A noise suppression method for electromagnetic detection data in semi-aircraft frequency domain based on wavelet debasing and notch fusion

Abstract

The present invention discloses a method for suppressing noise in semi-airborne frequency-domain electromagnetic detection data based on wavelet debasing and denoising combined with notch filtering. The method comprises the following steps: Step 1: performing spectral analysis on a measured signal containing noise; Step 2: processing the measured signal according to Step 2a, and then processing the resulting signal according to Step 2b; Step 3: applying frequency-domain wavelet threshold denoising to the signal data obtained after the processing in Step 2 to filter out other random noise. The present invention discloses a novel combined noise suppression method that combines wavelet debasing and denoising with notch filtering to process power frequency interference. This method can effectively suppress the main noise in the semi-airborne electromagnetic detection process, and the processed signal meets the requirements.

Description

Technical Field

[0001] The present invention belongs to the field of airborne geophysical exploration, and in particular relates to a noise suppression method for electromagnetic detection data in the semi-aeronautical frequency domain based on wavelet debasing, denoising and notch fusion in this field. The method can be applied to deep earth detection signal processing and terminal receiving signal processing in other fields. Background Art

[0002] Semi-aerial electromagnetic detection uses a high-power artificial electromagnetic source on the ground (a large loop or grounding wire) as the excitation field source, and collects aerial magnetic field signals through receiving equipment mounted on an unmanned aerial vehicle, thereby achieving rapid acquisition of earth electrical information within a large depth range.

[0003] Semi-airborne electromagnetic measurements are susceptible to noise interference, including: low-frequency noise generated by the receiving coil mounted on the aircraft moving in the air; power frequency interference caused by crisscrossing high-voltage lines, railway power grids, large industrial equipment, etc.; and other random electromagnetic noise.

[0004] Existing technologies mainly use wavelet algorithms and exponential adaptive fitting Kalman filtering algorithms to achieve low-frequency noise reduction processing of semi-aeronautical transient electromagnetic data, but almost no processing is done on power frequency interference and other random electromagnetic noise. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a noise suppression method for electromagnetic detection data in the semi-aeronautical frequency domain based on wavelet debasing, denoising and notch fusion.

[0006] The present invention adopts the following technical solutions:

[0007] A noise suppression method for electromagnetic detection data in the semi-aeronautical frequency domain based on wavelet debasing and denoising and notch fusion is improved in that it comprises the following steps:

[0008] Step 1: Perform spectrum analysis on the measured signal containing noise;

[0009] Step 2: Process the measured signal according to step 2a first, and then process the processed signal according to step 2b; or process the measured signal according to step 2b first, and then process the processed signal according to step 2a:

[0010] Step 2a: Decompose the signal into multiple scales using wavelet transform in the frequency domain to suppress low-frequency motion noise. Select coif5 as the wavelet basis in the wavelet baseline removal algorithm to suppress low-frequency motion noise below 16 Hz by performing multi-scale decomposition on the signal in the frequency domain.

[0011] Step 2b: The signal is filtered through a notch filter to remove the power frequency interference composed of 50 Hz and its harmonic components;

[0012] The transfer function H(z) of a single frequency notch filter is:

[0013]

[0014] In the above formula, z1 is the zero point of the filter, is the conjugate zero of the filter, p1 is the pole of the filter, is the conjugate pole of the filter, let:

[0015] but:

[0016]

[0017] In the above formula, w0=2πf0, f0 is the center frequency of the notch filter, T s is the sampling interval, r is the notch filter bandwidth parameter;

[0018] Step 3: The signal data obtained after processing in step 2 is denoised by wavelet threshold based on the frequency domain to filter out other random noise. The process of wavelet threshold denoising is as follows:

[0019] Step 31, converting the noisy signal into the frequency domain through Fourier transform to obtain the real part and imaginary part of the Fourier transform of the signal;

[0020] Step 32, performing wavelet transform on the real part and the imaginary part respectively, and performing wavelet soft threshold denoising on the coefficients of each layer;

[0021] According to the Mallat algorithm of wavelet transform, the signal is decomposed to obtain the wavelet coefficients C at different scales. n，k , then C n，k Perform threshold processing and then reconstruct the denoised signal. The soft threshold algorithm is as follows:

[0022]

[0023] In the above formula, is the wavelet coefficient after taking the threshold, λ is the threshold, N is the signal length, σ is the noise variance, σ=MAD n / 0.6745, MAD n is the median value of the high subband coefficients of the nth layer after wavelet decomposition;

[0024] Step 33, combining the processed real and imaginary components into a new complex number sequence;

[0025] In step 34, the sequence is converted to the time domain by inverse Fourier transform.

[0026] The beneficial effects of the present invention are:

[0027] The present invention discloses a novel joint noise suppression method, which combines wavelet de-radicalization and denoising with notch processing of power frequency interference, and can effectively suppress the main noise in the semi-airborne electromagnetic detection process, and the processed signal meets the requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a simulation diagram of the baseline removal effect of Coif5 wavelet;

[0029] Figure 2 It is a schematic diagram of wavelet decomposition;

[0030] Figure 3 This is the flow chart of wavelet threshold denoising;

[0031] Figure 4 is a schematic flow chart of the method disclosed in Example 1;

[0032] Figure 5 is a schematic flow chart of the method disclosed in Example 2;

[0033] Figure 6 is the time domain diagram of the original measured signal;

[0034] Figure 7 is the spectrum of the original measured signal;

[0035] Figure 8 is the time-frequency diagram of the original measured signal;

[0036] Figure 9 It is the time domain diagram of the signal after noise suppression;

[0037] Figure 10 is the spectrum of the signal after noise suppression;

[0038] Figure 11 It is the time-frequency diagram of the signal after noise suppression. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0040] This invention improves existing noise suppression technology and proposes a noise suppression method for semi-airborne frequency-domain electromagnetic detection data based on wavelet debasing and denoising combined with notch fusion. First, the noisy original signal is subjected to wavelet multi-layer decomposition to remove low-frequency motion noise; then, the notch method is used to remove power frequency interference; finally, frequency-domain wavelet soft threshold denoising is used to process other random electromagnetic noise, suppressing several major noises in the semi-airborne electromagnetic detection process. The processed signal meets performance requirements. Specifically, the method includes the following steps:

[0041] Step 1: Perform spectrum analysis on the measured signal containing noise;

[0042] Step 2: Process the measured signal according to step 2a first, and then process the processed signal according to step 2b; or process the measured signal according to step 2b first, and then process the processed signal according to step 2a:

[0043] Step 2a: Decompose the signal into multiple scales using wavelet transform in the frequency domain to suppress low-frequency motion noise. Select coif5 as the wavelet basis in the wavelet baseline removal algorithm to suppress low-frequency motion noise below 16 Hz by performing multi-scale decomposition on the signal in the frequency domain.

[0044] Select a suitable wavelet function. When selecting a wavelet basis, the main considerations should be the orthogonality, symmetry, compact support, and vanishing moment of the wavelet basis. Since the low-frequency motion noise signal oscillates unsteadily and the signal frequency is disorderly distributed, the vanishing moment and compact support of the function should be considered first when selecting a wavelet function. Several commonly used wavelet bases, haar, coif5, and sym8, were used to perform multiple MATLAB simulations on the actual data collected by semi-airborne electromagnetic detection. Finally, coif5 was selected as the wavelet basis in the wavelet baseline removal algorithm. Figure 1 This is a simulation diagram of the baseline removal effect of Coif5 wavelet. Figure 2 It is a schematic diagram of wavelet decomposition.

[0045] Step 2b: The signal is filtered through a notch filter to remove the power frequency interference composed of 50 Hz and its harmonic components;

[0046] The notch filter method can specifically suppress a certain noise frequency in a noisy signal without losing other useful frequency components in the mixed signal. The notch filter is usually an infinite impulse response digital filter. The transfer function H(z) of a single frequency notch filter is:

[0047]

[0048] In the above formula, z1 is the zero point of the filter, is the conjugate zero of the filter, p1 is the pole of the filter, is the conjugate pole of the filter, let:

[0049] but:

[0050]

[0051] In the above formula, w0=2πf0, f0 is the center frequency of the notch filter, T s is the sampling interval, r is the notch filter bandwidth parameter, the closer the r value is to 1, the more obvious the H(z) filtering effect is;

[0052] Step 3: The signal data obtained after processing in step 2 is denoised by wavelet threshold based on frequency domain to filter out other random noises, such as Figure 3 As shown in Figure 2, the process of wavelet threshold denoising is as follows:

[0053] Step 31, converting the noisy signal into the frequency domain through Fourier transform to obtain the real part and imaginary part of the Fourier transform of the signal;

[0054] Step 32, performing wavelet transform on the real part and the imaginary part respectively, and performing wavelet soft threshold denoising on the coefficients of each layer;

[0055] According to the Mallat algorithm of wavelet transform, the signal is decomposed to obtain the wavelet coefficients C at different scales. n，k , then C n，k Perform threshold processing and then reconstruct the denoised signal. The soft threshold algorithm is as follows:

[0056]

[0057] In the above formula, is the wavelet coefficient after taking the threshold, λ is the threshold, N is the signal length, σ is the noise variance, σ=MAD n / 0.6745, MAD n is the median value of the high subband coefficients of the nth layer after wavelet decomposition;

[0058] Step 33, combining the processed real and imaginary components into a new complex number sequence;

[0059] Step 34, converting the sequence into the time domain by inverse Fourier transform, so as to achieve the purpose of suppressing the magnetotelluric signal noise in the frequency domain by using the wavelet threshold method.

[0060] In summary, the method of the present invention uses a cascade approach, connecting different noise suppression methods in series, and sequentially processing the interference noise to obtain the desired signal data. The signal collected by the receiver is subjected to wavelet multiscale decomposition to suppress low-frequency motion noise below 16 Hz. The designed notch filter then removes power frequency interference. Finally, a designed frequency-domain wavelet soft threshold denoising algorithm is used to remove other random electromagnetic noise, resulting in the desired signal.

[0061] Example 1, as Figure 4 As shown, this embodiment discloses a method for suppressing noise in semi-aircraft frequency domain electromagnetic detection data based on wavelet de-basing and denoising and notch fusion, which processes the noisy signal measured on site, including the following steps:

[0062] Step 1: First, the measured signal containing noise ( Figure 6is the original measured signal time domain diagram) for spectrum analysis ( Figure 7 is the spectrum of the original measured signal, Figure 8 is the time-frequency diagram of the original measured signal), the measured signal is decomposed into multiple scales in the frequency domain using the wavelet transform method to suppress the low-frequency motion noise;

[0063] Step 2: The signal from which the low-frequency motion noise is filtered out in step 1 is filtered out through the notch filter designed by the present invention to remove the power frequency interference including 50 Hz and its harmonic components;

[0064] Step 3: The signal data obtained in step 2 is subjected to the frequency domain-based wavelet threshold denoising designed by the present invention to filter out other random noises and obtain the data we need.

[0065] Example 2, as Figure 5 As shown, this embodiment discloses a method for suppressing noise in semi-aircraft frequency domain electromagnetic detection data based on wavelet de-basing and denoising and notch fusion, which processes the noisy signal measured on site, including the following steps:

[0066] Step 1: First, the measured signal containing noise ( Figure 6 is the original measured signal time domain diagram) for spectrum analysis ( Figure 7 is the original measured signal spectrum, Figure 8 (This is the time-frequency diagram of the original measured signal). The power frequency interference including 50 Hz and its harmonic components is filtered out by the notch filter designed by the present invention.

[0067] Step 2: The signal filtered out of power frequency interference in step 1 is decomposed into multiple scales in the frequency domain using wavelet transform to suppress low-frequency motion noise;

[0068] Step 3: The signal data obtained in step 2 is subjected to the frequency domain-based wavelet threshold denoising designed by the present invention to filter out other random noises and obtain the data we need.

[0069] Figure 9 It is the time domain diagram of the signal after noise suppression; Figure 10 is the spectrum of the signal after noise suppression; Figure 11 It is the time-frequency diagram of the signal after noise suppression.

Claims

1. A method for suppressing noise in electromagnetic detection data in the semi-aeronautical frequency domain based on wavelet de-radicalization and notch fusion, characterized in that: The steps include: Step 1: Perform spectrum analysis on the measured signal containing noise; Step 2: Process the measured signal according to step 2a first, and then process the processed signal according to step 2b; or process the measured signal according to step 2b first, and then process the processed signal according to step 2a: Step 2a: Decompose the signal into multiple scales using wavelet transform in the frequency domain to suppress low-frequency motion noise. Select coif5 as the wavelet basis in the wavelet baseline removal algorithm to suppress low-frequency motion noise below 16 Hz by performing multi-scale decomposition on the signal in the frequency domain. Step 2b: The signal is filtered through a notch filter to remove the power frequency interference composed of 50 Hz and its harmonic components; The transfer function H(z) of a single frequency notch filter is: In the above formula, z1 is the zero point of the filter, is the conjugate zero of the filter, p1 is the pole of the filter, is the conjugate pole of the filter, let: but: In the above formula, w0=2πf0, f0 is the center frequency of the notch filter, T s is the sampling interval, r is the notch filter bandwidth parameter; Step 3: The signal data obtained after processing in step 2 is denoised by wavelet threshold based on the frequency domain to filter out other random noise. The process of wavelet threshold denoising is as follows: Step 31, converting the noisy signal into the frequency domain through Fourier transform to obtain the real part and imaginary part of the Fourier transform of the signal; Step 32, performing wavelet transform on the real part and the imaginary part respectively, and performing wavelet soft threshold denoising on the coefficients of each layer; According to the Mallat algorithm of wavelet transform, the signal is decomposed to obtain the wavelet coefficients C at different scales. n，k , then C n，k Perform threshold processing and then reconstruct the denoised signal. The soft threshold algorithm is as follows: In the above formula, is the wavelet coefficient after taking the threshold, λ is the threshold, N is the signal length, σ is the noise variance, σ=MAD n / 0.6745, MAD n is the median value of the high subband coefficients of the nth layer after wavelet decomposition; Step 33, combining the processed real and imaginary components into a new complex number sequence; In step 34, the sequence is converted to the time domain by inverse Fourier transform.

Images