In radio astronomy observations, radio frequency interference (RFI) is mixed into the telescope receiving system in various forms. The existence of RFI brings misjudgment to the observation or reduces the observation signal-to-noise ratio. In recent years, domestic and overseas radio astronomy has developed rapidly. Large-scale radio telescopes or arrays of large telescopes at home and abroad have been constructed successively. Observation sensitivity is greatly improved, and the influence of RFI is particularly prominent. With the development of science and technology and the intensification of human activities, RFI has become increasingly serious and irreversible. We propose to use the two-dimensional discrete wavelet transform method to analyze the data of radio astronomy observations, and to perform wavelet transform on the time-frequency sequence output by the telescope system. According to the wavelet coefficients, each component in the original signal is separated, and the corresponding threshold value is obtained by statistics of each component. Each component is compared with a threshold to identify the interference component and mark it for removal. This method is used to process the actual observation data. The results show that this method can well mark and eliminate the interference signal, and improve the observed signal-to-noise ratio.