Measurement-Driven Synthesis of Female Digital Mannequin Using Convex Sub-Volumes

In the context of computer-aided apparel-fitting simulation, the problem of generating (a) simulation-inexpensive and (b) tailor-measurement-driven digital mannequins is central. Three-dimensional scanning of human bodies produces high-fidelity datasets. However, this technique does not satisfy conditions (a) and (b) above. In addition, it requires extensive data cleaning and processing. Existing approaches to this problem broadly fall into these mainstreams: (i) biased scaling, interpolation, or morphing of template models; or (ii) bottom-up construction of anatomy (bone medial axis, kinematic joints, muscles, skin, and other layers). Both alternatives imply extensive scanning, application of heuristics, tuning, and storage, among other tasks. Both alternatives produce non-convex datasets that have to be processed further for cloth–body interaction simulation, as physics engines require some type of data convexity for realistic simulations. This manuscript presents a modeling methodology that partially overcomes these limitations by (1) coarsely approximating a template female body with sets of convex volumes (ellipsoids and <i>cushions</i>), (2) building a set of Reference Mannequins for a particular set of extreme and average tailor measurements, and (3) creating sets of functions that synthesize new individuals of digital mannequins as reunions of convex volumes that satisfy specified tailor measurements. These mannequins present a reasonable and realistic demeanor. At the same time, they are shown to be economical at the stage of simulation of garment fitting. Future work is encouraged to define kinematic chains for straightforward pose definition, modeling male bodies, and exploring the behavior of the synthesis functions with more parameters.