A Generic Method for Geometric Contraints Detection

Abstract

In this paper, we present a generic method to automatically detect geometric constraints on motion capture animations. At each frame, elementary geometric constraints are computed with respect to a reference which can either be the world coordinate system or any moving object in the scene. We then use constraint-related concepts of union and intersection to merge the elementary constraints together and/or to generate new ones. Finally, our algorithm provides an exhaustive list of geometric constraints with an accurate evaluation of their duration. The detected constraints can characterize virtual human motion (e.g. footprints) as well as interactions with moving objects of the scene (e.g. a hand touching a ball). While our approach also detects sliding geometric constraints, we focus our discussion on detecting positional geometric constraints.