A tracking-by-detection method with radio-frequency tomography network

This paper presents a tracking-by-detection method which can localize and track an unknown and time-varying number of moving objects with radio-frequency tomography network. First, a novel detection method based on scanning circles is proposed to preliminarily estimate the number of targets and determine the multi-target positions using the received signal strength measurements. Then, particle probability hypothesis density filter algorithm is applied to perform target tracking and eliminate false positives caused by multipath environment or system noise. Experimental results demonstrate that the tracking-by-detection method achieves high accuracy in estimating target number and positions.