Convex Optimization Approach to Identity Fusion for Multisensor Target Tracking

We consider the problem of identity fusion for a multisensor target tracking system whereby sensors generate reports on the target identities. Since sensor reports are typically fuzzy, incomplete, or inconsistent, the fusion of such sensor reports becomes a major challenge. In this paper, we introduce a new identity fusion method based on the minimization of inconsistencies among the sensor reports by using a convex quadratic programming (QP) formulation. In contrast to Dempster–Shafer's (D–S) evidential reasoning approach which suffers from exponentially growing complexity, our approach is highly efficient (polynomial time solvable). Moreover, our approach can fuse sensor reports of the form more general than that allowed by the evidential reasoning theory. Simulation results show that our method generates reasonable fusion results which are similar to that obtained via the evidential reasoning theory.