Enabling identity-aware tracking by vision-RFID fusion
Haofan Cai, Chen Qian · 2020
Person identification and tracking (PIT) is an essential research topic in computer vision (CV). A CV-based system typically needs to identify, locate, and track persons appearing in its sight. In this work, we propose RFTrack, an RFID and CV fushion system that enables cameras in public areas (like surveillance cameras) to 'recognize' the physical-identity(ID) of persons in the fields of view and track the persons with specific IDs with no training efforts. By asking the users to perform a simple authentication, the system will be aware of the targets' IDs in its sensing range. Later through comparing the motion trajectories derived from both camera videos and RF signal, we can associate RFID-tagged human objects in videos with their physical IDs. A preliminary study conducted shows that RFTrack can actively identify and track the RFID-tagged target objects using commercial RFID devices and cameras, in complex indoor environments where various multipath reflectors exist.