Markerless tracking and surface measurements in biomechanical applications

The need to capture human motion and to reconstruct a human body has been significantly recognized. Special interest lies in the description of changes as a result of pathologies compared to normal healthy subjects and short or long-term adaptations due to interventions. In this paper, we focus on a makerless vision based method to track 3-D motion information from video sequences. Although video sequences acquired from multiple static synchronized CCD cameras have been used as a tool for 3-D reconstruction of the movement of a human body, the accuracy and reliability remain as concerning issues, which are addressed in this paper. After a thorough self-calibration of the video system, the system is initialized manually with the help of a graphical user interface. The articulated soft objects are the approximations for the following constrained least squares matching procedure. Additionally, the least square matching algorithm takes into account the one to one relationship of a 3-D point to a corresponding limb, described as a quadric primitive, and therefore adds a topological information to every point. Graphical results for the 3-D motion capture data and the 3-D surface measurements will be presented.