In order to solve the problem of serious noise disturbance from signals in the process of the running rotor, as well as the problem of low rotor unbalance identification accuracy, a method combining the mathematical morphology filter and Fourier transform is applied that uses the characteristics of the mathematical morphology filter to effectively filter noise disturbance and improve the signal-to-noise ratio. This method could reduce the noise of the rotor vibration signal and accurately extract the rotation frequency amplitude and phase through the work state of the rotor. The calculation of the rotor unbalance is realized using the influence coefficient method, by adding and reducing weight to change the unbalance state of the rotor. Experimental results show that the proposed method can effectively filter a variety of noise disturbances and improve the precision of rotor unbalance identification.