A moving horizon estimation (MHE) method of the constrained system with uncertain measurement output and norm-bounded parameter uncertainties is proposed. Based on the probability of known error of the output signal, the initial state variables and disturbance are estimated by minimizing performance object at each sampling moment. The estimation values of the state variables are calculated by the optimization theory of model predictive control (MPC). Simulation results show that MHE method has better ability than the method of the robust filter.