A New Post-Processing Method for Ringing Noise Suppression of Magnetic Resonance Sounding Signal
Yang Zhang, Yue Zhou, Yingni Liu, Xuetao Zhang, Tingting Lin · IEEE Transactions on Geoscience and Remote Sensing · 2024
Magnetic resonance sounding (MRS) is an effective method for groundwater detection due to its direct sensitivity to the protons. However, in the actual measurement, the strong ringing current is always collected, which distorted the data samples in the early time. Different from other noise corrupting the signal, the ringing interference cannot be removed by the de-noising methods in the signal post-processing. The current processing strategy is to delete the distorted data samples, but the fitting error of the initial amplitude will be increased, which reduces the accuracy of the inversion result. Therefore, this article attempts to propose a new post-processing method for ringing noise suppression of MRS signal to address the problem that cannot be solved by the traditional method. Deep learning was applied to verify the idea. The mapping relationship between the signal containing ringing and the ringing noise is established thereby removing the noise to restore the ringing-free signal. The numerical simulations and the field experiments prove that the proposed method based on deep learning may suppress the ringing noise without losing valid information. Compared with the traditional processing strategy, the fitting error of the initial amplitude can be reduced to less than 10% and the root mean square error (RMSE) is less than 0.015 by using the new post-processing method, which significantly improves the accuracy of the measurement results.