Fully Automatic Skeleton Tracking in Optical Motion Capture

Recently, methods to accurately capture the motion of people gained increasing interest for variety of applications including interaction, animation, orthopedics, and rehabilitation. Compared to markerless approaches, marker-based methods are typically more accurate and are more robust against occlusions [8]. Major challenges in this context are to associate the observed markers with skeleton segments, to track markers between consecutive frames, and to estimate the underlying skeleton configuration for each frame. Existing solutions to this problem often assume fully labeled markers, which usually requires labor-intensive manual labeling. In our previous work [7], we propose a fully automated method to initialize and track the skeleton configuration of humans from optical motion capture data. This method applies a flexible T-pose-based initialization that works with a wide range of marker placements without additional manual effort. To this end, we scale a standard human skeleton, based on Contini [3], to the person’s size and align the skeleton to the person’s limbs. After initialization we robustly estimate the skeleton configuration through least-squares optimization. Initialization methods without an underlying known skeleton structure were investigated by Ringer and Lasenby [9], Kirk et al. [5] and de Aguiar et al. [4]. These methods require a certain number of markers associated to each segment and an additional manual labeling step. Assuming known marker labels, several authors estimate the joint positions of the skeleton segments while taking into account skin movement artifacts [1, 2]. The contribution to this workshop are recent enhancements in skeleton tracking methods that build on our previous method [7]. First, using a large database of known skeleton configuration, we are able to mitigate the requirements during initialization. Instead of T-Pose initialization, we are able to initialize tracking during natural walking movements. Second, we update the association of markers to segments and the corresponding relative positions online during the tracking process in order to cope with initialization errors.