A denoising method to improve the quality of electromagnetic exploration data in open-pit iron mines

Abstract

The present invention discloses a denoising method for improving the quality of electromagnetic exploration data for open-pit iron ore, comprising: segmenting magnetotelluric data into non-overlapping segments to form different data segments; utilizing multi-scale discrete entropy to divide the data segments into noisy data segments and valid data segments based on the data's clutter; utilizing a principal component analysis algorithm to separate valid signal components from noise components in the noisy data segments based on data variance differences, and utilizing a C-means clustering analysis algorithm to determine whether the separation is complete; after determining that the separation is complete, subtracting the feature space from the noisy data segments to obtain denoised data segments; and utilizing the valid data segments and the denoised data segments to reconstruct the denoised signal. The technical solution of the present invention avoids the problem of 'overprocessing' caused by loss of valid signals; can suppress long noise, even if the noise fills the entire data acquisition period; and can improve the data quality of data with low signal-to-noise ratio.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic exploration of open-pit iron mines, and in particular to a denoising method for improving the quality of electromagnetic exploration data of open-pit iron mines. Background Art

[0002] The disorderly mining of iron ore in the early stage has resulted in the existence of unknown underground goafs of varying sizes and uncertain depths in existing open-pit mines, posing a great threat to normal mining production and seriously restricting the mining and utilization of deep iron ore resources. Magnetotelluric sounding (MT) is commonly used and very suitable for mineral exploration and underground structure exploration. The field source of this method is not an artificial source, but a natural source, so MT has the characteristics of simple receiving equipment and rich frequency band information. However, during mine exploration, there are various noise sources around the surface or underground of the MT measuring point, such as communication towers or mining vehicles, which makes the quality of the collected MT data poor, resulting in the frequent occurrence of false underground anomalies and unreliable underground structures in the data interpretation results, greatly reducing the accuracy of goaf identification. Therefore, suppressing noise interference is crucial to improving the accuracy of mineral resource and goaf location and depth exploration.

[0003] In the aforementioned exploration areas, electromagnetic noise typically exhibits certain patterns, primarily categorized as strong noise and long-range noise. Long-range noise in iron ore mines primarily originates from geomagnetic pulsations. Geophysicists have developed a series of relatively mature theories and processes for dealing with strong noise, but research on methods for removing long-range noise is limited. Traditional methods, such as robust estimation and remote referencing, ignore the specific characteristics of long-range noise and typically perform noise removal on the entire data set. Robust estimation suppresses the impact of noise by downweighting deviations from the baseline (called "flying points"); remote referencing removes noise by replacing the local magnetic track data with the magnetic track data from a reference station. However, both methods impose high requirements on the signal-to-noise ratio (SNR). Robust estimation requires that no more than 30% of the data segment be affected by noise; remote referencing requires that the noise affecting the reference station be of a different origin than the noise affecting the local station (i.e., uncorrelated noise). When collecting MT data in mines, these areas are rife with strong interference sources, resulting in data quality failing to meet the requirements of these two methods, resulting in limited success in improving MT data. Other denoising methods, such as wavelet analysis and empirical mode decomposition, require some parameters to be preset. When the parameters are different, the improvement effect is different and the interference of human factors is greater.

[0004] In summary, existing technologies have the following main shortcomings: 1. They are primarily used to suppress strong noise and are less effective against long-duration noise; 2. They require a high signal-to-noise ratio (SNR) and cannot fully suppress noise when the observed data is of poor quality; 3. The denoising process can result in loss of effective signal. These shortcomings make existing methods difficult to apply to iron ore goaf exploration and unable to meet the country's growing demand for iron ore resources. Summary of the Invention

[0005] The present invention provides a denoising method for improving the quality of electromagnetic exploration data in open-pit iron ore, so as to solve the technical problems existing in the prior art, such as weak improvement effect on long noise, high requirement on data signal-to-noise ratio, and loss of effective signal during the denoising process.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In one aspect, the present invention provides a denoising method for improving the quality of electromagnetic exploration data of open-pit iron ore. The denoising method for improving the quality of electromagnetic exploration data of open-pit iron ore comprises:

[0008] S1, segmenting the magnetotelluric data into non-overlapping segments to form different data segments;

[0009] S2, using multi-scale discrete entropy, divide the data segment into noisy data segment and valid data segment according to the data clutter; wherein the valid data segment refers to the data segment that does not contain noise;

[0010] S3, using the principal component analysis algorithm, separate the effective signal component and the noise component in the noisy data segment according to the variance difference of the data, and calculate the eigenvector space;

[0011] S4, based on the C-means cluster analysis algorithm, calculating the distance D1 between the feature vector space and the cluster center of the noise component, and the distance D2 between the feature vector space and the cluster center of the effective signal component;

[0012] S5, determine whether D1 and D2 have overlapping parts. If D1 and D2 have overlapping parts, repeat S3 and S4 until D1 and D2 no longer have overlapping parts.

[0013] S6, subtracting the feature space from the noisy data segment to obtain a denoised data segment;

[0014] S7, reconstructing the denoised signal using the valid data segment and the denoised data segment.

[0015] Furthermore, the method of using multi-scale discrete entropy to divide the data segments into noisy data segments and valid data segments according to the clutter of the data includes:

[0016] Calculate the multi-scale discrete entropy of each data segment;

[0017] Compare the multi-scale discrete entropy of each data segment with the preset threshold;

[0018] The data segment whose multi-scale discrete entropy is lower than the preset threshold is regarded as the noisy data segment;

[0019] The data segments whose multi-scale discrete entropy is higher than the preset threshold are regarded as valid data segments.

[0020] Furthermore, the calculating of the multi-scale discrete entropy of each data segment includes:

[0021] Coarse-grained processing of data segments, dividing them into non-overlapping sequence segments, to obtain coarse-grained sequences:

[0022]

[0023] Where τ represents the multi-scaling factor, l is the length of each data segment set; represents the jth quantity in the coarse-grained sequence at the τ scale, x represents the amount of coarse-grained sequence data, Indicates the amount of raw data;

[0024] The normal cumulative distribution function is used to map the coarse-grained sequence to the interval from 0 to 1:

[0025]

[0026] Where rms and σ represent the root mean square and standard deviation of the coarse-grained series, respectively; represents the interval sequence after mapping, represents the jth quantity in the coarse-grained sequence at the τ scale, e represents the exponential function, and t represents the data length in each data segment;

[0027]

[0028] in, express The corresponding integer between 1 and c, Represents the interval sequence after mapping;

[0029] Introduce the embedding latitude m and delay parameter d, and reconstruct z j c for:

[0030] z i m,c ={z i c ,z i+d c …z i+(m-1)d c};

[0031] in, represents z j c The matrix formed after introducing m and d is, express The ud+1th data in the i-th row of the matrix, u=0,1,2,…,m-1;

[0032] Calculate each c m Potential dispersion patterns The relative probability is:

[0033]

[0034] Among them, c m Equivalent to Indicates distributed mode Assigned to The number after is divided by the total number of embedded signals with embedding dimension m. Number represents the number calculation function, and N represents the total number of embedded signals with embedding dimension m. hastype represents the discrete part corresponding function. Indicates a dispersed pattern;

[0035] Calculate the discrete entropy value at a single scale, that is:

[0036]

[0037] Among them, e(x,m,c,d) represents the discrete entropy value at a single scale, c m Equivalent to

[0038] Multi-scale discrete entropy is expressed as:

[0039] MDE i =E(x,τ,m,c,d)=[e1,e2…,e τ ];

[0040] Among them, MDE i Indicates data segment X i The multi-scale discrete entropy, e τ Represents the discrete entropy value at the τ scale.

[0041] Furthermore, the principal component analysis algorithm is used to separate the effective signal component and the noise component in the noisy data segment according to the variance difference of the data, and the characteristic vector space is calculated, including:

[0042] Write the one-dimensional noisy data segment into a square matrix form:

[0043]

[0044] Among them, a is an integer, MX q is the noisy data segment X q The corresponding square matrix form, x(a,a) represents MX q Data in

[0045] Decentralize the square matrix, that is, zero-mean, and obtain the zero-mean matrix:

[0046] Δ g =MX q -Ψ;

[0047] in, is the mean value of the noisy data segment Xq; Δ g represents the zero-mean matrix; x i Represents the noisy data segment X q The i-th magnetotelluric data in;

[0048] Calculate the covariance matrix C of the zero-mean matrix:

[0049]

[0050] Where T represents matrix transpose;

[0051] Perform eigendecomposition on the covariance matrix C and obtain the principal components:

[0052] C=eλe T ;

[0053] in, is the eigenvector matrix of C, 1≤s≤l2, is the eigenvalue matrix, whose main diagonal is arranged in descending order, e v represents the vth eigenvector in the eigenvector matrix e, λ v represents the vth eigenvalue in the eigenvalue matrix λ, v = 1, 2, ..., S;

[0054] Calculate the eigenvector space v(p,q)=[v1 v2 … v S ],in, MX q A square matrix representing a noisy data segment.

[0055] Furthermore, the distance D1 between the feature vector space and the cluster center of the noise component, and the distance D2 between the feature vector space and the cluster center of the effective signal component are calculated based on the C-means cluster analysis algorithm, including:

[0056] The number of cluster centers is preset to 2, and the mapping value between the eigenvector space v(p,q) and the two cluster centers is calculated to obtain the distance between the eigenvector space and the two cluster centers:

[0057]

[0058] Where D1 represents the distance between the feature vector space v(p,q) and the cluster center of the noise component, D2 represents the distance between the feature vector space v(p,q) and the cluster center of the effective signal component, || ||2 represents the norm function, w1 represents the cluster center of the noise component, and w2 represents the cluster center of the effective signal component.

[0059] On the other hand, the present invention further provides an electronic device, comprising a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the above method.

[0060] In yet another aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, and the instruction is loaded and executed by a processor to implement the above method.

[0061] The technical solution provided by the present invention can reduce the loss of effective signals, suppress the long-range noise in iron ore goaf, and improve the signal-to-noise ratio of data without the need for human intervention. Its beneficial effects include at least:

[0062] 1. The technical solution of the present invention utilizes MDE (Multiscale Dispersion Entropy) and Cmeans (C-means cluster analysis) to reduce the loss of valid signals before and during denoising. Since the valid signal and noise have different properties, MDE can retain the valid signal based on these differences, while only processing the noisy data in the subsequent denoising process. Cmeans serves as an evaluation criterion in the denoising process, avoiding the "over-processing" problem caused by PCA (Principal Component Analysis) when incompletely separating the valid signal and noise components in noisy data.

[0063] 2. When signals have variance differences, the technical solution provided by this invention can separate different signals using PCA. MT's effective signal and noise (e.g., rectangular, triangular, or pulse waves) have significant differences in shape and variance. Therefore, even if the noise persists for a long time, PCA can separate them as long as the variance difference exists.

[0064] 3. The presence of long noise will inevitably lead to poor data quality and low signal-to-noise ratio. The technical solution provided by the present invention can suppress long noise while reducing the loss of effective signals, thereby improving data quality and signal-to-noise ratio.

[0065] 4. The technical solution provided by the present invention will process data adaptively according to data characteristics, without the need for manually preset method parameters, thereby reducing interference from human factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0067] Figure 1 1 is a schematic diagram of an execution flow of a denoising method for improving the quality of electromagnetic exploration data for open-pit iron ore provided by an embodiment of the present invention;

[0068] Figure 2 This is a detailed execution flow chart of a denoising method for improving the quality of electromagnetic exploration data for open-pit iron ore provided by an embodiment of the present invention;

[0069] Figure 3 This is a comparison chart of the denoising effects of the present invention and the Robust method. DETAILED DESCRIPTION

[0070] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0071] First embodiment

[0072] In the case of low signal-to-noise ratio in iron ore goaf exploration data, which is affected by long-range noise and cannot accurately detect underground structures, this embodiment provides a denoising method for improving the quality of electromagnetic exploration data in open-pit iron ore mines. This method can be applied to magnetotelluric sounding, to open-pit iron ore goaf exploration, and to improving the reliability of electromagnetic methods for detecting mineral resources and underground structures, thus having significant application value.

[0073] First, it's important to note that long-duration noise, such as geomagnetic fluctuations, has a unique impact on signals: strong fluctuations have a significant impact on the data, while weak fluctuations have a smaller impact. Therefore, effectively identifying and filtering the collected data, classifying it, and removing noise from the areas strongly affected by geomagnetic fluctuations while retaining the areas less affected by the geomagnetic signal, and then reconstructing the signal, will significantly improve the signal-to-noise ratio of the valid signal. Based on this, this method utilizes multiscale dispersion entropy (MDE), principal component analysis (PCA), and C-means cluster analysis (Cmeans) to suppress noise in high-noise data, thereby improving the signal-to-noise ratio. In areas of strong interference, most data will be corrupted by noise, making it even more important to prevent the loss of the few valid signals unaffected by noise. MDE is used to separate noisy and valid signals. The properties of the effective signal and noise differ significantly. For example, the effective signal exhibits significantly higher levels of chaos and polarization than the noise (i.e., the effective signal exhibits a higher degree of chaos and lacks a distinct, fixed polarization direction). MDE distinguishes the effective signal from the noise based on the data's level of noise (the MDE entropy of the noise is significantly lower than that of the effective signal), thus avoiding the problem of over-processing the data. The subsequent denoising process only processes the noisy data, using PCA-Cmeans to suppress the noise. The MT effective signal and noise exhibit distinct morphological differences (noise typically takes the form of rectangular waves, triangular waves, or strong pulses). In contrast, the effective signal exhibits a small-scale waveform. Therefore, PCA is able to separate the effective signal from the noise in noisy data. Furthermore, because PCA separates the two signals based on differences in their variance, it can effectively handle long-duration noise, even if the noise permeates the entire observation period. Cmeans is used to verify that PCA has completely separated the effective signal from the noise. When the mean center values ​​of Cmeans differ greatly, it proves that they are completely separated, further reducing the loss of effective signals.

[0074] Specifically, the execution process of this method is as follows Figure 1 As shown, the following steps are included:

[0075] S1, segmenting the magnetotelluric data into non-overlapping segments to form different data segments;

[0076] Specifically, this embodiment converts the magnetotelluric data x1, x2, ... x n (x i represents the i-th magnetotelluric data, i = 1, 2…, n, n is the data length of the magnetotelluric data) and is segmented into non-overlapping segments to form different data segments X i ={x(i-1)*l+1 ,…,x i*l}; where X i Represents the i-th data segment, i = 1, 2, ..., m, m = [n / l] is the number of data segments, l is the set length of each data segment, and [n / l] represents the maximum integer not exceeding n / l.

[0077] S2, using multi-scale discrete entropy, divide the data segment into noisy data segment and valid data segment according to the data clutter; wherein the valid data segment refers to the data segment that does not contain noise;

[0078] Specifically, in this embodiment, the implementation process of the above S2 is: calculate each data segment X i Multiscale discrete entropy MDE i ; Compare the multi-scale discrete entropy of each data segment with the preset threshold M0 (M0 is set to 1); the data segment with multi-scale discrete entropy lower than M0 is regarded as the noisy data segment X q , denoising is required; the data segment with multi-scale discrete entropy higher than M0 is regarded as the valid data segment X p .

[0079] Furthermore, the calculating of the multi-scale discrete entropy of each data segment includes:

[0080] S21, coarse-grained processing of the data segment, dividing it into non-overlapping sequence segments, and obtaining a coarse-grained sequence:

[0081]

[0082] Where τ represents the multi-scaling factor, l is the length of each data segment set; represents the jth quantity in the coarse-grained sequence at the τ scale, x represents the amount of coarse-grained sequence data, Indicates the amount of raw data;

[0083] S22, use the normal cumulative distribution function (NCDF) to map the above coarse-grained sequence to the interval from 0 to 1:

[0084]

[0085] Where rms and σ represent the root mean square and standard deviation of the coarse-grained series, respectively; represents the interval sequence after mapping, represents the jth quantity in the coarse-grained sequence at the τ scale, e represents the exponential function, and t represents the data length in each data segment;

[0086]

[0087] in, express The corresponding integer between 1 and c, Represents the interval sequence after mapping;

[0088] S24, introduce the embedding latitude m and delay parameter d, reconstruct z j c for:

[0089] z i m,c ={z i c ,z i+d c …z i+(m-1)d c};

[0090] in, represents z j c After introducing m and d, the matrix is ​​formed. m, d and c are all arbitrarily settable values. express The ud+1th data in the i-th row of the matrix, u=0,1,2,…,m-1;

[0091] S25, calculate each c m Potential dispersion patterns The relative probability is:

[0092]

[0093] Among them, c m Equivalent to Indicates distributed mode Assigned to The number after is divided by the total number of embedded signals with embedding dimension m. Number represents the number calculation function, and N represents the total number of embedded signals with embedding dimension m. hastype represents the discrete part corresponding function. Indicates a dispersed pattern;

[0094] S26, calculate the discrete entropy value at a single scale, that is:

[0095]

[0096] Among them, e(x,m,c,d) represents the discrete entropy value at a single scale, c m Equivalent to

[0097] S27, the multi-scale discrete entropy is expressed as:

[0098] MDEi=E(x,τ,m,c,d)=[e1,e2…,eτ];

[0099] Among them, MDE i Indicates data segment X i The multi-scale discrete entropy, e τ Represents the discrete entropy value at the τ scale.

[0100] S3, using the principal component analysis algorithm, separate the effective signal component and the noise component in the noisy data segment according to the variance difference of the data, and calculate the eigenvector space;

[0101] Specifically, in this embodiment, the implementation process of the above S3 is as follows:

[0102] S31, the one-dimensional noisy data segment X q Written in square matrix form:

[0103]

[0104] Among them, a is an integer, MX q is the noisy data segment X q The corresponding square matrix form, x(a,a) represents MX q Data in

[0105] S32, decentralize the matrix, that is, zero-mean, and obtain the zero-mean matrix:

[0106] Δ g =MX q -Ψ;

[0107] in, is the mean value of the noisy data segment Xq; Δ g represents the zero-mean matrix; x i Represents the noisy data segment X q The i-th magnetotelluric data in;

[0108] S33, calculate the covariance matrix C of the zero-mean matrix:

[0109]

[0110] Where T represents matrix transpose;

[0111] S34, perform eigendecomposition on the covariance matrix C and obtain the principal components:

[0112] C=eλe T ;

[0113] in, is the eigenvector matrix of C, 1≤s≤l 2 , is the eigenvalue matrix, whose main diagonal is arranged in descending order, e v represents the vth eigenvector in the eigenvector matrix e, λ v represents the vth eigenvalue in the eigenvalue matrix λ, v = 1, 2, ..., S;

[0114] S35, calculate the eigenvector space v(p,q) = [v1 v2 … v S ],in, MX q A square matrix representing a noisy data segment.

[0115] S4, based on the C-means cluster analysis algorithm, calculating the distance D1 between the feature vector space and the cluster center of the noise component, and the distance D2 between the feature vector space and the cluster center of the effective signal component;

[0116] Specifically, in this embodiment, the implementation process of the above S4 is as follows:

[0117] The number of cluster centers o is preset (o=2), and the mapping value between the feature vector space v(p,q) and the two cluster centers is calculated to obtain the distance between the feature vector space and the two cluster centers:

[0118]

[0119] Where D1 represents the distance between the feature vector space v(p,q) and the cluster center of the noise component, D2 represents the distance between the feature vector space v(p,q) and the cluster center of the effective signal component, || ||2 represents the norm function, w1 represents the cluster center of the noise component, and w2 represents the cluster center of the effective signal component.

[0120] S5, determine whether D1 and D2 have overlapping parts. If D1 and D2 have overlapping parts, repeat S3 and S4 until D1 and D2 no longer have overlapping parts.

[0121] S6, subtract the feature space from the noisy data segment to obtain the denoised data segment; expressed as:

[0122]

[0123] in, represents the denoised data segment, and v(p,q) represents the feature vector space.

[0124] S7, reconstructing the denoised signal using the valid data segment obtained in S2 and the denoised data segment obtained in S6.

[0125] Based on the above, the specific implementation process of the method of this embodiment is as follows Figure 2As shown in the figure, x is the original electromagnetic data, ∩ represents the intersection symbol (meaning whether D1 and D2 intersect), y is the denoised signal, and || represents the reconstructed signal by connecting the data segments end to end in order. Figure 3 As shown. Among them, ρ xy represents the apparent resistivity response curve in TE mode; ρ yx represents the apparent resistivity-phase response curve in TM mode, represents the phase response curve in TE mode, The figure shows the phase response curve in TM mode. The triangles represent the apparent resistivity-phase response curve for the original measured data at station M1-90; the circles represent the apparent resistivity-phase response curve for station M1-90 improved by our method; and the squares represent the apparent resistivity-phase response curve for station M1-90 improved by the robust method denoising technique. A comparison shows that our method achieves the best improvement, making the curve more continuous and smooth, while the robust method still exhibits a few "jump points" in the results.

[0126] In summary, this embodiment provides a denoising method for improving the quality of electromagnetic exploration data from open-pit iron ore mines. This method utilizes MDE and Cmeans before and during denoising to reduce effective signal loss, minimizing the problem of over-processing. PCA is also used to suppress noise, particularly long-range noise, which has received little attention. PCA can also effectively process this noise. Therefore, this method requires a low signal-to-noise ratio and can improve data quality. By combining MDE, PCA, and Cmeans to suppress long-range noise in electromagnetic sounding data, the exploration results of iron ore goaf areas are more reliable.

[0127] Second embodiment

[0128] This embodiment provides an electronic device, which includes a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method of the first embodiment.

[0129] The electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) and one or more memories, wherein the memory stores at least one instruction, which is loaded by the processor to execute the above method.

[0130] Third embodiment

[0131] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method of the first embodiment described above. The computer-readable storage medium may be a ROM, random access memory, CD-ROM, magnetic tape, floppy disk, or optical data storage device. The instructions stored therein can be loaded by a processor in a terminal to execute the method described above.

[0132] Furthermore, it should be noted that the present invention may be provided as a method, apparatus, or computer program product. Thus, embodiments of the present invention may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code.

[0133] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0135] It should also be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or terminal device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or terminal device comprising the element.

[0136] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be noted that although the preferred embodiment of the present invention has been described, it is clear that those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered as within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A denoising method for improving the quality of electromagnetic exploration data in open pit iron ore, characterized in that: include: S1, segmenting the magnetotelluric data into non-overlapping segments to form different data segments; S2, using multi-scale discrete entropy, divide the data segment into noisy data segment and valid data segment according to the data clutter; wherein the valid data segment refers to the data segment that does not contain noise; S3, using the principal component analysis algorithm, separate the effective signal component and the noise component in the noisy data segment according to the variance difference of the data, and calculate the eigenvector space; S4, based on the C-means cluster analysis algorithm, calculating the distance D1 between the feature vector space and the cluster center of the noise component, and the distance D2 between the feature vector space and the cluster center of the effective signal component; S5, determine whether D1 and D2 have overlapping parts. If D1 and D2 have overlapping parts, repeat S3 and S4 until D1 and D2 no longer have overlapping parts. S6, subtracting the feature space from the noisy data segment to obtain a denoised data segment; S7, reconstructing the denoised signal using the valid data segment and the denoised data segment; Using the principal component analysis algorithm, the effective signal component and the noise component in the noisy data segment are separated according to the variance difference of the data, and the eigenvector space is calculated, including: Write the one-dimensional noisy data segment into a square matrix form: Among them, a is an integer, MX q is the noisy data segment X q The corresponding square matrix form, x(a,a) represents MX q The data in; l is the length of each data segment set; Decentralize the square matrix, that is, zero-mean, and obtain the zero-mean matrix: D g =MX q -Ψ; in, is the mean value of the noisy data segment Xq; Δ g represents the zero-mean matrix; x i Represents the noisy data segment X q The i-th magnetotelluric data in; Calculate the covariance matrix C of the zero-mean matrix: Where T represents matrix transpose; Perform eigendecomposition on the covariance matrix C and obtain the principal components: C=eλe T ; in, is the eigenvector matrix of C, 1≤s≤l 2 , is the eigenvalue matrix, whose main diagonal is arranged in descending order, e v represents the vth eigenvector in the eigenvector matrix e, λ v represents the vth eigenvalue in the eigenvalue matrix λ, v = 1, 2, ..., S; Calculate the eigenvector space v(p,q)=[v1 v2 … v S ],in, MX q A square matrix representing a noisy data segment.

2. The denoising method for improving the quality of electromagnetic exploration data of open-pit iron ore according to claim 1, characterized in that: The method of using multi-scale discrete entropy to divide data segments into noisy data segments and valid data segments according to the clutter of the data includes: Calculate the multi-scale discrete entropy of each data segment; Compare the multi-scale discrete entropy of each data segment with the preset threshold; The data segment whose multi-scale discrete entropy is lower than the preset threshold is regarded as the noisy data segment; The data segments whose multi-scale discrete entropy is higher than the preset threshold are regarded as valid data segments.

3. The denoising method for improving the quality of electromagnetic exploration data of open-pit iron ore according to claim 2, characterized in that: The calculating of the multi-scale discrete entropy of each data segment includes: Coarse-grained processing of data segments, dividing them into non-overlapping sequence segments, to obtain coarse-grained sequences: Where τ represents the multi-scaling factor, l is the length of each data segment set; represents the jth quantity in the coarse-grained sequence at the τ scale, x represents the amount of coarse-grained sequence data, Indicates the amount of raw data; The normal cumulative distribution function is used to map the coarse-grained sequence to the interval from 0 to 1: Where rms and σ represent the root mean square and standard deviation of the coarse-grained series, respectively; represents the interval sequence after mapping, represents the jth quantity in the coarse-grained sequence at the τ scale, e represents the exponential function, and t represents the data length in each data segment; The introduction of linear algorithms will linearly assigned to integer z between 1 and c j c ,Right now: in, express The corresponding integer between 1 and c, Represents the interval sequence after mapping; Introduce the embedding latitude m and delay parameter d, and reconstruct z j c for: With i m,c ={z i c ,With i+d c …With i+(m-1)d c }; in, represents z j c The matrix formed after introducing m and d is, express The ud+1th data in the i-th row of the matrix, u=0,1,2,…,m-1; Calculate each c m Potential dispersion patterns The relative probability is: Among them, c m Equivalent to Indicates distributed mode Assigned to The number after is divided by the total number of embedded signals with embedding dimension m, Number represents the quantity calculation function, N represents the total number of embedded signals with embedding dimension m; hastype represents the discrete part corresponding function, Indicates a dispersed pattern; Calculate the discrete entropy value at a single scale, that is: Among them, e(x,m,c,d) represents the discrete entropy value at a single scale, c m Equivalent to Multi-scale discrete entropy is expressed as: MDE i =E(x,τ,m,c,d)=[e1,e2…,e τ ]: Among them, MDE i Indicates data segment X i The multi-scale discrete entropy, e τ Represents the discrete entropy value at the τ scale.

4. The denoising method for improving the quality of electromagnetic exploration data of open-pit iron ore according to claim 1, characterized in that: The C-means cluster analysis algorithm is used to calculate the distance D1 between the feature vector space and the cluster center of the noise component, and the distance D2 between the feature vector space and the cluster center of the effective signal component, including: The number of cluster centers is preset to 2, and the mapping value between the eigenvector space v(p,q) and the two cluster centers is calculated to obtain the distance between the eigenvector space and the two cluster centers: Where D1 represents the distance between the feature vector space v(p,q) and the cluster center of the noise component, D2 represents the distance between the feature vector space v(p,q) and the cluster center of the effective signal component, || ||2 represents the norm function, w1 represents the cluster center of the noise component, and w2 represents the cluster center of the effective signal component.

Images