Human beings vary greatly in shape and size. Variations induced by body deformation accounts for significant differences during movement. While making human-centric products, designers often work with compromised anthropometric information. The two most common approaches are: using empirical anthropometric values and working with human models that scale unrealistically due to the lack of 3D body shape information: too much information about human shape is lost when defining segments of the body as geometric shapes such as cylinders, cones, and spheres. 3D scanners can provide accurate scans of humans but only static poses can be captured. To track movement, optical motion capture (mocap) systems that use reflective markers are the gold standard. This setup is inherently limited to indoor settings and the resulting human models do not incorporate soft tissue deformation. So-called ‘4D scanners’ exist only in a few expensive, specialized, dedicated laboratories.
Aim: Currently, if the scanning solution is dynamic and affordable, it is inaccurate; if it is accurate and affordable, it is static; if it is accurate and dynamic, it is not cheap. Our aim with InLocoMotion is to bridge this gap and make dynamic 3D models accurate and affordable and to make them available in CAD (computer aided design) environments.
Joint angles have been proved to be a parameter for determining pose. A systematic review will be conducted to determine the feasibility of off-site mocap systems that can be used to complement static scanning methods. We propose to scan a gender-balanced set of subjects in standing and cycling poses. Using these known subjects and poses, regression techniques and principal component analysis will help create algorithms to predict fresh poses of fresh subjects. To validate our methods, drag force of fresh subjects in cycling pose will be measured in a wind tunnel, while their CAD models will be subject to computational fluid dynamics techniques to determine ‘virtual’ drag force.
The review allowed us to identify the best system for capturing joint angles off-site. Previous results from our work using standing and sitting 3D scans has shown promising results. Our research with 3D skull modelling demonstrated high accuracy (root mean square error 2 mm) of our methods. The envisioned algorithm (InLocoMotion) could address broad problems in many sectors in the long run including but not limited to medical technology and performance sports.