Magnetotelluric signal denoising method based on sparse decomposition threshold value setting

Abstract

The invention discloses a magnetotelluric signal denoising method based on sparse decomposition threshold value setting. The method comprises the following steps that: carrying out equal-interval segmentation on a magnetotelluric signal time sequence, and selecting high-quality data segments; according to the mean square root values of the high-quality data segments, obtaining a decomposition threshold value; then, judging whether the mean square root value of each data segment is smaller than or equal to the decomposition threshold value or not, and if the mean square root value of each datasegment is smaller than or equal to the decomposition threshold value, proving that the data segment is a non-strong-interference data segment; if the mean square root value of each data segment is greater than the threshold value, proving that the data segment is a data segment to be subjected to decomposition processing; when a time-frequency atom is constructed, utilizing NPSO (Niche Particle Swarm Optimization) OMP (Orthogonal Matching Pursuit) to carry out spare decomposition on the data segment to be decomposed until a residual signal mean square root value is smaller than or equal to the decomposition threshold value; and finally, splicing each segment of decomposed residual signal with the non-strong-interference data segment obtained by judgment in sequence to obtain a denoised magnetotelluric signal. The method is high in reliability, does not decompose a high-quality signal, and keeps more useful information.

Description

technical field [0001] The invention belongs to the technical field of magnetotelluric signal processing, and in particular relates to a method for denoising magnetotelluric signals based on sparse decomposition threshold setting. Background technique [0002] Magnetotelluric detection technology is an important electromagnetic prospecting technology. It takes natural alternating electromagnetic field as the field source, collects magnetotelluric data by measuring the mutually orthogonal electric and magnetic fields on the surface, and then calculates the apparent resistivity value of the earth to distinguish the underground. The role of electrical structural properties. Due to the wide frequency range of the natural electromagnetic field, large detection depth, and too far away from the field source, the magnetotelluric signal received by the instrument on the ground is weak, and it appears as a random signal in the time domain. Therefore, the data collected in the field w...

AI Technical Summary

Problems solved by technology

Therefore, the data collected in the field will inevitably be interfered by various electrical and magnetic noise signals nearby.

For exploration in mine-concentrated areas and near cities, due to the influence of various factors such as high-voltage cables, wireless communication base stations, and large-scale mechanical equipment operations, the apparent resistivity-phase curve calculated from the collected magnetotelluric data is excessively distorted, which affects the The interpretability of the underground electrical structure and the reliability of the collected data itself have always troubled most geophysicists.

[0003] In recent years, new modern digital signal processing technology has been continuously developed. Magnetotelluric data processing is mainly based on methods such as signal-to-noise identification. The low-frequency part data has been greatly improved, but the processing of the strong interference segment has a certain degree of blindness, which has an impact on the low-frequency part. In the past two years, sparse decomposition has been applied to this field, mainly based on Gabor base and sine-cosine wavelet base. Due to its limitations, the solution to impact noise such as pulse, charge and discharge, etc. is not very applicable, and some useful signals are more or less removed

The particle swarm algorithm is the main method for sparse representation of impact atoms, but its algorithm is too simple. For the impact atom library with diverse forms and massive magnetotelluric data, too many useful signals are removed in the processing of massive magnetotelluric data.

At the same time, the above sparsity decompositions are all based on experience to set the sparsity. In order to ensure the accuracy, the sparsity of the general setting is greater than the actual sparsity, which inevitably removes too much useful information of magnetotellurics

Images