Hindawi Publishing Corporation

Journal of Construction Engineering

Volume 2015, Article ID 721380, 8 pages
http://dx.doi.org/10.1155/2015/721380

Research Article
Hard-Hat Detection for Construction Safety Visualization

Kishor Shrestha,1 Pramen P. Shrestha,1 Dinesh Bajracharya,2 and Evangelos A. Yfantis3

1
Department of Civil and Environmental Engineering and Construction, Howard R. Hughes College of Engineering,
University of Nevada, Las Vegas, NV 89154, USA
2
Cerner Corporation World, Kansas City, MO 64117, USA
3
Department of Computer Science, Howard R. Hughes College of Engineering, University of Nevada, Las Vegas, NV 89154, USA

Correspondence should be addressed to Pramen P. Shrestha; [email protected]

Received 30 September 2014; Accepted 17 January 2015

Academic Editor: F. Pacheco-Torgal

Copyright © 2015 Kishor Shrestha et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

In 2012, 775 fatalities were recorded, and many more were injured at construction sites in the United States. Of these, 415 fatalities
(54%) were due to fall, slips, and trips as well as being struck by falling objects. In order to decrease fatalities at construction
sites to these types of events, the Occupational Safety and Health Administration (OSHA) provides Fall Prevention and OSHA-10
trainings to construction workers. Moreover, safety personnel monitor whether the workers use personal protective equipment
(PPE) properly. Data shows that construction fatalities have decreased by 2% annually since 1994; however, the owners still are not
satisfied with this result. Various studies have shown that fall is the biggest contributor for construction fatalities. One study showed
that half of the fall fatalities were because the workers either had not used PPEs or had not used them properly. In addition, studies
showed that, with proper use of hard hats, the fatalities due to fall, slips, trips, and being struck by falling objects could be reduced.
This study developed and tested a hard-hat detection tool that uses image-processing techniques to identify whether workers are
wearing hard hats. The tool dispatches warning messages if the workers do not use hard hats.

1. Introduction double that of the United States [5]. This higher rate of
construction fatalities in developing countries had construc-
In the United States, many people work at jobsites under tion managers concerned. Furthermore, the construction
unsafe conditions, and thousands lose their lives every year. industry in the US, the largest industry, had a high rate
In the US, 4,383 fatal work injuries in workplaces were of construction fatalities and injuries [3]. Bureau of Labor
recorded in 2012; on average, this was 89 fatalities a week and Statistics data from 1990 onwards showed that the national
nearly 12 fatalities every day. Construction jobs are among fatality rate began to decrease (by 2% annually) from 1994;
the 10 deadliest jobs [1, 2]. The fatality rate—the ratio of however, it is increasing in Nevada. In 2011, construction
number of fatalities and total hours worked by all workers— industry had the second largest number of fatalities in the
at construction sites in 2012 was 3.4 [3]. The fatality rate United States [6, 7].
reflects the number of fatalities per 100,000 full time workers. Of the 4,383 fatal injuries in workplaces in the United
According to OSHA [4], the fatality rate is calculated as States in 2012, 4,175 were injured in private industry and
Number of fatal work injuries approximately one-fifth (19.30%) of total private industry
Fatality Rate = fatalities were recorded in the private construction industry
Total hours worked by all workers (1) [8]. The leading causes of the construction site fatalities
∗ 200,000. were falls, slips, being struck by objects, electrocution, and
being caught in/between objects [8]. Specifically, the deaths
The fatality rate in some developing countries is much higher in construction jobs due to falls were 34.6% of total deaths in
than in developed countries. For example, in Republic of construction [8]; this percentage was 49.9% during the 1980s
Korea, the fatality rate in construction industry is more than and first half of 1990s [9]. Table 1 shows the percentage of
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 Journal of Construction Engineering

Table 1: Fall fatality percentage by type of work [3].

Year Fatal falls to lower levels Roof Ladders Scaffolds and staging Subtotal Percentage
2012 570 124 133 58 315 55.3
2011 563 109 122 64 295 52.4
2010 522 117 132 44 293 56.1
2009 518 109 122 53 284 54.8
2008 593 123 119 68 270 52.3
2007 746 161 132 88 381 51.0

fatal falls by type of work for the last six years. The number of owners follow the safety rules to avoid injuries at the site. Due
fatal falls and the percentage of fatal falls from roofs, ladders, to various reasons, workers at construction sites sometimes
scaffoldings, and staging have increased since 2007. fail to obey the OSHA rules and regulations, for example,
In most of the fall incidents, the workers fall from heights under extreme weather conditions or due to stress in meeting
and hit their heads on hard floors. An investigation report work deadlines.
showed that half of the fall incidents were recorded from Because of OSHA’s involvement, most states have decreas-
a height of less than equal to 3 m; moreover, the origins of ing fatalities in the workplaces [3]; however, safety engineers
57% of the fall incidents were from ladders, roofs, buildings and construction managers are not satisfied with the current
under construction, and platforms or scaffolding [10]. Hard visual monitoring method to check workers on whether
hats are designed to resist shock and penetration by objects they are using PPEs. They are looking for an innovative
as well as contact with electrical hazards. If workers wear method to monitor the workers more comprehensively. In
hard hats properly, half of the fatalities due to falls and a the traditional supervision (visual monitoring) method, the
significant number of fatalities due to slips, trips, and struck safety engineers occasionally fail to enforce construction
by falling objects could be expected to decrease [11]. In one workers to use their hard hats because they could not check
study that investigated the number of construction fatalities the workers every hour and every day. However, if the safety
and the use of PPEs, the results showed that 47.3% of fatally engineers could monitor the workers in real time by using
injured victims either had not used PPEs or had not used video transmitted from the site, the safety rules regarding
them properly [5]. hard hats could be enforced more effectively. This could
Figure 1 shows a typical construction site. On the left, all reduce the fatality rates due to falls, slips, trips, and being
the workers are wearing hard hats except one, who is holding struck by falling objects. This study developed a tool of
his hard hat in his hand. On the right, two workers are not identifying the workers not wearing hard hat on sites.
wearing hard hats during their work on a scaffold. These types To develop the tool that automatically detects workers
of pictures transmitted from the construction sites can be who are not wearing hard hat in the real-time video, a
analyzed to determine whether the workers are following the visualization approach was used. The visualization approach
safety rules. In this study, a software tool was developed to is an innovative software tool that monitors the workers in
automatically detect whether the workers are wearing hard real time and dispatches warning messages to concerned
hats at construction site. personnel once the safety rules (wearing hard hats properly
The National Institute for Occupational Safety and Health at this stage) are violated. This approach, shown in Figure 2,
[9] investigated fatal injuries due to falls at construction sites. consists of
Data showed that, from 1980 to 1994, falls were the leading
cause of occupational fatalities. At construction sites during (1) a closed-circuit television (CCTV) camera installed at
that period, falls from ladders, scaffolding, and buildings and construction site,
other structures as well as falls from one level to another (2) a wired/wireless network to transfer videos taken by
were 12.3%, 13%, 34.7%, and 16.6%, respectively, during CCTV camera to the server (office computer),
1980 to 1994; from 1982 to 1997, they were 8.8%, 18.7%,
63.8%, and 8.8%, respectively. From 1980 to 1994, the dataset (3) a server at the nearest office,
of fatalities due to falls was compared among 11 divisions (4) speakers at the office and the site to warn against
(e.g., construction, agriculture, mining, manufacturing, and safety violations,
transportation). The data in this study showed that the
fatalities due to falls were 49.9% of the total fatalities, for a (5) cell phones with local area network (LAN) wireless to
fatality rate of 3.89. send warning messages to the concerned people.
To decrease injuries by enforcing work-related safety (6) Images of the construction site are displayed continu-
rules, the US government established the Occupational Safety ously on the office computers. Using real-time images,
and Health Administration (OSHA) in 1970. OSHA prepares the software program detects whether the workers are
guidelines for work safety and offers grants for safety training using their hard hats properly. When the program
to construction workers in order to train them about the detects a worker at the site without a hard hat, the
importance of using PPEs properly [12]. In addition, OSHA program triggers warning messages. This tool was
monitors construction sites to make sure that contractors and developed using image processing software.
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Construction Engineering 3

Figure 1: A typical construction site where some workers did not use hard hat and some used it improperly.

2.1. Computer Visualization. Teizer and Reynolds [17] stud-

ied and designed a “smart hat” to prevent construction
CCTV cameras workers from contact with heavy equipment, using radio
frequency (RF) technology. A wireless silicon microprocessor
Warning message on was attached to the hard hats as well as to the heavy
cell phone equipment. Once the workers ventured close to the heavy
A construction site equipment, as specified for the microprocessor, the system
broadcasted a high-pitched alarm to warn them. The authors
Server
(office computer) claimed that when construction workers and heavy equip-
ment share a workplace and safety concerns are high, this
technology is very useful.
Warning speakers The authors investigated the performance of an RGB-
Warning messages
on other computers D sensor to track a worker and or an object in terms of
three parameters [18]: (1) 3D movement of a worker’s body
Figure 2: A schematic diagram of construction safety visualization.
position, as tracked by the sensor; (2) 3D rotation angles
of the body parts; and (3) analysis of the sensor’s accuracy
regarding body-part movement. These three datasets were
This current study is an extension of previous studies in the
gathered during experimental studies. To compare these
field of construction safety visualization, computer vision,
datasets, another dataset also was taken, using six 4-mega-
pattern recognition, and Internet transmission [13–16]. The
pixel VICON sensors, which were considered to give the most
tool was developed using an image processing method to
accurate results. The comparative results showed that the
detect the face of a worker and then used edge detection
discrepancy of the location track of the RGB-D sensors was
program and segmentation methods to identify whether the
10.70 cm, and difference in rotation angles was 16.20 degrees.
worker had a hard hat. The software tool was developed by
However, the difference in results regarding movement anal-
a combined effort of construction engineers and computer
ysis was insignificant.
science professionals and then tested in the Construction
Management Laboratory at University of Nevada, Las Vegas
(UNLV). 2.2. Image Processing. Shrestha et al. [16] developed a frame-
The hard-hat detection tool could detect whether workers work for construction safety visualization and developed and
are wearing hard hats. At this initial stage of program devel- tested an edge detection algorithm. The framework included
opment, its scope is limited to hard hats. The use of this tool the installation of a set of CCTV cameras, a powerful server
primarily is applicable at building construction sites where at the site office, and equipment to send warning messages to
workers can be observed by cameras from the surrounding safety-related personnel once the program detected workers
environment. Technical difficulties in using this tool occur not wearing hard hats in the construction sites. The server was
in linear construction projects, such as road construction or a powerful office computer equipped with two algorithms,
buildings with limited range of the cameras. edge detection and segmentation. This study developed the
segmentation algorithm in order to identify the worker and
2. Literature Review hard hats.
In this study, Curio et al. [19] used an image-processing
The literature review to frame this study focused on two method to detect a walking pedestrian. The program con-
main areas, computer visualization and image processing. cluded that a walking pedestrian in the image needed to
The papers reviewed in these areas are summarized as follows. satisfy two conditions. The first condition was that the image
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 Journal of Construction Engineering

matched the outline of a human; the second condition was the location of the workers. To ensure that the image was a
the periodic movement of legs while walking. Taking an construction worker, the person detected by image process-
image using two cameras, stereovision could be produced. ing needed to have a hard hat on. The hard hat was detected
Stereovision could be used for short-range and midrange using pattern recognition.
distances from the camera to the pedestrian. Escorcia et al. [24] detected workers and their actions
Tsai et al. [20] studied the detection of defective traffic at construction sites using Kinect sensor technology. An
signals by using an image-processing technique. The recog- algorithm based on machine-learning techniques was used
nition of the traffic sign types, the exact location of the for this study. The video-log images were analyzed to
sign for inventory purposes, and identification of signboard determine the accurate actions of the construction workers.
conditions—retroreflectivity, faded sign colors, tilted signs, This program is useful to assess productivity, safety, and
and sign boards blocked by objects—could be identified using occupational health in indoor environments.
the program. In the detection process, traffic signs were
identified in terms of their shape, color, background, and
2.3. Gaps in the Literature. Many researches regarding the
legend. A crucial step for the image-processing algorithms
detection of workers had been conducted. Some researchers
was to separate the images that contained traffic signs from
used sensors attached to hard hats and other parts of body
those that did not. During the program execution, first, the
to track them; however, construction worker and hard-hat
traffic sign is detected and, second, recognition is achieved.
detection in real time using image-processing technique is a
This study dealt only with traffic detection.
new method. There is an advantage in using image processing
Canny [21] developed an edge detection algorithm, which
over sensor method to detect the workers and objects. The
had five distinct steps.
studies showed that the sensors used to track objects lacked
(1) Smoothing: this is a blurring of an image. Every image accuracies. However the image-processing technique can
has some amount of noise in it, and a Gaussian filter accurately detect the workers and objects. In addition, there
is used to smooth it. is no study conducted on hard-hat detection in real time.
(2) Finding gradients: these are edges in a grayscale image
where the grayscale intensity changes the most. This 3. Research Methodology
is identified by determining the gradients.
The visualization approach is a new technique. Nowadays,
(3) Nonmaximum suppression: the maxima in the gradi- visualization technique, which is used as a proactive tool,
ent image are preserved, and the rest is erased. can be a very useful to automatically detect construction
(4) Double thresholding: the pixels that remain after Step workers as well as their PPEs. It consists of a set of CCTV
(3) are marked with their strength, pixel by pixel. cameras installed at the site, a network connection between
cameras, an office computer (server), a server which acts as
(5) Edge tracking by hysteresis: strong edges and weak a data analysis computer in the office, display computers in
edges connected with strong edges considered “cer- the office, warning speakers installed at the site office, and
tain edges.” cell phones to deliver safety rule violation alert messages.
Park et al. [22] conducted research to determine whether the The visualization technique is divided into three parts: image
person at a construction site is actually a worker. A worker input, image analysis, and image output.
was confirmed by fulfilling two conditions, the outline of a Regarding image input, a set of high-resolution CCTV
person and the presence of PPEs. The program developed in cameras are installed at the site, which take real-time images
this study analyzed the video frames of the construction site and send them to the office computer by means of a wired
by using an image-processing technique to separate moving and or wireless network. Installing the cameras with proper
objects from background images in order to identify the overlap, stereovision can be achieved that can determine the
outline of the person. After the person was detected by exact location of the workers and/or objects. The cameras
analyzing the pixels of the images, the person was identified use charged-couple devices (CCDs) to capture high-quality
as a worker or not. In this experiment, a worker wore a vest video. The video is transmitted to the office computer using a
and a hard hat, which has a higher pixel rate than a person wired local area network (LAN) or else a wireless network.
wearing normal dress. From the differences in pixels, the Regarding image analysis, the transmitted video is split
system determined whether the person was a worker. into digital images. A video is a combination of several
Han et al. [18] used a vision-based motion detection to images. For this program, a one-second video is split into
track unsafe working behavior of construction workers, using 32 images; then, each image is analyzed with an image-
video camera images. A 3D model of a worker was developed processing technique. When 32 images are analyzed, this is
using images from two different cameras. The 3D model was the analysis of the one-second video. During image process-
analyzed to figure out whether the worker’s movement was ing, a worker can be detected using various methods, for
safe. However, this study did not apply a real-time image- example, walking pedestrian recognition or face detection
processing technique. [19]. For this study, the face-detection method is used. When
Tharindu et al. [23] detected workers at construction a worker faces the camera, his or her face is detected in
sites using an image-processing technique with a Kinect an image or a series of images. This, then, is considered a
sensor. Using video camera images, the program identified worker that is detected at the construction site. Because more
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Construction Engineering 5

Transfer video usually used. For the visualization technique using image
from construction
site at real time
processing, face detection is used. During data analysis, the
face detection program checks the presence of faces in each
and every image of the site surveillance. In this program,
Split video into an important assumption in face detection to be considered
number of
images is worker detection. Once the program detects one or more
faces, it calls the hard-hat detection program. If any faces or
workers were not detected in an image, it repeats the same
Analyze an image
using image
procedure in other images.
processing Figure 4 (left side) shows a worker’s face that is detected.
In this program development phase, in order to track whether
a face is detected, a black rectangular box around the face is
Is displayed, indicating that a face is detected. Figure 4 (right
a worker/s No
detected? side) shows a blue rectangular box at the upper head region
of the person, which indicates that the program is searching
Yes for a hard hat on head.
Is
Working Yes hard hat 5. Hard-Hat Detection Program Development
safely detected?
In the visualization technique (Figure 3), after the face
No detection program, the hard-hat detection program is imple-
mented. The objective of a hard-hat detection program is to
Dispatch warning messages Dispatch warning messages detect a hard hat by checking if a hard hat exists at the upper
to the office computers and to the cell phones of
speakers installed at site concerned personnel head region of that person. During implementation of this
program, the following steps are executed.
Figure 3: Framework for identifying hard hats with a worker.
(1) Focus on the upper head region, up to 12 inches above
the face.
than one camera is installed at a site, a worker’s face can be (2) Apply the edge detection program on the upper head
visualized as a front view to one of the cameras. region, which converts the digital images into outlines
As shown in Figure 3, once the program detects a worker, of the objects. The process of converting images into
the face-detection program calls a hard-hat detection pro- outlines of the objects has three stages. First, the
gram, which checks whether a hard hat can be identified program uploads an image; the hard hat image shown
in the image. Similar to the face-detection method, vari- in Figure 5 (left side) was taken in the Construction
ous hard-hat detection methods have been developed. Two Management Laboratory of the UNLV for this study.
methods—using a sensor chip in the hard hat and by means Second, the RGB image is converted into a grayscale
of edge detection algorithms—usually are used [16, 17, 23]. image, as shown in Figure 5 (middle). Third, the
In this study, the edge of objects inside the upper head program converts the grayscale image into outlines
region that obviously includes a hard hat is drawn using of the objects as a product of the edge detection
the edge detection program. If the program finds a hard algorithm, as shown in Figure 5 (right side).
hat, as shown in Figure 3, the program understands that a (3) Hard-hat detection involves two conditions being
“Safe Working Condition” exists. Then, the program executes fulfilled: first, the outline of hard hat is a semicircle,
the same procedure for other images; otherwise, it satisfies as shown in Figure 6, and the second, the color of the
conditions to execute image output. hard hat is red.
Regarding image output, if the workers are working (4) The hard hat tilt angle is found by a dot product of
without using hard hats, a warning message is triggered to the cord-line row vector with the row vector (1, 0);
three places: the office computer displays, the cell phones the cord line is the line joining the ends of semicircles
of concerned personnel (e.g., the construction manager and formed.
safety officer), and the speakers installed at the construction
site. To deliver these messages, the program may use any of (5) The results of dot product, calculated in Step Four,
three kinds of probable networks: LAN wired networks to refer to the degree of the hard-hat rotation with
the office computers, a LAN wireless network, or wireless respect to the 𝑥-axis; the values zero and nonzero
cell phone data. Details of the visualization technique are (in between −1 and 1) indicate that the hard hat is
explained in Figures 2 and 3. perfectly straight up and rotated with some angle,
respectively. If the hard hat is found rotated, then it
is rotated by −arccos𝜃 about the 𝑥 axis to convert into
4. Face Detection Program Development the normal position.
Face detection is an important step in the visualization (6) If logos or any special initials are presented in the hard
technique. Two methods, sensor chip and face detection, are hat, the hard hat is easier to detect.
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 Journal of Construction Engineering

Figure 4: Implementation of face detection program.

Figure 5: Implementation of the edge detection program.

first, the program splits up the real-time video sent from the
CCTV camera installed at the site at the rate of 32 images
per one second of video. Then, the face detection program is
applied in each image to detect a worker. The face detection
program detects faces using a Haar-like feature. If the pro-
gram detects a face, it calls the hard-hat detection program
to execute. This program detects a hard hat by fulfilling two
conditions, the red color of the hard hat and the outline of
the hard hat that is semicircle. To find the outline of the hard
hat, the program uses the edge detection algorithm. Since the
program is able to detect more than one face and hard hat,
if the number of hard hats detected is equal to the number
of faces detected, the program constitutes a “Safe Working”
message; otherwise, it dispatches a “Warning Message” to
office monitors, cell phones of concerned personnel, and
speakers installed at the site.
Figure 6: : Implementation of the hard-hat detection program.
The program development is in initial phase. Thus, under
some conditions, the program may fail to detect faces and
hard hats. A set of CCTV cameras need to be installed at sites
6. Discussion and Conclusion such that a camera is able to take an image of a worker from
the front; however, the cameras cannot take a front-facing
In this research, the software tool was developed in Microsoft images when the worker faces down while working. In such
Visual Studio 2012. During execution of the software tool, circumstances, the program cannot detect the faces and hard
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Construction Engineering 7

hats for the time being. Moreover, while executing the hard- the United States,” Injury Prevention, vol. 10, no. 4, pp. 199–205,
hat detection program, when the contrast between the hard 2004.
hats and background color is not adequately high, the edge [6] Bureau of Labor Statistics, Census of Fatal Occupational Injuries
detection program cannot give a clear outline of the hard hats. Summary, 2011, Bureau of Labor Statistics, US Department
When a worker moves quickly, sometimes, the program fails of Labor, Washington, DC, USA, 2012, http://bls.gov/news
to detect the worker adequately. .release/cfoi.nr0.html.
The development of this program will help to save time, [7] Centers for Disease Control and Prevention, “Announcements:
cost, and the life of workers at the site; however, at this stage, it national campaign to prevent falls in construction—United
States, 2014,” Morbidity and Mortality Weekly Report, vol. 63, no.
will not completely reduce construction injuries and fatalities.
16, p. 364, 2014.
Other factors, such as workers’ behavior, types of workers,
[8] Occupational Safety & Health Administration, Construction’s
types of hazardous work, and the level of skilled and trained
Fatal Four, 2013, https://www.osha.gov/oshstats/commonstats
workers, all influence the amount of construction injuries .html.
and fatalities [25, 26]. Therefore, it is necessary that safety [9] National Institute for Occupational Safety and Health, Worker
engineers at construction sites conduct in-depth analyses, Deaths by Fall: A Summary of Surveillance Findings and Inves-
including risk analysis, to reduce the amount of construction tigative Case Reports, DHHS (NIOSH) Publication no. 2000-
accidents. The tool developed from this research will help 116, U.S. Department of Health and Human Services, Public
safety engineers to achieve zero-fatality in construction sites. Health Services, 2000, http://www.cdc.gov/niosh/docs/2000-
116/pdfs/2000-116.pdf.
Further Study and Recommendation. In this initial stage of [10] Safe Work Australia, “Work-related injuries and fatalities
program development, ideal cases have been considered. This involving a fall from height, Australia,” October Report, 2013,
program was used with a laboratory scenario to test whether http://www.safeworkaustralia.gov.au/sites/SWA/about/Publi-
the program could detect the author’s hard hat in real time. cations/Documents/812/Falls-from-Height.pdf.
At present, efforts are underway to test this software at a [11] Occupational Safety & Health Administration, Personal Pro-
real construction site by installing the CCTV cameras and tective Equipment, 2003, https://www.osha.gov/Publications/
transferring the real-time images. Another extension of the osha3151.pdf.
program is to detect whether the workers are wearing their [12] Occupational Safety & Health Administration, 2014, https://
vests; a third extension involves adding a program to detect www.osha.gov/about.html.
a worker based on the outline of a worker and typical leg [13] E. A. Yfantis, M. Y. Au, and G. Miel, “Efficient image com-
movements while walking. Additionally, the use of database pression algorithm for computer animated images,” Journal of
management for all the workers such that the program Electronic Imaging, vol. 1, no. 4, pp. 381–388, 1992.
can record all the workers’ histories regarding safety-rule [14] A. Ritchie, J. Conradi, A. Prevot, and E. A. Yfantis, “Robot vision
and video transmission,” WSEAS Transactions on Communica-
violations is being explored. Further study will continue until
tions, vol. 9, no. 8, pp. 515–524, 2010.
the program is able to detect effectively all the PPEs (vests,
[15] P. P. Shrestha, E. A. Yfantis, and K. Shrestha, “Construction
boots, globes, etc.). It will be easier to detect vests when using
safety visualization,” in Proceedings of the 4th International
this program than to detect boots and gloves because the Multi-Conference on Engineering and Technological Innovation
cameras presently cannot correctly capture the shape of the (IMETI ’11), vol. 1, pp. 243–248, Orlando, Fla, USA, July 2011.
boots and gloves because of their small size compared to vests. [16] K. Shrestha, P. P. Shrestha, and E. A. Evangelos, “Framework
development for construction safety visualization,” in Pro-
Conflict of Interests ceedings of the Canadian Society for Civil Engineering Annual
Conference, CSCE, Montreal, Canada, May-June 2013.
The authors declare that there is no conflict of interests [17] J. Teizer and M. Reynolds, Hard Hat Alerts Workers to Dan-
regarding the publication of this paper. gerous Equipment , The Herald Sun, 2010, http://www.teizer
.com/a news 2010-08-23-HeraldSun.pdf.
[18] S. Han, M. Achar, S. Lee, and F. Peña-Mora, “Empirical
References assessment of a RGB-D sensor on motion capture and action
recognition for construction worker monitoring,” Visualization
[1] C. Kelly, “These are the top 10 most dangerous jobs in the U.S. in Engineering, vol. 1, article 6, 2013.
Time Newspaper,” 2014, http://newsfeed.time.com/2014/01/15/
[19] C. Curio, J. Edelbrunner, T. Kalinke, C. Tzomakas, and W. von
these-are-the-top-10-most-dangerous-jobs-in-the-u-s/.
Seelen, “Walking pedestrian recognition,” IEEE Transactions on
[2] Bureau of Labor Statistics, “Safety and health—dangerous jobs,” Intelligent Transportation Systems, vol. 1, no. 3, pp. 155–162, 2000.
Compensation and Working Conditions, 1997, http://www.bls
[20] Y. Tsai, P. Kim, and Z. Wang, “Generalized traffic sign detection
.gov/iif/oshwc/cfar0020.pdf.
model for developing a sign inventory,” Journal of Computing in
[3] Bureau of Labor Statistics, Rate of Fatal Work Injuries, Civil Engineering, vol. 23, no. 5, pp. 266–276, 2009.
2006–2012, Department of Labor, 2012, http://www.bls.gov/iif/ [21] J. Canny, “Canny edge detection,” 2009, http://www.scribd
oshwc/cfoi/cfch0011.pdf. .com/doc/88344259/Canny-09gr820#scribd.
[4] Safety Management Group, OSHA Incident Rate Calculator, [22] M. Park, E. Palinginis, and I. Brilakis, “Detection of construc-
2014, http://www.safetymanagementgroup.com/osha-incident- tion workers in video frames for automatic initialization of
rate-calculator.aspx. vision trackers,” in Proceedings of the Construction Research
[5] Y.-S. Ahn, J. F. Bena, and A. J. Bailer, “Comparison of uninten- Congress (ASCE ’02), pp. 940–949, West Lafayette, Ind, USA,
tional fatal occupational injuries in the Republic of Korea and May 2012.
 5394, 2015, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2015/721380 by Cochrane Philippines, Wiley Online Library on [16/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 Journal of Construction Engineering

[23] I. P. Tharindu, J. Y. Ruwanpura, E. J. Boyd, and A. Y. Habib,

“Application of microsoft kinect sensor for tracking construc-
tion workers,” in Proceedings of the Construction Research
Congress, ASCE, West Lafayette, Ind, USA, May 2012.
[24] V. Escorcia, M. A. Dávila, M. Golparvar-Fard, and J. C. Niebles,
“Automated vision-based recognition of construction worker
actions for building interior construction operations using
RGBD cameras,” in Proceedings of the Construction Research
Congress, pp. 879–888, ASCE, West Lafayette, Ind, USA, May
2012.
[25] N. N. Menzel and P. P. Shrestha, “Social marketing to plan a fall
prevention program for latino construction workers,” American
Journal of Industrial Medicine, vol. 55, no. 8, pp. 729–735, 2012.
[26] P. P. Shrestha and N. N. Menzel, “Hispanic construction workers
and assertiveness training,” Work: A Journal of Prevention,
Assessment and Rehabilitation, vol. 49, no. 3, pp. 517–522, 2014.