A Track Re-Association Approach for Continuous Target Tracking in Non-Line-of-Sight Environments Using mmWave Radar

In urban environments, where buildings are densely arranged with complex layouts, the reception of direct signals in non-line-of-sight (NLoS) areas is hindered by building obstruction, making accurate and continuous tracking of moving targets in these areas a significant challenge. Existing target tracking algorithms often misclassify reappearing targets as new ones after a brief disappearance due to imperfect data association strategies, leading to erroneous target counts and interruptions in tracking continuity. To address this issue, this paper proposes a track re-association method for tracking moving targets in NLoS areas based on labelled multi-Bernoulli (LMB) filtering. The method first employs frequency modulated continuous wave (FMCW) radar to acquire first-reflection signals for estimating the range, velocity, and angle of targets within the NLoS areas. Subsequently, within a multi-Bernoulli Bayesian recursive filtering framework based on random finite sets, the method quantitatively evaluates the probability of target reappearance after a brief disappearance by incorporating prior target motion information and the Mahalanobis distance metric. Global optimal matching is then utilized to achieve track re-association, effectively eliminating tracking discontinuities caused by mislabeling. Experimental results demonstrate that the proposed method can accurately estimate the number of targets in NLoS areas while ensuring tracking precision and significantly enhancing tracking continuity.