Measurement Accuracy and Precision Assessment of an Image-Based Clothing and Equipment Sizing System

摘要

In spite of highly standardized protocols designed to maximize the degree of repeatability and accuracy, traditional anthropometric data are not as reliable as they appear. Many factors come into play during the physical measurement of human subjects, resulting in numerous possible sources of error. Researchers have found the magnitude of these errors to be such that, even if measured by highly trained observers, comparison of two populations may be meaningless. Computerized image based systems can overcome some of the problems of traditional anthropometry, such as error due to instrument alignment, the pressure exerted on soft tissue by the various measurement instruments, or even transcription errors. However, all sources of error have not been eliminated. In image based systems, the sources of error take the form of perspective distortion, camera resolution, and inadequacy of the mathematical models used to estimate circumference measurements. The accuracy of measurements made by an image based clothing and equipment sizing system was estimated using a database of 349 subjects (male and female) who were also measured traditionally. The precision, or repeatability, of this system was estimated through repeated measurements of both a plastic mannequin and a human. Although the image based system did not exhibit systematic bias in the results, the standard deviations were somewhat smaller for some dimensions than those obtained by manual measurement. The repeatability results were comparable to those obtained by highly trained anthropometrists, as reported in recent large scale surveys. The reliability of the measurements needed for clothing, i.e. the proportion of error of measurement to biological variability, was greater than 99% in all cases.