A Shape-based Sizing System for Facial Wearable Product Design

Abstract

A sizing system of a multiple-size product have been conventionally generated based on anthropometric size of a human body part. But a product which fit to a complex-shaped body part such as the face need to have a sizing system generated with consideration of body shape characteristics. This study applied template registration and machine learning clustering methods in order to make a sizing system which can consider variations of size and shape of the face. A hybrid approach using the bounded biharmonic weights (BBW) and non-rigid iterative closet point (ICP) registration methods was applied in this study to generate template-registered face images. Then, the Self-Organizing Map (SOM), a type of artificial neural network model for large-data clustering was used in order to cluster the template-registered face images into multiple shape categories. The proposed methods can be usefully applied in design of a facial wearable product such as face mask.