Target tracking via a sampling stack-based approach

We propose a novel approach to tracking a target in clutter based on the stack algorithm for tree search. The proposed tracking approach reduces the size of the search tree by employing a coarse discretization of the target state space. To reduce the quantization error that results from coarse discretization, the representative value of each quantized region is sampled from an estimated importance sampling function. A forgetting factor is included in the likelihood metric to control the effect of previous decisions and to reduce algorithm complexity. Simulations reveal that the proposed algorithm provides significantly reduced complexity while suffering no performance degradation relative to stack-based tracking with finer quantization.