Advancements in Road Asset Management Systems: Leveraging Image

Processing and Machine Learning for Enhanced Decision Making – A

systematic review
Abstract
The Road Asset Management System (RAMS) has become an essential tool for government
agencies and decision-makers, allowing for efficient maintenance of aging pavement assets
while maximizing the use of budget resources. Recent data collection and analysis
advancements have greatly improved the estimation of remaining service life and the
selection of maintenance strategies. The integration of image processing (IP) and applied
machine learning (ML) artificial intelligence (AI) technologies has played a crucial role in
enhancing RAMS capabilities, enabling comprehensive analysis of extensive distress survey
data. This review paper will consolidate and present the latest RAMS advancements and
associated processes. Key focus areas include data collection methodologies, analytical
techniques, decision-making tools, and data processing methodologies. The transition from
manual to automated pavement condition surveys has significantly improved efficiency, with
various methods evaluated globally to ensure accurate and speedy data acquisition. These
methods include manual surveys, automated vehicle-based surveys, and cost-effective
alternatives. The paper also discusses the evolution of data analysis within the context of
RAMS, specifically in distress evaluation, creating homogeneous sections for maintenance
strategy selection, and prioritizing and optimizing maintenance approaches. It highlights the
vital role of IP, ML, and AI technologies in enhancing the effectiveness of RAMS in these
areas. Interdisciplinary research and multi-scale developments are also explored, emphasizing
the impact of AI and computer vision technologies.

Keywords: Road Asset Management System, Image processing, Machine Learning and
Artificial Intelligence.

1. Introduction
The evolution of Pavement Management Systems (PMSs) has been substantial since they
were first introduced in the late 20th century. Initially created to adapt to the changing focus
from construction-centric methods to maintenance-oriented strategies during the late 1960s
and early 1970s, PMSs have come a long way. As the nation's major freeway and highway
networks were completed mainly, highway agencies faced the critical challenge of preserving
these significant pavement investments. However, limited funding for maintenance, coupled
with factors like aging infrastructure and increased heavy truck traffic, led to a growing
backlog of pavement maintenance needs. Recognizing the need for a systematic approach to
optimize the allocation of limited resources, pavement engineers and planners conceptualized
PMSs [1, 2, 138]. In their early stages, PMSs primarily employed basic data-processing
methods to assess and prioritize pavement rehabilitation projects based on current pavement
conditions and traffic volume. These early systems did not incorporate forecasts of future
pavement conditions or conduct economic analyses to compare preventive and deferred
maintenance strategies. They mainly operated at the project level without addressing
network-level planning, such as budget allocation and achieving desired performance goals
for an entire roadway network [2].

By the start of the 1980s, Pavement Management Systems (PMSs) started to take on a
more expansive network view, marked by the creation of the inaugural system for Arizona's
Department of Transportation. As the 1990s rolled in, these systems further matured,
embracing comprehensive methods for forecasting performance, optimizing both network
and individual projects, prioritizing various components, and incorporating Geographic
Information Systems (GIS). This historical evolution of PMSs reflects their ongoing
adaptation to the changing needs and challenges highway agencies face in maintaining and
managing their pavement assets [2, 3]. As technology and methodologies continue to
advance, PMSs are expected to play an increasingly vital role in optimizing pavement
maintenance strategies and ensuring the longevity and sustainability of roadway networks [4,
137].

Historically, highway agencies have concentrated on developing RAMS to minimize

road deterioration and make informed maintenance decisions [5-7, 139]. Nevertheless, the
decisions derived from RAMS can be influenced by various factors, such as the methods
employed for data collection, the analytical procedures utilized, budget constraints, and
stakeholder input. Notably, data collection stands out as a pivotal element that significantly
impacts the system's overall efficiency [8]. Ensuring the effectiveness of RAMS necessitates
the maintenance of a dynamic database that can support other interconnected systems and
provide adequate data storage and retrieval capabilities. This, in turn, demands regular
updates to reflect any alterations in the pavement structure. Recently, highway agencies have
increasingly adopted dedicated Pavement Data Collection Vehicles (DCVs) to automate the
data-gathering process [9, 10]. These DCVs are equipped with high-speed digital cameras,
laser systems, and accelerometers, enabling them to generate high-quality data. However, it's
important to note that this automated approach can incur substantial costs.

Moreover, budget-friendly options like Unmanned Aerial Vehicles (UAVs) fitted with
camera systems, as well as data gathering via smartphones, have come into play. These
innovative approaches have demonstrated their efficacy in identifying visible issues on the
pavement surface, including cracks, potholes, repairs, surface wear, and unevenness.
Collectively, advancements in data collection methods have significantly enhanced the
efficiency and accuracy of RAMS. Dedicated data collection vehicles, UAVs, and
smartphone-based techniques have reduced costs and made data collection more accessible.
These innovative approaches excel in capturing pavement surface irregularities, thereby
contributing to the generation of more comprehensive and precise pavement condition data
[11-14].

Recent developments in IP and ML have provided cost-effective solutions for data

analysis in RAMS. These techniques can detect and quantify pavement distresses, compute
condition indices, and predict future performance. However, the computational costs
associated with IP and ML algorithms can be pretty high. Deep learning (DL), a subset of AI,
has emerged as a promising solution to address this issue [16, 17]. DL algorithms offer higher
statistical assistance and can significantly lower computational costs. Recent research
indicates that Deep Learning (DL) algorithms offer potential in identifying and measuring
cracks from 2D images of pavement surfaces. Nonetheless, these budget-friendly approaches
for data gathering and evaluation have yet to undergo extensive trials for network-level
(RAMS) activities on an international scale. Further research is needed to explore the
application of DL algorithms in network-level RAMS activities and assess their efficiency
and accuracy compared to traditional methods. These advancements can significantly
improve RAMS's analytical performance and reduce data analysis costs [19-21].

This review paper aims to gather and present the latest advancements in Road Asset
Management Systems (RAMS) and related processes. It specifically focuses on procedures
for data collection, processing, and analysis of various decision-making tools. The ultimate
aim is to provide a comprehensive narrative integrating interdisciplinary research and
developments across various scales, incorporating cutting-edge computer vision and artificial
intelligence technologies. The primary goal of this review is to contribute to advancing
future-oriented RAMS by leveraging these advanced technologies. It seeks to establish
connections among different aspects of RAMS, aiming to create a universally accepted
system that promotes scientific and analytical approaches. The paper's scope covers state-of-
the-art data collection procedures, an in-depth exploration of analytical techniques, including
performance prediction and management decision-making, insights into global RAMS,
challenges related to incorporating soft computing methods into RAMS, and identification of
critical research gaps. In summary, this paper serves as a framework and roadmap for the
future development of RAMS, with the overarching objective of enhancing the efficiency of
the traditional toolkit for managing roadway assets.

1.1 Guidance for Informed Pavement Management Decisions

Pavement management plays a pivotal role in guiding agency decisions across three distinct
levels, as outlined by AASHTO in 2012 [3]:

Strategic Level: At the strategic level, the primary focus is on determining the investment
levels and strategies that align with the overarching goals and objectives of the agency. To
effectively support decisions at this level, the pavement management system must be able to
forecast future pavement conditions and demonstrate the potential outcomes of deferring
investments. This decision-making level involves long-term planning and the allocation of
resources to ensure that the agency's pavement assets align with its strategic vision [1, 2].

Network Level: The network level of pavement management involves providing summary
information that covers the entire highway network. This data is instrumental in identifying
the most efficient combination of projects and treatment measures for a multiyear program.
Evaluating the costs and benefits associated with various project and treatment combinations
concerning current and future conditions is essential. The network-level perspective ensures
that decisions are made to maximize the overall health and performance of the highway
network [1, 2].

Project Level: At the project level, decision-making efforts are concentrated on specific
segments or sections of the pavement network. In contrast to the broader strategic or network
levels, project-level analysis necessitates a higher level of detail and specificity in the
information obtained from the PMS. This level of analysis emphasizes in-place conditions
and aims to address specific issues or challenges within pavement sections. An example of a
project-level analysis could involve investigating the underlying causes of underperforming
pavement sections and devising targeted solutions [1, 2].
By catering to these three distinct levels of decision-making, pavement management systems
enable agencies to make informed choices that encompass long-term strategic planning,
network-wide optimization, and targeted project-level interventions. This comprehensive
approach ensures the effective management and preservation of pavement assets while
aligning with the agency's broader objectives.

1.2 Elements of Pavement Management

A pavement management system comprises several key components that facilitate a wide
range of analyses. These components form the system's backbone and enable the analysis and
reporting modules to function effectively. Here's a breakdown of these essential components
[137]:

Database: The database is at the core of the pavement management system. It stores a diverse
set of inputs, including data related to pavement conditions, treatments, traffic volume, cost
data, and more. The sophistication of the database can vary, ranging from basic spreadsheets
to more complex relational databases or comprehensive agency-wide data warehouses. The
quality and comprehensiveness of the data stored in the database are crucial for accurate
analysis and decision-making.

Analysis Module: The analysis module is where the pavement management system performs
various analyses. This component utilizes the data stored in the database to customize
pavement performance prediction models, define treatment applicability under different
conditions, and adapt treatment costs and impacts to agency-specific circumstances. These
parameters are essential for determining funding levels needed to achieve performance goals,
identifying the most effective treatment combinations within budget constraints, and
projecting future pavement conditions under different investment strategies.

Reporting Module: The reporting module generates a wide array of outputs in various
formats. These outputs include reports, charts, graphs, and other types of information. The
reporting module plays a crucial role in communicating the results of the analyses to
decision-makers and stakeholders, making the data accessible and understandable for
informed decision-making.

Feedback Loop: The feedback loop is the final component of a pavement management
system. It serves as a mechanism for integrating field project data and performance trends
back into the system. This ongoing feedback loop helps keep the database up-to-date with
current information and ensures that the analysis parameters are continually adjusted to
support evolving decision-making processes. It allows the system to adapt to changing
conditions and improve its accuracy over time.

Together, these components create a robust and dynamic pavement management system that
assists agencies in making data-driven decisions to efficiently maintain and manage their
pavement assets. The system's effectiveness relies on data quality, analysis parameters'
sophistication, and the seamless integration of field data and performance trends.

1.3 The Development of Pavement Management over Time

The historical evolution of pavement management systems (PMS) underscores the
continuous progress in their complexity and significance in decision-making processes.
Initially, PMS primarily served as tools for documenting pavement conditions and estimating
immediate funding requirements, as highlighted in the 1987 synthesis "Pavement
Management Practices" by Peterson (1987) [22]. These early systems focused on generating
outputs such as pavement condition assessments, prioritized lists of maintenance needs,
deficiency identification, and associated treatment costs. However, they had limitations,
lacking the ability to estimate life-cycle costs, predict future pavement conditions, and
seamlessly integrate data with other agency systems.

In Gramling's 1994 synthesis, "Current Practices in Determining Pavement Condition," a

significant shift towards data collection and management for pavements was observed.
During that time, around 85% of agencies were using mainframe computers to archive their
pavement management information. Additionally, numerous agencies were either already
using or in the stages of creating their own pavement management systems. A key
advancement of that period was the rising acceptance of the International Roughness Index
(IRI). This change was largely driven by the reporting mandates of the Highway Performance
Monitoring System (HPMS) as established by the Federal Highway Administration (FHWA).
This highlighted a growing focus on the use of standardized methods for collecting data.

In the 1995 NCHRP synthesis called "Pavement Management Methodologies to Select

Projects and Recommend Preservation Treatments," authored by Zimmerman and ERES,
three unique approaches within pavement management were outlined: analysis of pavement
condition, models for assessing priority, and models for optimizing the network. However, it
was noted that only 12 agencies among the 56 responding prioritized projects using benefit–
cost analyses, emphasizing the need for more sophisticated decision-making tools. Despite
these advancements, a recurring challenge during this period was the inability to effectively
forecast future pavement conditions, highlighting a critical shortcoming in PMS [26-28].

1.4 Advantages of Implementing Pavement Management Systems

Agencies that have embraced PMS have recognized many benefits, as outlined in
AASHTO's 2012 [3] report. These benefits include effective resource utilization, justification
of funding needs, understanding of current and projected pavement conditions, improved
accessibility to pavement information, increased accountability and transparency, and
objective decision-making based on data. The adoption of PMS has proven to be instrumental
in optimizing budget allocation and enhancing pavement asset management practices,
contributing to more efficient and sustainable transportation infrastructure [28, 137].

Effective Resource Utilization: Pavement management systems enable agencies to allocate

available resources to enhance pavement performance efficiently. This optimization ensures
that funds are directed where they are needed most.

Justification of Funding Needs: These systems provide a structured and data-driven

approach to justify funding requirements, making it easier for agencies to secure necessary
financial support for pavement maintenance and improvement projects.

Understanding of Pavement Conditions: Agencies comprehensively understand current and

projected pavement conditions and requirements. This knowledge facilitates long-term
planning and strategic decision-making.

Improved Information Accessibility: Pavement management systems enhance access to

pavement-related information across the agency, fostering better communication and
coordination among relevant departments and stakeholders.

Enhanced Accountability and Transparency: By relying on data-driven insights, agencies

can improve accountability and transparency in their decision-making processes. This helps
stakeholders and the public better understand the rationale behind pavement management
decisions.

Objective Decision-Making: Using data and analysis tools promotes objective decision-
making, reducing the influence of subjective factors in determining pavement maintenance
and improvement strategies.
 1.5 Strategies for Managing Pavement Systems
Each transportation agency faces unique considerations when determining the most
suitable approach to pavement management. These decisions depend on various factors,
including the agency's organizational requirements, size, available resources for data
collection and analysis, compliance with FHWA (Federal Highway Administration)
regulations, and the level of support from executive leadership. After a careful evaluation of
these factors, an agency can formulate a plan for [3, 137]:

Collecting Pavement Inventory and Condition Data: Defining a strategy for gathering data
on pavement inventory and condition is crucial. This involves determining what information
needs to be collected, how it will be collected, and at what intervals.

Selecting Pavement Management Software: Choosing the appropriate pavement

management software is a critical decision. The selection should align with the agency's
specific needs and capabilities. Factors such as software cost, complexity, and flexibility
must be weighed.

Developing Treatment Rules and Costs: Gathering the necessary information to create
treatment rules and cost estimates is essential. This involves defining maintenance and
rehabilitation strategies based on the collected data and estimating associated costs.

In addition to in-house pavement management systems tailored to an agency's specific needs,

there are two primary categories of publicly available pavement management systems, each
with its own characteristics:

Public Domain Software: These software programs have been developed using public funds
and are typically available at little to no cost. They are often designed to provide basic yet
effective database and analysis tools. Public domain software is commonly used at the local
level, offering a cost-effective solution for smaller agencies with limited resources.

Commercial Software: Commercial pavement management software options are available

from private vendors. While they may involve licensing fees, these solutions often offer
greater complexity and flexibility, catering to the needs of larger agencies with more
extensive resources and requirements. They may include advanced features and support
services.

When deciding between public domain and commercial pavement management software,
transportation agencies must carefully assess their specific circumstances, budget constraints,
and long-term goals. The chosen approach should align with organizational priorities and
ensure effective pavement management practices. The actual benefits agencies realize depend
on several factors, including the quality and comprehensiveness of the data collected, the
extent to which agency decisions align with pavement management recommendations and the
software tools' capabilities.

Quantifying the benefits of pavement management strategies is crucial for demonstrating

their cost-effectiveness. Two notable studies provide concrete evidence of these advantages:

Hudson et al. (2000) [2]: In their research, Hudson and his team evaluated the financial
benefits of the Arizona Department of Transportation's investment in creating, deploying, and
maintaining pavement management software. The findings were remarkable, indicating that
for every dollar spent on the software, the agency saved a minimum of $30 in operational
costs. When user costs were taken into account, the savings escalated to an impressive $250
for every dollar invested. This study highlights the substantial financial benefits of effectively
implementing pavement management practices [2].

Ministry of Transportation in Alberta (Cowe Falls et al. 1994) [23]: A noteworthy

research effort by Alberta's Ministry of Transportation looked into the implications of
transitioning from a strategy focused solely on fixing the most deteriorated roads to one that
included a variety of more budget-friendly treatments. Over a 5-year period, this change in
approach yielded an exceptional return on investment, with a reported return rate of 100 to 1.
This finding underscores the substantial financial advantages associated with efficient
pavement management practices and optimizing treatment strategies.

These studies offer compelling evidence of the substantial cost savings and improved
resource utilization achievable through the effective implementation of pavement
management strategies. Some examples of public domain pavement management software
programs are listed below [137]:

PAVER: Created by the U.S. Army Corps of Engineers and made available through the
American Public Works Association (APWA), PAVER is a commonly used, open-source
software for evaluating and overseeing pavement conditions. [3]

StreetSaver: Originating from the Metropolitan Transportation Commission (MTC) in the

San Francisco Bay Area, StreetSaver is another open-source software aimed at aiding local
agencies in managing their pavement assets. [3]
RoadSoft: Formulated by the Center for Technology and Training at Michigan
Technological University, RoadSoft serves as a tool for gathering, scrutinizing, and
administering data related to the state of roads and maintenance tasks. [3]

These public domain software programs provide cost-effective solutions for pavement
management, particularly for smaller agencies or those with limited budgets, allowing them
to manage and maintain their pavement assets efficiently.

2. Methods of Pavement Data Collection

The database in RAMS is often called its core, which plays a crucial role throughout the
process [29]. Various stakeholders depend on this database to assess pavement conditions and
make well-informed decisions. Realistic data is necessary to perform various maintenance
activities and ensure optimal budget utilization to successfully implement PMS [30, 31]. The
data needed for PMS primarily includes information about pavement inventory and condition.
Road inventory data encompasses various details about road infrastructure within the right-
of-way, while pavement condition data comprises information related to pavement distress,
including its severity and extent [33]. Additionally, the international roughness index (IRI)
significantly contributes to the assessment of riding comfort. In addition to inventory and
condition data, secondary data such as traffic characteristics, axle load spectrum, climatic
conditions, material characterization, pavement design, and maintenance history data are also
required for in-depth analysis [32, 34].

Among the various data items, it is essential to periodically collect inventory, condition,
and auxiliary data to evaluate the service life of pavements [35, 36]. Recognizing that data
quality significantly influences decisions within the context of Road Asset Management
Systems (RAMS), research efforts have been dedicated to enhancing methods for collecting
condition data to meet the quality standards expected by stakeholders. Haas et al. [1, 2]
pointed out that pavement distress can manifest when the pavement is structurally sound but
functionally deficient. Pavements may exhibit structural and functional deficiencies [39].
When a pavement is structurally adequate but functionally deficient, minor maintenance
interventions are typically sufficient to keep it in service. On the other hand, structurally and
functionally deficient pavements necessitate substantial maintenance efforts to restore them
to a serviceable condition. Consequently, assessing pavements for their structural and
functional conditions has become imperative in determining appropriate maintenance
strategies [37, 38].
 2.1 Pavements functional data collection
Before the 1980s, pavement distress evaluation heavily relied on manual testing procedures,
which were notably inefficient [3, 4]. As a result, over the past decade, a persistent effort has
been to enhance, develop, and incorporate various Pavement Distress Detection (PDD)
technologies. This collaborative endeavour has engaged researchers from several esteemed
institutions, who have made substantial strides in advancing and implementing PDD
technologies [41, 42].

The first automated Pavement Distress Detection (PDD) technology to be introduced was a
radar system designed to collect surface topography data. However, as Wang pointed out in
2011, this technology had resolution limitations and eventually gave way to PDD
technologies that used 2D imaging methods. Despite this, 2D-based PDD technologies had
their own drawbacks, such as a lower rate of recognition and an inability to fully capture the
intricate features of pavement issues [43]. As a result, 3D PDD technologies that employ 3D
imaging have come into play, thanks to significant research contributions. [44, 45]

These 3D technologies operate on different principles, each with its own unique set of
advantages and drawbacks. The line-structured light-based 3D Pavement Distress Detection
(PDD) technology has garnered the most attention, primarily due to its outstanding features.
However, it's essential to acknowledge that the measuring resolution of this technology is
limited to the millimeter level, as observed in studies conducted by Gu et al. (1994) [46],
Mertz et al. (2002) [47], Pennington et al. (2001) [48], Rosati et al. (2009) [49], Wang and
Gong (2005) [53], and Okawa (1984) [56]. As part of an ongoing research initiative, a
concerted effort is to develop a 3D PDD system based on interference fringe technology. The
primary objective of this project is to attain higher accuracy and resolution when capturing
3D pavement profiles, as discussed in the works of Chu et al. (2019) [57], Lally (2010) [58],
and Wang et al. (2012) [59].

It is widely recognized that inherent errors are present in the 3D morphological measurement
of pavement distress. Consequently, obtaining the accurate 3D dimensions of pavement
distress at the measurement level is not possible without comprehensive calibration efforts at
the project level [50-52]. In essence, current Pavement Distress Detection (PDD)
technologies face a challenge in achieving metrological traceability, which refers to the
ability to trace measurements back to their actual values. The existing PDD technologies
yield results solely at the measurement level, which may not fully align with the stringent
metrological definition of actual measurement values. Addressing the absence of a
metrological traceability method in PDD measurements is an area that warrants attention and
further development [54, 55].

In the process of analyzing pavement distress, various types of issues affecting the pavement
are identified using pavement images or dimensional data. This identification is achieved
through image processing and machine learning techniques. A crucial initial step in this
process is marking and quantifying these distresses in terms of their type, severity, and
extent. This step is of utmost importance because the absence of standardized distress
definitions can lead to inconsistent classifications. Different organizations or institutions may
employ varying classification standards, as pointed out by Gopalakrishnan in 2018 [60].
When dealing with acquired 2D pavement images, it is essential to distinguish damaged areas
from undamaged ones, a process known as image segmentation. This segmentation helps in
effectively identifying distress boundaries and types. On the other hand, when utilizing 3D
Pavement Distress Detection (PDD) technology, the integration of the acquired 3D data is
critical. This integration allows for creating a comprehensive 3D map of the pavement
surface, facilitating the determination of distress boundaries and types. ANN belongs to a
category of supervised algorithms often employed for image segmentation tasks, aiding in
separating distressed and non-distressed areas in pavement images. Once the distress is
identified, the subsequent phase involves distress assessment, wherein the severity of the
damaged sections is evaluated.

In the 1990s, Phoenix Scientific was at the forefront of creating a revolving laser
radar system specifically tailored for pavement assessments. This ground-breaking system
employed Light Detection and Ranging (LIDAR) technology to generate 3D representations
of pavement surfaces. The LIDAR system was composed of three essential elements: a laser
scanner, a GPS receiver, and an inertial measurement unit (IMU). Data from the laser scanner
was gathered through a spinning mirror, which allowed for the capture of surface profiles
over an expansive area. However, over several decades, it became apparent that this method
struggled to significantly improve its resolution significantly, leading to its gradual
replacement by other Pavement Distress Detection (PDD) technologies, as discussed by
Wang in 2004 [50]. After years of refinement, methods for capturing 3D information of
pavements using LIDAR technology have become more advanced. In 2013, a Chinese
research and design firm unveiled an innovative technique for isolating features of specific
road sections. This approach began with the initial collection of point cloud data for the
targeted section via a LIDAR sensor. A guideline for road features was then formulated based
on this point cloud data, and data points near this guideline were isolated. A detailed slice of
the section, containing the relevant features, was chosen to form a corresponding point cloud
template. Parameters for transitioning between this feature point cloud template and the
original section point cloud were established. This facilitated the computation of the road's
spatial coordinates, allowing for the extraction of the specific road features in question. This
method offered several advantages, including rapid processing, high work efficiency, precise
feature extraction, and minimal disruption to traffic flow, ensuring the safety of personnel
involved in the process, as outlined by Chen et al. in 2015a and b [40, 41].

Research in structural damage detection primarily centers on leveraging image processing

technology (IPT), often incorporating pre-processing and post-processing techniques to
assess its suitability for real-world applications. While some researchers have integrated
machine learning techniques to enhance the efficiency and robustness of IPT, they often rely
on time-consuming pre-processing and post-processing methods, which may limit their
ability to detect specific types of damage. An emerging and increasingly popular alternative
approach involves utilizing machine learning algorithms (MLA) that gather signals from non-
destructive testing to identify defects. However, these methods sometimes inherit sensor
complexity and encounter challenges related to false feature extraction, similar to those
encountered in IPT. Cha et al. [61,62] introduced an encouraging approach that utilizes deep
learning, specifically convolutional neural networks (CNN), to detect cracks in asphalt
surfaces, achieving a remarkable recognition accuracy of approximately 98%. Other
researchers have also proposed effective methods, such as Peng et al.'s (2020) [63] triple-
threshold pavement crack detection, Zheng et al.'s [64-68] cross-layer manifold invariance-
based pruning for road type recognition, Gao et al.'s [69-71] pothole detection method and
Tang et al.'s (2020) [72] iteratively optimized patch label inference network for detecting
various pavement distresses. Each of these methods has demonstrated improved performance
in their respective domains.

Researchers have addressed the challenge of detecting thin cracks in pavement

distresses by enhancing traditional image segmentation with binary methods. This involves
setting gray values below a threshold to black and values above it to white, making cracks
and potholes more visible. ANN, machine learning algorithms, and deep learning have
recently revolutionized autonomous road crack detection, offering improved object
recognition and feature extraction capabilities. Notable approaches include Chu et al.'s feed-
forward neural network classifier, Huang et al.'s [73-75] wavelet neural network, and Koch
and Brilakis' pothole detection method. Yin et al. [76] introduced a deep convolutional neural
network for pavement distress detection. Tong et al. [77, 78] integrated convolutional
networks with Gaussian-conditional random fields for high-quality pavement distress
extraction. These advancements have significantly improved the accuracy of pavement
distress identification.

In the past two decades, developing Pavement Distress Detection (PDD) technology
worldwide has continuously improved measurement accuracy for assessing pavement
conditions. Initially, radar-ranging-based PDD technology faced limitations in enhancing
measurement resolution, prompting a transition to 2D-image-based PDD technology.
Subsequently, 3D PDD technology employing laser ranging gained prominence for profiling
pavement surfaces in three dimensions. In response to the demand for high-precision
measurements requiring extensive point cloud data, structured light-based 3D PDD
technology has garnered significant attention. Despite these advancements, PDD technology's
resolution remains at the millimeter level, and researchers are actively working to enhance it
further.

Wang et al. [79] of Virginia Tech achieved a breakthrough by reconstructing

pavement aggregate in 3D using interference fringe and Fourier transform interferometry
(FTI), achieving a remarkable resolution of ±10 micrometers. Building on this research, Chu
et al. of Beijing University of Science and Technology work on a 3D PDD system with even
higher accuracy and resolution, targeting errors less than 1 mm [80]. Additionally,
researchers have explored crowd-sourcing-based systems like CRSM and Smart Road Sense
to identify pavement distresses and assess road roughness levels. CRSM demonstrated a 90%
accuracy rate in recognizing pavement distresses, even in the presence of tiny bumps or
potholes [40]. Smart Road Sense, equipped with GPS receivers and triaxial accelerometers,
has effectively identified critical pavement distresses through collaborative data collection.
These developments signify ongoing efforts to improve PDD technology and enhance its
practical applications [41].

The discussion underscores the limitations of existing 3D Pavement Distress

Detection (PDD) systems, which suffer from measurement errors and cannot provide precise
3D dimensions of pavement distresses. Ongoing research aims to establish metrological
traceability for pavement distress measurements. Chu et al. (2019) [80] have advocated for
the importance of metrological traceability in the PDD field and introduced a traceability
chain to calibrate 3D PDD systems, aligning their measurements with metrological standards.
This calibration process involves high-accuracy 3D pavement profilometers and ultimately
traces measurements back to the mete-wand, enhancing the accuracy and scientific validity of
standard 3D PDD systems [80].

Over the past decade, various Pavement Distress Detection (PDD) technologies have
advanced, contributing to progress in the field. While 2D image-based PDD offers a 1 mm
resolution, it is prone to missed detections and miss-judgments and cannot fully capture
pavement distress, especially in 3D. Laser triangulation, radar ranging, and laser sensor
ranging have limited measurement resolutions. Structured light-based 3D pavement profiling
technology, while achieving a 1 mm resolution, faces challenges during dynamic
measurements. Interference fringe-based 3D PDD holds promise for better accuracy. Despite
the advancements in 3D PDD, there are still shortcomings, such as limited resolution, lack of
metrological traceability, and challenges with structured light-based systems. The
development of higher-precision PDD systems and achieving metrological traceability are
critical research areas, focusing on realizing intelligent and efficient PDD using laser
technology in the future.

2.2 Structural Evaluation Methods

In the mid-1970s, the Transport and Roads Research Laboratory (TRRL) in the United
Kingdom introduced a modified parameter known as the structural number (SNC). This
modification aimed to account for the influence of the subgrade and the pavement layers
when assessing pavement performance. Researchers observed that the original structural
number (SN) tended to overestimate pavement capacity, especially for thicker pavements
exceeding 700 mm in thickness. To address this limitation, they introduced the adjusted
structural number (SNP), which has found application in models like the Highway
Development and Maintenance Management System - 4 (HDM4) for economic evaluations.
SNP incorporates a weighting factor for the base, subbase, and subgrade layers in a general
context. It serves as a valuable parameter for evaluating pavement quality. The structural
number (SN) concept gained prominence from the AASHO road tests conducted in the late
1950s. Subsequently, the AASHTO Guide for Pavement Design [81] provided an empirical
equation for calculating SN, further establishing its use in pavement engineering and design.
 In the realm of non-destructive pavement assessments, two prevalent methods for gauging
surface deflections in operational pavements are static and impulse loading. Reddy and
Veeraragavan conducted various tests [83], such as Benkelman beam (BB) rebound
deflection evaluations, rut-depth gauging, and assessments of cracked areas, to scrutinize the
structural integrity of flexible pavements. It's crucial to recognize that the BB technique falls
short in simulating the influence of moving vehicular loads. This limitation led to the
adoption of the falling weight deflectometer (FWD), equipped with geophones to record
surface deflections at different radial distances under a partial equivalent standard axle load
(ESAL). This mimics the effect of moving traffic and allows for a deeper understanding of
the pavement's structural capabilities. By analyzing the deflection pattern created by the
impact, valuable insights into the condition of the layers beneath the surface can be gleaned,
building on prior research [84]. FWD-derived data is useful for back-calculating layer
stiffness, which is crucial for overlay design. Various software solutions like Pavement ME
(or MEPDG), ILLIPAVE®, and KENPAVE™, as well as specialized tools like
MODCOMP®, ELMOD®, and BACKCALC®, are used for designing pavement sections.
To streamline the process and enhance reliability, artificial neural networks (ANNs) have
been developed to emulate these software packages, offering a more efficient and automated
approach to pavement design and analysis.

A notable relationship has been found between the structural number (SNC), Falling
Weight Deflectometer (FWD) readings, and layer stiffness. However, it's important to point
out that gathering data with FWD can be time-consuming and potentially disruptive to traffic
flow. To mitigate these issues, the Rolling Weight Deflectograph (RWD) was developed by
Applied Research Associates (ARA), Inc., supported by the Federal Highway Administration
(FHWA). The RWD captures pavement reactions to half an equivalent standard axle load
(ESAL) at common highway speeds, closely mimicking the dynamic impact of moving
vehicles. It provides average deflection values and standard deviations over 0.1-mile
segments [85]. The laser sensors on the RWD enable a much quicker rate of surface
deflection measurement compared to the FWD.

However, it's crucial to understand that RWD results don't allow for the direct calculation of
layer stiffness. So, while RWD has the advantage of faster data collection and less traffic
interference, it may not deliver the same level of detailed analysis that FWD can offer.

2.3 Trends in Prediction Modeling

 Addressing the differences in pavement condition survey procedures and distress
definitions used as the basis for Mechanistic-Empirical Pavement Design Guide (MEPDG)
models presents a significant challenge. As pavement performance modeling evolves, a
growing focus is on mitigating uncertainty within pavement management performance
modeling. In a pertinent study, authors discuss the limitations of characterizing pavement
sections solely through average condition ratings. They emphasize the risk of losing valuable
information and generating potentially inaccurate or misleading results (Kadar et al. 2015)
[86]. Their research advocates utilizing the complete dataset and treating each dataset as a
distribution to estimate outcome probabilities and predicted values (Kadar et al. 2015) [86].
Furthermore, the literature explores the adoption of Bayesian approaches for updating expert-
based Markov transition probability matrices as historical data becomes available within
pavement management systems (Tabatabaee and Ziyadi 2013) [87]. This approach,
developed by researchers, incorporates uncertainty into the initial transition probability
matrices and the pavement condition survey method. It was validated using data from the
Minnesota DOT's MnROAD test facility, underscoring the significance of considering
variability in these factors. In summary, addressing differences in survey procedures and
distress definitions and incorporating uncertainty into pavement performance modeling are
essential trends in pavement management, aiming to improve the accuracy and reliability of
predictions and decisions.

In addition to predicting performance and prioritizing maintenance, ML techniques can

also be used for optimizing maintenance strategies. For example, Li et al. [88] used a genetic
algorithm to optimize the maintenance schedule of pavement sections, considering budget
constraints and performance indicators. The results showed that the proposed approach can
effectively reduce the life-cycle cost of pavement maintenance compared to traditional
methods. Furthermore, ML techniques can be used to analyze and interpret large amounts of
data from various sources such as sensors, cameras, and social media. For example, Wang et
al. [50-53] developed an artificial neural network (ANN) model combined with a grey
dynamic model (GDM) for monitoring key indices in construction projects using limited time
series. This helps improve construction project management and enables predictive analytics
for identifying potential risks or problems before they occur.

Cluster-based deterioration models are particularly useful for low-volume rural roads
where limited resources may be available for maintenance. By clustering similar road
segments together based on specific characteristics, ML techniques can provide more
accurate predictions of deterioration rates and help allocate resources effectively. However, it
is essential to note that the effectiveness of ML techniques depends heavily on the quality and
availability of data. Data collection efforts should focus on gathering relevant information
about road conditions, traffic volume, climate conditions, and previous maintenance activities
to train ML algorithms accurately [89, 90].

In conclusion, machine learning techniques have emerged as valuable tools for predicting
future performance in transportation infrastructure and prioritizing maintenance decisions.
They offer an alternative approach that overcomes challenges related to data insufficiency by
utilizing soft computing methods like artificial neural networks (ANN), genetic algorithms
(GA), fuzzy systems, and cluster analysis. It’s essential to continuously evaluate these
techniques’ applicability, suitability, and limitations within specific contexts while leveraging
their potential benefits

2.4 Trends in the Use of Pavement Management Data

Traditional pavement management system applications involve recommending
treatments, developing multi-year work plans, and assessing the effects of different
investment strategies on road conditions over time [91, 92]. As the focus shifts towards
performance-based investment decisions globally, pavement management is crucial in
providing analytical results demonstrating various options' outcomes. Internationally, many
transportation agencies prioritize a service-based approach, considering safety, travel
reliability, comfort, and liveability over a purely condition-based approach [93, 94]. This shift
influences data collection and performance targets. There's growing interest in utilizing
pavement management in the United States to address customer-driven priorities, including
safety and environmental concerns. For instance, friction and crash data are being combined
to predict crash locations and reduce accidents [96]. Pavement management data is
increasingly used to assess and mitigate risks, aligning with federal requirements for risk-
based asset management plans [97, 98]. Lastly, environmental considerations are integrated
into pavement management, reducing energy usage and greenhouse gas emissions while
maintaining pavement performance through various adaptation strategies [99, 100].

2.5 Problems faced during the implementation of PMS

The practical implementation of Pavement Management Systems (PMS) is defined as
adopting and using the system constructively to achieve an organization's goals. Smith and
Hall emphasize that PMS implementation is only complete when it becomes critical for
making pavement decisions. While larger organizations have been better at overcoming
institutional barriers and sustaining PMS use, smaller organizations often need help
implementing it effectively. Several factors contribute to the challenges of PMS
implementation, including limited technical and financial resources and the need to avoid
disrupting daily business activities within an agency. PMS is typically used briefly each year
and is just one of several agency functions. Immediate obstacles include resource availability,
while long-term obstacles include the inability to keep up with rapidly evolving technology
and a need for a clear organizational philosophy. The complexity of software programs and
financial analysis can deter users from adopting and continuing to use PMS. This perceived
complexity can hinder decision-making processes, especially in agencies dependent on
federal funding. PMS may be maintained for funding purposes but could be more effectively
utilized for pavement decision-making.

In summary, effective PMS implementation involves adopting and using the system
constructively, but many organizations need help with resources, complexity, and
organizational philosophy, which can hinder its successful implementation.

2.6 Application of AI/ML in decision making

The following paragraphs discuss the historical use of knowledge-based expert systems in
pavement management and provide information on several specific expert systems developed
for this purpose:

DESIGNER: An expert system, rooted in knowledge-based principles, aids road engineers in

choosing appropriate pavement design elements. This system leverages a knowledge
repository sourced from the Recommended Pavement Design Features Catalog and is built on
the CLIPS expert system framework. Known as DESIGNER, the system offers feature
recommendations based on various input parameters such as traffic load, the support capacity
of the subgrade, and climatic conditions. [101].

OVERDRIVE: A rule-based expert system for designing the structural thickness of flexible
pavement overlays. It determines existing pavement structure thickness, asphalt-concrete
construction thickness, and overlay requirements. OVERDRIVE contains over 100 rules and
uses the EXSYS knowledge engineering shell [102].

PARADIGM: An integrated set of expert systems and algorithmic models, including

OVERDRIVE and SCEPTRE. SCEPTRE evaluates pavement surface distress and helps
engineers plan cost-effective rehabilitation strategies. It considers various surface distress
types and factors like traffic levels and climate. Cost-effectiveness analysis is based on
pavement-performance curves [102, 103].

Expert System for Project Selection: Developed for the Huntington Town Highway
Department, this system recommends rehabilitation or maintenance strategies based on
pavement conditions and user inputs. It searches a pavement database and considers factors
unavailable, such as geometric design requirements and pavement history [104].

Expert System for Pavement Maintenance in Developing Countries: Designed to assist

local highway agencies in managing pavement maintenance, it combines an interactive
algorithmic program and a rule-based expert system [105]. It ranks maintenance alternatives,
including recurrent maintenance, periodic maintenance, and strengthening, and estimates
costs. Users input distress data and other relevant information [106].

PADMA (Pavement Deterioration and Maintenance): Developed in India, this rule-based

expert system identifies causes of pavement deterioration and recommends appropriate
treatments. It categorizes 20 types of defects and requires input about surface conditions,
material characteristics, and construction practices. The system uses the DEKBASE expert
system shell and has been calibrated with field data.

These expert systems aim to enhance pavement management by assisting engineers in

making informed decisions about design, maintenance, and rehabilitation strategies based on
various factors and expert knowledge [106].

Fuzzy: Grivas and colleagues (2016) [107] researched utilizing a combination of fuzzy theory
and knowledge graphs as a decision support tool for pavement engineers to handle uncertain
information when selecting maintenance or rehabilitation strategies effectively. They
employed knowledge graphs to represent expert knowledge regarding the connections
between road condition symptoms and potential treatments, with the graph's connection
weights indicating the strength of belief in specific treatments. These relationships were
expressed using fuzzy set theory, and the method was applied to a practical problem
involving highway pavement preservation based on rut depth data and engineering judgment.
Three types of knowledge graphs were created to model different relationships, and the
decision-making process considered the degree of truth derived from connection weights in
each graph [107].
ANN: The literature review highlights the extensive use of Artificial Neural Networks
(ANNs) in pavement management systems (PMSs) since the 1990s. ANNs offer an
alternative approach to expert systems for providing decision support to pavement managers
[108-112]. Unlike rule-driven expert systems, ANNs are data-driven models. They are
employed in various aspects of PMSs, including estimating current pavement conditions,
predicting future deterioration, and selecting maintenance or rehabilitation actions [113-118].
ANNs have been applied to estimate pavement condition indices, such as the Visual
Condition Index (VCI) and the International Roughness Index (IRI). They have also been
used to predict future pavement conditions and assess maintenance needs. ANNs consistently
outperform traditional linear regression techniques in predicting pavement deterioration,
making them valuable tools for optimizing maintenance strategies in PMSs [119-120].
Additionally, ANNs have been used for automating the selection of pavement preservation
projects, reducing subjectivity, and improving efficiency in decision-making processes. In
some cases, ANNs are combined with other techniques like genetic algorithms to enhance
their performance in pavement management tasks [121-122].

Genetic Algorithms (GAs): In this set of articles, Kwasi and colleagues [123] introduce the
concept of genetic algorithms (GAs) and their application in estimating roughness
progression in flexible pavement. They emphasize the importance of crossover probabilities
in GA models for this problem. However, the focus on GA fundamentals overshadows
practical application. Other articles explore GAs in programming maintenance activities for
pavement systems, with Fwa and colleagues [124] providing a detailed GA model for long-
term planning. Hoque and colleagues [125] use the penalty method to handle constraints in
maintenance planning for various highway types. The simulations show GA's effectiveness in
generating near-optimal solutions with resource allocation optimization. FWA and colleagues
[126] delve further into GA fundamentals, while Yuge and colleagues [127] aim to minimize
pavement maintenance costs but lack clarity in their approach.

Hybrid Systems: Chou and colleagues [128, 129] present a comprehensive image-processing
system for assessing road pavement conditions, combining fuzzy set theory and Artificial
Neural Networks (ANNs) in five stages: image acquisition, segmentation, feature extraction
using Hu moments [130], Bamieh moments [131], and Zernike moments [132], classification
of distress types (e.g., longitudinal cracks, transverse cracks), and computation of crack
severity and extent. The system achieves perfect classification performance on the training
and validation sets. Ritchie and colleagues [133] propose an intelligent system for processing
flexible pavement images in real-time, incorporating ANNs and knowledge-based expert
systems to determine distress type, severity, and extent from video images and suggest
rehabilitation strategies based on historical data [134]. The ANN classifies 32x32 pixel sub-
images with impressive accuracy, though it faces challenges with combination distress
patterns. Goh's research [135] combines an Artificial Neural Network (ANN) with a
knowledge-based expert system to choose repair strategies for aging road surfaces. The
system takes into account factors like the type and extent of damage, as well as the type of
road. However, the efficacy of the ANN in this context has not been proven. Prechaverakul's
theoretical study [136] explores the use of knowledge-based expert systems and fuzzy logic
for smaller-scale road repair projects in Ohio, with a focus on multi-criteria decision-making.

AI technologies are utilized in various stages of the decision-making process, including

pavement diagnosis, deterioration modeling, rehabilitation assessment, maintenance action
selection, and priority programming. The choice of AI technology depends on the problem's
nature and the specific configuration of the decision support system, with artificial neural
networks suitable for numerical data processing and expert systems efficient for complex
expert knowledge-based problem-solving. While many advantages are associated with AI-
based systems, it's noted that real-data applications are a significant challenge. Integrating AI
into pavement management aims to enhance productivity and decision quality. Still, it doesn't
replace the role of pavement engineers, emphasizing the need for collaboration and
awareness of model limitations within their domain of validity.

3. Conclusions
In conclusion, integrating advanced AI techniques such as image processing, machine
learning, and deep learning can significantly enhance the field of RAMS. However, several
gaps in research and application still need to be addressed. These include the need for cost-
effective data collection methods, standardization of data collection and analysis techniques,
and further exploration of the application of deep learning techniques in RAMS beyond crack
and pothole detection. Addressing these gaps will contribute to the development of more
accurate and efficient methods for managing roadway assets on a global scale.

Overall, there is a need for further research and development in multiple areas related to
RAMS. These include assessing the potential use of cost-effective data collection methods,
such as smartphones and UAVs, for network-level applications. Additionally, it is crucial to
explore the application of deep learning techniques beyond crack and pothole detection and
utilize them for tasks like distress quantification, performance prediction, pavement
delineation, and prioritizing pavements for maintenance decisions. Developing a unified
pavement condition index incorporating region-specific parameters would also improve
maintenance decision-making. Finally, assessing the reliability and scalability of proposed
cost-effective studies and deep learning technologies at a larger scale is crucial to ensure their
applicability in real-time agency decision-making. Ultimately, bridging these gaps will help
transform RAMS from a data-driven approach to a technology-driven approach, aiding in
achieving sustainable transportation goals.

Reference:

[1] Haas, R., W.R. Hudson, and L. Cowe Falls, Pavement Asset Management, John
Wiley & Sons, Inc., Hoboken, N.J. and Scrivener Publishing LLC, Salem, Mass.,
2015.
[2] Hudson, W.R., S.W. Hudson, G. Way, and J. Delton, “Benefits of Arizona DOT
Pavement Management System After 16 Years’ Experience,” Pre-Print CD-ROM,
79th Annual Meeting of the Transportation Research Board, Jan. 9–13, 2000,
Washington, D.C.
[3] American Association of State Highway and Transportation Officials (AASHTO),
Pavement Management Guide, Second Edition, AASHTO, Washington, D.C., 2012.
[4] American Association of State Highway and Transportation Officials (AASHTO),
Standard Practice for Definition of Terms Related to Quality and Statistics as Used in
Highway Construction, AASHTO, Washington, D.C., 2011.
[5] Cowe Falls, L., S. Khalil, W.R. Hudson, and R. Haas, “Long-Term Cost-Benefit
Analysis of Pavement Management System Implementation,” Conference
Proceedings—Volume 2, Third International Conference on Managing Pavements,
May 22–26, 1994, San Antonio, Tex., 1994.
[6] McGhee, K.H., NCHRP Synthesis 334: Automated Pavement Distress Collection
Techniques, Transportation Research Board of the National Academies, Washington,
D.C., 2004.
[7] K. Gopalakrishnan, Deep learning in data-driven pavement image analysis and
automated distress detection: a review, Data 3 (3) (2018) 1–19,
https://doi.org/10.3390/data3030028.
[8] K.R. Benson, G.E. Elkins, W. Uddin, W.R. Hudson, Comparison of methods and
equipment to conduct pavement distress surveys, Transp. Res. Rec. 1196 (1988) 40–
50 http://onlinepubs.trb.org/Onlinepubs/trr/1988/1196/1196-005.pdf
[9] M.J. Markow, Highway management systems: state of the art, J. Infrastruct. Syst. 1
(3) (1995) 186–191, https://doi.org/10.1061/(ASCE)1076-0342(1995)1:3(186).
[10] R.B. Kulkarni, R.W. Miller, Pavement management systems: past, Present, and
future, in: Transportation Research Record 2003 (1853) 65–71,
https://doi.org/10.3141/1853-08.
[11] R. Haas, W.R. Hudson, L.C. Falls, Pavement Asset Management, Scrivener
Publishing, Beverly, MA, 978-1-119-03870-2, 2015,
https://doi.org/10.1002/9781119038849.
[12] ASCE, 2017 ASCE Infrastructure Report Card - Roads, American Society of Civil
Engineers, https://www.infrastructurereportcard.org/, (2017).
[13] S.A. Dewan, Pavement Management and Asset Management Side-by-Side, 6 th
International Conference on Managing Pavements. 34 (2004) pp. 1–12. http://
citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.652.3576&rep=rep1&type=pdf.
[14] L. Amador-Jiménez, N. Matout, A low cost solution to assess road's roughness
surface condition for Pavement Management, Transportation Research Board 2014
Annual Meeting 2014, pp. 1–16.
https://pdfs.semanticscholar.org/ad95/d721ac58555dda3e67e45eef0344c8b05f6e.pdf
[15] B. Becerik-Gerber, S.F. Masri, M.R. Jahanshahi, An Inexpensive Vision-Based App.
roach for the Autonomous Detection, Localization, and Quantification of Pavement
Defects, Innovations Deserving Exploratory Analysis Programs 169,2015.
http://onlinepubs.trb.org/onlinepubs/IDEA/FinalReports/Highway/NCHRP169_Final
_Report.pdf.
[16] Y. Du, C. Liu, D. Wu, S. Li, Application of Vehicle Mounted Accelerometers to
Measure Pavement Roughness, International Journal of Distributed Sensor Networks.
(2016), pp. 1–8. doi:https://doi.org/10.1155/2016/8413146.
[17] W. Aleadelat, K. Ksaibati, Estimation of pavement serviceability index through
android-based smartphone application for local roads, Transp. Res. Rec. 2639 (2017)
129–135, https://doi.org/10.3141/2639-16.
[18] M.M. Aydın, M.S. Yıldırım, L. Forslof, THE USE OF SMART PHONES TO
ESTIMATE ROAD ROUGHNESS: A CASE STUDY IN TURKEY, E-Journal of
New World Sciences Academy 13(3) (2018), pp. 247–257.
doi:10.12739/nwsa.2018.13.3.1a0416.
[19] K. Gopalakrishnan, S.K. Khaitan, A. Choudhary, A. Agrawal, Deep convolutional
neural networks with transfer learning for computer vision-based data-driven
pavement distress detection, Constr. Build. Mater. 157 (2017) 322–330,
https://doi.org/10.1016/j.conbuildmat.2017.09.110.
[20] H. Maeda, Y. Sekimoto, T. Seto, T. Kashiyama, H. Omata, Road Damage Detection
and Classification Using Deep Neural Networks with Smartphone Images, Computer-
Aided Civil and Infrastructure Engineering (2018), pp. 1–15.
doi:https://doi.org/10.1111/mice.12387.
[21] H. Zakeri, F.M. Nejad, A. Fahimifar, Image based techniques for crack detection,
Classification and Quantification in Asphalt Pavement: A Review, Archives of
Computational Methods in Engineering 24 (4) (2017) 935–977,
https://doi.org/10.1007/s11831-016-9194-z.
[22] Peterson, D.E., NCHRP Synthesis 135: Pavement Management Practices,
Transportation Research Board, National Research Council, Washington, D.C., 1987,
135 pp.
[23] Cowe Falls, L., S. Khalil, W.R. Hudson, and R. Haas, “Long-Term Cost-Benefit
Analysis of Pavement Management System Implementation,” Conference
Proceedings—Volume 2, Third International Conference on Managing Pavements,
May 22–26, 1994, San Antonio, Tex., 1994.
[24] Gramling, W.L., NCHRP Synthesis 203: Current Practices in Determining Pavement
Condition, Transportation Research Board, National Research Council, Washington,
D.C., 1994.
[25] Zimmerman, K.A. and ERES Consultants, Inc., NCHRP Synthesis 222: Pavement
Management Methodologies to Select Projects and Recommend Preservation
Treatments, Transportation Research Board, National Research Council, Washington,
D.C., 1995.
[26] Zimmerman, K.A. and D.G. Peshkin, “Issues in Integrating Pavement Management
and Preventive Maintenance,” Transportation Research Record: Journal of the
Transportation Research Board, No. 1889, Transportation Research Board of the
National Academies, Washington, D.C., 2004.
[27] Zimmerman, K.A., J. Corley-Lay, J.B. Wlaschin, and R.M. Tetreault, “Findings from
the International Scan on Managing Pavements and Monitoring Performance,”
Transportation Research Record: Journal of the Transportation Research Board, No.
2366, Transportation Research Board of the National Academies, Washington, D.C.,
2013, pp. 43–49.
[28] Zimmerman, K.A., L.M. Pierce, and J. Krstulovich, Pavement Management
Roadmap, Report FHWA-HIF-11-011, Federal Highway Administration,
Washington, D.C., 2010.
[29] G. Ping, R. Ong, Automated Pavement Condition Data Collection Quality Control,
Quality Assurance and Reliability, FHWA/IN/JTRP-2009/17. Joint Transportation
Research Program Indiana Department of Transportation and Purdue University, West
Lafayette, Indiana, 2010. doi:https://doi.org/10.5703/1288284314288
[30] AASHTO, Guide for Design of Pavement Structures, AASHTO, Washington, D.C.,
1993. https://habib00ugm.files.wordpress.com/2010/05/aashto1993.pdf. ISBN: 1-
56051-055-2.
[31] C. Rolt, J Parkman, CHARACTERISATION OF PAVEMENT STRENGTH IN
HDM III AND CHANGES ADOPTED FOR HDM-4, 10th International Conference
on Road Engineering Association of Asia and Australia 2000, (2000), pp. 1–11.
https://pdfs.semanticscholar.org/e587/ca3ce8c0257de4b3627de3bbef240fd546b4.pdf.
[32] B.B. Reddy, A. Veeraragavan, Structural performance of inservice flexible
pavements, J. Transp. Eng. 123 (1997) 156–167, https://doi.org/10.1061/(ASCE)
0733-947X(1997)123:2(156).
[33] E. Horak, ASPECTS OF DEFLECTION BASIN PARAMETERS USED IN A
MECHANISTIC REHABILITATION DESIGN PROCEDURE FOR FLEXIBLE
PAVEMENTS IN SOUTH AFRICA, Univesity of Pretoria, Ertoria, PhD Thesis
(1987). http://hdl.handle.net/2263/23960.
[34] M.A. Elseifi, A.M. Abdel-Khalek, K. Gaspard, Z. Zhang, S. Ismail, Evaluation of
continuous deflection testing using the rolling wheel deflectometer in Louisiana, J.
Transp. Eng. 138 (4) (2012) 414–422, https://doi.org/10.1061/(ASCE)TE.1943-
5436.0000349.
[35] L. Zhou, Q. Wu, J. Ling, Comparison of FWD and Benkelman Beam in evaluation of
pavement structure capacity, in: Geotechnical Special Publications, 203, 2010, pp.
405–411. doi:https://doi.org/10.1061/41104(377)51.
[36] E. Horak, Benchmarking the structural condition of flexible pavements with
deflection bowl parameters, J. South Afr. Inst. Civil Eng. 50(2) (2008) pp. 2–9.
http://www.scielo.org.za/pdf/jsaice/v50n2/01.pdf.
[37] O. Elbagalati, M.A. Elseifi, K. Gaspard, Z. Zhang, Prediction of in-service pavement
structural capacity based on traffic-speed deflection measurements, J. Transp. Eng.
142 (11) (2016) 1–8, https://doi.org/10.1061/(ASCE)TE.1943-5436.0000891.
[38] W. Zaniewski, J; Hudson, S., Hudson, Pavement condition data analysis, J. Transp.
Eng. 114, (1987) pp. 413–421.https://doi.org/10.1061/(asce)0733-
947x(1987)113:4(413).
[39] T.B.J. Coenen, A. Golroo, A review on automated pavement distress detection
methods, Cogent Engineering 4 (1) (2017) 1–23,
https://doi.org/10.1080/23311916.2017.1374822.
[40] Chen, C., Ming, Y., Xu, S., et al., 2015a. A Road Feature Extraction Method Based
on Dection Division. China Patent No.103605135B. National Intellectual Property
Administration, Beijing.
[41] Chen, K., Tan, G., Lu, M., et al., 2015b. CRSM: a practical crowd sourcing-based
road surface monitoring system. Wireless Networks 22 (3), 1e15. Chen, L., Wu, J.,
Liu, Y., et al., 2005. A new design for speedway road inspecting system. Electronic
Measurement Technology 2, 32e34.
[42] Chen, X., Dong, Q., Zhu, H., et al., 2016. Development of distress condition index of
asphalt pavements using LTPP data through structural equation modeling.
Transportation Research Part C: Emerging Technologies 68, 58e69.
[43] Cheng, H., Miyojim, M., 1998. Automatic pavement distress detection system.
Information Sciences 108 (1e4), 219e240.
[44] Nejad, F.M., Zakeri, H., 2011. A comparison of multi-resolution methods for
detection and isolation of pavement distress. Expert Systems with Applications 38 (3),
2857e2872.
[45] Lu, G., Fan, Z., Sun, Z., et al., 2020. Improving the polishing resistance of cement
mortar by using recycled ceramic. Resources, Conservation and Recycling 158,
104796.
[46] Gu, R., Yoshizawa, T., Otani, Y., 1994. One-step phase shift 3-D surface profilometry
wi Mertz, C., Kozar, J., Miller, J.R., et al., 2002. Eye-safe laser line
[47] Mertz, C., Kozar, J., Miller, J.R., et al., 2002. Eye-safe laser line triper for outside use.
In: Intelligent Vehicle Symposium 2002, Versailles, 2002.
[48] Pennington, T.L., Xiao, H., May, R., et al., 2001. Miniaturized 3-D surface
profilometer using a fiber optic coupler. Optics & Laser Technology 33 (5), 313e332.
[49] Rosati, G., Boschetti, G., Biondi, A., et al., 2009. Real-time defect detection on highly
reflective curved surfaces. Optics and Lasers in Engineering 47 (3e4), 379e384.
[50] Wang, K.C.P., 2004. Automated Pavement Distress Survey through Stereovision.
NCHRP-IDEA Program Project Final Report. Transportation Research Board,
Washington DC.
[51] Wang, K.C.P., 2011. Elements of automated survey of pavements and a 3D
methodology. Journal of Southwest Jiaotong University (English Edition) 19 (1),
51e57.
[52] Wang, P., Bai, X., 2018. Regional parallel structure based CNN for thermal infrared
face identification. Integrated Computer- Aided Engineering 25 (3), 247e260.
[53] Wang, K.C.P., Gong, W., 2005. Real-time automated survey system of pavement
cracking in parallel environment. Journal of Infrastructure Systems 11 (3), 154e164.
[54] Wang, J., Ma, J., Ma, R., et al., 2010. Road three-dimensional shape measurement
method with multi-laser-sensors. Journal of Wuhan University of Technology:
Transportation Science & Engineering 34 (6), 1202e1205.
[55] Wang, L., Sun, W., Lally, E.M., et al., 2012. Application of LADAR in the Analysis
of Aggregate Characteristics. NCHRP Report 724. Transportation Research Board,
Washington DC.
[56] Okawa, Y., 1984. Automatic inspection of the surface defects of cast
metals.Computer Vision, Graphics,and Image Processing 25 (1), 89e112.
[57] Chu, C., Yang, H., Wang, L., 2019. Design of a pavement scanning system based on
structured light of interference fringe. Measurement 145, 410e418.
[58] Lally, E.M., 2010. Fourier Transform Interferometry for 3D Mapping of Rough and
Discontinuous Surfaces (PhD thesis). Virginia Polytechnic Institute and State
University, Blacksburg.
[59] Wang, L., Sun, W., Lally, E.M., et al., 2012. Application of LADAR in the Analysis
of Aggregate Characteristics. NCHRP Report 724. Transportation Research Board,
Washington DC.
[60] Gopalakrishnan, K., 2018. Deep Learning in Data-driven Pavement Image Analysis
and Automated Distress Detection: a Review. Iowa State University, Ames.
[61] Cha, Y.J., Choi, W., Buyukozturk, O., 2017. Deep learning-based crack damage
detection using convolutional neural networks. Computer-Aided Civil and
Infrastructure Engineering 32 (5), 361e378.
[62] Cha, Y.J., Choi, W., Suh, G., et al., 2018. Autonomous structural visual inspection
using region-based deep learning for detecting multiple damage types. Computer-
Aided Civil and Infrastructure Engineering 33 (9), 731e747.
[63] Peng, C., Yang, M., Zheng, Q., et al., 2020. A triple-thresholds pavement crack
detection method leveraging random structured forest. Construction and Building
Materials 263,120080.
[64] Zheng, Q., Tian, X., Yang, M., et al., 2020a. CLMIP: cross-layer manifold invariance
based pruning method of deep convolutional neural network for real-time road type
recognition. Multidimensional Systems and Signal Processing 32 (4), 239e262.
[65] Zheng, Q., Tian, X., Yang, M., et al., 2020b. PAC-Bayesian framework based drop-
path method for 2D discriminative convolutional network pruning. Multidimensional
Systems and Signal Processing 31 (5), 793e827.
[66] Zheng, Q., Yang, M., Tian, X., et al., 2020c. A full stage data augmentation method in
deep convolutional neural network for natural image classification. Discrete
Dynamics in Nature and Society 2, 1e11.
[67] Zheng, Q., Yang, M., Yang, J., et al., 2018. Improvement of generalization ability of
deep CNN via implicit regularization in two-stage training process. IEEE Access 6,
15844e15869.
[68] Zheng, Z., Washington, S., 2012. On selecting an optimal wavelet for detecting
singularities in traffic and vehicular data. Transportation Research Part C: Emerging
Technologies 25, 18e33.
[69] Gao, M., Wang, X., Zhu, S., et al., 2020. Detection and segmentation of cement
concrete pavement pothole based on image processing technology. Mathematical
Problems in Engineering 6, 1e13.
[70] Gao, Y., Jiang, W., Xiao, J., 2007. Pavement management system. Communications
Standardization 9, 73e76.
[71] Gao, Y., Zhong, R., Tang, T., et al., 2017. Automatic extraction of pavement
markings on streets from point cloud data of mobile LiDAR. Measurement of Science
and Technology 28 (8), 085203.
[72] Tang, W., Zhao, Q., Huang, S., et al., 2020. Iteratively optimized patch label
inference network for automatic pavement disease detection. arXiv,
arXiv:2005.13298v1.
[73] Huang, J., 2013. Research on the Key Technologies of Pavement Crack Inspection
Based on 2D Image and Depth Information (PhD thesis). Harbin Institute of
Technology, Harbin.
[74] Huang, W., Xiao, W., Lu, X., et al., 2005. A study on pavement surface distress
image feature extraction based on distributive characteristics of image tiles. China
Civil Engineering Journal 38 (10), 54e60.
[75] Huang, Y., Tsai, Y., 2011. Dynamic programming and connected component analysis
for an enhanced pavement distress segmentation algorithm. Transportation Research
Record 2225, 89e98.
[76] Yin, J., Peng, T.H., Kuan, J.L., et al., 2019. Towards perspective-free pavement
distress detection via deep learning. In: IEEE 8 th Global Conference on Consumer
Electronics (GCCE), Tokyo, 2019.
[77] Tong,Z.,Yuan,D.,Gao, J., etal.,2020. Pavement detection with fully convolutional
network and an uncertainty framework. Computer-Aided Civil and Infrastructure
Engineering 35 (8), 832e849.
[78] Tong, Z., Gao, J., Sha, A., et al., 2018. Convolutional neural network for asphalt
pavement surface texture analysis. Computer-Aided Civil and Infrastructure
Engineering 33 (12), 1056e1072.
[79] Wang, K.C.P., Gong, W., 2005. Real-time automated survey system of pavement
cracking in parallel environment. Journal of Infrastructure Systems 11 (3), 154e164.
[80] Chu, C., Yang, H., Wang, L., 2019. Design of a pavement scanning system based on
structured light of interference fringe. Measurement 145, 410e418.
[81] AASHTO, Guide for Design of Pavement Structures, AASHTO, Washington, D.C.,
1993. https://habib00ugm.files.wordpress.com/2010/05/aashto1993.pdf. ISBN: 1-
56051-055-2.
[82] C. Rolt, J Parkman, CHARACTERISATION OF PAVEMENT STRENGTH IN
HDM-III AND CHANGES ADOPTED FOR HDM-4, 10th International Conference
on Road Engineering Association of Asia and Australia 2000, (2000), pp. 1–11.
https:// pdfs.semanticscholar.org/e587/ca3ce8c0257de4b3627de3bbef240fd546b4.pdf
[83] B.B. Reddy, A. Veeraragavan, Structural performance of inservice flexible
pavements, J. Transp. Eng. 123 (1997) 156–167, https://doi.org/10.1061/(ASCE)
0733-947X(1997)123:2(156).
[84] E. Horak, ASPECTS OF DEFLECTION BASIN PARAMETERS USED IN A
MECHANISTIC REHABILITATION DESIGN PROCEDURE FOR FLEXIBLE
PAVEMENTS IN SOUTH AFRICA, Univesity of Pretoria, Ertoria, PhD Thesis
(1987). http://hdl.handle.net/2263/23960.
[85] M.A. Elseifi, A.M. Abdel-Khalek, K. Gaspard, Z. Zhang, S. Ismail, Evaluation of
continuous deflection testing using the rolling wheel deflectometer in Louisiana, J.
Transp. Eng. 138 (4) (2012) 414–422, https://doi.org/10.1061/(ASCE)TE.1943-
5436.0000349.
[86] Kadar, P., T. Martin, M. Baran, and R. Sen, “Addressing Uncertainties of
Performance Modeling With Stochastic Information Packages—Incorporating
Uncertainty in Performance and Budget Forecasts,” Conference Proceedings, Ninth
International Conference on Managing Pavements, Alexandria, Va., May 18–22,
2015.
[87] Tabatabaee, N. and M. Ziyadi, “Bayesian Approach to Updating Markov-Based
Models for Predicting Pavement Performance,” Transportation Research Record:
Journal of the Transportation Research Board, No. 2366, Transportation Research
Board of the National Academies, Washington, D.C., 2013, pp. 34–42.
[88] Li, J., L.M. Pierce, and J. Uhlmeyer, “Calibration of Flexible Pavement in
Mechanistic-Empirical Pavement Design Guide for Washington State,”
Transportation Research Record: Journal of the Transportation Research Board, No.
2095, Transportation Research Board of the National Academies, Washington, D.C.,
2009.
[89] S. Bemanian, P. Polish, G. Maurer, Pavement management system based on financial
consequence, in: Transportation Research Record 1940, 2005, pp. 32–37.
doi:https://doi.org/10.3141/1940-04.
[90] V. Sunitha, A. Veeraragavan, K.K. Srinivasan, S. Mathew, Cluster-based pavement
deterioration models for low-volume rural roads, ISRN Civil Engineering. 2012
(2012) 1–8, https://doi.org/10.5402/2012/565948.
[91] Zimmerman, K.A. and ERES Consultants, Inc., NCHRP Synthesis 222: Pavement
Management Methodologies to Select Projects and Recommend Preservation
 Treatments, Transportation Research Board, National Research Council, Washington,
D.C., 1995.
[92] Zimmerman, K.A. and D.G. Peshkin, “Issues in Integrating Pavement Management
and Preventive Maintenance,” Transportation Research Record: Journal of the
Transportation Research Board, No. 1889, Transportation Research Board of the
National Academies, Washington, D.C., 2004.
[93] Zimmerman, K.A., J. Corley-Lay, J.B. Wlaschin, and R.M. Tetreault, “Findings from
the International Scan on Managing Pavements and Monitoring Performance,”
Transportation Research Record: Journal of the Transportation Research Board, No.
2366, Transportation Research Board of the National Academies, Washington, D.C.,
2013, pp. 43–49.
[94] Zimmerman, K.A., L.M. Pierce, and J. Krstulovich, Pavement Management
Roadmap, Report FHWA-HIF-11-011, Federal Highway Administration,
Washington, D.C., 2010.
[95] Wu, H., Z. Zhang, K. Long, and M.R. Murphy, “Considering Safety Impacts of Skid
Resistance in Decision-Making Processes for Pavement Management,” Transportation
Research Record: Journal of the Transportation Research Board, No. 2455,
Transportation Research Board of the National Academies, Washington, D.C., 2014,
pp. 19–27.
[96] de León Izeppi, E., S. Katicha, G.W. Flintsch, and K.K. McGhee, “Pioneering the Use
of Continuous Pavement Friction Measurements to Develop New Safety Performance
Functions, Improve the Accuracy of Crash Count Predictions, and Evaluate Possible
Treatments for the Roads in Virginia,” TRB 95th Annual Meeting Compendium of
Papers, Jan. 10–14, 2016, Washington, D.C.
[97] Cambridge Systematics Inc., et al., NCHRP Report 632: An Asset–Management
Framework for the Interstate Highway System, Transportation Research Board of the
National Academies, Washington, D.C., 2009, 71 pp.
[98] Faghih-Imani, A. and L. Amador-Jimenez, “Toward Sustainable Pavement
Management: Incorporating Environmental Impacts of Pavement Treatments into a
Performance-Based Optimization,” Transportation Research Record: Journal of the
Transportation Research Board, No. 2366, Transportation Research Board of the
National Academies, Washington, D.C., 2013, pp. 13–21.
[99] Muench, S. and T. Van Dam, Climate Change Adaptation for Pavements, Report
FHWA-HIF-15-015, Federal Highway Administration, Washington, D.C., 2015.
[100]Pellecuer, L., G. Assaf, and M. St-Jacques, “Towards the Incorporation of
Environmental Impacts into Pavement Management Systems,” Transport Reviews,
Vol. 36, No. 3, Taylor and Francis, UK, 2016, pp. 361–382.
[101]Jiang, Y. J., Darter, M. I., Von Quintus, H. & Killingsworth, B. M., Designer: A
prototype knowledge-based expert system for selecting recommended highway
pavement design features, in Third International Conference on Road and Airfield
Pavement Technology, Proceedings, Vol. 1, Beijing, China, April 1998, pp. 157–65.
[102]Ritchie, S. G., Expert systems in pavement management, Transportation Research,
21A (2) (1987), 145–52.
[103]Ritchie, S. G., Che, I. Y., Mahoney, J. P. & Jackson, N. C., Surface condition expert
system for pavement rehabilitation planning, ASCE Journal of Transportation
Engineering, 113 (2) (1987), 155–67.
[104]Lee, H. & Galdiero, V., Enhancing the existing pavement management systems by
using expert system techniques, in First International Conference on Applications of
Advanced Technologies in Transportation Engineering, San Diego, CA, February
1989, pp. 76–81.
[105]Kotb, A. S. & Moore, C. J., A decision support expert system for maintenance and
rehabilitation needs, in World Transport Research, Proceedings of the 7th World
Conference, Vol. 4: Transport Management, 1996, pp. 397–410.
[106]Kuncheria, P. I. & Veeragavana, A., PADMA: Expert system for flexible pavement
deterioration and maintenance, Indian Highways, 24 (9) (1996), 27–47.
[107]Grivas, D. A. & Shen, Y. C., Use of fuzzy relations to manage decisions in preserving
civil infrastructure, Transportation Research Record, 1497 (1995), 10–26.
[108]Attoh-Okine, N. O., Predicting roughness progression in flexible pavements using
artificial neural networks, in Third International Conference on Managing Pavements,
San Antonio, TX, 1994.
[109]Eldin, N. N. & Senouci, A. B., A pavement condition rating model using
backpropagation neural networks, Microcomputers in Civil Engineering, 10 (6)
(1995), 433–41.
[110]Van der Gryp, A., Bredenhann, S. J., Henderson, M. G. & Rohde, G. T., Determining
the visual condition index of flexible pavements using artificial neural networks, in
Fourth International Conference on Managing Pavements, Durban, South Africa, May
17 to 21, 1998, Vol 2, pp. 115–29.
[111]Kaseko, M. S. & Ritchie, S. G., Pavement image processing using neural networks, in
Proceedings of the Second International Conference on Applications of Advanced
Technologies in Transportation Engineering, American Society of Civil Engineers,
New York, 1991, pp. 238–42.
[112]Banan, M. R., & Hjelmstad, K. D., Neural networks and AASHO road test, Journal of
Transportation Engineering, 122 (5) (1996), 358–66.
[113]George, K. P., El-Rahim, A. M. A. & Shekharan, A. R., Updates of pavement
performance modeling, in Third International Conference on Road and Airfield
Pavement Technology, Proceedings, Vol. 1, Beijing, China, April 1998, pp. 402–10.
[114]Kaseko, M. S., Lo, Z. P. & Ritchie, S. G., Comparison of traditional and neural
classifiers for pavement-crack detection, ASCE Journal of Transportation
Engineering, 120 (4) (1994), 552–69.
[115]Kaseko, M. S. & Ritchie, S. G., A neural network–based methodology for automated
distress classification of pavement images, in Proceedings of the International
Conference on Artificial Intelligence Applications in Transportation Engineering,
1992, pp. 195–214.
[116]Kaseko, M. S. & Ritchie, S. G., A neural network–based methodology for pavement
crack detection and classification, Transportation Research, Part C: Emerging
Technologies, 1C (4) (1993), 275–91.
[117]Huang, Y. & Moore, R. K., Roughness level probability prediction using artificial
neural networks, Transportation Research Record, 1592 (1997), 89–97.
[118]Nallamothu, S. & Wang, K. C. P., Experimenting with recognition accelerator for
pavement distress identification, Transportation Research Record, 1536 (1996), 130–
5.
[119]La Torre, F., Domenichini, L. & Darter, M. I., Roughness pre-diction model based on
the artificial neural network approach, in Fourth International Conference on
Managing Pavements, May 1998, pp. 599–612.
[120]Roberts, C. A. & Attoh-Okine, N. O., A comparative analysis of two artificial neural
networks using pavement performance prediction, Conference Proceedings, Neural
Network Applications in Highway and Vehicle Engineering, 1996, pp. 59–73.
[121]Sundin, S., Predicting rut depth in pavements using artificial neural networks, in
Nimes ’98 Conference on Complex Systems, Intelligent Systems, & Interfaces, La
Lettre de l’IA Nimes 98, May 26–28, 1998, pp. 263–6.
[122]Hajek, J. J., Haas, R. C. G. & Phang, W. A., ROSE: A knowledge-based expert
system for routing and sealing, in Proceedings of the Second North American
Pavement Management Conference, Toronto, Canada, Nov. 1987, pp. 2.301–2.341.
[123]Kwasi, A. A. & Attoh-Okine, N. O., Effects of cross probabilities in evolutionary
algorithm trained to predict pavement performance, presented at the 78th
Transportation Research Board Annual Meeting, Washington, January 1999.
[124]Fwa, T. F., Chan, W. T. & Tan, C. Y., Genetic algorithm programming of road
maintenance and rehabilitation, Journal of Transportation Engineering, 122 (3)
(1996), 246–53.
[125]Hoque, K. Z., Fwa, T. F. & Chan, W. T., Programming of pavement maintenance
based on genetic algorithm, in Third International Conference on Road and Airfield
Pavement Technology, Proceedings, Vol. 2, Beijing, China, April 1998, pp. 1108–16.
[126]Fwa, T. F., Chan, W. T. & Hoque, K. Z., Network level programming for pavement
management using genetic algorithms, in Fourth International Conference on
Managing Pavements, Durban, South Africa, May 17 to 21, 1998, Vol. 2, pp. 815–29.
[127]Yuge, F., Kameyama, S., Ishitani, M., Himeno, K. & Kasahara, A., Optimization of
pavement maintenance and rehabilitation strategy considering bearing capacity, in
Third International Conference on Road & Airfield Pavement Technology,
Proceedings, Vol. 2, Beijing, China, April 1998, pp. 1047–55.
[128]Chou, C. J. & Liau, T., Development of automated algorithms for pavement condition
survey, Transportation Research Record, 1536 (1996), 103–9.
[129]Chou, J., O’Neill, W. A. & Cheng, H., Pavement distress evaluation using fuzzy logic
and moments invariants, Transportation Research Record, 1505 (1995), 39–46.
[130]Hu, M. K., Visual pattern recognition by moment invariants, IRE Transactions on
Information Theory, 8 (1962), 179–87.
[131]Belkasim, S. O., Shridhar, O. M. & Ahmadi, M., Pattern recognition with moment
invariants: A comparative study and new results, Pattern Recognition, 24 (12) (1991),
1117–38.
[132]Teague, M. R., Image analysis via the general theory of moments, Optical Society
American Journal, 70 (8) (1980), 920–30.
[133]Ritchie, S. G, Kaseko, M. & Bavarian, B., Development of an intelligent system for
automated pavement evaluation, Transportation Research Record, 1311 (1991), 112–
19.
[134]Ritchie, S. G., A knowledge-based approach to pavement overlay design,
Transportation Research Record, 1145 (1987), 61–8.
[135]Goh, A. T. C., A hybrid neural network based pavement management system, Road &
Transport Research, 6 (4) (1997), 62–71.
[136]Prechaverakul, S. & Hadipriono, F. C., Using a knowledge-based expert system and
fuzzy logic for minor rehabilitation projects in Ohio, Transportation Research Record,
1497 (1995), 19–24.
[137]National Academies of Sciences, Engineering, and Medicine. 2017. Pavement
Management Systems: Putting Data to Work. Washington, DC: The National
Academies Press. https://doi.org/10.17226/24682.
[138]Gowda, S., Kavitha, G. & Gupta, A. Economic Analysis and Prioritisation of Non-
core Roads in India: A Case Study. Int. J. Pavement Res. Technol. (2022).
https://doi.org/10.1007/s42947-022-00250-2.
[139]N.S.P. Peraka, K.P. Biligiri, Pavement asset management systems and technologies:
A review, Automation in Construction, 119 (2020), Article 103336,
10.1016/j.autcon.2020.103336