A new sizing system based on 3D shape descriptor for morphology clustering

Abstract

The variety of human morphologies is an important issue for the textile-apparel industry. Indeed, sizing systems currently used by companies have to be continuously updated or adapted to the population target.For this reason, the Textile-Apparel-Industry requires a very accurate sizing system to minimize their costs and satisfy their customers. However, the specific constraints of human morphotologies complicate the sizing system definition procedure and distributors prefer to use standard sizing system rather than an intelligent system suitable to their customers.Until now, the morphotypes of a population are extracted from measurement charts. However, new technologies such as 3D body scanning open new opportunities to enhance the morphotype generation from a sample of population especially with the 3D data of bodies.The aim of this research is to define an exhaustive methodology to obtain a clustering of human morphology shapes representative of a population and to extract the most significant morphotype of each class. A two-level clustering method (SOM+K-means) based on 3D scans to define 3D adaptive morphotypes mannequins is implemented and the performances are evaluated using real data from the French Sizing Survey conducted in 2003 by the French Institute of Textiles and Clothing.The description of the 3D scans is performed with a computation of the geodesic distributions based on anthropometrics feature points of the human torso which enables a quantitative comparison of the morphologies. These geodesic distributions are then used as inputs for the clustering methods. Finally, a geometrical model associated with reverse engineering techniques has been realized to generate the 3D virtual parametric mannequins from the 3D body scanned of the morphotypes. Based on these morphotypes mannequins, we define an intelligent system for virtual try-on and a new sizing system can be defined in the future.