Generation of a midsize-male headform by statistical analysis of shape data

Abstract

Human surrogates for safety applications, such as crash dummies, are often constructed to represent the mean or average of individuals within a population. Generating accurate mean headforms is particularly challenging, even with three-dimensional (3D) surface measurement technology, because hair limits the ability to accurately represent the shape of the head. This report presents a new midsizemale headform generated using statistical analysis of head dimensions and landmark data, combined with a reference shape obtained by averaging the head shapes of two men close in size to the target stature and body weight. The locations of 26 head and face landmarks from 1747 men were extracted from the 1988 U.S. Army Anthropometric Survey (ANSUR) and registered using a Procrustes superimposition. A linear regression predicted landmark locations as a function of stature and body mass index, using target values of 1755 mm and 27.3 kg/m2. To construct a reference head shape, two bald men whose head and face dimensions were within 2% of the target mean values were extracted from a large database of head scans. The scans were visually aligned using landmark locations, then resampled cylindrically and averaged. The reference head shape was then morphed to match the target landmark locations using a radial-basisfunction (RBF) morphing method with a multiquadric kernel. A second morphing step was performed to match the desired head length, head breadth, and the distance from the tragion landmarks to the top of the head. In contrast with headforms obtained solely by averaging individuals close to the desired body size, the current headform is generated from a statistical model that represents a large range of head size and shape, and hence a broad set of statistically consistent headforms can be generated programmatically without additional analysis. Approved for public release.