Multi-marker tracking for large-scale X-ray stereo video data

Xiaoyan Jiang, Marcel Simon, Yuan Yang, Joachim Denzler

Abstract: Highlights •Efficient and effective tracking of many objects in big data scale videos is tackled.•A fast two-stage graph-based multi-object tracking-by-detection approach is proposed.•Local and global features are adopted in the data association of each graph and tracks are found by optimizing a global objective.•We apply the proposed approach to the tracking of implanted markers in stereo X-ray cardiac recordings and achieve human-level accuracy.