Adaptable digital human models from 3D body scans

Abstract

The human body comes in many sizes and shapes. So, realistic virtual mannequins, which represent body shapes that occur in a target population, are valuable tools for product developers who design near-body products. Statistical shape modeling is a promising approach to map out the variability of a population using 3D scans. A statistical shape model (SSM) can therefore be used as a virtual mannequin. By adapting the parameters of the SSM, a new, realistic surface can be obtained. In this chapter, we present a registration-based framework to build such a statistical shape model. The framework consists of three parts. First, the input surfaces are brought into correspondence with each other by elastic surface registration. Next, a statistical shape model is built. Then, the relationship between the body shape and body measurements is described. Furthermore, a posture model is constructed to filter out posture variances that occur in the database and generate a posture-invariant shape model. The statistical shape model is convertible to a digital human model (DHM) that contains much more 3D information than traditional DHMs (e.g., Jack). The surface prediction from features is an intuitive extension of the traditional anthropometry percentiles and can also be the basis of a new sizing system for near-body products. In this chapter, we propose a fast, skeleton-less, marker-less, data-driven method to generate statistical shape models in a posture-invariant way.