Coopération entre un robot mobile et des caméras d'ambiance pour le suivi multi-personnes

In this paper, we present a cooperative perception system between wall mounted cameras and sensors on-board a mobile robot to : (i) track and follow a given target person wearing a Radio Frequency Identification (RFID) tag, and (ii) facilitate navigation in the presence of passers-by during the execution of this mission. We employ a "trackingby-detection" approach within a Particle Filtering strategy fusing visual detections from the wall mounted cameras and detections from the various on-board sensors (laser, active vision, RFID). The multi-person tracker’s ability to track passers-by near the robot distinctively is demonstrated through qualitative and quantitative off-line experiments. Finally, the designed perceptual modalities are deployed on our robotic platform, controlling its actuators via visual servoing techniques and free space diagrams in the vicinity of the robot, to illustrate the robot’s ability to follow a given target person in human crowded scenes.