Skill measurement through real-time 3D reconstruction and 3D pose estimation

Skill measurement for sports analysis is still a challenging issue for sports scientists. In recent years, video based analysis of body movements and its corrections has become a widespread tool within modern athlete training centres throughout the different training stages of the athletes' pre-competition preparation. The challenges imposed by this supporting service are manifold and range from high resolution capturing of the athlete to a precise real-time reconstruction of body movements. By augmenting ideal virtual poses to the real motions, a coach is able to analyze and communicate incorrect or even incomplete motions of the athlete. For the precise measurement of body motions, typically motion capture systems (MoCap) are used. During the last decade marker based motion capture systems matured and did find their way into the area of medical and sports science. Markerless motion capturing systems, however, are still in its infancy. This paper describes a new approach to markerless motion capturing and its use for sports skills measurement and analysis