Abstract: In the exploration of radio wave perspective in coal mines, metal bodies and electrical equipment have strong interference with radio wave signals, which affects the amplitude preservation and imaging accuracy of radio wave perspective data. Based on the measured data, this paper focuses on the analysis of the noise signal characteristics of the four common interference sources of drainage equipment, communication equipment, locator equipment and large electrical equipment in coal seam working face. Four kinds of interference source noises are added to the theoretical data of radio waves. The median filtering, Savitzky-Golay filtering and wavelet filtering methods are used to quantitatively analyze the denoising effects with different parameters. After median filtering （window 5）, the signal-to-noise ratios of the data are increased to 6.72, 6.42, 6.33 and 6.15, respectively. After Savitzky-Golay filtering （window 9）, the signal-to-noise ratio of the data is increased to 7.32, 8.13, 9.51 and 7.67, respectively. After wavelet filtering （db6, 3 layer）, the signal-to-noise ratio of the data is increased to 8.49,7.68, 8.21 and 11.63, respectively. After a large number of filtering experiments, it is recognized that S-G filtering is suitable for processing the noise of communication equipment and locator equipment, and wavelet filtering is suitable for processing the noise of drainage equipment and large electrical equipment. The best moving median window value of median filtering is 5, the best fitting window value of S-G filtering is 9, and the best wavelet basis of wavelet filtering is db6.These three filtering methods are further applied to the measured data. After filtering by the three methods, the signal-to-noise ratio of the measured data is effectively improved. In the attenuation coefficient tomographic results, the YC4 abnormal area is less affected by noise, and the location of the abnormal area is more accurate after filtering. It is proved that the three filtering methods can effectively improve the signal-to-noise ratio and detection accuracy of radio wave data by selecting appropriate filtering parameters for different interference sources.