Optimizing retroreflective marker set for motion capturing props

Abstract

One of the most widely used motion capture (MoCap) methods depends on retroreflective markers placed on an object. Through cameras, the MoCap system captures the moving prop’s (3D object) motion. However, noise in MoCap data caused by the shape of the captured volume, motion between frames, ghost points, and markers’ self-occlusion could impact the quality of the captured data. To improve the quality of capturing the motion of props, we tackle the problem of finding an optimal marker set configuration for a given input prop while considering various constraints. By “props,” we mean any objects or handheld items of different shapes and sizes on which markers can be precisely placed to reduce MoCap errors. We propose an approach to optimize the placement of optical (retroreflective) markers over props while encountering various constraints, such as the visibility of markers, the number of makers used, the symmetry of the marker set, and markers’ overlapping. We solve the marker set configuration problem using an optimization-based method, the reversible-jump Markov chain Monte Carlo. We provide marker set configurations for various props and constraints obtained through several simulations we ran to evaluate the performance of our method.