Figure 10

Output from the multiobject tracking algorithm working with input from six camera views. (a) Shows four camera views of a scene with several humans walking. Each camera independently detects/tracks the humans using a simple background subtraction scheme. The center location of the feet of each human is indicated with color-coded circles in each view. These estimates are then fused together taking into account the relationship between each view and the ground plane. (b) shows the fused trajectories overlaid on a top-down view of the ground plane.