transportation agencies: automated evaluation (machine
observation and machine analysis) and manual evaluation
(human observation and human analysis). Most state or local
transportation agencies use one of these two aforementioned
methods or a combination of them. Each method has advan-
tages and disadvantages.
Only a few state and local transportation agencies are still
using manual methods to survey the surface distress of high-
way pavement (Bandini
et al.,
2012). Using this method, data
are collected by inspectors walking along a section of pave-
ment and rating the level of distress. These data are primarily
handwritten data and attached to archived images acquired by
inspectors on the ground. Manual evaluation methods can col-
lect detailed information for various types of distresses, and it
is the reason that this method is still used. However, this meth-
od is expensive, extremely labor intensive, time-consuming,
and data collected by different inspectors can exhibit a high
degree of variability (Bogus
et al.,
2010). Manual evaluation is,
therefore, sometimes unable to provide meaningful quantita-
tive information, and eventually leads to inconsistencies in
distress conditions over space and across evaluation (Cheng
et
al.,
1999; Hudson and Uddin, 1987; Wang and Li, 1999; Wang,
2000). In addition, manual evaluation relies on the subjective
evaluation of distress type, extent, and severity by a trained
inspector based on visual observation (Hudson and Uddin,
1987), which means the evaluation results are prone to subjec-
tive bias. Another problem with manual evaluation is that it
is potentially dangerous to inspectors. Crews must walk along
the side the road to perform their evaluation and, despite
safety precautions (e.g., safety training and high-visibility gar-
ments), are exposed to significant risk of personal injury.
In an attempt to address the shortcomings of manual evalu-
ation, many transportation agencies have adopted automated
technology to conduct distress surveys (Bandini
et al.,
2012).
The automated methods typically include the use of vehicle-
mounted electronic sensors at a fine enough spatial resolu-
tion to detect individual distress measures (e.g., cracks) in
the pavement surface. The application of automated surveys
based on a variety of electronic sensors (e.g., video cameras
and laser sensors) became common in the 1980s (Curphey
et
al.,
1985; Haas
et al.,
1985; Hudson and Uddin, 1987). These
various types of sensors are designed to detect and assess
either a specific type of individual distress such as trans-
verse cracks or a specific type of pavement such as concrete
(Uddin
et al.,
1987; Hudson
et al.,
1987; Hudson and Uddin,
1987; Mahler
et al.,
1991; Georgopoulos
et al.,
1995; Pynn
et al.,
1999; Lee and Kim, 2005; Huang and Xu, 2006; Zhou
et al.,
2006; Oliveira and Correia, 2008; Nguyen
et al.,
2009;
Coudray
et al.,
2010; Gavilan
et al.,
2011; Koch and Brilakis,
2011; Adarkwa and Attoh-Okine, 2013).
Although automated evaluations can collect detailed
information quickly and safely, and technological advances in
computer hardware and imaging recognition have improved
the performance of the automated evaluation methods, seri-
ous problems still remain in the areas of implementation
costs, processing speed, and accuracy (Wang, 2000). Automat-
ed methods require significant time to process data to extract
useful information, since it requires very complicated ana-
lytical models and algorithms (Wang, 2000). These methods
require substantial technical expertise and are expensive to
deploy, requiring specially trained operators. In addition, data
are collected on the ground as a single task and cannot be
shared with other partner agencies to reduce the cost because
a single image can only cover a small area which is usually
less than five square meters (McGhee, 2004). For example,
the Vermont Agency of Transportation reported costs of up
to $170
USD
per mile in urban areas for automated evaluation
(McGhee, 2004).
Pavement Surface Distress Evaluation from Aerial Photography
Pavement evaluation from aerial photography is not a new
idea, but is also not used for operational evaluation of pave-
ment surface distresses yet. The application of an aerial
photography-based evaluation method to pavement surface
distress was first implemented in the 1950s. Several studies
(McMaster and Legault, 1952; Stoeckeler, 1968; Stoeckeler,
1970) focused on visually comparing analog panchromatic
aerial photographs to determine pavement surface distress.
They concluded that untreated cracks and other pavement
defects (e.g., patching and bleeding) can be identified through
the visual analysis. Although they concluded that visual
analysis of panchromatic analog aerial photography is a prac-
tical means of conducting pavement condition surveys, it is
not used for operational pavement surface distress evaluation.
This is because cracks are distinguishable only in large scale
(e.g., 1:100) analog panchromatic aerial photographs and the
associated cost is extremely high
.
Chen
et al
. (2011) proposed a method of analyzing very
high spatial resolution (
VHSR
) multispectral digital aerial
photography to detect large cracks on bridge pavement. Crack
type, length, width, and severity were measured from the
post-processed
VHSR
multispectral digital aerial photography,
and then these measurement results were used in a bridge
surface condition index (
BSCI
) rating system to calculate the
distress conditions. However, in their research, the collec-
tion of the
VHSR
multispectral digital aerial photography was
customized to fulfill the specific bridge pavement evaluation
purpose. In other words, the collected aerial photos, like the
pavement evaluation methods that are currently used opera-
tionally, serve only a single purpose and therefore represent
an expensive option for routine distress evaluation
.
There are many programs to routinely collect
HSR
mul-
tispectral digital aerial photography. For example, with the
support of the National Agricultural Imagery Program (
NAIP
),
the United States Geological Survey (
USGS
) and the United
States Department of Agriculture (
USDA
) regularly acquire
digital, color-infrared, ortho-corrected aerial photography
which covers all states at 1-meter spatial resolution, and they
provide the data to the public for free. Many counties and cit-
ies now routinely acquire natural color 6-inch (0.1524-meter)
and even 3-inch (0.0762-meter) spatial resolution, ortho-
corrected aerial photos. In addition, some states have initiated
the program to regularly acquire statewide aerial photos. For
example, the State of Missouri images the entire state regular-
ly with 2-foot (0.6096-meter) spatial resolution multispectral
digital aerial photographs (Wright, 2014). It would not be hard
to imagine more states to moving to do so because the uses for
these data continue to expand.
Past and current research for pavement distress evaluation
has focused on the detection of individual distresses (e.g., an
individual crack). This information is commonly aggregated to
determine the overall level of pavement distress, which is then
used by transportation agencies for planning purposes. Ac-
cording to Stoeckeler (1970), cracks are only distinguishable
in large-scale (e.g., 1:100) aerial photographs. A key limitation
of the routinely collected
HSR
multispectral aerial photography
is that its spatial resolution is too coarse to enable the detec-
tion and quantification of individual distresses. As a result,
this research does not focus on assessing individual distresses,
but rather, on estimating the overall condition by analyzing
the brightness and variation of resolution cells. Specifically,
the research presented here is focused on analyzing
HSR
multispectral digital aerial photographs to determine overall
pavement distress rates through pixel-based spectral response
assuming an L-resolution scene model (Strahler, 1986).
710
September 2015
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
