Data association and tracking using hidden Markov models and dynamic programming

The problem of target tracking from distributed sensors in a cluttered environment is addressed. An algorithm that achieves target tracking and target motion analysis is introduced. This technique uses the formalism of hidden Markov models (HMMs) and is based on two successive steps: the first one consists of a spatial fusion of the measurements obtained at a given time, and the second achieves temporal association, thus leading to the target trajectory. This approach basically differs from classic ones (PDA, JPDA, etc.) because it requires no initialization and no a priori hypothesis for target motion. Simulation results are shown in which the multiple target and maneuvering target cases are given particular attention