Efficient Marker Matching Using Pair-Wise Constraints in Physical Therapy

In this paper, we report a robust, efficient, and automatic method for matching infrared tracked markers for human motion analysis in computer-aided physical therapy applications. The challenges of this task stem from non-rigid marker motion, occlusion, and timing requirements. To overcome these difficulties, we use pair-wise distance constraints for marker identification. To meet the timing requirements, we first reduce the candidate marker labels by proximity constraints before enforcing the pair-wise constraints. Experiments with 38 real motion sequences, our method has shown superior accuracy and significant speedup over a semi-automatic proprietary method and the Iterative Closest Point (ICP) approach.