Coal and rock small fault feature extraction method based on wavelet transform and principal component analysis fusion

Abstract

The application discloses a y rock small fault feature extraction method based on wavelet transform and principal component analysis, and belongs to the technical field of seismic signal processing. The method comprises the following steps: pre-processing obtained seismic data, performing multi-scale decomposition on the seismic signal by using discrete wavelet transform; calculating the energy proportion of each decomposition level, selecting the level with the highest energy proportion and the approximate component to perform signal reconstruction, so as to enhance the effective signal; extracting various seismic attributes of the reconstructed seismic data and constructing an attribute set; and performing dimension reduction and fusion on the attribute set by using principal component analysis, so as to obtain a principal component feature vector. The application can effectively improve the signal-to-noise ratio of the seismic signal, eliminate attribute redundancy, significantly enhance the small fault response feature and improve the recognition accuracy.

Description

Technical Field

[0001] This invention relates to a seismic signal processing technique, and more specifically, to a method for extracting features of small coal and rock faults by fusing wavelet transform and principal component analysis. Background Technology

[0002] In coal mining, precise detection of micro-faults is crucial for preventing gas outbursts, roof collapses, and ensuring safe production. Currently, 3D seismic technology is the primary means of identifying underground geological structures, with wavelet transform for seismic signal processing and analysis becoming an important technique. However, in wavelet decomposition and reconstruction, the selection of effective signal levels often relies on manual experience or qualitative observation of the spectrum, lacking objective quantitative standards. This subjective selection method easily leads to insufficient extraction of effective fault information and incomplete noise removal, making it difficult to adapt to complex and variable geological conditions. Analyzing fault-sensitive seismic attributes is another important method for fault interpretation. A single seismic attribute can reflect the discontinuity of strata to a certain extent, but when facing complex micro-faults, a single attribute often has multiple interpretations and cannot comprehensively characterize fault features. To improve identification accuracy, researchers have begun to try to comprehensively utilize multiple seismic attributes, but simple attribute superposition has failed to effectively solve the problems of information redundancy and coherence interference between multi-dimensional attributes. Therefore, how to eliminate the correlation between attributes and achieve data dimensionality reduction while retaining effective fault information is key to achieving precise identification of small faults under complex geological conditions in coal mines. Summary of the Invention

[0003] One of the objectives of this invention is to address the aforementioned shortcomings by providing a method for extracting features of small coal and rock faults by fusing wavelet transform and principal component analysis. This method aims to solve the technical problems in existing technologies, such as the strong subjectivity caused by relying on manual experience in selecting wavelet decomposition levels for small fault identification, and the defects of information redundancy and dimensionality curse in the joint interpretation of multiple seismic attributes.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0005] Step A: Obtain the original 3D seismic data volume of the coal mining area.

[0006] Step B: Perform multi-scale decomposition of the acquired seismic signal using discrete wavelet transform to obtain approximate components (low-frequency components) and detail components (high-frequency components) at different levels. The formula for the aforementioned wavelet transform is:

[0007]

[0008]

[0009] In the formula, These are approximate coefficients. For detail coefficients; For scale parameters, These are translation parameters; This is an earthquake signal; It is a low-pass filter. It is a high-pass filter.

[0010] Step C: Calculate and compare the energy proportions of the detailed components at each decomposition level, and select the detailed component corresponding to the decomposition level with the highest energy proportion as the effective signal containing the seismic excitation frequency.

[0011] Step D: Superimpose the effective level detail components selected in Step C with the corresponding approximate components of the same level to form a frequency domain seismic characteristic signal. Then, generate high signal-to-noise ratio time domain seismic data through inverse wavelet transform. The inverse wavelet transform formula is:

[0012]

[0013] In the formula, This is the reconstructed high signal-to-noise ratio seismic signal; The maximum scale for wavelet decomposition. For the first Approximation coefficients of the layer; The effective level with the highest energy percentage selected in step C. For the first The detail factor of the layer; To reconstruct the low-pass filter, To reconstruct the high-pass filter.

[0014] Step E: Based on the obtained reconstructed seismic data, extract various seismic attributes that are sensitive to faults to form an attribute sample matrix.

[0015] Step F: Standardize the attribute sample matrix, calculate the covariance matrix based on the standardized data and perform eigenvalue decomposition. Calculate the principal component contribution rate according to the magnitude of the eigenvalues ​​corresponding to each principal component, and select the principal components whose cumulative contribution rate reaches a preset threshold. The standardization process and characteristic equation formula are as follows:

[0016]

[0017] In the formula, For standardized attribute values, The first element in the original attribute matrix The first attribute Each sample value, For the first The mean of each attribute, For the first The standard deviation of each attribute.

[0018]

[0019] In the formula: Let covariance matrix be the variance matrix. For eigenvalues, These are the eigenvectors.

[0020] Step G: Use the eigenvectors corresponding to the principal components to perform a weighted summation of the standardized attributes to obtain the comprehensive characteristic attributes of the small coal and rock faults. Cumulative contribution rate of each principal component :

[0021]

[0022] In the formula, For cumulative contribution rate, For the first The eigenvalues ​​of the principal components The number of eigenvalues.

[0023] A further technical solution is that, in step A above, the target data volume is seismic data extracted within a time window containing the target coal seam, thus avoiding interference from non-target layer data on the response of small faults and reducing redundant data.

[0024] A further technical solution is that the basis function selected for discrete wavelet transform in step B above is a mother wavelet suitable for the characteristics of seismic signals, and the maximum number of decomposition layers of discrete wavelet transform is determined by the signal length of the seismic data.

[0025] A further technical solution is that the energy of the detailed components of the decomposed layer in step C above is obtained by squaring and summing the wavelet coefficients corresponding to the decomposed layer, which is used to characterize the magnitude of the seismic signal energy contained in the decomposed layer within the corresponding frequency band.

[0026] A further technical solution is that, in step C above, the energy ratio is the ratio between the energy of a detail component at a certain decomposition level and the total energy of all detail components at all decomposition levels. By comparing the energy ratios corresponding to each decomposition level, the degree of contribution of different decomposition levels to the energy of the original seismic signal can be measured.

[0027] A further technical solution is that the high signal-to-noise ratio time-domain seismic data in step D above is reconstructed from the frequency-domain seismic characteristic signal through inverse wavelet transform. According to the reconstruction formula of inverse wavelet transform, new detail components and corresponding new approximate components of the reconstructed seismic data are selected respectively, and the new detail components and new approximate components are combined and reconstructed to obtain the result.

[0028] A further technical solution is that, before using principal component analysis to reduce the dimensionality of the high-dimensional initial attribute set in step F above, the attribute data needs to be standardized to obtain a standardized attribute matrix. .

[0029] A further technical solution is that, in step G above, the contribution rate of each principal component is calculated based on the magnitude of its corresponding eigenvalue, and the principal components whose cumulative contribution rate reaches a preset threshold are selected. This involves sorting the principal components according to the magnitude of their corresponding eigenvalues ​​and selecting the top-ranked ones. Each principal component, before calculation Cumulative contribution rate of each principal component .

[0030] Compared with the prior art, the beneficial effects of the present invention may be one of the following:

[0031] By introducing a wavelet decomposition energy-proportion-based hierarchical selection mechanism, the problem of relying on human experience or fixed frequency bands to select decomposition levels in traditional methods is avoided, thus achieving adaptive selection of the effective frequency band of the seismic signal. By combining the decomposition level with the highest energy proportion with its corresponding approximate component at the same level and reconstructing it using inverse wavelets, not only is noise in the original seismic data suppressed, but the response characteristics of small faults are also effectively enhanced.

[0032] Principal component analysis (PCA) was used to perform dimensionality reduction and fusion of multiple fault-sensitive attributes, effectively eliminating redundant correlations between attributes and reducing feature dimensionality while preserving key fault information. Seismic data processed by wavelet transform was then combined with PCA.

[0033] By combining wavelet transform with principal component analysis, wavelet transform is used to effectively filter and reconstruct seismic signals, suppressing noise and enhancing the response of small faults. Then, multiple attributes are extracted from the reconstructed seismic data and dimensionality reduction and fusion are performed using principal component analysis. This allows principal component analysis to be performed on high signal-to-noise and low-redundancy data, thus avoiding the instability caused by directly reducing the dimensionality of the original high-noise and multi-correlated attribute data. This achieves a synergistic gain effect between wavelet transform and principal component analysis. Attached Figure Description

[0034] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0035] Figure 2 This is the original seismic profile before wavelet transform in one embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the multi-scale decomposition result of discrete wavelet transform in one embodiment of the present invention.

[0037] Figure 4This is a comparison of seismic profiles before and after discrete wavelet transform in one embodiment of the present invention.

[0038] Figure 5 This is a multi-attribute response map extracted from wavelet-reconstructed seismic data volume in one embodiment of the present invention.

[0039] Figure 6 This is a schematic diagram comparing the original attributes, wavelet-enhanced attributes, and attribute feature responses after the fusion of wavelet transform and principal component analysis in one embodiment of the present invention. Detailed Implementation

[0040] The invention will now be further described with reference to the accompanying drawings.

[0041] One embodiment of the present invention is a method for extracting small fault features in coal and rock by fusing wavelet transform and principal component analysis. In this embodiment, reference is made to... Figure 1 As shown, this method is performed according to the following steps:

[0042] S1: Select 3D seismic data of the coal mining area as the original input data.

[0043] In this step, the 3D seismic data volume is acquired through field seismic exploration and processed using conventional seismic data processing procedures. The 3D seismic data volume is imported into a seismic interpretation software platform. Based on the burial depth range of the target coal seam, a corresponding time window interval is determined. Time window data containing the target coal seam is extracted from the 3D seismic data volume. This method not only ensures a sufficient training dataset but also reduces interference from non-target area data on the extraction of small fault features. Within the time window, 2D seismic profiles are extracted along the Inline or Xline direction as the analysis object, such as... Figure 2 As shown, this is the basic data used for subsequent wavelet transform processing to analyze the fault characteristic response.

[0044] S2: Multi-scale decomposition of seismic signals is performed using discrete wavelet transform.

[0045] In this step, in order to improve Figure 2 The original seismic profile shown represents the response characteristics of small faults. Discrete wavelet transform is used to perform multi-scale decomposition processing on the acquired seismic signal, mapping the original time-domain signal to a multi-frequency band space to achieve the separation and expression of features at different scales.

[0046] Seismic signals in the time domain are typically composed of a superposition of multiple frequency components. Tectonic anomalies related to small faults mainly manifest as amplitude abrupt changes, phase discontinuities, and phase axis discontinuities within specific frequency bands. Based on this, a suitable mother wavelet function is selected, and before implementing discrete wavelet transform, the number of decomposition layers is reasonably determined according to the sampling frequency and signal length of the seismic data. In order to achieve a balance between time resolution and frequency resolution.

[0047] According to the Nyquist sampling theorem, the effective frequency range of a seismic signal is: ,in The sampling frequency is denoted as . The discrete wavelet transform performs a bisection of the current frequency band at each decomposition level: the low-frequency component is used to enter the next decomposition level, while the high-frequency component is preserved as the detail component of that level. Therefore, the ... The physical frequency range corresponding to the layer detail components is The final approximation component corresponds to the lowest frequency band. This results in a multi-scale hierarchical structure in the frequency domain, from high to low.

[0048] Based on this, discrete wavelet decomposition is performed on the seismic data volume, such as... Figure 3 The diagram shows the decomposition of the original signal into approximate components (low-frequency components) and detail components (high-frequency components) at multiple scale levels. As shown in the following formula, the discrete wavelet transform essentially decomposes the original seismic signal at different scale levels using a set of variable-size low-pass and high-pass filters. During the decomposition process, the scale parameter... As the frequency band of the approximate component increases, the smoothness of the approximate component gradually increases, while the frequency band of the detail component gradually shifts to lower frequencies, so that the different levels of detail components form a hierarchical structure from high to low in the frequency domain.

[0049]

[0050]

[0051] In the above formula, These are approximate coefficients. For detail coefficients; For scale parameters, These are translation parameters; This is an earthquake signal; It is a low-pass filter. It is a high-pass filter.

[0052] S3: Calculate the energy ratio of the detailed components of each decomposition layer, and select the decomposition layer with the highest energy ratio as the effective layer.

[0053] In this step, compared to traditional Fourier spectral analysis which only provides the global frequency domain energy distribution, wavelet domain energy modeling has the advantage of joint time-frequency localization, revealing the dynamic evolution of signal energy in each frequency band within a specific time period. Since the discrete wavelet transform maintains the energy conservation law under orthogonal conditions, according to Parseval's theorem, the total energy of the original seismic signal in the time domain can be equivalently expressed as the sum of the energies of the detail coefficients and approximation coefficients at each scale. Therefore, for the ... Detail coefficient sequence obtained from layer decomposition The energy of this subband is defined as:

[0054]

[0055] Correspondingly, approximation coefficient Energy is defined as:

[0056]

[0057] This constitutes the completed energy vector. To eliminate the impact of amplitude scale differences between different data segments on energy statistics, a normalized energy proportion index is introduced to convert the energy of each layer into a relative proportion:

[0058]

[0059]

[0060] In the formula, Indicates the first The proportion of layer detail components in the total energy across the entire frequency band.

[0061] Based on this, the energy proportion of the detailed components of each decomposition layer is... The decomposition layers with the highest energy percentage are selected through ranking and comparison. This serves as the effective scale level for the target seismic signal. The physical frequency band corresponding to this level is determined by the Nyquist frequency division relationship in step S2. The frequency components contained within it dominate in terms of energy contribution, indicating that the seismic response within this frequency band most concentratedly reflects fault-related tectonic anomaly information.

[0062] S4: Combine the effective level detail components with the corresponding approximate components of the same level and reconstruct them through inverse wavelet transform to generate high signal-to-noise ratio time-domain seismic data.

[0063] In S3 above, by quantitatively analyzing the energy proportion of the detail components in each decomposition layer, the target scale layer with the largest energy contribution is selected. As a valid level, and the valid level Corresponding detail coefficient sequence With approximation coefficient The coefficients are then combined. Subsequently, inverse discrete wavelet transform (IDWT) is performed to map the filtered multi-scale coefficients back to the time domain, thereby obtaining the reconstructed seismic signal. Its expression is:

[0064]

[0065] In the formula, For effective level detail coefficients, These are the approximation coefficients for the corresponding level.

[0066] Figure 4 The comparison between the original and reconstructed seismic profiles is shown, where (a) is the original seismic profile, (b) is the reconstructed seismic profile, (c) is the normalized amplitude attribute of the coal seam in the original seismic profile, and (d) is the normalized amplitude attribute of the coal seam in the reconstructed seismic profile. From (a) and (b), it can be seen that the in-phase axis misalignment characteristics at the fault location are clearer in the reconstructed profile. From (c) and (d), it can be seen that after wavelet transform, the amplitude value at the fault location is significantly smaller than that at the non-fault location, significantly improving the response characteristics at the fault location.

[0067] S5: Extract various seismic attributes sensitive to small coal and rock faults based on reconstructed seismic data, and construct an attribute sample matrix.

[0068] In this step, after completing the inverse wavelet reconstruction, a high signal-to-noise ratio time-domain seismic data volume is obtained. To further explore the small fault structure information contained in the reconstructed seismic data, various seismic attributes that are sensitive to small coal and rock faults are extracted to construct the attribute sample matrix required for subsequent feature fusion and intelligent identification.

[0069] Seismic attributes are extracted from seismic signals through a series of mathematical calculations, revealing geometric, dynamic, and statistical characteristics. The fusion of seismic attributes with three-dimensional data volumes enhances the accuracy of seismic data interpretation. This invention selects multidimensional seismic attribute parameters, specifically including but not limited to:

[0070] Optionally, the variance volume attribute is used to characterize the variation in similarity between adjacent seismic traces and is highly sensitive to lateral discontinuities. The variance volume attribute can be calculated using the following formula:

[0071]

[0072] In the formula, The variance value. It is a triangular weighted function. This represents the amplitude value of the seismic trace. The average amplitude within the time window, The number of adjacent channels involved in the calculation.

[0073] Optionally, chaotic volume properties identify formation discontinuities by the degree of disorder in reflection features. The algorithm includes gradient tensor calculation, local covariance matrix calculation, and eigenvalue calculation. , , Solve for the chaos coefficients, defining them as follows:

[0074]

[0075] Optionally, the reflection intensity attribute describes the local energy distribution of the seismic signal. When the strata are continuous, the phase axis of the reflected wave is continuous, and the energy is stable. Conversely, if the seismic reflected wave field undergoes phase changes and energy attenuation, the reflected energy decreases, and a fault may occur. The calculation formula is as follows:

[0076]

[0077] In the formula, Indicates the sampling interval. Indicates the total number of sampling points. Indicates the number of non-zero sampling points. This represents the amplitude value at the sampling point.

[0078] Optionally, the instantaneous phase attribute reflects the formation continuity through the local phase angle change of the signal, and its calculation formula is as follows:

[0079]

[0080] In the formula, Represents the imaginary part of the seismic signal. This represents the real part of the seismic signal.

[0081] Optionally, the instantaneous frequency attribute reflects the time-varying characteristics of the seismic signal phase, revealing energy attenuation and tectonic disturbance characteristics. Its calculation formula is defined as the first derivative of the instantaneous phase with respect to time:

[0082]

[0083] Optionally, the dip angle deviation attribute is used to characterize the degree of deviation of the local dip angle of the target coal seam from the regional background dip angle. It can be used to reflect the discontinuity of the strata. The calculation formula is as follows:

[0084]

[0085] In the formula, Indicates the direction of the main lateral line. This is for connecting the flanks.

[0086] Optionally, the root mean square amplitude can represent the vibration intensity during the propagation of a seismic signal. If a seismic wave encounters a discontinuity during propagation, its energy will significantly attenuate, causing the amplitude to decrease significantly. The calculation formula is as follows:

[0087]

[0088] Optionally, curvature attributes describe the degree of bending of strata. By calculating the curvature attributes of structural surfaces, the geometric shape of geological bodies can be reflected, effectively characterizing geological structures such as faults and folds. First, least-squares fitting is performed to construct the surface equations:

[0089]

[0090] The coefficients are solved by the apparent dip angle of the formation and its first derivative. :

[0091]

[0092]

[0093]

[0094]

[0095]

[0096] In the formula, and for and The apparent tilt angle attribute of the direction. It is a constant. Maximum positive curvature and minimum negative curvature The calculation formula is as follows:

[0097]

[0098]

[0099] During implementation, attribute values ​​obtained at the same spatial sampling location are concatenated column-wise to construct an attribute sample vector, and the attribute sample vectors corresponding to all sampling locations are stacked row-wise to form a multidimensional attribute sample matrix.

[0100] S6: Standardize the attribute sample matrix, calculate the covariance matrix based on the standardized data, and perform eigenvalue decomposition.

[0101] In this step, after constructing the multi-attribute sample matrix in S5, since the scales of different seismic attributes are different, it is necessary to standardize the sample matrix to obtain the output matrix. :

[0102]

[0103] In the formula, For standardized attribute values, The first element in the original attribute matrix The first attribute Each sample value, For the first The mean of each attribute, For the first The standard deviation of each attribute.

[0104] After standardization, the standardized attribute matrix is ​​used. As input, calculate the covariance matrix and solve for the eigenvalues:

[0105]

[0106] In the formula: Let covariance matrix be the variance matrix. For eigenvalues, These are the eigenvectors.

[0107] S7: Select principal components based on their contribution rates and perform weighted summation to output the comprehensive characteristic attributes of small coal and rock faults.

[0108] After completing the eigenvalue decomposition of the covariance matrix in S6, the principal component sets and their corresponding eigenvalues ​​are sorted according to the magnitude of the eigenvalues. To reduce feature dimensionality and suppress attribute redundancy while ensuring the integrity of fault information, this invention filters principal components based on their cumulative contribution rate. The standardized attributes are then weighted and summed using the eigenvectors corresponding to the principal components to obtain the comprehensive characteristic attributes of small coal-rock faults. Cumulative contribution rate of each principal component :

[0109]

[0110] In the formula, For cumulative contribution rate, Let i be the eigenvalue of the i-th principal component. The number of eigenvalues.

[0111] By setting a cumulative contribution rate threshold, it is determined that the following conditions are met. Minimum number of principal components not less than the threshold Thus, the first Each principal component serves as an effective fusion feature.

[0112] Figure 6The paper presents a comparison of the multi-attribute characteristic responses at fault locations, including the original data, the reconstructed data after wavelet processing, and the data after fusion of wavelet transform and principal component analysis. It can be seen that the attribute characteristic responses at fault locations are significantly different from those in non-fault areas after discrete wavelet transform processing of the original seismic data. Furthermore, combining wavelet transform with principal component analysis of the original seismic data not only enhances the characteristic responses at fault locations but also greatly reduces the amount of data in the sample.

[0113] As can be seen from the above embodiments of the present invention, by introducing a multi-scale decomposition and energy proportion analysis mechanism of wavelet transform, adaptive filtering of the main frequency information of seismic data is achieved. On this basis, combined with principal component analysis, the seismic response to small faults is significantly improved, the data redundancy caused by strongly correlated seismic attributes is reduced, and the loss of fault-related information during the dimensionality reduction process is minimized, ensuring comprehensive analysis of the selected attributes.

[0114] In addition to the above, it should be noted that the terms "one embodiment," "another embodiment," and "embodiment" used in this specification refer to specific features, structures, or characteristics described in connection with that embodiment, which are included in at least one embodiment described in the general description of this application. The appearance of the same expression in multiple places in the specification does not necessarily refer to the same embodiment. Furthermore, when a specific feature, structure, or characteristic is described in connection with any embodiment, the intention is to suggest that implementing such a feature, structure, or characteristic in conjunction with other embodiments also falls within the scope of this invention.

[0115] Although the invention has been described herein with reference to several illustrative embodiments, it should be understood that many other modifications and implementations can be devised by those skilled in the art, which will fall within the scope and spirit of the principles disclosed herein. More specifically, various variations and modifications can be made to the components and / or layout of the subject matter arrangement within the scope of the disclosure, drawings, and claims. Besides variations and modifications to the components and / or layout, other uses will be apparent to those skilled in the art.

Claims

1. A method for extracting small fault features in coal and rock by fusing wavelet transform and principal component analysis, characterized in that... The method includes the following steps: Obtain the original 3D seismic data volume of the coal mining area; Discrete wavelet transform is used to decompose seismic signals in 3D seismic data volume into multiple scales, obtaining approximate components and detail components at different decomposition levels. Calculate and compare the energy proportions of detail components at each decomposition level of the seismic signal, and then select the detail component corresponding to the decomposition level with the highest energy proportion as the effective signal containing the seismic excitation frequency. The detailed components of the effective signal are superimposed with the corresponding approximate components at the same level to form a seismic feature signal in the frequency domain, and high signal-to-noise ratio time-domain seismic data are generated through inverse wavelet transform. Based on high signal-to-noise ratio time-domain seismic data, various fault-sensitive seismic attributes are extracted to form an attribute sample matrix; The attribute sample matrix is ​​standardized, the covariance matrix is ​​calculated based on the standardized data, eigenvalue decomposition is performed, and the contribution rate of the principal components is calculated based on the eigenvalues ​​of each principal component. The principal components whose cumulative contribution rate reaches a preset threshold are selected. The standardized attribute sample matrix is ​​weighted and summed using the eigenvectors corresponding to the principal components to obtain the comprehensive characteristic attributes of small coal and rock faults.

2. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1, characterized in that: The three-dimensional seismic data volume is seismic data extracted within a time window containing the target coal seam.

3. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1, characterized in that: The basis function selected for the discrete wavelet transform is the mother wavelet, and the maximum number of layers of the discrete wavelet transform decomposition is determined by the signal length of the seismic data.

4. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1 or 3, characterized in that: The energy of the detail components at each decomposition level is obtained by squaring and summing the wavelet coefficients corresponding to the decomposition level, and is used to characterize the magnitude of seismic signal energy contained in the corresponding frequency band of the decomposition level.

5. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1, characterized in that: The energy ratio is the ratio between the energy of a detail component at one decomposition level and the total energy of all detail components at all decomposition levels. By comparing the energy ratios corresponding to each decomposition level, the contribution of different decomposition levels to the energy of the original seismic signal can be measured.

6. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1, characterized in that: The high signal-to-noise ratio time-domain seismic data is reconstructed from the frequency-domain seismic characteristic signals through inverse wavelet transform. New detail components and corresponding new approximate components of the same level are selected according to the reconstruction formula of inverse wavelet transform, and the new detail components and new approximate components are combined and reconstructed to obtain the data.

7. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1 or 6, characterized in that: The attribute sample matrix is ​​standardized, and the standardized attribute matrix Z is obtained by calculating the following formula: In the formula, For standardized attribute values, The first element in the original attribute matrix The first attribute Each sample value, For the first The mean of each attribute, For the first The standard deviation of each attribute.

8. The method for extracting coal and rock small fault features by fusing wavelet transform and principal component analysis according to claim 1, characterized in that: The step of calculating the principal component contribution rate based on the magnitude of the eigenvalues ​​corresponding to each principal component involves sorting the principal components according to the magnitude of their corresponding eigenvalues ​​and then calculating the contribution rate using the following formula. Cumulative contribution rate of each principal component And select the one that makes the cumulative contribution rate The smallest positive integer that reaches the preset threshold Corresponding principal components: In the formula, For cumulative contribution rate, For the first The eigenvalues ​​of the principal components The number of eigenvalues.

Images