Sizing Us Up
New 3-D body scanners are reshaping clothing, car seats, and more
If you've been having trouble shopping for clothes that fit well, you're not alone. In developed countries from Australia to Finland, people are getting bigger--much bigger. Decades of better nutrition and health care, coupled with sedentary lifestyles, have had predictable effects on the human form: we're getting taller, wider, and heavier.
Not surprisingly, this situation presents, well, sizable problems for the companies that make everything from clothes to coffins, airplane seats to bicycles. When designing anything that's intended for use by lots of people, whether it be a desk or the air bags in your car, engineers must rely on databases of anthropomorphic measurements to make sure things are neither too large nor too small for most of us. For major manufacturers, substantial sums can hinge on such seemingly trivial matters as the exact width of an airline seat or the precise dimensions of a men's ”medium” shirt.
Unfortunately, measuring by hand a statistically useful sampling of thousands of people to create these databases is time-consuming, error-prone, and expensive--which explains why the last major sizing survey in Britain, done for the clothing industry, was conducted half a century ago.
Now, however, technology is at last coming to the rescue. Light projectors, optical sensors, and modeling software were recently combined in the service of major new sizing surveys carried out in the United Kingdom and the United States. The surveys, known as SizeUK and SizeUSA, took 140 body measurements of some 11 000 subjects in two stances, standing and seated, in both countries. SizeUK, which began in 2001 and was the model for SizeUSA, was the first national sizing survey ever to use body scanners as the principal means of capturing measurements.
These three-dimensional whole-body scanners drastically reduced the cost and duration of the study. For each person measured, the entire process took mere minutes. After completing a questionnaire, a subject stood and sat in his or her underwear in an enclosed booth set up with light projectors and optical sensors. Automatically, the booth then extracted all the needed measurements, except for height, weight, and a few others that were easier to take by hand.
SizeUK is a collaboration of the British government, major UK clothing companies, academic institutions, and technology companies such as the nonprofit Textile Clothing Technology Corp. This last organization, known as (TC)2, in Cary, N.C., provided much of the hardware and software required for the survey.
With SizeUK serving as a model for future surveys, the European Commission and the European Textile Services Association, both in Brussels, Belgium, are encouraging other European states to conduct similar national surveys so that a common European size-designation standard can be developed. France has already begun, and following Europe's lead, China, Korea, Mexico, and Brazil are also conducting or planning similar body scanner based national surveys.
To measure a body, a 3-D scanner generates a so-called point cloud from the subject's frame, a constellation of 100 000 200 000 points located on the body's surface in three dimensions. Researchers can choose any of several scanners using different techniques to determine the point locations, including techniques that use beams of ordinary or laser light. Systems are starting to appear that use millimeter-wave radiation to scan through clothing, eliminating the need for subjects to undress.
SizeUK used a light-stripe system built by (TC)2. Within the enclosed booth, a series of stripes of white light is projected onto a subject via six projectors. The curves of the subject's body distort the pattern of the stripes, and these distortions are captured by six CCD cameras. A computer then processes the images to produce the point cloud.
Once it has the point cloud, the computer joins all the points to create a digital skin. Next, sophisticated software locates key landmarks on the digital body, such as the shoulders, breasts, navel, and so on. From these landmarks, such measurements as those for the bust, waist, and hips can be generated. It's similar to placing a virtual tape measure around different parts of the body and reading off the measurements [see ”Highlighting the Body”].
The body scanner approach has many advantages over traditional sizing surveys. The scanners are more precise and consistent than even trained human beings, requiring only 8 10 seconds per stance to capture the needed data. Also, the entire point cloud becomes available almost instantly, allowing it to be mined and analyzed along with thousands of other clouds to yield information about the changing shape of our bodies as well as about changes in size.
Because the scanning system captures shapes rather than measurements, the most important element of the body scanner is not the method used to obtain the point cloud but the software that extracts the size measurements. Some older systems, for example, cannot determine landmarks automatically, and so technicians must make the measurements manually, albeit digitally, after the body is scanned. Other systems require that reflective markers be placed on key locations, such as the shoulders or the hips, to cue their location to the scanning system--a procedure subjects generally dislike. Other equipment, originally designed for issuing uniforms and the like, can extract only 10 12 measurements; a good sizing survey requires over 100.
My institution, University College London, developed for SizeUK a suite of software tools for analyzing the body-shape data. These verify the integrity of captured data, correct for variations in posture, extract measurements, and provide statistical analysis of what's measured. They also allow market-research data mining based on correlations between the subject's body measurements and answers to a 40-item questionnaire about such topics as shopping and lifestyle preferences.
Besides taking the body measurements, SizeUK also attempted to completely automate the survey process from subject registration through data collection to data analysis.
Of course, data analysis is only as good as the data that is captured in the first place. The accuracy of the scanners compared with that of hand measurement has been the subject of some contention, due to discrepancies in some measurements. Often measurements done by hand use bony points, such as the hips or the ribs, as reference markers. Scanners, however, can't find bony points consistently, so differences between measurements taken on the digital model and those taken by hand may occur.
This inconsistency is forcing developers to consider some deceptively subtle questions, such as: where exactly is your waist, anyway? The traditional definition of the waist is that it is the circumference around the torso at the midpoint between the iliac crest of the hips and the bottom rib. These bony points are reliably located by the human hand. But scanner systems can't prod you, and their guess as to the exact location of the waist can be different as a result.
To deal with the problem, new reference markers for various body measurements will have to be adopted to bring the human and machine measurements into sync. For example, the computer can identify the small of the back easily, and if that spot is used as a reference marker for the waist, then human and machine measurements will tally reliably.
But measuring individuals accurately wasn't the only goal of SizeUK. Before we started the survey, we also had to determine what would be a representative sample of the UK population--across age, gender, and ethnic groups--to produce meaningful average measurements within each group. Fortunately, we could rely on the pioneering body scanning work of the joint U.S., Italian, and Dutch CAESAR (Civilian American and European Surface Anthropometry Resource) Project in 2000, which gathered digital data on nearly 8000 subjects from these three countries to generate a database of human physical dimensions.
For the SizeUK project, we divided the UK into three regions and determined that we would need more than 2600 subjects in each region, evenly distributed by age and gender, or about 7800 subjects in total, to produce average sizing measurements accurate to within 1 centimeter.
The response to our recruitment for volunteers demonstrated a massive public interest in the sizing survey. About 17 000 people registered through the SizeUK Web site, another 11 000 returned questionnaires distributed through retail outlets and mass mailings, and 20 000 called the SizeUK telephone line. The enthusiasm of the public was palpable: volunteers believed that clothing companies were finally starting to address complaints about ill-fitting clothes; they appeared to enjoy the noninvasive, high-tech process; and they generally felt that they were personally contributing to a major national project.
Volunteers were selected to be scanned from their answers to a short preliminary demographic questionnaire. The subjects recruited were asked to come to one of the eight data-collection locations, each based at a leading UK clothing and fashion college.
In the end, more than 11 000 subjects were scanned. The data on each subject is stored in the SizeUK database, hosted within a secure server farm in London's Docklands section. Each subject's data occupies about 12 15 megabytes and contains sitting and standing point clouds, as well as anonymized market research information.
For UK clothing companies, SizeUK created a CD-ROM containing what it calls the Standard Data Analysis. This collection of data supplies enough shape information to allow new clothes dummies to be created in the standard clothing sizes and to break down sizing information across various demographics.
That this new sizing information was sorely needed is shown by the numbers. Today's British women are bigger in every way than their mothers, grandmothers, and great-grandmothers. On average, they're 2 cm taller than the women of 80-odd years ago, and their busts have grown by 10 cm and gone from a B cup to a C. Where the average woman of the 1920s had bust-waist-hips measurements of 32-20-32 inches (81-51-81 cm), the 21st-century woman measures 36-28-38. Comparable data are not available for men--earlier surveys measured only women, because they tended to buy more clothes than men--but all indications are that men have expanded similarly.
For companies that need to mine the original data itself for finer-grained breakdowns or novel measurements, SizeUK provides secure online access to the database.
The discovery that large-scale, accurate, and relatively low-cost surveys of large numbers of people are now possible should lead not just to better-fitting clothing in the UK but to safer seat belts, more accommodating airplanes, and more ergonomic offices around the world. What's more, the increasing use of 3-D body scanning systems may bring the cost of the equipment down to the point where you could go to your local shopping center, get scanned, order online, and receive perfectly fitting, custom-made clothes--an ironic twist on a technology designed for off-the-rack retailers.
Body scanners will also open up new applications in health and medicine. Gyms could track the effects of patrons' diets or exercise regimens, while hospitals and clinics could monitor public health problems such as obesity or keep tabs on a child's growth.
Also in the pipeline are scanners that would be able to extract not just geometric information about the subjects but color scans of their skin as well. In medicine, for example, doctors could use this technology to screen for any changes in a patient's moles, an indicator for skin cancer. In the games industry, already players frequently paste photographs of their faces onto the stock digital avatars used in online combat arenas. Body scanners could allow them to paste accurate models of their entire bodies online (perhaps with a few tweaks here and there to better meet the heroic mold of most of these games). Pretty soon, both online and off, we should all be finding the world a better fit.