Proper dynamic models and estimation methods are crucial factors to achieve a nice tracking performance for a reentry target. First, the dynamic model of maneuver reentry vehicle was chosen for reentry target tracking, and the tracking problem was equivalent to the simultaneous estimation of the state and the parameter. Then, the relativity of the state and the parameter were discussed by numerical examination. Next, for the information lost caused by the assumption of a deterministic input vector in the original dual unscented filter, an improved dual unscented filter based on a random input vector was proposed, and a detail accuracy analysis was achieved in theory. Finally, the accuracy, applicability and ability for processing incomplete data were validated by simulations.