Abstract: Accurate detection of tide level is a very important topic for ocean monitoring, circulation analysis, and so on, and it is also a crucial link for the development of China’s marine industry. Previous studies have shown that the ready-made global positioning system（GPS） can use the characteristics of multipath interference to monitor the tide level. Lomb scalar periodogram（LSP） is usually used to extract multipath frequency from signal-to-noise ratio（SNR） sequences. In order to improve the resolution of the data, continuous wavelet transform can be used to extract the instantaneous frequency of the SNR sequence and obtain the instantaneous tide level. Based on this, the instantaneous tidal level is obtained. Because the instantaneous tide level data is very noisy, there are many outliers and errors due to the noise of the tide data. This paper tries to use different types of SNR data of the same satellite from multi-frequency and multi-mode stations at the same time, and perform a certain regularization noise reduction processing on the basis of preserving continuity. The numerical results show that the regularization method can effectively reduce noise level and improve the accuracy of tide level analysis.