Iterative joint probabilistic data association for avoiding track coalescence and track swap in multi-target tracking

Abstract

The iterative technique developed in this paper achieves simultaneous reduction of track coalescence and track swap by computing a better prior for the joint association event used in estimating target states. This prior is computed using expectation maximization (EM). Using the computed prior the iterative method avoids track coalescence and track swap while preserving the robustness of JPDA towards clutter and missed detection. Compared to other coalescence avoidance schemes, the proposed method avoids coalescence and swap without pruning the joint association events. Monte Carlo simulations verify the advantage of the proposed method over other approaches in a cluttered multi-target environment.