Target Identity Probability Analysis for Multi-Target Tracking

This paper proposes a novel target identity probability analysis algorithm for multi-target tracking (MTT) problems. By introducing a target identity probability matrix, finding the identity probability is converted into the problem of target-to-target association. The issue in solving the equivalent problem is that the computational complexity increases exponentially with the increase of the problem size. We propose a Gibbs sampling approach to find approximate polynomial-time solutions. Theoretical analysis reveals that the proposed algorithm provides a performance-guaranteed approximation. Numerical simulations are conducted to support the analytical findings.