A method for coherent calculation of gradient structure tensors based on Gaussian differential operators

Abstract

This invention relates to a coherent calculation method for gradient structure tensors based on Gaussian differential operators. It includes: preprocessing a 3D seismic data volume; performing a Hilbert transform on the processed data volume and calculating the instantaneous phase, using the calculated instantaneous phase information as input data for constructing the structure tensor; using multi-scale Gaussian differential operators to calculate the gradients in the traverse, trace, and time directions to construct the gradient structure tensor; using Gaussian smoothing in the approximate fault and stratigraphic directions to enhance the lateral discontinuities in the gradient structure tensor corresponding to fault and stratigraphic features; performing matrix eigenvalue decomposition; and finally, performing different coherence solutions. This invention's gradient structure tensor coherent calculation method based on Hilbert transform and Gaussian smoothing can highlight narrow channels and small-impact faults; by utilizing the structure tensor, it reduces the sensitivity of coherence algorithms to noise, reduces signal clutter and strong noise interference at the intersections of channels and small-scale faults, and improves the accuracy of coherence identification.

Description

Technical Field

[0001] This invention relates to the field of seismic data coherence algorithm technology, specifically to a gradient structure tensor coherence calculation method based on Gaussian differential operators. Background Technology

[0002] In oil exploration and development, determining the boundaries of geological information such as paleochannels, sand bodies, faults, and fractures / cavities is crucial for improving the accuracy of oil and gas reservoir prediction. This is especially true in areas like the Shunbei region of the Tarim Basin, where Ordovician carbonate reservoirs are generally buried at depths greater than 7000m. Due to the influence of surface dunes and strata absorption, the energy attenuation of high-frequency components in seismic waves is rapid, resulting in generally low signal-to-noise ratios in seismic data. Furthermore, the small vertical displacement and strong heterogeneity of strike-slip faults lead to poor descriptive ability of coherent data for small-scale faults, fractures, and fractures / cavities.

[0003] In terms of algorithms, coherence algorithms have a relatively clear mathematical meaning. They utilize the similarity or coherence of adjacent traces to extract commonalities (coherence or incoherence) through mathematical means, thereby improving the interpretation ability of special geological bodies.

[0004] Currently, coherence volume has become an essential technique for reservoir interpretation and prediction. Coherence technology is widely used to detect channels and faults, playing an increasingly important role in fault identification and the interpretation of special lithologies. Commonly used coherence volume algorithms mainly include first-generation algorithms based on cross-correlation, second-generation algorithms utilizing seismic trace similarity, and third-generation algorithms based on eigenvalues. Many researchers both domestically and internationally have made significant contributions to the improvement and application analysis of these three generations of algorithms. For example, Wang Yonggang et al. were among the first in China to introduce second-generation coherence volume technology; Zhang Junhua et al. conducted quantitative analysis of the resolution and noise resistance of third-generation coherence volumes; Wang Xiwen et al. introduced wavelet transform into coherence volume calculation; Song Weiqi et al. proposed a method for calculating seismic multi-vector attribute coherence data volumes; Lu Wenkai et al. proposed an estimation algorithm based on higher-order statistics; and Yuan Shujin provided a comprehensive review of coherence volume technology. Although coherence technology has undergone three generations of development, all still suffer from averaging effects, making it difficult to highlight narrow channels and small-displacement faults. Meanwhile, coherent algorithms are sensitive to noise. At the intersections of river channels and faults, seismic signals are chaotic and noisy. Existing coherent technologies are unable to accurately identify narrow river channels and small faults, and can only identify the general development areas of river channels and faults.

[0005] Structure tensor and orientation field techniques have been widely used in the determination of texture direction in data. In three-dimensional seismic volume data, the volume data table is a layered structure. The longest axis (the eigenvector corresponding to the largest eigenvalue) corresponds to the direction of the largest gradient change, which is the average gradient direction; the orientation field of the volume data is perpendicular to the gradient direction and is represented by the shortest axis (the eigenvector corresponding to the smallest eigenvalue). Gradient structure tensor (GST) is an effective tool for fine characterizing and analyzing the reflection structure features in seismic data. Based on the GST algorithm and its derived multiple attributes, the discontinuous reflection features caused by geological bodies such as faults and channels can be qualitatively and quantitatively described. At the same time, it also provides a good indication of the characteristics of parallel, inclined, and wavy sedimentary bedding structures. For example, the patent document [1] published by CN 110749924 Method B, named a fault zone identification method, utilizes preprocessed seismic data to calculate a structural guide volume; applies the structural guide volume to calculate an edge detection data volume; based on the edge detection data volume, performs fault enhancement processing to obtain a fault enhancement data volume; calculates the seismic gradient structure tensor eigenvalue data volume and obtains eigenvalue data volume characterizing the chaotic reflection features of the fault zone; based on the obtained eigenvalue data volume characterizing the chaotic reflection features of the fault zone, obtains the fault zone structure data volume; finally, performs data volume fusion processing on the obtained fault enhancement data volume and the obtained fault zone structure data volume to obtain the fault zone prediction result. This method can improve the identification accuracy of fault zone contours and internal features. This reduces the risks of seismic interpretation in the exploration and development of fractured and fractured oil and gas reservoirs. It directly processes instantaneous amplitudes and calculates the gradient of the instantaneous phase along the x, y, and t directions. It uses a uniform standard deviation Gaussian kernel for smoothing and applies Gaussian smoothing to each element of the structural tensor. However, when analyzing areas such as the Tarim River and Shunbei regions, the presence of caves or fissures leads to the development of scattered waves, extremely low signal-to-noise ratios in seismic records, and poor continuity of phase axes. Furthermore, the strong heterogeneity of the reservoirs means that noise in the seismic signal has a significant impact on the processing results, often creating many "traps" in the interpretation. In these cases, the identification of narrow channels and small faults is limited, and only the approximate development areas of the channels and faults can be identified.

[0006] Furthermore, since the edge features of discontinuous information such as fault or fracture development zones also extend in the depth direction, there are multi-scale features in the depth direction. Different aperture sizes selected in the depth direction highlight different geological body information. Therefore, existing coherent technologies are difficult to accurately identify narrow channels and small faults, and can only identify the general development areas of channels and faults.

[0007] Therefore, how to accurately distinguish different types of reservoirs, delineate their boundaries, and highlight the subtle discontinuities in seismic reflection wave groups, so as to enhance the research on seismic attribute processing and interpretation methods, is a key issue that needs to be addressed in automatic seismic attribute detection technology. Summary of the Invention

[0008] To address the issues of averaging effects in coherent techniques, which make it difficult to highlight narrow channels and small faults, and the sensitivity of coherent algorithms to noise, making it difficult to accurately identify narrow channels and small faults, this invention provides a gradient structure tensor coherent calculation method based on Gaussian differential operators.

[0009] This invention claims protection for the following technical solution: This invention provides a method for calculating the gradient structure tensor of a three-dimensional geological body based on Hilbert transform. First, the data undergoes fine preprocessing such as standardization, normalization, and denoising. Then, the three-dimensional post-stack data volume is processed using Hilbert transform to obtain the signal envelope. The instantaneous phase of the three-dimensional seismic data is calculated using the processed data volume, and this instantaneous phase information is used as input data for constructing the structure tensor. A constructed multi-scale Gaussian differential operator is used to calculate the gradient of the data in the x-axis, y-axis, and t-time direction. A Gaussian smoothing filter is used to smooth the gradient structure tensor, and the gradient structure tensor of the seismic body data is constructed using the smoothed data. A Gaussian smoothing filter based on the three-dimensional geological structure and stratum dip angle is constructed to smooth the data volume. Matrix eigenvalue decomposition is performed on the new gradient structure tensor, and the meaning of the eigenvalues ​​is related to the spatial structure features in the corresponding data volume. Finally, coherent solutions are obtained using the eigenvalues.

[0010] This method mainly includes:

[0011] First: Instantaneous phase calculation method based on Hilbert transform

[0012] The Hilbert transform is one of the core technologies of this invention. This technology relies on the fact that the Hilbert transform of a continuous-time signal x(t) is equal to the output response xh(t) of the signal after passing through a linear system with an impulse response h(t) = 1 / πt. After the Hilbert transform, the amplitude of each frequency component in the frequency domain remains unchanged, but the phase will shift by 90°. That is, it lags by π / 2 for positive frequencies and leads by π / 2 for negative frequencies; therefore, the Hilbert transform is also called a 90° phase shifter. To address this, the Hilbert transform is used for preprocessing to obtain the signal envelope, improving data continuity and signal-to-noise ratio, and reducing the effects of aliasing and noise.

[0013] In 3D seismic data, the gradient vector at each point consists of three elements, representing the x, y, and t directions. At the location point (x, y, t), the 3D gradient vector describes the dip and azimuth angles of the seismic phase axis. Instantaneous phase is used as input data to estimate the data structure tensor. The instantaneous phase undergoes a Hilbert transform, and the arctangent of the ratio of its real part to the original seismic trace is used to obtain the instantaneous phase at each point. Preprocessing using the Hilbert transform to obtain the instantaneous phase of the signal improves data continuity and signal-to-noise ratio, and reduces the influence of aliasing and noise.

[0014] Second: Construction of the tri-directional structure tensor based on Gaussian differential operators and Gaussian smoothing for denoising.

[0015] The Gaussian differential operator weights the effect of pixel position, reducing edge blurring. When using instantaneous phase as input for coherent calculations of seismic data, the Hibert transform on the time axis suffers from positive and negative π-phase folding. Tensor calculations are unaffected by coordinate systems and can typically be represented by three-dimensional matrices, with the order representing the exponential magnitude being described. For example, a zero-order tensor represents a scalar, a first-order tensor represents a vector (both can be represented by a first-order matrix), and a two-dimensional matrix can represent a second-order tensor. In high-dimensional cases like 3D seismic data, the gradient vector at each point consists of three elements, representing the x, y, and t directions. The average gradient structure tensor S is a 3×3 positive semi-definite symmetric matrix with eigenvalues ​​greater than or equal to 0. To reduce the impact of noise interference in seismic data, the structure tensor method can be used to smooth the gradient vector. However, the structure tensor is highly unstable due to various types of seismic noise; therefore, a Gaussian function is used to mitigate instability during structure tensor calculations.

[0016] Third: Calculation of coherence properties of seismic data based on the eigenvalues ​​of a positive semidefinite matrix.

[0017] The eigenvectors and eigenvalues ​​can be easily calculated using the eigenvalue decomposition method. The eigenvectors and eigenvalues ​​(λ1, λ2, λ3) corresponding to the largest eigenvalue are perpendicular to the reflection interface and indicate the normal direction of the seismic phase axis.

[0018] Since the gradient structure tensor matrix is ​​a real symmetric matrix, satisfying λ1, λ2, λ3 > 0. For curved layered texture units, generally λ1, λ2, λ3 = 0. Therefore, by sorting all the eigenvalues ​​from largest to smallest, we can use the eigenvalues ​​to construct different structural properties or identify or distinguish the type of image texture unit based on the relative magnitude of the above three eigenvalues.

[0019] Compared with existing technologies, this invention establishes a new gradient solution using Gaussian differential operators in the approximate fault and stratigraphic directions under a novel Hilbert transform. It then utilizes these Gaussian differential operators in the approximate fault and stratigraphic directions to smooth the gradient structure tensor coherence calculation method. This method is used to highlight narrow channels and small-displacement faults. At the intersections of channels and small-scale faults, seismic signals are often cluttered and noisy. By using the structure tensor to reduce the noise sensitivity of the coherence algorithm, the accuracy of coherence identification is improved. This provides support for the optimal selection of exploration and development well locations and the optimization of well trajectory design, improves the quality of coherence maps, enhances the representation of weak reflections and weak features, and improves the accuracy of coherence identification. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a gradient structure tensor coherence calculation method based on Gaussian differential operators provided by the present invention.

[0021] Figure 2 This diagram illustrates the directional derivative in this embodiment. I represents different phases, II represents the gradient transformation direction of the phase magnitude along one direction, and III represents the gradient transformation direction of the phase magnitude along two directions.

[0022] Figure 3 This diagram shows the correspondence between feature values ​​and spatial structure in this embodiment.

[0023] Figure 4 Table 1 compares the results of various algorithms in the Shunbei region. In this table, 'a' represents the coherence calculated by the conventional C3 algorithm, and 'b' represents the slice calculated by the gradient structure tensor used in this invention. As indicated by the red arrows in the actual comparison, the coherence algorithm used in this paper better displays secondary faults. In other regions, the coherence algorithm calculated in this paper better displays detailed information about seismic traces, including the display of secondary faults and minor faults. Fault continuity is richer, and the information on faults and cracks is more comprehensive. In areas with well-developed anisotropy, such as Shunbei, this paper can perform better coherence calculations.

[0024] Figure 5 This invention provides a schematic diagram of the program modules and model of a seismic exploration coherence measurement method device for implementing a gradient structure tensor coherence calculation method based on Gaussian differential operators. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] Currently, coherent seismic technology is widely used to detect river channels and faults. Although coherent technology has undergone three generations of development, it still suffers from averaging effects, making it difficult to highlight narrow river channels and faults with small displacements. At the same time, coherent algorithms are sensitive to noise. At the intersections of river channels and faults, seismic signals are cluttered and noisy, making it difficult for existing coherent technologies to accurately identify narrow river channels and small faults; they can only identify the general development areas of river channels and faults.

[0027] This paper proposes a method for quantitatively estimating phase by directly calculating the first derivative of a 3D seismic data volume using the Hilbert transform. Taking the instantaneous phase information of the seismic data as input, and based on the structural tensor method, quantitative geological features are extracted through structural tensor analysis. A Gaussian function is introduced to mitigate instability issues during the structural tensor calculation process. The structural tensor feature method can smooth gradient estimation and reduce noise interference in the seismic data. The eigenvalue decomposition method can conveniently calculate the eigenvectors and eigenvalues ​​(λ1, λ2, λ3). The eigenvector corresponding to the largest eigenvalue is perpendicular to the reflection interface, indicating the normal direction of the seismic phase axis, thereby depicting and predicting fault and fracture discontinuities.

[0028] This invention discloses a gradient structure tensor coherence calculation method based on Gaussian differential operators. The obtained seismic coherence attribute profile effectively depicts subtle vertical variations in stratigraphic deposition. It enhances fault continuity, suppresses interference from pseudo-phase axes in low-coherence strips, and highlights subtle stratigraphic discontinuities. Furthermore, it maintains the original vertical resolution and signal-to-noise ratio of the seismic data. Utilizing Hilbert's transformation on the time axis, preprocessing is performed to obtain the signal envelope, improving data continuity and signal-to-noise ratio while reducing the impact of aliasing and noise. A 3D model three-directional structure tensor is constructed using Gaussian differential operators. The influence of pixel position on the Gaussian differential operator is weighted, reducing edge blurring. This method is sensitive not only to the waveform and lateral variations of 3D seismic amplitude in seismic traces but also to dip angle, compensating for the shortcomings of the C3 algorithm. Based on existing data, this method has good applicability and scalability. The research approach and technical methods are highly feasible, guiding the improvement of 3D structure tensor coherence calculation in the Shunbei area.

[0029] To illustrate the structure and function of this invention, the following description, in conjunction with the accompanying drawings, takes a three-dimensional work area in the Shunbei region of the Tarim Basin (hereinafter referred to as S-3D) as an example to further describe a gradient structure tensor coherence calculation method based on Gaussian differential operators.

[0030] like Figure 1 The diagram illustrates the steps of a gradient structure tensor coherence calculation method based on Gaussian differential operators provided by this invention, including the following steps:

[0031] Step S1: Data preprocessing: The 3D seismic data volume is preprocessed to remove noise, outliers and other interference, and the data is regularized.

[0032] Step S2: Calculate the instantaneous phase: After performing a Hilbert transform on the preprocessed 3D seismic data volume, the data obtained is used to calculate the instantaneous phase of each point in the 3D seismic data volume and serve as the input data for the next step.

[0033] Step S3: Construct the gradient structure tensor: Using the instantaneous phase as input, construct a Gaussian differential function using the Gaussian differential operator, and calculate the directional derivative of the instantaneous phase in different strata and in the direction perpendicular to the strata using the Gaussian differential; combine the directional derivatives to construct the gradient structure tensor.

[0034] Step S4: Smooth the gradient structure tensor using a Gaussian function: Use a Gaussian function to smooth the gradient structure tensor in the approximate fault and stratigraphic directions to enhance the lateral discontinuities in the gradient structure tensor corresponding to fault and stratigraphic features;

[0035] Step S5: Calculate eigenvalues: Perform matrix eigenvalue decomposition on the smoothed gradient structure tensor;

[0036] Step S6 uses eigenvalues ​​to perform different coherence solutions: Based on the eigenvalue relationships, construct measures based on different types of anisotropic parameters, and construct relational attributes for small fractures, cracks and cavities, thus obtaining the coherence attributes constructed for small fractures, cracks and cavities.

[0037] In this preferred embodiment, the method for preprocessing the three-dimensional seismic data volume in step S1 includes: first performing amplitude energy compensation and noise reduction normalization processing on the data volume to remove extreme values, and then performing edge-preserving smoothing filtering preprocessing to ensure that the seismic reflection of discontinuities such as faults and cracks is free from noise and other interference, and finally obtaining the preprocessed three-dimensional seismic data volume free from noise and other interference.

[0038] Furthermore, in step S2 when calculating the instantaneous phase: performing a Hilbert transform on the preprocessed three-dimensional seismic data volume means performing a Hilbert transform on the three-dimensional seismic data volume along the time axis, and using the transformed data to calculate the instantaneous phase of each point in the three-dimensional seismic data volume.

[0039] In 3D seismic data, the gradient vector at each point consists of three elements, representing the x, y, and t directions. At the location point (x, y, t), this is a 3D gradient vector describing the dip and azimuth of the seismic strata reflection surface. Instantaneous phase is used as input data to estimate the data structure tensor. The instantaneous phase undergoes a Hilbert transform, and the arctangent of the ratio of its real part to the original seismic trace is used to obtain the instantaneous phase at each point.

[0040] In this embodiment, the instantaneous phase is calculated using the processed signal, which can be obtained by the following formula:

[0041] (6)

[0042] In the formula: It is the instantaneous phase at that point. This refers to the Hilbert transform of u with respect to the time axis t. Instantaneous phase information is used as input data for coherent calculations of seismic data. After the preprocessed seismic data volume signal undergoes the Hilbert transform, the amplitude of each frequency component remains unchanged in the frequency domain, but the phase will exhibit a 90° phase shift. That is, it lags by π / 2 for positive frequencies and leads by π / 2 for negative frequencies; therefore, the Hilbert transform is also called a 90° phase shifter. For this purpose, the Hilbert transform is used for preprocessing to obtain the signal envelope, improve data continuity and signal-to-noise ratio, and reduce the influence of aliasing and noise.

[0043] Based on the instantaneous phase point data obtained above, the gradient structure tensor is constructed in step S3. The Gaussian differential operator weights the influence of pixel position, which can reduce edge blurring. When using the instantaneous phase as the data input for coherent calculation of seismic data, the Hibert transform on the time axis will have the problem of positive and negative π phase folding. Tensor calculation is not affected by the coordinate system and can usually be represented by a three-dimensional matrix. Its order represents the exponential size that needs to be described. For example, a zero-order tensor represents a scalar, a first-order tensor represents a vector, which can be represented by a first-order matrix, and a two-dimensional matrix can represent a second-order tensor. In the case of high-dimensional data such as three-dimensional seismic data, the gradient structure tensor S is a 3×3 positive semi-definite symmetric matrix with eigenvalues ​​greater than or equal to 0. To reduce the influence of noise interference in seismic data, the structure tensor method can be used to smooth the gradient vector. The structure tensor is affected by various types of seismic noise, which can make the calculation results extremely unstable. Therefore, a Gaussian function is used to reduce the instability in the structure tensor calculation process.

[0044] Using Gaussian differential operators to calculate gradients in the x-axis, y-axis, and time-t direction involves: taking the instantaneous phase as input u(x,y,t), and calculating the directional derivatives ∇u(x,y,t) of the instantaneous phase in directions perpendicular to and parallel to the seismic structure (reflecting surface). This allows for better capture of lateral seismic discontinuities in the structural tensor, particularly the subtle stratigraphic features within dipping structures. Combining these directional derivatives constructs a gradient structure tensor, yielding gradients in the x-axis, y-axis, and time-t direction; for example... Figure 2 The diagram shown is a schematic of the directional derivative in this embodiment. I represents different phases, II represents the gradient transformation direction of the phase magnitude along one direction, and III represents the gradient transformation direction of the phase magnitude along two directions.

[0045] The Gaussian differential operator is mainly used to calculate the gradient derivative. Technically, it is a discrete difference operator. Applying this operator to any point in an image will produce a corresponding grayscale vector or its normal vector. It calculates the derivative of the instantaneous phase at each point and constructs three convolution factors in three directions, resulting in three kernels.

[0046] Define a seismic signal volume u(x,y,t) of three-dimensional data, where the directional derivative vector at each point is ∇u(x,y,t);

[0047] The gradient structure tensor is given by the following equation:

[0048] (1).

[0049] Then, in step S4, the gradient structure tensor is smoothed using a Gaussian function. In this embodiment, instead of using a uniform standard deviation Gaussian kernel for smoothing, or applying Gaussian smoothing to each element of the constructed tensor, smoothing is performed using a Gaussian function in the approximate fault and stratigraphic directions. Specifically, this includes:

[0050] A Gaussian kernel Gρ function with standard deviations ρ1, ρ2, and ρ3 in the x-axis, y-axis, and time-t directions is used to smoothly average the three components of the gradient vector; where Gρ is a three-dimensional standard Gaussian function with the form:

[0051] (2)

[0052] Where U represents the original seismic data; U σ For smoothed seismic data, σ is a scale parameter whose value depends on the noise intensity of the data to be analyzed, i.e., it is defined as the noise scale, which can reduce the sensitivity of gradient numerical calculations to noise and also establishes the smallest tectonic scale for measurable gradient structure tensors; G σThe Gaussian kernel is used for the scale parameter σ; the choice of standard deviation ρ depends on the dip angle, dip direction, tectonic measurement, and resolution of the analysis, i.e., it is defined as the tectonic scale. The gradient tensor is smoothed in the Gaussian neighborhood of ρ to make it a symmetric positive semi-definite matrix, which improves the robustness of the structural analysis.

[0053] Furthermore, in step S5, eigenvalues ​​are calculated by: [calculating the gradient structure tensor]. Perform matrix eigenvalue decomposition

[0054] (3)

[0055] Let the gradient structure tensor be the tensor.

[0056] Wherein: λ １ ,λ ２ ,λ ３ for The three non-negative eigenvalues ​​are v1, v2, and v3, which are the corresponding eigenvectors. They form a local orthogonal coordinate system. v1 represents the direction of maximum contrast in the local neighborhood, i.e., the gradient direction of the signal. v2 and v3 form a local plane perpendicular to v1. １ ≥λ ２ ≥λ ３ ≥0, The meanings of the three characteristic values ​​correspond to the spatial structural characteristic relationships in the seismic data volume.

[0057] The eigenvectors and eigenvalues ​​can be easily calculated using the eigenvalue decomposition method. The eigenvectors and eigenvalues ​​(λ1, λ2, λ3) corresponding to the largest eigenvalue are perpendicular to the reflection interface and indicate the normal direction of the seismic phase axis.

[0058] Since the gradient structure tensor matrix is ​​a real symmetric matrix, satisfying λ1, λ2, λ3 > 0. For curved layered texture units, λ1, λ2, λ3 generally satisfy λ1, λ2, λ3 = 0. Therefore, by sorting all the eigenvalues ​​from largest to smallest, we can use the eigenvalues ​​to construct different structural properties or identify or distinguish the type of image texture unit based on the relative magnitude of the above three eigenvalues.

[0059] like Figure 3 The diagram shown illustrates the correspondence between eigenvalues ​​and spatial structure in this embodiment. Wherein, if λ １ =λ ２ =λ ３ If λ = 0, then all eigenvalues ​​are approximately zero, indicating no signal change in the local neighborhood, which corresponds to a typical seismic profile 1; if λ １ >0, λ ２ =λ ３=0, eigenvalue λ ２ , λ ３ If λ is approximately zero, it indicates a planar structure in the local neighborhood, corresponding to a typical seismic profile 2; if λ １ >0, λ ２ >0, λ ３ =0, eigenvalue λ ３ Approximately zero indicates that the local neighborhood has a linear-like structure. Furthermore, if λ １ ≈λ ２ This represents a uniform linear feature, corresponding to a typical seismic profile 3; if λ １ >0, λ ２ >0, λ ３ >0, all eigenvalues ​​are greater than zero, indicating that the signal structure in the local neighborhood deviates from the linear model. This may be caused by the appearance of curves or noise, which corresponds to a typical seismic profile 4.

[0060] Finally, through step S6: Based on the eigenvalue relationships, construct chaos metrics based on different types. For faults, construct the following chaos metrics:

[0061] (4)

[0062] Chaos metrics based on GST can effectively reflect the regularity of local structures and are independent of amplitude variations. That is: →1, corresponding to the lateral discontinuity structure; →0 corresponds to a region with irregular and chaotic reflections; →-1, corresponding to a regular layered structure.

[0063] The coherence properties calculated using the structural tensor method of seismic data are as follows:

[0064] (5)

[0065] In the formula: C is the coherence coefficient; J is the number of eigenvalues, J = 3.

[0066] As described in the above embodiments, this invention employs a coherent calculation method for gradient structure tensors constructed along fault and stratigraphic dip angles using Gaussian differential operators. Subsequent calculations are performed using instantaneous phase obtained through Hilbert transform. The calculated instantaneous phase is represented by the directional derivative ∇u(x,y,t) in directions perpendicular and parallel to the geological strata, rather than directly calculating the gradient along the x, y, and t directions. This method better captures lateral seismic discontinuities in the structural tensor. Instead of using a uniform standard deviation Gaussian kernel for smoothing, or applying Gaussian smoothing to each element of the structural tensor, a Gaussian function smoothing approximating fault and stratigraphic directions is used to enhance the lateral discontinuities in the structural tensor corresponding to fault and stratigraphic characteristics. A Gaussian kernel Gρ function with standard deviations ρ1, ρ2, and ρ3 in three directions is used to smoothly average the three components of the gradient vector. The obtained seismic coherence property profile effectively characterizes the subtle vertical variations exhibited by stratigraphic deposition. It enhanced the continuity of faults, suppressed the interference of pseudo-phase axes in low-coherence strips, and highlighted subtle discontinuities in the strata. Furthermore, it maintained the vertical resolution and signal-to-noise ratio of the original seismic data.

[0067] like Figure 4 The table shown is a comparison of the results of various algorithms in the Shunbei area. Here, 'a' represents the coherence calculated by the conventional C3 algorithm, and 'b' represents the slice calculated by the gradient structure tensor used in this invention. As indicated by the red arrows in the actual comparison, the coherence algorithm used in this embodiment can better display secondary small faults. In other areas, the coherence algorithm calculated in this paper better displays the detailed information of seismic traces, including the display of secondary faults for small faults. The fault continuity is richer, and the information on faults and cracks is more comprehensive. In areas with well-developed anisotropy, such as Shunbei, this method can perform coherence calculations better.

[0068] Furthermore, this invention is the first to propose the application of Hilbert transform in coherent calculation methods, particularly using Hilbert's transformation on the time axis to obtain the instantaneous phase, thereby improving data continuity and signal-to-noise ratio, and reducing the impact of spurious frequencies and noise. This method utilizing Hilbert transform demonstrates excellent effectiveness in improving data continuity and signal-to-noise ratio, and reducing the impact of spurious frequencies and noise, when applied to the exploration and development of other fractured and fractured oil and gas reservoirs, and in addressing the urgent need for geophysical techniques for fault zone identification and description.

[0069] Furthermore, in the field of coherent property calculation for seismic exploration interpretation, the gradient structure tensor coherent calculation method based on Gaussian differential operators provided by this invention has excellent application value. Therefore, this invention is not limited to the examples described above, and all applications involving the field of coherent detection should be within its protection scope.

[0070] A seismic exploration interpretation coherence calculation device is used to implement the above-mentioned gradient structure tensor coherence calculation method based on Gaussian differential operators, such as... Figure 5 As shown, it specifically includes the following program modules and models:

[0071] Data preprocessing module 100: used to preprocess the 3D seismic data volume to remove noise and other interference;

[0072] Instantaneous phase calculation module 200: used to calculate the instantaneous phase of each point in the three-dimensional seismic data volume after performing Hilbert transform on the preprocessed three-dimensional seismic data volume, and use it as input data for the next step;

[0073] Model 300 for constructing gradient structure tensor: It takes the instantaneous phase as input, uses the Gaussian differential operator to construct the Gaussian differential function, calculates the directional derivative of the instantaneous phase in different strata and in the direction perpendicular to the strata through Gaussian differentiation, and uses the directional derivatives to combine to construct the gradient structure tensor;

[0074] Smoothing model 400 for gradient structure tensor: used to enhance the lateral discontinuities in the gradient structure tensor corresponding to fault and stratigraphic features by utilizing Gaussian smoothing in the approximate fault and stratigraphic directions;

[0075] Eigenvalue calculation module 500: used to perform matrix eigenvalue decomposition on the smoothed gradient structure tensor;

[0076] Different coherence solution modules 600: are used to construct measures based on different types of anisotropic parameters according to the eigenvalue relationships, and to construct relational attributes for small fractures, cracks and cavities, that is, to obtain the coherence attributes constructed for small fractures, cracks and cavities.

[0077] The device or instrument also includes: other methods for seismic coherence calculation, and coherence calculation for fields such as detection, inspection, and image processing.

[0078] As described in the above embodiments, the method of this invention and the application instruments or devices used to implement it utilize Hilbert's transformation on the time axis to process and obtain the instantaneous phase, thereby improving the continuity and signal-to-noise ratio of the data and reducing the influence of aliasing and noise. A 3D model three-directional structure tensor is constructed using Gaussian differential operators. Instead of using a uniform standard deviation Gaussian kernel for smoothing, or applying Gaussian smoothing to each element of the constructed tensor, Gaussian functions approximating fault and stratigraphic directions are used to enhance the lateral discontinuities in the constructed tensor corresponding to fault and stratigraphic features. A Gaussian kernel Gρ function with standard deviations ρ1, ρ2, and ρ3 in three directions is used to smoothly average the three components of the gradient vector, reducing edge blurring. This method is sensitive not only to the waveform and lateral variations of the three-dimensional seismic amplitude of the seismic trace, but also to dip angle, compensating for the shortcomings of the C3 algorithm. It provides support for the optimal selection of exploration and development well locations and the optimization of well trajectory design, improves the quality of coherence maps, enhances the representation of weak reflections and weak features, and improves the accuracy of coherence identification. Based on existing data, this method has good applicability and generalizability. The research ideas and technical methods are highly feasible, and it has guided the improvement of tensor coherence calculation of three-dimensional structures such as small fractures, small faults, fissures and small channels with directional anisotropy.

[0079] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A gradient structure tensor coherence calculation method based on a Gaussian differential operator, characterized in that, Includes the following steps: S1 Data Preprocessing: Preprocessing the 3D seismic data volume to remove noise, outliers and other interference, and regularize the data; S2 Calculate Instantaneous Phase: After performing a Hilbert transform on the preprocessed 3D seismic data volume, the data obtained is used to calculate the instantaneous phase of each point in the 3D seismic data volume and serve as the input data for the next step; S3 constructs the gradient structure tensor: using the instantaneous phase as input u(x,y,t), a Gaussian differential function is constructed using the Gaussian differential operator. The directional derivatives ∇u(x,y,t) of the instantaneous phase in the directions perpendicular to and parallel to the seismic structure are calculated using the Gaussian differential. The directional derivatives are combined to construct the gradient structure tensor, obtaining different gradients in the x-axis direction, y-axis direction, and time t direction. S4 uses a Gaussian function to smooth the gradient structure tensor: the smoothing of the gradient structure tensor in the approximate fault and stratigraphic direction by the Gaussian function is used to enhance the lateral discontinuities in the gradient structure tensor corresponding to fault and stratigraphic features. S5 calculates eigenvalues: Performs matrix eigenvalue decomposition on the smoothed gradient structure tensor; S6 utilizes eigenvalues ​​to perform different coherent solutions: Based on the eigenvalue relationships, it constructs metrics based on different types of anisotropic parameters, and builds relational attributes for small fractures, cracks, and cavities, thus obtaining the coherent attributes constructed for small fractures, cracks, and cavities.

2. The gradient structure tensor coherence calculation method based on Gaussian differential operators as described in claim 1, characterized in that, The method for preprocessing the three-dimensional seismic data volume includes: first, performing amplitude energy compensation and noise reduction normalization on the data volume to remove extreme values; then, performing edge-preserving smoothing filtering preprocessing to ensure the removal of seismic reflection noise, including discontinuities such as faults and cracks, as well as the influence of other interferences; and finally obtaining the preprocessed three-dimensional seismic data volume free of noise and other interferences.

3. The gradient structure tensor coherence calculation method based on Gaussian differential operators as described in claim 1, characterized in that, Performing a Hilbert transform on the preprocessed 3D seismic data volume means: performing a Hilbert transform on the 3D seismic data volume along the time axis, and using the transformed data to calculate the instantaneous phase of each point in the 3D seismic data volume.

4. The gradient structure tensor coherence calculation method based on Gaussian differential operators as described in claim 1, characterized in that, In step S3, the gradient structure tensor is expressed as follows: （1） Where U represents the original seismic data; U σ The data is smoothed seismic data, where σ is the scale parameter; G σ σ is the Gaussian kernel with scale parameter σ; the standard deviation ρ is defined as the construction scale; Gρ is the standard Gaussian function in three dimensions; Let the gradient structure tensor be the tensor of the gradient structure. In step S4, smoothing the gradient structure tensor using a Gaussian function, the smoothing using a Gaussian function in the approximate fault and stratigraphic directions specifically includes: The three components of the gradient vector are smoothly averaged using a Gaussian kernel Gρ function with standard deviations ρ1, ρ2, and ρ3 in the three directions of line direction, track direction, and time direction. The resulting average is as follows: （2）。 5. The gradient structure tensor coherence calculation method based on Gaussian differential operators as described in claim 4, characterized in that, In step S5, calculating eigenvalues, performing matrix eigenvalue decomposition on the gradient structure tensor means: （3） Where: λ1,λ ２ ,λ ３ for The three non-negative eigenvalues ​​are v1, v2, and v3, which are the corresponding eigenvectors. They form a local orthogonal coordinate system. v1 represents the direction of maximum contrast in the local neighborhood, i.e. the gradient direction of the signal. v2 and v3 form a local plane perpendicular to v1.

6. The gradient structure tensor coherence calculation method based on Gaussian differential operators as described in claim 5, characterized in that, The calculation method for constructing coherence for small fractures, cracks, and cavities is as follows: First, based on the eigenvalue relationships, we construct chaos metrics for different types, as follows: （4） Where: If →1, corresponding to a transverse discontinuity structure; if →0 corresponds to a region with irregular and chaotic reflections; if →-1, corresponding to a regular layered structure; Secondly, the coherence properties calculated using the structural tensor method of seismic data are: （5） Where: C is the coherence coefficient; J is the number of eigenvalues, J = 3.

7. The application of the gradient structure tensor coherence calculation method based on Gaussian differential operators as described in any one of claims 1-6 in the extraction of three-dimensional seismic attribute volumes in seismic exploration coherence calculation.

8. A coherence calculation device for seismic exploration interpretation, characterized in that, This method is used to implement the gradient structure tensor coherence calculation method based on Gaussian differential operators as described in any one of claims 1-6.

9. The seismic exploration interpretation coherence calculation device as described in claim 8, characterized in that, It includes the following program modules and models: Data preprocessing module: used to preprocess 3D seismic data volumes, remove noise, outliers and other interference, and regularize the data; Instantaneous phase calculation module: used to calculate the instantaneous phase of each point in the preprocessed 3D seismic data volume after performing a Hilbert transform, and use the data as input data for the next step; The model for constructing the gradient structure tensor is as follows: the instantaneous phase is used as input, a Gaussian differential function is constructed using the Gaussian differential operator, the directional derivatives of the instantaneous phase in different strata and perpendicular to the strata are calculated by Gaussian differentiation, and the directional derivatives are combined to construct the gradient structure tensor. Smoothing model of gradient structure tensor: used to enhance the lateral discontinuities in the gradient structure tensor corresponding to fault and stratigraphic features by using Gaussian function smoothing in the approximate fault and stratigraphic directions; The eigenvalue calculation module is used to perform matrix eigenvalue decomposition on the smoothed gradient structure tensor. Different coherence solution modules: These are used to construct metrics based on different types of anisotropic parameters according to eigenvalue relationships, and to construct relational attributes for small fractures, cracks, and cavities, thus obtaining coherence attributes constructed for small fractures, cracks, and cavities.

Images