Abstract:

For the maneuvering target tracked via the dual-station bearing tracker, to solve the problem of the bounded mixture likelihood caused by the incomplete information and unstructured environment, an unscented particle filter with spatiotemporal soft constraint (SCUPF) algorithm is proposed. For the unknown target prior, the target's position is achieved via the spatial intersect measurement method with the epipolar geometric constraints, so as to predict the center of the camber to calculate the turn rate. To cover the multi-area likelihood, the target state is updated via the unscented transformation, and the importance distribution is modulated via a fuzzy measure. To keep the property of dimensionality invariant of the estimation variance, the posterior distribution of the target state is approximated by drawing the importance samples from the Dirac approximation. Simulation results demonstrate that for the typical scenario that the point target tracked by the dual-station theodolites, compared with the Rao-Blackwell multiple model particle filter (MMRBPF) algorithm, the computational complexity of the proposed SCUPF algorithm is on the same order as the single-model particle filter. Meanwhile, compared with the constrained auxiliary particle filter (CAPF) algorithm, the filtering accuracy is improved by 27%-41%.

Key words: dual-station bearing tracking, maneuvering target, bounded mixture likelihood, spatiotemporal soft constraint, unscented particle filter