Fusing RFID and Computer Vision for Occlusion-Aware Object Identifying and Tracking

Abstract

Real-time identifying and tracking monitored objects is an important application in a public safety scenario. Both Radio Frequency Identification (RFID) and computer vision are potential solutions to monitor objects while faced with respective limitations. In this paper, we combine RFID and computer vision to propose a hybrid indoor tracking system, which can efficiently identify and track the monitored object in the scene with people gathering and occlusion. In order to get a high precision and robustness trajectory, we leverage Dempster-Shafer (DS) evidence theory to effectively fuse RFID and computer vision based on the prior probability error distribution. Furthermore, to overcome the drift problem under long-occlusion, we exploit the feedback from the high-confidence tracking results and the RFID signals to correct the false visual tracking. We implement a real-setting tracking prototype system to testify the performance of our proposed scheme with the off-the-shelf IP network camera, as well as the RFID devices. Experimental results show that our solution can achieve 98% identification accuracy and centimeter-level tracking precision, even in long-term occlusion scenarios, which can manipulate various practical object-monitoring scenarios in the public security applications.