Marker-less 3D video motion capture in cluttered environments

Measuring the joint kinematics of an athlete is a key component for sports biomechanical studies. In our work, we used only video cameras and do not need the subject to wear any markers or sensors on their body The motivation of using a marker-free motion capture setup is driven by the need to acquire video images non-intrusively in cluttered and outdoor environments, e.g. during tournaments. We proposed a new analysis-by-synthesize method that built on the concept of collaboration between computer vision and computer graphics to capture human movements. Unlike common vision-based tracking, we do not use the noise-sensitive image segmentation and skeletonization. Our algorithm automatically realizes the colour or texture onto an animatable 3D human model of our subject. Our computation will synthesize the 3D puppet postures such that it minimizes the differences between the synthesized movements and real athlete's motion. This is achieved by using the simulated-annealing algorithm to compute the degree-of-freedoms of the joint kinematics. The results show that our method is able to track the motion of the arms that appear highly articulated and qualitatively small in the images. It can also operate in cluttered environments.