Error Propagation in Multi-Camera Tracking

Multi-camera tracking often involve the projection of image data onto a ground plane. In this paper, we analyze the propagation of object tracking errors after trajectory transformation from multiple views onto the ground plane. In particular, we contrast a deterministic and a probabilistic algorithm and present an empirical study of their multiple object tracking results on dataset of 18000 frames. By measuring tracking accuracy, we highlight the processes that generate the most significant errors and how these errors impact the estimation of the final object location. Ultimately, the propagation of these errors from the image plane to the ground plane trajectories gives insights for future enhancement of the algorithms employed.