu W, Mantao Xu (2012) 2012 Pavement crack char-
Salman M, Mathavan S, Kamal K, Rahman M (2013) Pavement crack acteristic detection based on sparse representation. EURASIP J
detection using the gabor filter. In: Proceedings of 16th inter- Adv Signal Process 1:1–11
national IEEE annual conference on intelligent transportation Talab AMA, Huang Z, Xi F, Hai Ming L (2016) Detection crack in
systems pp 2039–2044 image using Itsu method and multiple filtering in image process-
Sandra AK, Sarkar AK (2013) Development of a model for estimat- ing techniques. Opt-Int J Light Electon Opt 127(3):1030–1033
ing international roughness index from pavement distresses. Int Tarefder RA, Rahman MM (2016) Development of dynamic system
J Pavement Eng 14(8):715–724. https://​doi.​org/​10.​1080/​10298​ approaches to airport pavements maintenance. J Transp Eng
436.​2012.​703322 142(8):04016027. https://​doi.​org/​10.​1061/​(ASCE)​TE.​1943-​
Setyawan A, Nainggolan J, Budiarto A (2015) Predicting the remaining 5436.​00008​56
service life of road using pavement condition index. Procedia Tsai YC, Kaul V, Mersereau RM (2010) Critical assessment of pave-
Eng 125:417–423 ment distress segmentation methods. J Transp Eng 136(1):11–19.
Shah S, Deshmukh C (2019) Pothole and bump detection using con- https://​doi.​org/​10.​1061/​(ASCE)​TE.​1943-​5436.​00000​51
volution neural networks. In: 2019 IEEE transportation electri- Vidal M, Ostra M, Imaz N, Garcı´a-Lecina E, Ubide C (2016) Analysis
fication conference pp 0–3. https://​doi.​org/​10.​1109/​ITEC-​India​ of SEM digital images to quantify crack network pattern area
48457.​2019.​ITECI​ndia2​019-​186 in chromium electrodeposits. Surf Coat Technol 285:289–297
Shah A, Tighe S, Stewart A (2004) Development of a unique dete- Vyas V, Singh AP, Srivastava A (2021a) Entropy-based fuzzy SWOT
rioration index, prioritization methodology, and foreign object decision-making for condition assessment of airfield pavements.
damage evaluation models for Canadian airfield pavement man- Int J Pavement Eng 22(10):1226–1237. https://​doi.​org/​10.​1080/​
agement. Can J Civ Eng 31(4):608–618 10298​436.​2019.​16715​90
Shah YU, Jain SS, Tiwari D, Jain MK (2013) Development of overall Vyas V, Singh AP, Srivastava A (2021b) Prediction of asphalt pavement
pavement condition index for the urban road network. Procedia condition using FWD deflection basin parameters and artificial
Soc Behav Sci 104:332–341. https://​doi.​org/​10.​1016/j.​sbspro.​ neural networks. Road Mater Pavement Des 22(12):2748–2766.
2013.​11.​126 https://​doi.​org/​10.​1080/​14806​29.​2020.​17978​55
Shahin MY (1989) PAVER Manual: Asphalt Surfaced Airfield Manal, Wang H, Thakkar C, Chen X, Murrel S (2016) Life-cycle assessment
US Army Construction Engineering Research Laboratory of airport pavement design alternatives for energy and environ-
Shahin MY (2005) Pavement management for airports, roads, and park- mental impacts. J Clean Prod 133:163–171. https://​doi.​org/​10.​
ing lots. Springer Science Business Media LLC, New York, NY 1016/j.​jclep​ro.​2016.​05.​090
Shahin MY, Becker JM (1984) Development of performance prediction Wang X, Hu Z (2017) Grid-based pavement crack analysis using deep
models for airfield pavements. Transp Res Rec 985:25 learning. In: 2017 4th international conference on transportation
Shahnazari H, Tutunchian MA, Mashayekhi M, Amini AA (2012) information and safety-proceedings pp 917–924. https://​doi.​org/​
Application of soft computing for prediction of pavement con- 10.​1109/​ICTIS.​2017.​80478​78
dition index. J Transp Eng 138(12):1495–1506 Wesołowski M, Iwanowski P (2020) APCI evaluation method for
Shan B, Zheng S, Jinping Ou (2016) A stereovision-based crack width cement concrete airport pavements in the scope of air operation
detection approach for concrete surface assessment. KSCE J Civ safety and air transport participants’ life. Int J Environ Res Public
Eng 20(2):803–812 Health 17(5):1663. https://​doi.​org/​10.​3390/​ijerp​h1705​1663

13
1190 Iranian Journal of Science and Technology, Transactions of Civil Engineering (2024) 48:1171–1190

Wolf J, Pirskawetz S, Zang A (2015) Detection of crack propagation Zakeri H, Nejad FM, Fahimifar A (2017) Image-based techniques
in concrete with embedded ultrasonic sensors. Eng Fract Mech for crack detection, classification, and quantification in asphalt
146:161–171 pavement: a review. Arch Comput Methods Eng 24(4):935–977.
Xu C, Xie J, Chen G, Huang W (2014) An infrared thermal image https://​doi.​org/​10.​1007/​s11831-​016-​9194-z
processing framework based on superpixel algorithm to detect Zhang S, Lippitt CD, Bogus SM, Neville PR (2016b) Characterizing
cracks on metal surface. Infrared Phys Technol 67:266–272. pavement surface distress conditions with hyper-spatial resolu-
https://​doi.​org/​10.​1016/j.​infra​red.​2014.​08.​002 tion natural color aerial photography. Remote Sens 8(5):392.
Xue-jun Xu, Zhang X-N (2013) Crack detection of reinforced concrete https://​doi.​org/​10.​3390/​rs805​0392
bridge using video image. J Cent South Univ 20(9):2605–2613 Zhang A, Wang KCP, Li B, Yang E, Dai X, Peng Y, Fei Y, Liu Y,
Yamaguchi T, Hashimoto S (2010) Fast crack detection method for Li JQ, Chen C (2017) Automated pixel-level pavement crack
large-size concrete surface images using percolation based image detection on 3D asphalt surfaces using a deep-learning network.
processing. Mach vis Appl 21(5):787–809 Comput Aided Civ Infrastruct Eng 32:805–819. https://​doi.​org/​
Yang Y-S, Yang C-M, Huang C-W (2015) Thin crack observation in a 10.​1111/​MICE.​12297
reinforced concrete bridge pier test using image processing and Zhang L, Yang F, Daniel Zhang Y, Zhu YJ (2016) Road crack detection
analysis. Adv Eng Softw 83:99–108 using deep convolutional neural network. In: Proceedings—inter-
Yang XX, Li H, Yu Y, Luo X, Huang T, Yang XX (2018) Automatic national conference on image processing pp 3708–3712. https://​
pixel-level crack detection and measurement using fully convo- doi.​org/​10.​1109/​ICIP.​2016.​75330​52, 2016-Augus
lutional network. Comput Aided Civ Infrastruct Eng 33:1090– Zhao X, Xue L, Xu F (2021) Asphalt pavement paving segregation
1109. https://​doi.​org/​10.​1111/​mice.​12412 detection method using more efficiency and quality texture fea-
Yang F, Zhang L, Yu S, Prokhorov D, Mei X, Ling H (2020) Feature tures extract algorithm. Constr Build Mater. https://​doi.​org/​10.​
pyramid and hierarchical boosting network for pavement crack 1016/j.​conbu​ildmat.​2021.​122302
detection. IEEE Trans Intell Transp Syst 21:1525–1535. https://​
doi.​org/​10.​1109/​TITS.​2019.​29105​95 Springer Nature or its licensor (e.g. a society or other partner) holds
Yiyang Z (2014) The design of glass crack detection system based on exclusive rights to this article under a publishing agreement with the
image pre-processing technology. In: proceedings of information author(s) or other rightsholder(s); author self-archiving of the accepted
technology and artificial intelligence conference pp 39–42 manuscript version of this article is solely governed by the terms of
Yuan J, Mooney MA (2003) Development of adaptive performance such publishing agreement and applicable law.
models for Oklahoma Airfield pavement management system.
Transp Res Rec 1853(1):44–54

13