An Improved Particle Filter Based Track-Before-Detect Method for Underwater Target Bearing Tracking

Track-before-detect (TBD) methods have been shown to greatly abate measurement-to-track association (MTA) challenges which could cost plenty of operator workload in many detection systems. In the field of underwater acoustic signal processing, the low signal-to-noise ratio, random missing measurements, multiple interference scenarios, and merging-splitting contacts in measurement space are challenging for common target tracking algorithms. As a result, particle filter (PF) based track-before-detect methods that compute posterior density distribution directly using beamformer output instead of bearing measurements are effective in these cases. However, the general PF always suffers from the particle impoverishment problem which can lead to the misleading state estimation results due to system process error. To cope with above problems, an improved particle filter based track-before-detect method is proposed in this paper. The proposed PF-TBD method adopts crossover and mutation operators from genetic algorithm to evolve particles with small weight. And the quasi-Monte Carlo method is applied into resampling procedure of particle filter. The effectiveness of the method is verified by using experimental data obtained at sea, which can continuously and accurately track target bearings in the case of temporary signal disappearance and multi-target intersection. The estimated result is close to the real value even when the motion model suffers mismatch.