As a result of modeling uncertainties, data decade unbalanceand noise, Fault Estimation (FE) for real quadrotor is rough with low accuracy when model observer based FE scheme - Adaptive Thau Observer(ATO) is applied. Based on these issues, this paper proposes a comprehensive optimization approach on FE accuracy for quadrotor using ATO. First, the ATO observer is improved by adding unmolded terms with parameters derived from identification. Secondly, a data adjustment based on amplifying and reducing is introduced to overcome the data decade unbalance. Thirdly, the Infinite Impulse Response (IIR) filter is utilized to decrease the noise in measurement signals. Finally, experiments for injecting the rotor fault have been carried out to show the effectiveness of the proposed optimization method and the results are discussed.