Marker Correspondence Initialization in an IR Motion Capturing System

Marker-based motion capturing (MoCap) systems are of great importance among the different visual referencing systems, because of their robustness, speed, and precision. Due to the lack of highly descriptive visual features in captured images in such systems, establishing the correspondences between the multi-view observations of the markers is a major challenge. This problem is even more challenging when there is no preceding or low reliable data on the markers’ positions. Under these conditions, the correspondences have to be found from a large space of possible matches. This paper aims to provide a solution for this correspondence initialization problem in multi-view tracking systems based on graph theory and Girvan-Newman community detection. The validity and robustness of the algorithm are verified via experimental analysis.