Fracture system depiction characterization method
By combining navigation pyramid, diffusion equation and tensor voting technology, the ambiguity and multi-solution problems in fault feature identification are solved, and the fine characterization of the fault system and the clear identification of the geological body boundaries are achieved.
Patent Information
- Application Number
- CN202511128803.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-13
- Publication Date
- 2025-10-14
AI Technical Summary
The existing fault feature identification methods are difficult to accurately characterize the fault system due to the unclear seismic signals, which leads to the fragmentation and discontinuity of the main faults, the strong multi-solution of small faults, and the loss of detailed information.
The navigation pyramid technology is used for frequency division processing, combined with high-angle fracture identification based on the diffusion equation and tensor voting technology. Through consistency enhancement processing and discontinuity feature extraction, combined with drilling loss matching and other technical means, a detailed characterization of the fracture system is achieved.
It improves the accuracy and continuity of fault identification, clearly presents the trunk and branches of the fault system, enhances the accuracy of geological body boundary identification, and reduces the ambiguity of the results.
Smart Images
Figure CN120779474A_ABST
Abstract
Description
Technical Field
[0001] The present invention provides a fracture system characterization method, belonging to the technical field of seismic signal processing. Background Art
[0002] Existing fault feature identification methods often face difficulties in accurately characterizing fault systems due to the ambiguity of seismic signals, which can lead to fragmented and discontinuous main faults, multiple interpretations of minor faults, and loss of detailed information. Therefore, innovative research on fracture system characterization techniques is needed, based on basic data such as geology, seismology, well logging, and reservoir development.
[0003] For example, the ant-based algorithm currently used in technology mimics the behavior of ants in nature. Ants that choose the shortest path during foraging will arrive earlier than those that take longer routes. Shorter paths are marked with more pheromones than longer paths. Influenced by these pheromones, the next ant is more likely to choose the shorter path, thus achieving the overall optimization goal of the ant colony. Based on the principles of the ant algorithm, seismic attribute volumes that reflect fault information appear as local maxima or minima of attribute values at fault locations. This results in a new data volume with a high signal-to-noise ratio and clear fracture tracing, thus achieving the goal of characterizing faults or fractures. However, the disadvantage of this technique is that the results can be highly subjective and open to multiple interpretations through parameter adjustments. Summary of the Invention
[0004] To characterize fault systems, this paper designs a technical process encompassing three key steps: comprehensive seismic data processing, fault system identification, and fracture skeleton carving. This process addresses the multi-solution problem of ant-body techniques while simultaneously enabling geological boundary characterization. This process primarily incorporates three core technologies: a navigation pyramid, high-angle fracture identification based on diffusion equations, and tensor voting. Furthermore, it incorporates techniques such as discontinuity processing, geological analysis, and drilling loss matching. This method significantly improves fracture identification results compared to ant-body techniques and enables the identification of river channel sandbody boundaries.
[0005] The fracture system characterization method includes the following steps:
[0006] Step 1: After obtaining the post-stack seismic data, the navigation pyramid technology is first used to perform frequency division processing to obtain sub-band seismic data of different scales. At the same time, the sub-band data of different scales are combined as needed to achieve the purpose of highlighting the target information.
[0007] The navigation pyramid seismic data processing process mainly includes the creation of pyramid structure (image decomposition) and the reconstruction of the decomposed image.
[0008] The navigation pyramid structure combines the Laplacian pyramid data structure, commonly used in image processing, with steerable filters. The core of this method for seismic signal processing is to decompose the seismic signal into a steerable pyramid data structure using a multi-scale and multi-directional analysis method. During this decomposition process, steerable filters are used to extract geological information from each sub-band. The entire processing process primarily involves the creation of the pyramid structure and the reconstruction of the decomposed signal. This method can operate in different directions and at different scales, enabling clearer characterization and identification of subsurface geological bodies of diverse shapes and sizes compared to other methods.
[0009] Step 2: Based on the combined seismic data obtained in Step 1, the high-angle fracture identification technology based on the diffusion equation is used to perform fracture detection or boundary identification on the seismic data.
[0010] In order to make the identified fault information more consistent with the original seismic data, a consistency enhancement processing method was introduced based on the first technique.
[0011] Highlighting the differences between faulted and non-faulted conditions, a fracture system treatment based on the diffusion equation is used:
[0012]
[0013] Where D(x) is the diffusion tensor of the analysis point, which is constructed by the eigenvectors of the seismic data structure tensor. The structure tensor T of seismic data can be constructed as the smoothed outer product of the seismic data gradient. The structure tensor T can be expressed by eigendecomposition as follows:
[0014] T=λ u uu T +λ v vv T +λ w ww T
[0015] Among them, λ u ,λ v ,λ w are the eigenvalues corresponding to the orthogonal eigenvectors u, v, and w respectively. If the mark λ u ≥λ v ≥λ w ≥0, the corresponding eigenvector u will be parallel to the direction where the u image changes most significantly, and the corresponding eigenvector w will be parallel to the direction where the u image changes least significantly, based on which the original seismic data is diffused.
[0016] The defined diffusion rate in the diffusion process is:
[0017]
[0018] Where α is the threshold used to detect discontinuity features at the analysis point, and d(x) is the directional derivative of the analysis point. The relationship between the directional derivative and the threshold is used to distinguish between faults and noise, and to control the diffusion direction of the analysis point. When d(x) > α, the point is considered to be a discontinuity feature and is retained; otherwise, it is considered to be noise and is suppressed.
[0019] Voting is performed in parameter space to determine the shape of an object, determined by the local maximum in the ρ-θ spectrum. The basic process is: ① Calculate the ρ-θ energy spectrum of the image data; ② Determine the parameters ρ and θ of the line based on the local maximum in the ρ-θ energy spectrum; ③ Finally, determine the line based on the ρ and θ parameters.
[0020] ρ=xcosθ+ysinθ
[0021] Where x and y are the weight coefficients of the line equation. Geometric relationships provide the direction vector along the line, while the direction of a point not on a line can be determined by interpolation.
[0022]
[0023] Then the diffusion tensor that diffuses only along the direction of this vector is
[0024] The obtained diffusivity does not show a zero value line along the fault, but corresponds to many low value points. These low value points are filtered along the fault direction. The consistency enhancement process can detect points on the same line and determine their linear direction. The consistency enhancement process is then performed on the diffusivity based on the linear direction. The diffusivity of the constructed diffusion tensor is:
[0025]
[0026] Among them, μ u 、μ v represents the diffusion rate in the u and v directions, H represents the filter guided by the diffusion rate when performing consistency enhancement processing, α is the threshold used to detect discontinuity features at the analysis point, and d(x) is the directional derivative of the analysis point. The relationship between the directional derivative and the threshold is used to distinguish between faults and noise, and to control the diffusion direction of the analysis point. When d(x)>α, the location is considered to be a discontinuity feature and is retained; otherwise, it is considered to be noise and is suppressed.
[0027] On this basis, the diffusivity attributes are further refined and ridge-extracted to obtain fault identification results that are consistent with the original seismic data.
[0028] Step 3: Using the discontinuity feature extraction technology based on the navigation pyramid filtering technology, the discontinuity features in the seismic data are highlighted, and the fault system is predicted on this basis.
[0029] Step 4: Fuse the fracture detection results obtained in Step 2 with the discontinuity feature extraction results obtained in Step 3 to obtain preliminary fracture detection results.
[0030] Step 5: Use tensor voting technology to process the fusion results obtained in Step 4, effectively improving the continuity of the fault system and effectively removing the recognition artifacts caused by noise during the recognition process, making the fault direction and structure clearer.
[0031] In a way that simulates the vision of geologists, the properties of the fracture body are enhanced and hollowed out to enhance the continuity and integrity of the fracture system, thereby achieving the purpose of clear trunk and branches and realizing the optimization of the fracture system characterization.
[0032] The navigation pyramid technology adopted in the present invention can realize the frequency division processing of seismic data, highlight the required target information while obtaining multi-scale seismic data, greatly improve the resolution of seismic data, and reduce the difficulty of further processing; on this basis, seismic attribute extraction operations such as non-continuous seismic data processing or high-angle fracture identification processing based on diffusion equations are performed. Finally, tensor voting technology is used to perform structural hollowing enhancement on the obtained attribute results, the purpose of which is to enhance the continuity and visibility of the attribute results, making the geological structure in the calculated attribute results clearer. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 This is a flowchart of the seismic data processing implementation of the navigation pyramid method of the present invention;
[0034] Figure 2 Schematic diagram of the conceptual structure of the image pyramid of the present invention;
[0035] Figure 3 This is a schematic diagram of the voting rules of the present invention;
[0036] Figure 4 is the voting domain under the scale of σ=3 of the present invention;
[0037] Figure 5 is the voting domain under the scale of σ=10 in the present invention;
[0038] Figure 6 The voting domain of the present invention at different scales of σ=50;
[0039] Figure 7 Schematic diagram of the voting process of the present invention;
[0040] Figure 8 This is a schematic diagram of two-dimensional tensor voting of the present invention;
[0041] Figure 9 Original image for example;
[0042] Figure 10 Tensor voting linear feature probability map for example;
[0043] Figure 11 Technical flow chart for the invention;
[0044] Figure 12 High angle fracture identification method based on diffusion equation for example;
[0045] Figure 13 Tensor voting further processing result for example;
[0046] Figure 14 Ant body calculation result for example;
[0047] Figure 15 Seismic data-raw ant body stack profile for example;
[0048] Figure 16 Seismic data-diffusion ridge stack profile for example;
[0049] Figure 17 Fracture system characterization technology flow processing result for example;
[0050] Figure 18 Sand body, river channel geological body boundary identification result and sand body distribution result in the work area for example. DETAILED DESCRIPTION
[0051] The fracture system characterization method aims to identify faults or boundaries of geological bodies, and comprises the following steps:
[0052] Step 1: After obtaining the stacked seismic data, first use the navigation pyramid technology for frequency division processing to obtain sub-band seismic data of different scales. At the same time, according to the needs, the different scale sub-band data is combined to highlight the purpose information. At the same time, it can improve the signal-to-noise ratio of seismic data, and further improve the calculation effect of conventional fracture attributes.
[0053] The technical principle is as follows: the technology is a multi-scale and multi-directional decomposition technology, which has translation invariance and direction controllability, which is beneficial to the accurate identification of complex underground discontinuity characteristics, and different scale combinations and direction filtering can be designed according to geological characteristics, further mining the hidden discontinuity characteristics in seismic attributes.
[0054] For example, the seismic data is divided into different scales, and the seismic data of different scales is combined to highlight the purpose information. At the same time, it can improve the signal-to-noise ratio of seismic data, and further improve the calculation effect of conventional fracture attributes. Figure 1The navigation pyramid seismic data processing process mainly includes the creation of pyramid structure (image decomposition) and the reconstruction of decomposed image, as shown in the figure. The image pyramid is an image decomposition technique to explain the multi-scale structure of image in multi-resolution, which decomposes information into a series of top-down, gradually changing in scale images.
[0055] As Figure 2 the pyramid goes from bottom to top, the scale of the constituent image changes from small to large. Small scale can reflect the detailed information of the image, and large scale can better reflect the macro information of the image. The pyramid decomposition is realized in the frequency domain, and the pyramid structure is realized by recursively calling the low-pass radial filter in the frequency domain, and then the band-pass filter image of different base direction is obtained by the direction controllable filter at each layer of the pyramid. Therefore, the image can be decomposed into sub-band information under different scales, and each layer can be decomposed into different directions to obtain sub-band images containing direction information.
[0056] The direction controllable filter is composed of a group of base filters, which has the property of direction rotation. The filter kernel of the direction controllable filter has controllability, and the processed result has the characteristics of low calculation amount and high precision. In seismic data, the direction controllable filter filters by using the direction of the event. The function expression of the controllable filter is as follows:
[0057]
[0058] In the formula, f θ (x,y) is the function of the controllable filter in the θ direction, which can be obtained by linear combination of the interpolation function k j (θ) in the θ direction and the base function, j is the number of decomposition layers, and N is the logarithm of the base function and the interpolation function.
[0059] In order to solve the problem of how to select the functions of k j (θ) and the number of filters, the rectangular coordinate system is converted into the polar coordinate system, in which φ = arg (x, y). The expression of f θ (x, y) in polar coordinates is as follows:
[0060]
[0061] In the formula, g i (r, φ) is the base function of f θ (r, φ).
[0062] The navigation pyramid structure combines the Laplacian pyramid data structure, commonly used in image processing, with steerable filters. The core of this method for seismic signal processing is to decompose the seismic signal into a steerable pyramid data structure using a multi-scale and multi-directional analysis method. During this decomposition process, steerable filters are used to extract geological information from each sub-band. The entire processing process primarily involves the creation of the pyramid structure and the reconstruction of the decomposed signal. This method can operate in different directions and at different scales, enabling clearer characterization and identification of subsurface geological bodies of diverse shapes and sizes compared to other methods.
[0063] Step 2: Based on the combined seismic data obtained in Step 1, a high-angle fracture identification technique based on the diffusion equation is used to detect fractures or identify their boundaries. This step can be a key step in the overall fault identification process. This method is characterized by its ability to extract detailed information within large fault zones. However, due to its focus on extracting fracture details, the calculated results can be visually fragmented, which can easily clutter the results interface.
[0064] The technical principle is as follows: In order to make the identified fault information more consistent with the original seismic data, this method introduces a consistency enhancement processing method based on technology one.
[0065] Highlighting the differences between faulted and non-faulted conditions, a fracture system treatment based on the diffusion equation is used:
[0066]
[0067] Where D(x) is the diffusion tensor of the analysis point, which is constructed by the eigenvectors of the seismic data structure tensor. The structure tensor T of seismic data can be constructed as the smoothed outer product of the seismic data gradient. The structure tensor T can be expressed by eigendecomposition as follows:
[0068] T=λ u uu T +λ v vv T +λ w ww T
[0069] Among them, λ u ,λ v ,λ w are the eigenvalues corresponding to the orthogonal eigenvectors u, v, and w respectively. If the mark λ u ≥λ v ≥λ wIf ≥ 0, the corresponding eigenvector u will be parallel to the direction of the most significant image variation, and the corresponding eigenvector w will be parallel to the direction of the least significant image variation, based on which the original seismic data is diffused.
[0070] The defined diffusion rate in the technique diffusion process is:
[0071]
[0072] Wherein, α is a threshold value for detecting discontinuity characteristics of the analysis point, d(x) is a directional derivative of the analysis point, and the directional derivative and the size of the threshold value are used to distinguish faults and noises and control the diffusion direction of the analysis point; when d(x) > α, it is considered that the place is a discontinuity feature and feature preservation is performed, otherwise it is noise and suppression is performed.
[0073] The basic principle is to perform voting in the parameter space to determine the shape of the object, and the local maximum value in the p-theta spectrum is used to determine. The basic flow is: ① calculate the p-theta energy spectrum of the image data; ② determine the parameters p and theta of the straight line according to the local maximum value in the p-theta energy spectrum; ③ finally determine the straight line according to the p and theta parameters.
[0074] p = xcos theta + ysin theta
[0075] Wherein, x, y are weight coefficients of the straight line equation. According to the geometric relationship, the direction vector along the straight line can be obtained, and the direction of the point on the non-straight line can be obtained by interpolation.
[0076]
[0077] Then the diffusion tensor of the diffusion only along the vector direction is
[0078] From the perspective of seismic data filtering, the filter should be stopped at the fault position, that is, the diffusion rate should be zero. However, the actually obtained diffusion rate does not present a zero value line along the fault at the fault position, but corresponds to many low value points. Therefore, it is necessary to filter these low value points along the fault direction. The above consistency enhancement processing can detect points on the same straight line and determine their linear direction, and then perform consistency enhancement processing on the diffusion rate according to the linear direction, so that the diffusion rate along the fault is more continuous and the direction is more consistent with the actual fault direction. At this time, the diffusion rate of the diffusion tensor constructed is:
[0079]
[0080] Wherein, μ u , μ vrepresents the diffusion rate in the u and v directions, H represents the filter guided by the diffusion rate when performing consistency enhancement processing, α is the threshold used to detect discontinuity features at the analysis point, and d(x) is the directional derivative of the analysis point. The relationship between the directional derivative and the threshold is used to distinguish between faults and noise, and to control the diffusion direction of the analysis point. When d(x)>α, the location is considered to be a discontinuity feature and is retained; otherwise, it is considered to be noise and is suppressed.
[0081] On this basis, further refinement and ridge-extraction processing of the diffusivity attributes can obtain fault identification results that are consistent with the original seismic data.
[0082] Step 3: Using a discontinuity feature extraction technique based on navigation pyramid filtering, we highlight discontinuities in the seismic data. This allows us to predict fault systems and enhance their identification accuracy. This step can reveal structural features of discontinuities, such as the general direction of the fault, but it cannot capture detailed information about the fault, such as the structural phase.
[0083] Step 4: The fracture detection results obtained in Step 2 are combined with the discontinuity feature extraction results obtained in Step 3 to obtain preliminary fracture detection results. However, the calculation results at this time have poor continuity and the fracture direction is unclear, which requires further processing and optimization.
[0084] Step 5: Use tensor voting technology to process the fusion results obtained in Step 4, effectively improving the continuity of the fault system and effectively removing the recognition artifacts caused by noise during the recognition process, making the fault direction and structure clearer.
[0085] The technical principle is: to enhance and hollow out the properties of the fracture body in a way that simulates the vision of geologists, enhance the continuity and integrity of the fracture system, thereby achieving the purpose of clear trunk and branches and realizing the optimization of fracture system characterization.
[0086] Tensor voting is a new and highly robust method for extracting prominent image features. It is a theory for inferring significant structural features, proposed by Guy of the University of Southern California based on the Gestalt principle of "the whole is greater than the sum of its parts." It utilizes tensor superposition to enhance the features to be extracted and interprets the voting results using the eigenvalue decomposition of tensor matrices. It can infer implicit structural features from point clouds with strong noise and outliers, visualizing information perceived by humans through machine algorithms. In this algorithm, domain information is aggregated through voting, while orientation information is estimated using vector field calculation rules to achieve inter-data communication. Tensor voting is used to mine seismic attribute features and enhance boundaries within seismic data. It can also be applied to understanding and skeletonizing geological structures in seismic data.
[0087] The first step of the tensor voting algorithm is to construct the tensor. In two-dimensional seismic data, let I(x, y) represent the seismic data, then The gradient vector of each pixel is represented by G(x, y), and the tensor is constructed using the gradient vector as shown in Equation 10:
[0088]
[0089] According to the second-order matrix, any second-order symmetric non-negative tensor can be decomposed as:
[0090]
[0091] The main eigenvalue λ1 is used as the input data for the voting process to improve the accuracy of the tensor information. According to the Gestalt principle, the voting strength received by the receiving point in the voting process is determined by the attenuation function, as shown in Equation 12:
[0092]
[0093] where, represents the arc length of OP; represents the curvature of the arc; σ is the voting scale factor, which determines the size of the voting domain; controls the degree of curvature attenuation. To avoid the amplitude characteristic disorder phenomenon of the attenuation function near the voting point, artificially set the range is set to 0.
[0094] As Figure 3 , let θ be the angle between the tangent of the curvature circle of the voter O and the straight line OP, and No and Np be the normal vectors of O and P points, respectively, then the voting operator V s (p) is defined as:
[0095]
[0096] The tensor voting scale σ is the only parameter in the entire voting process, as shown in Figures 4 to 6 , the horizontal and vertical coordinates represent the coordinates of the receiving point, and the brightness information represents the voting strength. When the voting scale σ is selected with a small value, the voting domain range is small, the curve smoothing degree is weak, and the local information is more prominent. As the value of the voting scale σ gradually increases, the voting domain range increases, and the curve smoothing degree becomes higher and higher. However, the larger the voting scale, the more the calculation amount in the processing process, and it presents an exponential growth.
[0097] The voting process is as follows Figure 7As shown, the center of the voting area is placed at the voting point. Each point accumulates the votes of the neighborhood, calculates the number and size of votes for each pixel, and finally forms a new tensor representation at each pixel. The expression is as follows:
[0098]
[0099] Among them, V represents the cumulative vote, V s (p) is the received vote tensor, and K represents the number of voters in the receiver’s neighborhood.
[0100] The tensor after voting can be decomposed into:
[0101]
[0102] in, is a stick tensor, λ1-λ2 represents the significance of the stick tensor, is the spherical tensor, and λ2 represents the significance of the circular tensor. If λ1-λ2>λ2, the dominant tensor is the rod tensor, and the point is most likely located on the curve with e1 as the normal. If λ1≈λ2>0, it means that the point is likely located at the intersection of multiple curves, or that the point is located in a region where all directions are equally likely. When both λ1 and λ2 are extremely small, it means that the point is likely an outlier.
[0103] During the tensor voting process, the neighboring point p i Cast your own vote (tensor matrix) to point p, the two tensor components vote separately, and point p accumulates the stick tensor votes and ball tensor votes from each neighboring point to infer its own structural characteristics, such as Figure 8 shown.
[0104] After the voting is completed, tensor decomposition is performed to obtain the geological body boundary feature map. Further voting is carried out on this basis, and finally the structural and geological unit boundary feature map will be obtained. Figure 9 and Figure 10 This is the tensor voting effect diagram for earthquake attributes.
[0105] Step 6: Combine the results obtained in Step 5 with the drilling loss data and regional geological analysis results to analyze the geological significance of the calculation results and evaluate the reliability of the calculation results. Ultimately, a multi-scale fault system that conforms to geological understanding is formed to guide actual production.
[0106] The technical flow chart of the present invention is as follows: Figure 11 Analysis of the technical process processing effect of the present invention:
[0107] 1. Fracture identification (compared with ant body technology)
[0108] Based on the processing of seismic data by the navigation pyramid, the fault attributes are identified and the results are processed using the high-angle fracture identification technology based on the diffusion equation. Figure 12 (Stpe3 results).
[0109] Then the tensor voting technique is used to perform structure enhancement processing, and the results are as follows Figure 13 (Stpe5 results).
[0110] The processing results of the same data ant body are as follows Figure 14 .
[0111] The cross-section of the ant body calculation results and the technical process calculation results proposed by the present invention is as follows: Figure 15 and Figure 16 ,It can be seen that the faults identified by this ,technique process have stronger linearity on the profile and are more ,conformable to actual geological knowledge.
[0112] 2. Application effect of sand body and river geological body boundary identification, such as Figure 17 .
[0113] The result is superimposed with the sand body distribution result of the work area, such as Figure 18 .
[0114] It can be seen from the figure that the attribute recognition results obtained using the technical process proposed in the present invention are highly consistent with the sand body identification results, and the boundaries of multi-period and single-period river channels are clearly identified.
Claims
1. A fracture system characterization method, characterized in that: The following steps are involved: Step 1: After obtaining the post-stack seismic data, the navigation pyramid technique is first used to perform frequency division processing to obtain sub-band seismic data of different scales. At the same time, the sub-band data of different scales are combined as needed to highlight the target information. Step 2: Based on the combined seismic data obtained in Step 1, the high-angle fracture identification technology based on the diffusion equation is used to perform fracture detection or boundary identification on the seismic data. Step 3: Using the discontinuity feature extraction technology based on the navigation pyramid filtering technology, the discontinuity features in the seismic data are highlighted, and the fault system is predicted based on this; Step 4: Fuse the fracture detection results obtained in Step 2 with the discontinuity feature extraction results obtained in Step 3 to obtain preliminary fracture detection results. Step 5: Use tensor voting technology to process the fusion results obtained in Step 4 to improve the continuity of the fracture system and remove the recognition artifacts caused by noise during the recognition process.
2. The fracture system characterization method according to claim 1, characterized in that: In Step 1, the navigation pyramid seismic data processing process includes the creation of the pyramid structure and the reconstruction of the decomposed image; The navigation pyramid structure combines the Laplacian pyramid data structure in the field of image processing with the steerable filter; The navigation pyramid processes seismic signals by using a multi-scale and multi-directional analysis method to decompose the seismic signals into a direction-controllable pyramid data structure. During the decomposition process, the information related to the geological body in each sub-band is extracted through a direction-controllable filter.
3. The fracture system characterization method according to claim 1, characterized in that: In Step 2, in order to make the identified fault information more consistent with the original seismic data, a consistency enhancement processing method is used; Highlighting the differences between faulted and non-faulted conditions, a fracture system treatment based on the diffusion equation is used: Where D(x) is the diffusion tensor of the analysis point, which is constructed by the eigenvector of the seismic data structure tensor; g(x; t) is the gradient value of the calculation point; the structure tensor T of the seismic data is constructed as the smoothed outer product of the seismic data gradient. The eigendecomposition of the structure tensor T is expressed as: T=λ u oh T +λ v vv T +λ w wow T Among them, λ u ,λ v ,λ w are the eigenvalues corresponding to the orthogonal eigenvectors u, v, and w respectively; if the mark λ u ≥λ v ≥λ w ≥0, the corresponding eigenvector u will be parallel to the direction where the u image changes most significantly, and the corresponding eigenvector w will be parallel to the direction where the u image changes least significantly, based on which the original seismic data is diffused; The defined diffusion rate in the diffusion process is: Where α is the threshold used to detect discontinuity features at the analysis point, and d(x) is the directional derivative of the analysis point. The relationship between the directional derivative and the threshold is used to distinguish between faults and noise, and to control the diffusion direction of the analysis point. When d(x) > α, the point is considered to be a discontinuity feature and is retained; otherwise, it is considered to be noise and is suppressed. Voting is performed in parameter space to determine the shape of the object, which is determined by the local maximum in the ρ-θ spectrum. The basic process is: ① Calculate the ρ-θ energy spectrum of the image data; ② Determine the parameters ρ and θ of the line based on the local maximum in the ρ-θ energy spectrum; ③ Finally, determine the line based on the ρ and θ parameters. ρ=x cosθ+y sinθ Where x and y are the weight coefficients of the line equation. The direction vector along the line is obtained based on the geometric relationship, and the direction of the point not on the line is solved by interpolation. The diffusion tensor that diffuses only along this vector direction is The obtained diffusivity does not present a zero-value line along the fault, but corresponds to many low-value points. These low-value points are filtered along the fault direction. The consistency enhancement process detects points on the same line and determines their linear direction. The diffusivity is then subjected to consistency enhancement based on the linear direction. The diffusivity of the constructed diffusion tensor is: Among them, μ u 、μ v represents the diffusion rate in the u and v directions, H represents the filter guided by the diffusion rate when performing consistency enhancement processing, α is the threshold used to detect discontinuity features at the analysis point, and d(x) is the directional derivative of the analysis point. The relationship between the directional derivative and the threshold is used to distinguish between faults and noise, and to control the diffusion direction of the analysis point. When d(x)>α, the location is considered to be a discontinuity feature and is retained; otherwise, it is considered to be noise and is suppressed. The diffusivity attributes are further refined and ridge-extracted to obtain fault identification results that are consistent with the original seismic data.
4. The fracture system characterization method according to claim 1, characterized in that: In Step 5, the properties of the fracture body are enhanced and hollowed out in a way that simulates the vision of geologists, thereby enhancing the continuity and integrity of the fracture system, thereby achieving the purpose of clear trunk and branches and realizing the optimization of the fracture system characterization.