Professional Documents
Culture Documents
Paula S. Joe , Yasushi Ito , Alan M. Shih , Riedar K. Oestenstad & Claudiu T.
Lungu
To cite this article: Paula S. Joe , Yasushi Ito , Alan M. Shih , Riedar K. Oestenstad & Claudiu
T. Lungu (2012) Comparison of a Novel Surface Laser Scanning Anthropometric Technique
to Traditional Methods for Facial Parameter Measurements, Journal of Occupational and
Environmental Hygiene, 9:2, 81-88, DOI: 10.1080/15459624.2011.640557
Download by: [NWFP University of Engineering & Technology - Peshawar] Date: 20 February 2017, At: 20:34
Journal of Occupational and Environmental Hygiene, 9: 81–88
ISSN: 1545-9624 print / 1545-9632 online
Copyright c 2012 JOEH, LLC
DOI: 10.1080/15459624.2011.640557
This study was designed to determine if three-dimensional illness and/or injury. When exposure to these environments
(3D) laser scanning techniques could be used to collect accu- cannot be adequately reduced by engineering and/or adminis-
rate anthropometric measurements, compared with traditional trative controls, the use of respiratory protection is required.
methods. The use of an alternative 3D method would allow
for quick collection of data that could be used to change As a result, an estimated 5 million U.S. workers are legally
the parameters used for facepiece design, improving fit and required to use respiratory protective equipment while
protection for a wider variety of faces. In our study, 10 facial working.(1)
dimensions were collected using both the traditional calipers Key factors in the performance and protective value of a
and tape method and a Konica-Minolta Vivid9i laser scanner. respirator are its fit and the integrity of the face seal it affords.
Scans were combined using RapidForm XOR software to create
a single complete facial geometry of the subject as a triangu- The lack of an intact face seal leaves the wearer vulnerable to
lated surface with an associated texture image from which potentially hazardous exposures. In addition, if the respirator
to obtain measurements. A paired t-test was performed on does not fit properly when donned, or the users perceive it to not
subject means in each measurement by method. Nine subjects properly fit, the user will be less likely to comply with its use.
were used in this study: five males (one African-American and One of the reasons for the lack of compliance could be
four Caucasian females) and four females displaying a range
of facial dimensions. Five measurements showed significant the poor fit of respirators, to some wearers, that could be
differences (p < 0.05), with most accounted for by subject attributed to the limited size/form options available.(2) Simply
movements or amended by scanning technique modifications. stated, there are face shapes that cannot be properly fitted
Laser scanning measurements showed high precision and ac- by the currently available small, medium, and large respirator
curacy when compared with traditional methods. Significant facepieces. To design better respirators, new fit test panels need
differences found can be very small changes in measurements
and are unlikely to present a practical difference. The laser to be developed, and the National Institute for Occupational
scanning technique demonstrated reliable and quick anthropo- Safety and Health (NIOSH) is collecting anthropometric facial
metric data collection for use in future projects in redesigning parameters from a sample population representative of the
respirators. current U.S. work force.(3)
To collect these data, NIOSH is using both traditional
Keywords facial anthropometry, personal protective equipment de- caliper and ruler measures as well as advanced 3D scanning
sign, respirators, surface laser scanning, three-
dimensional facial measurement techniques. Over 1000 subjects were scanned using a Cyber-
ware rapid 3D digitizer (Cyberware, Inc., Monterey, Calif.)
and its associated data processing software. Ten relevant facial
Correspondence to: Claudiu T. Lungu, Department of Environ- dimensions were obtained for each scanned subject, the same
mental Health Sciences, University of Alabama at Birmingham, as those measured using the traditional method.(4) The use
RPHB 530, 1530 3rd Ave. S., Birmingham, AL 35294; e-mail: of 3D methods in anthropometry has several advantages over
clungu@uab.edu. the traditional use of calipers and steel tape. Measurements
in 3D lack some of the inherent errors of manual methods,
INTRODUCTION particularly the compression of soft tissues of the face. This
type of compression can occur while closing the calipers or
pulling the tape across the face to obtain a measurement,
M illions of workers in the United States are employed
in hazardous environments that can cause respiratory and results in a decreased parameter value. Three-dimensional
of subjects in 3D as triangulated surfaces with associated investigator conducted all the manual and 3D anthropometric
texture images (Figure 2). Using the scanning controls avail- data collection.
able in RapidForm XOR software, three scans were performed
on each subject to obtain data on the entire face. These scans Analysis
were performed with one scan for the subject directly facing
Means and standard deviations (SD) were calculated per
the scanner and two scans at approximately 45◦ to the left and
measurement in each subject, testing for accuracy and preci-
right (Figure 3).
sion. A paired t-test was performed on each type of anthro-
RapidForm XOR was also used to composite the three scans
pometric measurement using the subject means of the manual
of each subject into a complete representation of the face.
measurements and the means of the 3D measurements. This
The three raw scans of the subject are aligned both manually
comparison will also test accuracy of the 3D laser scanning
(“picked point”) and automatically (“global and fine”). Once
method to the traditional manual collection of facial measure-
aligned, the individual scans are trimmed to remove larger
ments.
polygons and areas of overlap between scans, while retaining
areas needed to fill gaps in individual scans (Figure 4). The
three trimmed scans were then decimated (smoothing and RESULTS
compressing data) prior to merging to form a single facial
surface (Figure 5a). Anthropometric measurements were col-
lected in triplicate from the merged data using distance and
3D spline functions in RapidForm to determine linear and arc
O ver the course of data collection, several observations
were made dealing with the sensitivity of the exper-
imental setup. Observations included changes in head and
measurements, respectively (Figures 5b and 5c). The same facial position affecting scan completeness, data processing
FIGURE 3. Three scans collected from a single subject (45◦ right, frontal, 45◦ left).
DISCUSSION to data collection and processing. First were the issues dealing
with the alignment and capture of the three individual subject
would move their heads, resulting in incomplete image cap- landmarks were more difficult than others to identify due to
tures. Subjects could also change their facial positioning, in shadows occurring on the sides of the face and changes to the
between individual scans, altering certain dimensions of the color of the subject’s skin during digitization. The initial study
face. An additional issue with alignment resulted from the design used washable black marker to place dots at landmark
rotation of the subject between individual scans. Often, the locations. These marker dots were often difficult to locate after
angle of the scanner relative to the subject in the initial frontal scanning due to insufficient contrast to the scanner’s percep-
view scan would not be suitable for the 45◦ angled view scans, tion of the subject’s skin color. A second method was tested
leading to the exclusion of the upper portion of the face. using bright white, triangle-shaped stickers. While providing
Another issue dealt with the ambient lighting in the area. sufficient contrast, it was difficult to determine the true location
The laboratory space used in the study employed overhead of the landmark, as the scanner’s perception of the sticker’s
fluorescent lighting, which was at an angle to the experimental color and thickness would not occur in the same location
setup. While the individual scans would appear to be fine, this (Figure 8). Two types of white theatrical makeup were also
angling of light relative to the subject being scanned caused tested for use in identifying landmarks. A white cream makeup,
significant discolorations to the texture of the compressed and applied with a cotton swab, presented the same issues with
merged scan (Figure 7). These changes to the color of the contrast as the black marker method. A white grease makeup,
facial surface made it impossible to detect facial landmarks also applied with a cotton swab, proved the best method of
identified in these areas. those tested.
In an effort to deal with these observed issues, amendments The second amendment to the scanning protocol dealt with
were made to the scanning protocol that was initially employed the method of compositing the individual trimmed scans. Ini-
to improve the quality of the data collected. First, several meth- tially, the individual scans were decimated to smooth and
ods were tested to improve landmark identification. Certain compress the data in order to reduce the time needed to render