Based on the global satellite navigation system(GNSS), the angle signal of the grader during the leveling process was detected by using paddy field rotary tiller field test. The wavelet hard threshold method was used to obtain the low frequency signal, and real-timely estimate the noise variance of the angle signal as the corrective information of Kalman filter. And then the second correction on the signal was performed using Kalman filtering with the system input of the low-frequency signal. The experimental results show that the wavelet hard threshold-Kalman fusion algorithm has better filtering effect than the single wavelet threshold method and Kalman filtering, respectively. When the inclination signal is processed by the fusion algorithm, the signal-to-noise ratio of the signal is increased from 21.704 to 39.116, and the root mean square error reduced from 0.035 1 to 0.012 6.