Noise removal for semi-airborne data using wavelet threshold and singular value decomposition

Abstract

Field data from the semi-airborne transient electromagnetic method (SATEM) contain various noises. Therefore, performing denoising before data inversion and interpretation is important. Currently, different types of noise are suppressed separately, not considering the variance in the signal-to-noise ratio at different time ranges. The signal details after denoising are lost and the residual noise is large. We propose a denoising scheme that uses wavelet thresholding (WT) and adaptive singular-value decomposition (SVD) filtering. We first intercepted the decay curves into several segments according to the statistical characteristics of the noise. Subsequently, a WT denoising method was used to remove the Gaussian noise and suppress noise in different signal segments. Finally, the denoised datasets were decomposed and filtered using the SVD method. The number of singular values was determined based on the optimal difference spectrum peak value, which was used to adaptively restore the responses in each segment. The proposed denoising scheme can effectively suppress multi-source noise and provide solid datasets for inversion, which was confirmed through a field survey in Shanxi Province.