A multi-target track-before-detect algorithm in the scenario of bearing trajectory crossing

Traditional detect-then-track approaches encounter challenges when the bearings of multiple targets occlude each other, leading to trajectory interruption or mismatch. To address this issue, a multi-target track-before-detect algorithm is proposed based on the generalized labeled multi-Bernoulli filtering. This method directly uses the raw data of the array elements as measurement in the form of covariance matrices, without pre-processing such as beamforming. Probabilistic models for trajectory birth, death, evolution, and observation are established. Multidimensional integration in the update step is eliminated by using a theoretical approximation. Joint multi-target detection, bearing tracking, and trajectory management are implemented. Simulation results show that the proposed method can accurately estimate the number of targets and produce consistent trajectories even when the bearings of multiple targets occlude each other. Additionally, it achieves high tracking accuracy under a low signal-to-noise ratio scenario (−18 dB). Sea trial data from seabed linear array also verify the performance of the proposed method.