A Simple and Standardized Method for Analysing Head and Face Morphology of a Population Sample

In order to design products that will interact with head and face, engineers need to build a solid knowledge about their target customer’s morphology. Classical methods allow only a definition of different size groups derived from specific 2D anthropometric measurements such as head girth, bizygoma-tic breadth, mouth breadth ... (Snyder 1978 [2]). Acquiring 3D geometry is another method that is being used increasingly for the past 20 years. From Bradtmiller in 1993 [3] to Mochimaru in 2008 [8], 3D scans offer the possibility to carry a full analysis of morphological details. Indeed, using 3D point clouds instead of classical 2D anthropometric measurements allows taking into account the curvature variations in specific areas such as cheek or chin. Some methods are using mathematical functions (Biasotti 2008 [7]), others are working with an analogous 3D model associated with anatomical landmarks (Mochimaru 2008 [8] ; Ma 2005 [6]). Most of those methods are trying to classify population into size groups which are based on global 3D variations such as depth, width and height of the model. Those groups are then represented with average 3D model (Kouchi 2004 [5]). Our goal here is to compute a simple and standardized method for analysing head and face morphology. It had to give results that directly take into account more details than just global dimension variations through an average 3D model and that could be directly usable to validate the right interface geometry of a computer aided designed product. The sample population was composed of 51, 15 to 47 year old women and 64, 16 to 55 year old men. The digitalisation was performed using a Konica Minolta VI-910 laser scanner. Every subject’s face was digitalized using three automatically registered views. The reconstructed scanned faces were then oriented along a reference plane computed from 3 anatomical landmarks (Projection of the Nasion on the face’s symmetry plane, Mental protuberance and Exocanthion). The same plane was then used to extract parallel level curves of the face (Figure 1). After normalizing all the curves in polar coordinates, every mean level curve was calculated on our scanned population. Using the whole population, we manage to compute a mean 3D face mapped with the geometric standard deviation (STD) (Figure 8). This preliminary result is a first step toward an easy face / product interface design for manufacturers as it provide them with the zone of less significant morphology disparity (The green zones). However, the use of a common reference point was necessary to register every face of our sample, and moving this point to another location changed the global standard deviation mapping which is easily explained by a null STD at that point -. The anatomical landmark picking required (in this method) are also a source of inaccuracy. In order to avoid those two main limits, further work is necessary to use an automatic registration and spherical normalisation (allowing not working with a reference point and landmarks). Finally, this method should be validated in regard to traditional anthropometric methods (2D measurements).