improved apparel sizing -fit and anthropometric 3d scan data.pdf

NTC Project: S04-CR01 1

SO4-CR01 Improved Apparel Sizing: Fit and Anthropometric 3D Scan Data http://www.ntcresearch.org/projectapp/index.cfm?project=S04-CR01

http://www.bodyscan.human.cornell.edu

Primary Investigators: Susan Ashdown (leader) & Suzanne Loker, Cornell; Margaret Rucker, UC-Davis

Project Managers: Lindsay Lyman-Clarke, Erica Carnrite Graduate Students: Adriana Petrova, Sanchit Tiwari, Fatma Baytar, Hwa Kyung Song,

Tasha Lewis, Eui Choi, Jian-Guo Cao, Jennifer Cohen, GOAL To extend and improve our fit analysis process, establish our sample of 203 women as statistically representative of the USA population using SizeUSA scan data, to further develop the mathematical model of fit analysis derived from our sample, and to recommend a process by which apparel firms can interpret and apply body scan data in the development and assessment of their pattern making, grading, and sizing systems. ABSTRACT This project extends earlier research by developing methodologies for applying a combination of fit data and anthropometric population data to the problem of developing effective sizing systems for apparel products. We identify sets of critical measurements for effective sizing for specific target markets and the process to apply them in the development of base patterns, grade rules, and sizing systems. These objective measures of the target population can be used to refine the traditional method of pattern and sizing system development based on one fit model, standard grade rules, and descriptive demographic data, such as age and income. We have focused our work on the last two objectives of this study this year:

1. Link our mathematical model of fit analysis based on scan data to anthropometric data of the U.S. population.

A. Establish relationships between body characteristics of that portion of the target market that are poorly fitted and a statistically representative sample of the target market in the U.S. population.

B. Determine each possible pattern making and grading decision variable in the sizing system that improves fit for the greatest number of targeted individuals from the population.

2. Extend and improve our fit analysis process based on scans of different target markets and apparel styles.

3. Identify and generalize critical scan measurements, anthropometric data, analysis methods, and strategies in order to develop a process by which apparel firms can interpret body scan data to optimize an existing sizing system for a specific target market.

BACKGROUND One of the greatest challenges facing apparel companies today is finding a cost-effective method to provide quality fit in apparel. Repeatedly studies of degree of satisfaction with apparel have found that about 50% of women cannot find satisfactorily fitting clothes [1,2]. Lack of good fit is often the reason given by consumers for deciding not to purchase clothing, and it is estimated

National Textile Center Annual Report: November 2007


that as much as 35% of clothing purchased from catalogs is returned because of problems with fit [3].

The creation of an effective ready-to-wear sizing system is a complex undertaking. The great variation in sizes and shapes of people in the population and the need to keep stock keeping units (SKU’s) to a minimum in order to control inventory costs are in direct conflict with one another. Because of the variation in the population, a change in the base pattern shape or the grading of a pattern has the potential to improve the fit of the garment for one segment of the population at the expense of another segment. Two issues have limited the ability of apparel companies to make informed decisions about their sizing systems. First, there is a lack of data on fit characteristics of garments for a variety of different body sizes and shapes. Second, there has been a lack of current anthropometric data to describe the civilian population. Apparel companies typically only attempt to fit one body type, developing base patterns and grade rules matching the proportions of their fit model. The fit model is chosen to represent the target market but little information is available to help choose a fit model with the appropriate body characteristics. The marketing of apparel typically focuses on the age, income, and lifestyle choices of the target market, which are not necessarily a predictor of body size and shape. Three-dimensional scanning systems can provide both anthropometric and fit information but the tools and processes to analyze and apply these data are still needed. Developing quantitative models of fit applicable to multiple target markets and styles, and new analysis processes that link fit data and anthropometric data will ultimately result in better sizing and fit methodologies. The creation of better sizing systems for the apparel industry will result in a reduction in unsold or discounted garments. The 3-D body scanner is contributing to research for the apparel industry and holds promise to revolutionize the way apparel is manufactured and sold. Two anthropometric surveys of the civilian U.S. population (CAESAR and Size USA) [4,5] using this technology have been conducted. These are the first attempts to collect anthropometric data from a representative U.S. adult population relevant to apparel since the 1940’s and are made possible with body scanning technology. These data have the potential to provide new insights into issues of sizing and fit of apparel. The apparel industry has not had access to reliable, representative data from body scans, so tools and methodologies to harrness, apply, and interpret this information are critically needed 6]. Our current work, funded primarily with two NTC-funded projects [7,8], is distinctive in focus as we:

a) apply multiple types of measurements--circumference (linear), surface and slice areas, and volumes--only available with 3D scan data to the study of apparel fit,

b) merge scans to visualize misfit and to consider both visual and statistical analyses of fit by comparing scans with minimal clothes and test garments, and

c) concentrate on target markets and improving fit with current ready-to-wear sizing systems used by individual apparel firms.

Our approach has applied both the numeric and visual data acquired from 3D body scanners to improve the fit of ready-to-wear apparel using the model presented in Figure 1.



METHODOLOGY AND PROGRESS We are building on our previous work to improve current ready-to-wear sizing systems by developing protocols and methodologies to analyze fit based on fit analysis of garments from an existing sizing system. In doing so, we are using new measurements—surface and slice area and volume—that garner the power of scan data. Our three major are:

• refining our mathematical model of apparel fit using 3D computer modeling that interprets scan data in order to apply it directly to the pattern making and grading processes used by individual apparel firms;

• scanning, analyzing scans, and testing new protocols

with different target markets and active body positions; and

• testing the reliability of expert fit assessments using virtual scan visualizations;

Figure 1: Use of body scan data to improve apparel fit Mathematical Model Using Scan Data to Test Apparel Fit We tried several approaches to develop a mathematical model that describes the acceptable fit of a pair of test pants using mathematical formulas to describe body and pants measurements derived from our data base of body scans. The final model is based on a series of general equations derived from 3D geometric shapes to model body and garment configurations. For example, the crotch height required four equations, one of which described the front rise using an elliptical arc length to define the waist height to abdomen height arc. A series of equations were combined to craft a mathematical model describing the lower body measurements contributing to acceptable apparel fit (Figure 2). Based on our initial modeling results, we were able to decrease the number of linear and 3D body and pant measurement variables from 34 to 18. The mathematical model was developed using Matlab software and identifies the changes in pant measurements required to increase the number of target market members that will achieve acceptable fit using

1. pattern specifications of a graded set of test pants 2. acceptable ease measurements and tolerances of 18 different body measurements,

including surface area, slice area, and volume data that are derived from fit data from the initial study

3. body measurements from a set of scans in a database of target market members a. one set for whom acceptable fit was achieved with the current specifications b. one set for whom acceptable fit was not achieved.



The mathematical model is designed to indicate the percentage of people, within a given scan data base, that will fit into a company’s sizing system. The model compares the 18 body measurements from a person’s scan, that we have determined are most important to achieving acceptable pant fit, with the same 18 measures from the test pants to determine the quality of fit. The program generates the percentage of scans in the data base that fit well in the pants. After an initial run with the original inputted pants specifications, measurement specifications at targeted areas can be changed in order to find the pant specifications that allow the largest percentage of the target market scan data base to find acceptable fit within the sizing system. The program iteratively runs through the whole process until an acceptable percentage of the target market is fit with the chosen measurement specifications. The results can then be applied to alter the production pants patterns. The goal is to identify the optimum change in pant measurement specifications for improved fit of the target market based on these results and other business considerations such as variation in style, fit and number of sizes in a firm’s sizing system.

Figure 2. Ideal pant model based on 3D geometric equations Our 3D geometric modeling method shows promise and confirms the feasibility of this approach. It will be tested on other test pants for the same target market and revised. Comparing Body Measurements for Active Positions: Standing and Seated Scans We conducted a study to measure the change in body measurements between sitting and standing positions. Following a pilot test to develop research protocols, forty-nine subjects in the target market, aged 34-55, were scanned six times in minimal clothing. Four scans were taken in the sitting position to capture all necessary breadth and thigh landmarks and two scans were required to capture data and landmarks in the standing position. The 3D scans were transferred from the Human Solutions scan system to Innovmetric’s Polyworks software suite for processing.



Polywork’s IMEdit was used to set planes and create cross sections and Polywork’s IMInspect was used to align cross sections into one plane for measurement.

Figure 3. Standing and seated postures used for comparison. A variety of methods for categorizing body shape were tested. The method that divided the study participants into the most distinct categories of body circumference and body breadth dimensions was body mass index (BMI = mass/height2 [kg/m2]). Therefore, subjects were divided into four groups based on their BMI values (we had no subjects in underweight category). Table 1 displays the results of the seven body measurements. Increases in the waist, hip, and thigh circumferences in the seated position were significantly different between the normal group and obese II group and show considerable variation in the mean scores. The crotch length is the only measurement to decrease in the seated position, though with no significant differences among groups. Hip and crotch measurement results as well as significant differences in waist and thigh measurements point to the difficult fitting issues when considering the variety of body positions and movements. The results of these analyses will be used to adjust existing patterns and pattern systems to improve fit for pants during wear in many positions. General conclusions that can be drawn include.

• Circumferences and breadths increase between the standing and seated position, while crotch lengths decrease.

• The data generally shows greater differences in measurements as BMI increases with the normal weight classification significantly different from one or more of the overweight classifications in all measures except crotch length.



• Significant differences in measurement changes from the standing to seated positions occur at all measurement areas with the most significant differences, p≤0.01, at waist circumference, hip circumference, thigh circumference and thigh breadth

5.170** 0.418 0.175normal 12.7 8.2 a

overweight 6.4 2.6 b

obese I 30.9 34.1obese II 42.1 35.4 ab

10.330** 0.52 0.27normal 64.7 64.2 ab

overweight 110.8 107.2 a

obese I 93.5 80.6obese II 126 119 b

7.070** 0.535 0.286normal 56.5 60.2 a

overweight 85.5 102.1obese I 99.9 90.2

obese II 125.5 151.1 a

2.838* -0.251 0.063normal -42 -39.9

overweight -28 -31.8obese I -58 -60.2obese II -51 -52.4

4.438* 0.383 0.146normal 1.9 0.2 a

overweight -1.5 -2.3 b

obese I 0.9 4obese II 7 10.4 ab

4.725* 0.222 0.049normal 21.6 20.7 a

overweight 34.7 35.9 a

obese I 28.1 22.3obese II 33.8 29.1

17.605** 0.712 0.507normal 29.5 30.4 abc

overweight 56 52.5 a

obese I 58.3 56.4 b

obese II 74.3 69.2 c

Median R2Mean F PearsonDifferences

in millimeters

Waist circumference

Hip circumference

Thighs circumference

Crotch length

Waist breadth

Hip breadth

Thighs breadth

**p £ 0.01, *p £ 0.05

N = 48 [underweight: N=0; normal: N=24; overweight: N=12; obese I: N=5; obese II: N=7]

Results of multiple ANOVA comparisons are indicated using a common superscript to show pairs of groups with significantly different means.

Table 1. Significant differences in change in body measurements between seated and standing positions.

Scanning New Target Markets

A number of scans representing new target markets were taken and added to our scan data bases. Analyses of the scan process, participant reactions, body type,and size, and garment style and fit have contributed to greater understanding of diverse target markets.

Fifty female teens between the ages of 10 and 15 were scanned during several on-campus events for teenagers sponsored by the Cornell Cooperative Extension 4-H Office. The teens were scanned in the Human Solutions scanner in the standard position wearing a Lycra scan suit over their underwear garments. These scans will be analyzed to develop a better understanding of this target market’s body shape, especially as compared to our older female scan data.



Eleven male and female subjects were scanned using our portable TC2 scanner in conjunction with measurements made by a professional tailor, from Macway’s Tailors, to create patterns for custom fitted suits, shirts, pants, and sport coats. The differences between the scan measurements and those taken by a professional tailor will be generated from these data, and the body scans have been added to the main database. Methods of ensuring that scans taken using scanners from two different manufacturers, Human Solutions and TC2, are valid for analysis using the same range of software tools are also being developed. A scan study was conducted with an industry partner, Endurance LLC, of New York. Endurance manufactures Joseph Abboud Jeanswear. For this study 120 men between the ages of 25 and 55 years were scanned in a local retail store using a portable TC2 scanner. Appropriate consent was collected so that these scan measurements can also be added to our data base. These and the Macways scans are the first male scan data in our project and will be analyzed to develop a preliminary understanding of the significant body measurements for men in this targeted age group. In preparation for a scanning event focused on protective coverall fit in active working positions, a total of 57 subjects, 18 in California and 39 in New York, from different occupational groups were interviewed, photographed and given questionnaires addressing clothing fit and function to complete. Data have been compiled and the photographs have been analyzed to gain a better understanding of typical working positions and how coverall fit issues may differ with common deviations from the standard anthropometric standing position. A set of typical working positions have been identified to be used in a scan study of coverall fit in active positions. Women aged 26 to 55 in a range of sizes were scanned for a study of the fit of women’s jackets. A custom-fitted princess style jacket was created for each participant in the study from a twill fabric, perfecting the fit of each jacket with multiple fittings. Scans of the women minimally clothed and in the custom fitted jackets will add to our knowledge of ease values in garments for the upper body. In a study conducted with collaborator Nike Apparel, 73 college women age 18 to 29 were scanned minimally clothed and in a Nike ready-to-wear jacket. Custom jackets were made for 37 of the participants of the study to investigate style and fit preferences of this demographic for an upper body garment. Participants preferred a close fitting jacket and a longer length than that preferred by the older demographic that was Nike’s original target market. Visual Analysis of Fit We developed methods to visually analyze fit using 3-D scans of clothed subjects. Then, we determined the reliability of the fit ratings at different body areas to establish the number of judges needed for reliable results overall. Analysis of the fit of women’s test pants was conducted on 153 scans of women aged 35 to 55 years old. Five judges rated the fit of test pants at 15 different lower body areas.



Figure 4. Visual analysis of 3-D scans can be used to asses fit, as stress folds and the balance of the garment are clearly visible in the scanned image. Data were analyzed to find 1) subject to subject variance and 2) variance due to the judges. To determine how many judges are needed to perform a reliable fit analysis, the reliability based on subject to subject variability (σs

2) and variability between the judges (σj2) were calculated

using the equation: Reliability = σs

2 σs

2 + ( σj2 / # judges )

A reliability analysis using Cronbach’s alpha was performed to determine consistency or reliability across judges’ ratings of fit: overall, front, back, and individual area ratings. Subject to subject variance An analysis was conducted using a multilevel model to assess non-independent repeated ratings by judge and subject. The same group of 153 subjects was rated at 15 different body areas. Table 2. Variance and reliability results from the multilevel model

without

covariates with covariates

residual 2.006 1.457 judge 0.119 0.119 VARIANCE

subject 0.370 0.308

5 judges 0.940 0.928 4 judges 0.926 0.912 3 judges 0.903 0.886 RELIABILITY

2 judges 0.862

0.838

Without considering covariates (i.e., size, area, front or back), reliability of the judges was acceptable (above .80); even with only two judges, the reliability was high at 0.862. When covariates were taken into account, reliability decreased only slightly for two judges to 0.838.



Our conclusion is that only two judges are needed for reliable fit tests if visual fit parameters are established and clearly defined for the judges. Variance due to judge Cronbach’s alpha values were used to determine reliability across judges’ ratings. Table 3. Cronbach’s alpha values for each area.

AREA All Front Back

Cronbach’s Alpha with 5 Judges 0.831 0.824 0.828

Judge 1 0.757 0.749 0.750 Judge 2 0.859 0.855 0.858 Judge 3 0.792 0.790 0.789 Judge 4 0.783 0.771 0.779

Alpha if item deleted

Judge 5 0.789 0.779 0.785 Cronbach’s alpha values were evaluated to test whether retaining any specific judge was better than another. Taken as a whole, the judges’ reliability scores (Cronbach’s alpha = 0.831) were acceptable (above .80). Alpha values for the front and back ratings respectively, 0.824 and 0.828, were also acceptable. Judges were consistent in their ratings overall, ranging from 0.757 to 0.859, so any combination of two judges would have the same fit analysis results. Cronbach’s alpha was used for further analysis to identify the most consistently rated body areas. Table 4. Cronbach’s alpha values for specific body areas.

AREA CRONBACH'S

ALPHA Waist Front 0.912 Waist Back 0.832

Waist Placement Front -0.040 Waist Placement Back -0.060

Abdomen Front 0.857 Abdomen Back 0.856

Hip Front 0.739 Hip Back 0.849

Crotch Front 0.680 Crotch Back 0.510

Below Butt 0.871 Thigh Front 0.835 Thigh Back 0.849

Overall Front 0.898 Overall Back 0.737

Ratings for most body areas indicated high reliability ranging from 0.737 for overall back to 0.912 for waist front. Front and back crotch ratings were less consistent (0.680 and 0.510 respectively). Rating crotch fit is difficult, as visual cues to crotch misfit are subtle and complicated by interactions with fit at other areas. Front and back waist placement ratings were



negative which is highly unusual (-0.04 and -0.06 respectively). Subsequent discussions with the judges revealed that the scale was interpreted differently by each judge indicating the need for discussion and agreement on the meaning of scales among judges prior to rating. We found that two judges are sufficient for reliable ratings for most body areas if fit parameters and the instrument scale are established and clearly defined for the judges. Based on the unreliability of the crotch area ratings, we concluded that different assessment methodology may be required for the crotch and other complex areas of misfit that are difficult to rate visually.

NEXT STEPS Although this is the final year for and S04-CR01 Improved Apparel Sizing: Fit and Anthropometric 3D Scan Data, our research will continue on the development of a mathematical model, scanning new target markets, and visual fit analysis of body scans. We will test our mathematical model by changing garment patterns, scanning the original subjects to test the fit and see whether it improved, modifying the model, and testing it on other pant styles and other garments. Once the model works acceptably, the focus will be on developing software that can be easily used by apparel firms to improve apparel fit of their specific products for their target markets using the model. Our visual fit assessment research will focus on designing, testing, and implementing a tool for apparel firms to collect and analyze visual data on the fit of their products across the full range of sizes of their target market customers using the 3D scanner and other image capture tools. Visual modes appropriate for apparel professionals will be assessed and selected, visual data collected and process code written for software development. We will also continue to scan new target markets to refine our protocols, visual and statistical analysis techniques, and applications to pattern making and grading procedures in the industry. REFERENCES 1. LaBat, K. L. Improving garment fit - A challenge for industry. American Society for Quality Control Congress

Transactions, Toronto, Canada, 1989;377-382. 2. Goldsberry, E.; Shim, S.; and Reich, N. "Women 55 years and older: Part II. Overall satisfaction and

dissatisfaction with the fit of ready-to-wear." Clothing and Textiles Research Journal 14 (1996): 121-132. 3. "Sizing up virtual fit technology." http://www.techexchange.com/thelibrary/VirtualFit.html June 30, 2001. 4. Society of Automotive Engineers. "CAESAR: Civilian American and European Surface Anthropometry Resource

Project." http://www.sae.org/technicalcommittees/caesarhome.htm August 29, 2003. 5. [TC]

2. "Info." http://www.sizeusa.com/info.html August 29, 2003.

6. DesMarteau, K. "CAD: Let the fit revolution begin." Bobbin, vol 42, 2000;42-56. 7. Ashdown, S. P.; Loker, S.; and Adelson, C. "Use of body scan data to design sizing systems based on target markets." Annual Report S01-CR01 (formerly S01-B01): National Textile Center, 2002. 8. Ashdown, S. P.; Loker, S.; and Rucker, M. " Improved Apparel Sizing: Fit and Anthropometric 3D Scan Data." Annual Report S04-CR01: National Textile Center, 2005. ACKNOWLEDGMENTS Undergraduate Students & Staff: Katherine Schoenfelder, Charlotte Coffman, Fran Kozen, Janaki Parthasarathy, Emily Calhoun, Arnab Bose, Orren Wexler, Eve Cahill, Heather Burkman, NatalieWalsh, Tanya Garger Industry Partners: [ TC]2, Liz Claiborne, Inc., MacWays, Joseph Abboud, Nike Inc


improved apparel sizing -fit and anthropometric 3d scan data.pdf

Documents