Joint data association using importance sampling

Data association, which involves the assignment of one collection of objects to another, is an important problem in multiple target tracking. Exact computation of data association probabilities is not always computationally feasible, in particular when many targets are in close proximity and share many measurements. In this paper a Monte Carlo method for approximation of data association probabilities in such situations is proposed. The proposed method is a refinement of an existing importance sampling method for matrix permanent approximation. It is shown via numerical simulations that the proposed method can accurately approximate data association probabilities in dense multiple target scenarios with reasonable computational expense.