NTC Project: S01-CR01 (formerly S01-B01)

1
Use of Body Scan Data to Design Sizing Systems Based on Target Markets
http://www.human.cornell.edu/units/txa/research/ntc/S01-CR01-03.pdf
http://www.explore.cornell.edu/bodyscanner
Susan Ashdown & Suzanne Loker, Cornell U.; Carol Adelson, Fashion Institute of Technology
Project Manager: Katherine Schoenfelder
Graduate Students: Adriana Petrova, Eui Choi
GOAL
To develop a prototype mathematical process using body scan data to improve sizing systems for
specific target market populations of an apparel firm.
ABSTRACT
Two hundred and three female subjects were scanned twice, once in minimal clothing and once
wearing test pants, using the protocols and instruments developed by the research team. Subjects
were selected to fit the target market of the industry partner, 34-55 years of age and sizes misses
4-16 or women’s 14-24. After cleaning and organizing scan data, a variety of statistical analyses
were conducted to compare 1) body shapes based on circumference, surface area, slice area, and
volume measurements, 2) ease values by body proportions, 3) misfit based on expert fit ratings
and ease values. Results are promising in our approach in using unique measurements and
procedures provided by body scan technology to identify fit problems, i.e., misfit. Areas of
special interest in the ongoing development of a mathematical algorithm to adjust existing
pattern systems are 1) waist to abdomen and abdomen to hip surface area and volume
measurements, 2) focus on certain sizes of misfit (i.e., women scanned in size 12 and 14 pants)
rather than all sizes, and 3) combining ease values and body proportion data for analysis.
BODY/APPAREL RELATIONSHIP
As the relationship between the body and clothing is complex and often ambiguous, garment fit
is a difficult concept to research and analyze. Current objective methods of analyzing garment fit
involve comparing the garment measurements to the body measurements using linear
measurements and utilizing pressure gauges to measure the amount of pressure a garment places
on a specific body location. Although these traditional and other manual methods are useful for
evaluating simple garment fit issues, they are not adequate to investigate the complexities of the
multifaceted relationship that exists between the body and clothing. Both provide only limited
information about the human body shape and proportion. On the other hand the body scanner has
the ability to obtain three dimensional data of the surface of the human body, providing valuable
information to improve garment fit (DesMarteau, 2000). Body scanning technology collects data
that can be analyzed using non-linear measures such as surface area, volume, or data from body
slices that may be better able to comprehensively analyze the human body and address problems
of garment fit.
Anthropometric research using body scan data provides information on consumer body types and
measurements to designers and manufacturers to improve the fit of garments (CAESAR uses
latest laser technology, 2002). A Kurt Salmon Associates study reported that 50 percent of
women claim they cannot find apparel that fits; other studies have indicated that fit problems are
the reason for 50 percent of catalog returns (DesMarteau, 2000). Clearly, utilizing body scanning
technology to improve garment fit is a critical need. Retailers such as Lands’ End, Levi Strauss,
and Brooks Brothers have taken initial steps to help consumers find better fitting garments by

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
2
using customer measurement information to either create custom fit garments or give consumers
the ability to visualize their potential purchases in virtual try-on software programs (Wellington,
2001). While these methods are innovative, they are merely a first step in solving garment fit
problems for consumers. Research analyzing the complex relationship between the consumer’s
body and the way a garment fits on his/her body is needed to make significant strides in
improving garment fit for apparel consumers. This research is being undertaken by both
university and government research facilities. Some examples are projects on measurement
extraction and use (Pargas, Staples, & Davis, 1997), shape analysis and quantification (Connell,
Ulrich, Gray, Woronka, & Ashdown, 2001; Jones, Li, Brooke-Wavell, & West, 1995), fit
analysis (Paquette, 1996; Whitestone & Robinette, 1997), use of scan data in a mass
customization scenario (Fralix, 2001), and automated pattern making from body measurements
(Kang & Kim, 2000).

METHODOLOGY
Our industry partner, Liz Claiborne, provided test pants in two size ranges (misses 4-16 and
women’s 14-24), size specifications and grade rules for the pant styles, and fit models’
measurements and body scans. Liz Claiborne’s target market age ranges were used to recruit
subjects wearing these sizes. Demographic and other personal characteristics were requested
from subjects with a questionnaire in order to evaluate the sample against Liz Claiborne target
market descriptions and to describe the sample’s comfort with the body scan process and interest
in scan applications. Our 203 female subjects (not including two fit models) were representative
of the Liz Claiborne target market with 60% married, 82% full-time workers, 66% holding a
bachelor’s degree or higher, and the majority with annual household income over $50,000. The
mean age of our subjects was 44.9 years and subjects’ ages were relatively equally divided
across the four five-year age spans─34-39 years (n=41), 40-44 (n=51), and 45-49 (n=65), and
50-55 (n=46).
Subjects were scanned with a Human Solutions/Tecmath scanner on campus or in New York
City (with a rented scanner). Each subject was scanned twice, once with a Lycra scanning suit
designed for this project and a second time wearing test pants in the size with the best fit at the
hip as determined by the researcher. A second pant scan was taken if the subject wore a size that
overlapped the misses and women’s pant size ranges, i.e., 14 and 16, or when the best fitting pant
was difficult to determine. For analysis purposes, subjects who were scanned in multiple pants
were assigned either the misses or women’s size group by a panel of experts who judged the fit
of each scan to choose the best fitting size. Table 1 displays the results of the ratings by the best-
fit pant size and size category, 156 in the misses and 49 in the women’s pant size categories.
Although the numbers per cell in the women’s sizes were small, the sample provided a relatively
balanced representation of all available sizes. Following the scan, subjects completed a
questionnaire to indicate demographic information and their level of comfort and interest in the
body scan process, preference for access and storage of personal body scans, and interest in
potential applications of body scans. They were shown their body scans on the computer in still
and moving formats and asked to indicate their comfort with seeing their body scans in both
formats.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
3
205 Study
Subjects

Number of Number of
Size Misses Women's
Scans Scans
04 32
06 21
23 subjects fit in both 08 20*
138 subjects 44 subjects
Misses and Women’s
fit in Misses sizes fit in Women’s sizes 10 21
sizes
12 24
14 28 13
16 10 15
18 9*
20 5
22 3
24 4
156 49
* fit models included
Misses Women’s
Subjects Subjects

Figure 1. Sample population description.

Data Cleaning and Organizing

The Human Solutions/Tecmath VitusSmart scanner uses eight cameras and four laser light
sources to record xyz coordinate data from the surface of the scanned object. Data from the eight
camera views were assembled and visualized on the screen. Scan data consisting of over 300,000
data points were transferred to Polyworks, a software package developed by Innovmetric, for
processing, measuring, and displaying data in three dimensions. Data from the eight camera
views were merged and re-triangulated, eliminating redundant points. Any missing data (holes)
in scans were then patched, as the volume measures required a complete, closed model. The
patching of scans prior to measurement required manual manipulation and was a labor intensive
process that we hope to shorten in the future. Polyworks software allows 3D visualization and
measurement of circumference, surface area, volume, and slice area or shape. These dimensional
measurements were similar to the Volume Index, Cross Sectional Index, and Signature Curve
described in previous research on developing 3D models of clothed figures (Ng, Chan, Pong, &
Au, 1997).
Twenty measurement variables were selected and evaluated for effectiveness in identifying the
differences between body and apparel measurements. Planes were set parallel to the floor at the
abdomen, hip, crotch, and thigh based each subject’s minimally clothed scan, locating each plane
according to specific visual landmarks. Tilted planes were also set to capture the natural waist
angle and the waistband of the pant. The following landmarks were chosen to set the planes for
each subject: center of the back, natural waist indent in front, abdomen at the greatest protrusion
to the anterior of the frontal plane, crotch based on manual measurement at time of scan, hip at
the greatest protrusion to the posterior of the frontal plane, and thigh set 25mm below the crotch
plane. The planes from each subject were placed on the pant scan and the minimally clothed scan
at the same level so that measures of the two scans would occur at the same location. Figure 2
illustrates this process of setting horizontal planes on a scan at the waist, abdomen, crotch, and
hip as well as frontal and sagittal planes that will be used in future analysis.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
4

Figure 2. Critical Fit Points Defined by Planes.

Data Analysis
Cluster Analysis
To analyze the differences among body shapes or proportions of the misses sized subjects in the
sample, K-Means cluster analysis in SPSS was used to group subjects with similar measurement
characteristics. Twenty variables were selected that contribute to the differences among the
subjects (see Table 2). K-Means cluster analysis requires the selection of number of clusters
before the analysis. We ran eight separate cluster analysises with 156 misses subjects for two,
three, four, five, six, seven, eight, and nine clusters to test which solution was most appropriate.
The four and five cluster solutions proved most interesting. The four cluster solution had one
very large cluster with a large range of values. The five cluster solution essentially broke that
single cluster into two (now Cluster 1 and 5) and provided five clusters of more equal size and
smaller range of values. We report ANOVA results from the five cluster solution in Table 2 in z-
scores in order to have a common unit to compare across linear circumference, surface area, slice
area, and volume measurements.
We expected and found that there would be significant differences among most variables across
clusters. The clustering variables were body measurements that would largely reflect a
proportional increase in size. We also were looking for irregularities in a proportional body size
increase on variables. Table 2 displays significant differences in the superscripts unique along
rows (across clusters). For example, when each variable in a cluster has a unique superscript (i.e.,
a, b, c, d, e) there are significant differences across all variables. When two variables share a
superscript, such as in Cluster 1 and 5 where the hip and thigh circumference both have the
superscript a, there is no significant difference across those clusters on those variables. Most z-
scores increased in the following order: from Cluster 2, with the lowest negative values from the
sample mean (z = 0) on most variables, to Cluster 4, Cluster 1, Cluster 5, to Cluster 3 with the
highest positive values from the sample mean, roughly following the mean ordinal pant sizes
from smallest to largest.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
5
Table 1
Clusters Based on 20 Scan Measurements (n=156)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5

Mean StDev Mean StDev Mean StDev Mean StDev Mean StDev
a b c d e
Natural Waist 0.25 0.37 -1.09 0.44 1.73 0.46 -0.34 0.37 1.08 0.35
Circumference

a b c d e
Top of Waistband 0.31 0.36 -1.11 0.45 1.76 0.43 -0.34 0.35 1.03 0.38
a b c d e
Bottom of Waistband 0.40 0.38 -1.17 0.44 1.72 0.29 -0.30 0.30 0.98 0.44
a b c d e
Abdomen 0.40 0.40 -1.15 0.43 1.67 0.57 -0.27 0.48 0.91 0.36
a b c d a
Hip 0.83 0.26 -1.25 0.44 1.62 0.41 -0.22 0.33 0.58 0.42
a b c d a
Thigh 0.87 0.45 -1.17 0.55 1.44 0.63 -0.21 0.44 0.49 0.55
a b c d e
Natural Waist 0.21 0.35 -1.06 0.39 1.79 0.53 -0.37 0.36 1.08 0.40
a b c d e
Top of Waistband 0.27 -1.07 -0.38 1.85 -0.37 1.02
Slice Area

0.33 0.51 0.32 0.41

a b c d e
Bottom of Waistband 0.36 0.37 -1.14 0.38 1.80 0.33 -0.33 0.28 0.98 0.47
a b c d e
Abdomen 0.34 0.42 -1.10 0.37 1.76 0.64 -0.31 0.45 0.92 0.40
z -scores

a b c d a
Hip 0.75 0.27 -1.22 0.39 1.73 0.45 -0.27 0.33 0.63 0.43
a b c d e
Thigh 0.87 0.49 -1.15 0.50 1.48 0.69 -0.24 0.43 0.48 0.57
Top of Waistband to
Surface Area

a b c d e
Bottom of Waistband 0.31 0.33 -1.12 0.37 1.85 0.33 -0.35 0.28 1.01 0.40
Top of Waistband to
a b c d e
Abdomen 0.20 0.51 -0.97 0.44 1.65 0.99 -0.31 0.55 0.92 0.48
a b a c d
Abdomen to Hip 0.94 0.51 -1.11 0.58 1.25 1.07 -0.15 0.45 0.33 0.57
a b c d a
Hip to Thigh 0.66 0.39 -1.15 0.47 1.70 0.62 -0.27 0.37 0.63 0.54
Top of Waistband to
a b c d e
Bottom of Waistband 0.33 0.36 -1.11 0.38 1.82 0.42 -0.35 0.31 0.99 0.40
Volume

Top of Waistband to
a b c a d
Abdomen 0.08 0.66 -0.78 0.47 1.45 1.38 -0.27 0.63 0.80 0.73
a b acd c bcd
Abdomen to Hip 0.90 0.73 -0.65 0.65 0.35 1.88 0.01 0.60 -0.13 0.83
a b c d a
Hip to Thigh 0.57 0.46 -1.07 0.48 1.72 0.81 -0.29 0.42 0.62 0.61
a b c d a
Ordinal Pant Size 12.74 1.13 4.44 0.84 15.33 0.98 8.13 1.53 12.54 1.56
Size 4 n = 32 - 32 - - -
Subject Count

Size 6 n = 21 - 9 - 12 -
Size 8 n = 20 - - - 20 -
Size 10 n = 21 1 - - 15 5
Size 12 n = 39 15 - - - 24
Size 14 n = 28 11 - 5 - 12
Size 16 n = 10 - - 10 - -
n = 156 27 41 15 47 26

Note. Superscripts unique to one cluster indicate a significatn difference between that cluster and all others at p.<.05. A set of common
superscripts indicates that the cluster is not significantly different from those clusters with any of the common superscripts.

One interesting departure from proportional increases was in Clusters 1 and 5 between which
there were no significant differences in the hip and thigh circumference, hip slice area, hip to
thigh surface area, and hip to thigh volume measurements. These clusters included subjects who
were scanned in best-fit pants sizes 10, 12, and 14, indicating that there were two different body
types of women in these size ranges based on lower body measurements. This indicates a
possible difficulty in fitting women in these current size ranges with proportional sizing.
Two other interesting departures from proportional increases across clusters related to surface
area and volume measurements, measurements only available with body scanning tools. Cluster
3, which is composed of subjects who fit best into pant sizes of 14 and 16, is not significantly
different from Cluster 1 (pant sizes 10, 12, 14) on abdomen to hip surface area and abdomen to
hip volume measurements. The relatively large values for these measures compared to other
clusters’ values suggest a protruding abdomen that might need a special fitting approach. Also,
Cluster 5’s relatively small measure for these measures suggests a small hip or buttocks body
shape that also requires special fitting. The relatively low volume measure from abdomen to hip
for Cluster 3, the cluster with the overall largest measurements, also identifies a fitting issue.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
6
These findings confirmed that surface area, slice area, and volume body measurements available
with body scan data can add important new information to sizing system development.

Figure 3. Misses size 12 subjects in Cluster 1 and 5 show different abdomen and buttocks
shapes.
Linear Ease Values Compared to Body Proportions
Body and pant scan measurements were compared using data from 160 subjects (n = 160 before
placing subjects with multiple scans into either misses or women’s categories) whose best-fit
pants were misses sizes 4-16. We focused on hip to waist ratios for body and pant scans,
combining ease values and proportional information to characterize fit and misfit. The waist and
hip measurements from the minimally clothed scans and were converted to ratios by dividing the
hip circumference measure by the waist circumference measure. In Figure 3, the values for each
subject from the scan measurements (the circles) were plotted by the percent change between
body and pant circumference at the waist, hip, and abdomen on the Y axis and ordered by the
subject rank order, low to high, of body hip to waist circumference ratio on the X axis for clearer
comparisons.

Figure 4. Subjects ranked by body proportions showing trends in ease values (n = 160).
Note. The star on the x-axis indicates the fit model proportions used to develop this pant style.

Double polynomial best fit lines indicate fit variation due to ease values based on body
proportions. Essentially, the very straight body silhouettes have negative waist ease values (i.e.,

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
7
pants too tight) as indicated by the red downward curve line in the lower left while the very
curved body silhouettes have relatively large positive waist ease values (i.e., pants too loose) as
indicated by the red upward curve in the upper right. The abdomen ease values indicated in the
yellow best-fit trend line are similar but not as pronounced. The hip trend line is flattest as
expected since pant size was chosen by best hip fit. The most important misfit characterization
from this analysis was that body proportion variations were frequently too tight or too loose in
the waist. Results suggest that a combination of ease values and body shape are important in
pattern development and adjustments in existing sizing systems to accommodate both curvy and
straight body figures are recommended.
Misfit Analysis
One of the unique approaches to analyzing body scan data in this research is comparing the
minimally clothed and best-fit pant scan of each subject to analyze fit or, actually, misfit. Using
the visualization tools of Polyworks, three experts in apparel fit visually analyzed the pant scans
of 160 subjects on a computer screen and rated the front and back overall fit of the pants, the
simple rating method. The numerical rating system was based on scoring each view for fit from 1
to 3 and averaging the raters’ scores for each subject to determine an overall acceptable (2.51-3),
marginal (1.51-2.5), and unacceptable (1-1.5) fit rating. The experts used discussion and
negotiation to arrive at a common decision when no two raters were in agreement. There were 54
(34%) pant scans classified as acceptable fit, 83 (52%) pant scans as marginal fit (33%), and 23
(14%) scans as unacceptable fit. To characterize fit in more specific locations, a more complex
rating method was conducted by the experts as well. They rated 13 critical fit locations in the
front and back. The experts’ ratings were again averaged over each subject for each different
location. Two approaches were explored using these two rating methods to investigate the
relationship between body and pants, or misfit, in order to find predictive strategies that could be
used to identify acceptable fit: 1) analysis of fit ratings (13) by size and 2) analysis of fit ratins
(2) based on ease values.

Fit ratings by size

Cross-tabulations of the complex fit ratings by size were run for the standard critical fit locations
used in pant pattern making: waist, waist location, abdomen, hip, crotch, below buttocks, and
thigh. Significant differences across sizes were calculated using the chi-square statistic in order
to identify problem fit areas by size. The crotch and below buttocks ratings were mostly
unacceptable for all sizes. The basic pant pattern is at fault for the crotch misfit and should be
redesigned while the below buttocks misfit may be caused by the crotch misfit at least in part.
The smaller sizes (4 & 6) were rated with high acceptable fit across all critical fit locations
except the crotch and below buttocks ratings. These ratings indicate that the smaller sizes of
bodies were more closely proportional to the fit model than the larger sizes but were also
probably somewhat residual as in our culture a looser fit tends to be preferred.
Most interesting were the significant differences across the waist and abdomen, pinpointing fit
problems in sizes 12 and 14 with many more marginal and unacceptable ratings than the smaller
sizes. Figures 3 & 4 illustrate the ratings across sizes for these critical fit locations. The hip fit
also had significantly more marginal ratings for size 14 indicating the potential for transforming
marginal fit to acceptable fit with some minor pattern making or grading adjustments. Size 4 also
had more marginal and unacceptable hip ratings but these could be due to a lack of available size
2 pants rather than a pattern or grading problem.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
8

Waist Abdomen
80 80
70 70
60 60
50 50 Acceptable
Percent

Percent
40 40 Marginal
30 30 Unacceptable
20 20
10 10
0 0
4 6 8 10 12 14 16 4 6 8 10 12 14 16
Size Size

Figure 5. Percent fit ratings within size at waist and abdomen (n=156).
To conclude, rating fit and misfit using a 3D visualization of a scanned body in pants provides a
useful tool and process to evaluate how well garments fit the target market. The method could
easily be used on subsets of the target market during pattern development or adjustment,
expanding the fitting process from one fit model to a set of fit models for each size. In addition,
once virtual try-on software becomes available for body scans, it will be possible to use a set of
scans to test the fit of garments of every size in a firm’s sizing system using this process and
visualization system.

Ease values based on comparison of body and pant scans

Data were organized based on the percentage change in surface area from body measurements to
pant measurements to represent the fitting ease value at these body locations. For pants, the
waist, abdomen, and hip measurements and relationships are the most difficult to fit. Analyzing
the surface area of these locations, a measurement only available by body scan, allows us to
characterize the necessary ease for an acceptable fit. Surface area measures are particularly
interesting as they have a direct relationship with the surface area of a 2D pattern. Some
variation in the data occurred due to the different size categories. To eliminate this variation so
that the ease measurements could be viewed independently of size, the differences for each
variable were calculated as a percentage increase of the unclothed variable.
Sizes 4, 12, 14, and 16 categories included more unacceptable fit garments than sizes 6, 8, and
10. This can be explained by the proportional method of pattern making based on a size 8 fit
model’s measurements. MANOVA found significant differences between the acceptable and
unacceptable and marginal and unacceptable rated scans for the means of the percent change in
surface area of the waist to abdomen. These preliminary results concur with the findings of the
cluster analysis by identifying the same variables to focus on for adjusting the existing sizing
system to better fit. Significant interactions between size and fit rating were found by MANOVA
in the abdomen to hip surface area analyses. This confirms that existing sizing systems work
better for some sizes than others and suggests that adjustments to current sizing systems may
focus on some rather than all sizes.

Additional Analytical Approaches to be Tested

The next step will be to continue the analyses reported above adding the data for subjects whose
best-fit was in women’s pant sizes and comparing the results across size categories. Two
additional mathematical approaches from other disciplines will also be experimented with for

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
9
appropriateness with these data and our objective of adapting an apparel firm’s current sizing
system based on scan data from its target market. In the biological sciences, Gielis’s
“superformula” seems to offer great potential in mathematically defining pseudo-elliptical shapes
(Gielis, 2003). Using this procedure, we may be able to analyze shape parameters relevant to fit
in cross sections generated from body scans. In addition, Statistical Parameter Mapping (voxel
based statistical normalization/analysis) used in the medical research field for determining
volumetric statistical comparisons seems directly applicable to body scan research aims
(Ashburner & Friston, 2000). The first step will be to evaluate how data are stored and the
variety of approaches taken to account for breathing and body variability among other issues.

CONCLUSIONS
The objective of our project is to adjust existing apparel sizing systems based on target markets
using body scan data. Using body and pant scans from a sample of women meeting the target
market characteristics of our industry partner, we can evaluate whether the size range,
measurement specifications, and grading rules accommodate the target market. The Polyworks
software allowed manual extraction of circumference, slice area, surface area, and volume data
and 3D visualizations to describe body and apparel relationships in ways never before possible.
Following the development of scanning protocols, the scanning of over 200 women, and
cleaning and preparation of the scan data, we conducted a variety of analyses to characterize
specific problems of pant misfit in an apparel firm’s target population. Cluster analysis of body
scans identified five clusters distinguished by general size and some unexpected (based on
proportional theory) measurements in the abdomen. Experts used 3D visualizations to rank fit
acceptable, marginal, or unacceptable in a simplified and then more complex manner. These
rankings were used to identify problem areas or misfit based on size and to categorize the sample
by fit for further analysis. Ease values or body to pant comparisons in linear, area, and volume
measurements proved to be a valuable method to identify and characterize misfit. Displaying
ease values to show misfit trends based on body proportions was a third approach. Statistically
significant patterns across sizes and fit ratings will focus our attention as these results are applied
to the development of a mathematical algorithm to apply to existing patterns of our industry
partner. By making the pattern adjustments, we expect to more closely reflect the body
measurements of our partner’s target market and, therefore, increase the number of target market
member who can find good fit.

REFERENCES
Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometry - The methods. NeuroImage,
11, 805-821.
CAESAR uses latest laser technology. Retrieved September 27, 2002, from
http://www.sae.org/technicalcommittees/caelaser.htm
Connell, L. J., Ulrich, P. V., Gray, S., Woronka, D., & Ashdown, S. P. (2001). Body scan
analysis for virtual fit models based on body shape and posture analysis (New project
proposal No. IO1-A27): National Textile Center.
DesMarteau, K. (2000, October). CAD: Let the fit revolution begin. Bobbin, 42, 42-56.
Fralix, M. T. (2001). From mass production to mass customization. Journal of Textile and
Apparel, Technology and Management, 1(2), 1-7.

National Textile Center Annual Report: November 2003

 NTC Project: S01-CR01 (formerly S01-B01)
10
Gielis, J. (2003). A generic geometric transformation that unifies a wide range of natural and
abstract shapes. American Journal of Botany, 90(3), 333-338.
Jones, P. R. M., Li, P., Brooke-Wavell, K. F., & West, G. (1995). Format for human body
modelling from 3-D body scanning. International Journal of Clothing Science and
Technology, 7(1), 7-17.
Kang, T. J., & Kim, S. M. (2000). Optimized garment pattern generalization based on three-
dimensional anthropometric measurement. International Journal of Clothing Science and
Technology, 12(4), 240-254.
Ng, R., Chan, C. K., Pong, T. Y., & Au, W. M. (1997). Objective measurement of the 'fit' of
apparel. Paper presented at the World Conference of the Textile Institute, Manchester,
England.
Paquette, S. P. (1996). 3D scanning in apparel design and human engineering. IEEE Computer
Graphics and Applications, 16(5), 11-15.
Pargas, R. P., Staples, N. J., & Davis, J. S. (1997). Automatic measurement extraction for
apparel from a three-dimensional body scan. Optics and Lasers in Engineering, 28(2),
157-172.
Wellington, E. (2001, February 5). For good measure. News & Record, p. B10.
Whitestone, J. J., & Robinette, K. M. (1997). Fitting to maximize performance of HMD systems.
In J. E. Melzer & K. W. Moffit (Eds.), Head-Mounted Displays: Designing for the User.
New York: McGraw-Hill.

ACKNOWLEDGEMENTS
Graduate Students: Adriana Petrova, Eui Choi, Lucy Dunne, Kirk Mayer, Jeong-Ju Yoo
Undergraduate Students: Luisa Avila, Stephanie Chan, Fred Fang, Amy Kinateder, Erica
Lastufka, Kira Luxon, Ewunike Patterson, Janaki Parthasarathy, Jaclyn Popeil, Tara Taff,
Frankie Tsang, Natalie Walsh
Industry Partners: Liz Claiborne, Inc., Saint Laurie Merchant Tailors

National Textile Center Annual Report: November 2003