Color-based human tracking by a robot equipped with a panoramic camera

The goal of this work is to make a robot able to track a person as both the robot and the person move.
The user of the system provides to the robot the initial location of the human in the filed of view of a panoramic camera.
Then, the robot tracks the human by matching color information within consecutive frames in time. Color information is computed
for the head, body and legs of the human. Moreover the spatial relations of these regions are monitored.

  • Contributions: Damien Muti, Olivier Bantiche, Antonis Argyros, Panos Trahanias, Dimitris Tsakiris
  • Current status: Method ported on the robotic wheelchair of CVRL, preliminary results available.
  • Sample results

You may download
a video
(2,7 Mbytes) showing a person being tracked by the robotic
wheelchar. Notice that the person remains trackable even though at
some time it is occluded by another moving person. You may
download another video
(4,2 Mbytes) which shows the same results with the
difference being that panoramic image frames are unfolded and therefore,
results are more intuitive.

Available documents:

  • Technical report under preparation