A seismic image recognition method, system and device based on random theory

Abstract

The present invention relates to a seismic image recognition method, system and device based on random theory, and the method includes: calculating the seismic spectral moment element matrix of the second-order energy spectral moment of the seismic image; filtering the seismic spectral moment element matrix to obtain the filtered seismic spectral moment element matrix; using the filtered seismic spectral moment element matrix to construct spectral moment statistical parameters for characterizing the local energy in the seismic image; performing boundary extraction on the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image. The present invention suppresses the interference wave, scattered wave and noise signals well by denoising the seismic image spectral moment element matrix, and can improve the recognition ability of the seismic spectral moment image; constructing new spectral moment statistical parameters of the local energy of the seismic image through the seismic spectral moment element matrix, and the new spectral moment statistical parameters can better identify the boundary of the tiny abnormal body in the seismic image; boundary extraction can effectively enhance the boundary of the abnormal body, and the extracted boundary is more continuous and accurate.

Description

Technical Field

[0001] The present invention relates to the field of seismic image recognition, and in particular to a seismic image recognition method, system and device based on random theory. Background Art

[0002] Using reflection seismic images to identify underground spatial structures will help in the subsequent study of the seismic state of the deep crust. Since seismic data are inevitably interfered by noise during field collection, the signal-to-noise ratio of the seismic image is affected, and the limitations of conventional processing methods lead to the generation of subtle errors in the seismic image. The existing seismic spectral moment method is an effective seismic image recognition method; the seismic spectral moment method calculates the spectral moment elements in different directions of the seismic image, constructs seismic spectral moment parameters such as statistical invariant spectral moment sum and notch coefficient, and thus identifies the edge features of the anomaly in the seismic image. However, some anomalies in the seismic image have weak edge energy and are accompanied by strong noise. The seismic spectral moment method is difficult to separate the edges of the anomalies blurred by noise in the seismic image, resulting in poor recognition of the complex underground structural area, which cannot meet the urgent needs of studying the seismic state of the crust. Summary of the invention

[0003] The present invention provides a seismic image recognition method, system and device based on random theory, which can suppress the noise at the edge of the seismic image anomaly, enhance the contour clarity of the seismic image anomaly, effectively identify the underground space structure, and provide support for the study of crustal seismic activity.

[0004] The technical solution of the present invention to solve the above technical problem is as follows: a seismic image recognition method based on random theory, comprising:

[0005] S1, acquiring a seismic image, and calculating a seismic spectral moment element matrix of a second-order energy spectral moment of the seismic image;

[0006] S2, filtering the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix;

[0007] S3, constructing spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix;

[0008] S4, performing boundary extraction on the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image.

[0009] Based on the above technical solution, the present invention can also be improved as follows.

[0010] Further, the S1 is specifically:

[0011] Acquire a seismic image, and calculate a spectral moment element of a local second-order energy spectral moment of each point in the seismic image;

[0012] The spectral moment element set of the local second-order energy spectral moment of all points in the seismic image is used to form the seismic spectral moment element matrix;

[0013] Among them, the seismic spectral matrix includes a first seismic spectral matrix, a second seismic spectral matrix and a third seismic spectral matrix; the spectral elements in the first seismic spectral matrix represent the variance of the slope in the lateral change direction of the seismic wave; the spectral elements in the second seismic spectral matrix represent the variance of the slope in the longitudinal change direction of the seismic wave; the spectral elements in the third seismic spectral matrix represent the covariance of the slope in the lateral and longitudinal change directions of the seismic wave.

[0014] Furthermore, the discrete spatial domain convolution form of the spectral element in the seismic spectral element matrix is ​​expressed as:

[0015]

[0016] Among them, m 20 represents the spectral element in the first seismic spectral element matrix; m 02 represents the spectral element in the second seismic spectral element matrix; m 11 represents the spectral moment element in the third seismic spectral moment element matrix; S(x j ,y i ) represents the local facet of the S wave field intercepted by the sliding window in the seismic image; j=1, 2, ..., n; i=1, 2, ..., l; n and l are the numbers of calculation points in the x-direction and y-direction in the sliding window, respectively; specifically, the x-direction represents the lateral change direction of the seismic wave, and the y-direction represents the longitudinal change direction of the seismic wave.

[0017] Furthermore, in S2, a two-dimensional Gaussian smoothing filtering method is specifically used to filter the seismic spectrum element matrix.

[0018] Furthermore, the formula for filtering the seismic spectrum element matrix using a two-dimensional Gaussian smoothing filter method is:

[0019] M′(x,y)=M(x,y)*G(x,y,δ g );

[0020] Wherein, M′(x, y) represents the filtered seismic spectrum element matrix; M(x, y) represents the seismic spectrum element matrix; G(x, y, δ g ) represents a two-dimensional Gaussian smoothing filter function, and δ g represents the standard deviation; x and y represent the number of points of the seismic image in the lateral change direction of the seismic wave and the longitudinal change direction of the seismic wave, respectively.

[0021] Further, in S3, the spectral moment statistical parameters include a first spectral moment statistical parameter, a second spectral moment statistical parameter and a third spectral moment statistical parameter; the first spectral moment statistical parameter is used to characterize the sum of energy changes in all directions of the seismic image; the second spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is most intense; the third spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is the largest after eliminating the background energy.

[0022] Further, the seismic spectral moment element matrix includes a first seismic spectral moment element matrix, a second seismic spectral moment element matrix and a third seismic spectral moment element matrix; the spectral moment elements in the first seismic spectral moment element matrix represent the variance of the slope in the lateral change direction of the seismic wave; the spectral moment elements in the second seismic spectral moment element matrix represent the variance of the slope in the longitudinal change direction of the seismic wave; the spectral moment elements in the third seismic spectral moment element matrix represent the covariance of the slope in the lateral change direction and the longitudinal change direction of the seismic wave;

[0023] The filtered seismic spectral matrix comprises a first filtered seismic spectral matrix, a second filtered seismic spectral matrix and a third filtered seismic spectral matrix; the first filtered seismic spectral matrix is ​​obtained by filtering the first seismic spectral matrix; the second filtered seismic spectral matrix is ​​obtained by filtering the second seismic spectral matrix; the third filtered seismic spectral matrix is ​​obtained by filtering the third seismic spectral matrix;

[0024] The expression of the spectral moment statistical parameter is:

[0025] M s =m′ 20 +m′ 02 +m′ 11 ;

[0026]

[0027] Among them, M s represents the first spectral moment statistical parameter; Δ 22 represents the second spectral moment statistical parameter; Λ2 represents the third spectral moment statistical parameter; m′ 20 represents the filtered spectral element in the first filtered seismic spectral element matrix; m′ 02 represents the filtered spectral element in the second filtered seismic spectral element matrix; m′ 11 Represents the filtered spectral moment element in the third filtered seismic spectral moment element matrix.

[0028] Further, in S4, the formula for performing boundary extraction on the spectral moment statistical parameter is:

[0029]

[0030] G=G x +G y ;

[0031] Wherein, F(x,y) represents the spectral moment statistical parameter; G x represents the gradient result of the spectral moment statistical parameter in the lateral change direction of the seismic wave; G y represents the gradient result of the spectral moment statistical parameter in the longitudinal change direction of the seismic wave; G represents the boundary result of the spectral moment statistical parameter.

[0032] Based on the above-mentioned seismic image recognition method based on random theory, the present invention also provides a seismic image recognition system based on random theory.

[0033] A seismic image recognition system based on random theory, comprising:

[0034] A seismic spectrum moment element matrix calculation module, which is used to obtain a seismic image and calculate a seismic spectrum moment element matrix of a second-order energy spectrum moment of the seismic image;

[0035] A filtering module, which is used to filter the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix;

[0036] A spectral moment statistical parameter construction module, which is used to construct spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix;

[0037] The boundary extraction module is used to extract the boundary of the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image.

[0038] Based on the above-mentioned seismic image recognition method based on random theory, the present invention also provides a seismic image recognition device based on random theory.

[0039] A seismic image recognition device based on random theory comprises a processor, a memory and a program code stored in the memory. When the program code is executed by the processor, the seismic image recognition method based on random theory as described above is implemented.

[0040] The beneficial effects of the present invention are as follows: in a seismic image recognition method, system and device based on random theory of the present invention, by using two-dimensional Gaussian filtering to perform denoising preprocessing on the seismic image spectral moment element matrix, the interference wave, scattered wave and noise signals are well suppressed, and the recognition ability of the seismic spectral moment image can be improved; at the same time, the present invention constructs new spectral moment statistical parameters of the local energy of the seismic image through the seismic spectral moment element matrix, and the new spectral moment statistical parameters can better identify the boundaries of tiny anomalies in the seismic image than the existing spectral moment statistical parameters; in addition, the boundary extraction of the present invention can effectively enhance the boundary of the anomaly body, and the extracted boundary is more continuous and accurate, which is conducive to the characterization of underground anomalies and subsequent crust structure and state research. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of a seismic image recognition method based on random theory of the present invention;

[0042] Figure 2 This is a schematic diagram of a 2D pre-stack migration seismic image;

[0043] Figure 3 is the spectral moment element m 20 Schematic diagram of

[0044] Figure 4 is the spectral moment element m 02 Schematic diagram of

[0045] Figure 5 is the spectral moment element m 11 Schematic diagram of

[0046] Figure 6 is the original spectral moment statistical parameter M s Schematic diagram of

[0047] Figure 7 It is a schematic diagram of the original spectral moment statistical parameter Δ2;

[0048] Figure 8 It is a schematic diagram of the original spectral moment statistical parameter Λ;

[0049] Fig. 9 is the spectral element m′ after filtering 20 Schematic diagram of

[0050] Fig.10 is the spectral element m′ after filtering 02 Schematic diagram of

[0051] Fig.11 is the spectral element m′ after filtering 11 Schematic diagram of

[0052] Fig.12 is the new spectral moment statistical parameter M s Schematic diagram of

[0053] Fig.13 is the new spectral moment statistical parameter Δ 22 Schematic diagram of

[0054] Fig.14 It is a schematic diagram of the new spectral moment statistical parameter Λ2;

[0055] Fig.15 is the new spectral moment statistical parameter M s Schematic diagram of the boundary result G;

[0056] Fig.16 is the new spectral moment statistical parameter Δ 22 Schematic diagram of the boundary result G;

[0057] Fig.17 Schematic diagram of the boundary result G of the new spectral moment statistical parameter Δ2;

[0058] Fig.18 It is a structural block diagram of a seismic image recognition system based on random theory of the present invention;

[0059] Fig.19 This is a structural block diagram of a seismic image recognition device based on random theory in the present invention. DETAILED DESCRIPTION

[0060] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0061] Embodiment 1:

[0062] like Figure 1 As shown, a seismic image recognition method based on random theory includes:

[0063] S1, acquiring a seismic image, and calculating a seismic spectral moment element matrix of a second-order energy spectral moment of the seismic image;

[0064] S2, filtering the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix;

[0065] S3, constructing spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix;

[0066] S4, performing boundary extraction on the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image.

[0067] The seismic image recognition method of the present invention calculates the seismic image spectral moment elements, filters the seismic image spectral moment elements, obtains the filtered seismic spectral moment element matrix, constructs new spectral moment statistical parameters, obtains the seismic spectral moment statistical parameter matrix, performs edge calculation (i.e., boundary extraction) on the seismic spectral moment statistical parameter matrix, and obtains edge results. The method effectively suppresses noise, better identifies fuzzy boundaries, more accurately depicts underground structural features, enhances the boundaries of underground anomalies, and provides effective maps for subsequent underground structure and crust state research.

[0068] In some embodiments:

[0069] The S1 is specifically:

[0070] S11, obtaining a seismic image S(x,y); wherein x represents the number of points of the seismic image in the lateral change direction of the seismic wave, that is, the CDP (common depth point) number of the survey line direction of the offset seismic image; y represents the number of points of the seismic image in the longitudinal change direction of the seismic wave, that is, the travel time of the offset seismic data or the depth after time-depth conversion;

[0071] S12, calculating the spectral element of the local second-order energy spectral moment of each point in the seismic image; specifically, in the seismic image, a sliding window is selected to calculate the local surface element S(x j ,y i )’s second-order energy spectral moment;

[0072] S13, forming the seismic spectral moment element matrix by taking the spectral moment element set of the local second-order energy spectral moments of all points in the seismic image.

[0073] In S1, the seismic spectral element matrix includes a first seismic spectral element matrix, a second seismic spectral element matrix and a third seismic spectral element matrix; the spectral elements in the three seismic spectral element matrices are three spectral element of the second-order spectral moment of the S wave field, which represent different physical meanings of the seismic wave field; specifically, the spectral element in the first seismic spectral element matrix represents the variance of the slope in the lateral change direction of the seismic wave; the spectral element in the second seismic spectral element matrix represents the variance of the slope in the longitudinal change direction of the seismic wave; the spectral element in the third seismic spectral element matrix represents the covariance of the slope in the lateral change direction and the longitudinal change direction of the seismic wave. The discrete spatial domain convolution form of the spectral element in the seismic spectral element matrix is ​​expressed as:

[0074]

[0075] Among them, m 20 represents the spectral element in the first seismic spectral element matrix; m 02 represents the spectral element in the second seismic spectral element matrix; m 11represents the spectral moment element in the third seismic spectral moment element matrix; S(x j ,y i ) represents the local facet of the S wave field intercepted by the sliding window in the seismic image; j=1, 2, ..., n; i=1, 2, ..., l; n and l are the numbers of calculation points in the x-direction and y-direction in the sliding window, respectively; specifically, the x-direction represents the direction of the seismic survey line, that is, the lateral change direction of the seismic wave; the y-direction represents the propagation direction of the seismic wave from the surface to the deep underground, that is, the longitudinal change direction of the seismic wave.

[0076] In some embodiments:

[0077] In S2, a two-dimensional Gaussian smoothing filter method is specifically used to filter the seismic spectrum element matrix.

[0078] The formula for filtering the seismic spectrum element matrix using the two-dimensional Gaussian smoothing filter method is:

[0079] M ′ (x,y)=M(x,y)*G(x,y,δ g );

[0080] Among them, M ′ (x, y) represents the filtered seismic spectrum matrix; M (x, y) represents the seismic spectrum matrix; G (x, y, δ g ) represents a two-dimensional Gaussian smoothing filter function, and δ g represents the standard deviation, which determines the size of the suppressed noise; x and y represent the number of points of the seismic image in the lateral change direction of the seismic wave and the longitudinal change direction of the seismic wave, respectively.

[0081] Since the seismic spectral matrix in the present invention includes the first seismic spectral matrix, the second seismic spectral matrix and the third seismic spectral matrix, the filtered seismic spectral matrix also includes three corresponding filtered seismic spectral matrixes, which are the first filtered seismic spectral matrix, the second filtered seismic spectral matrix and the third filtered seismic spectral matrix; the first filtered seismic spectral matrix is ​​obtained by filtering the first seismic spectral matrix; the second filtered seismic spectral matrix is ​​obtained by filtering the second seismic spectral matrix; the third filtered seismic spectral matrix is ​​obtained by filtering the third seismic spectral matrix.

[0082] Specifically, when M(x,y) represents the first seismic spectrum element matrix, then M ′ (x, y) represents the first filtered seismic spectrum element matrix; when M(x, y) represents the second seismic spectrum element matrix, then M ′(x, y) represents the second filtered seismic spectrum element matrix; when M(x, y) represents the third seismic spectrum element matrix, then M ′ (x,y) represents the seismic spectrum element matrix after the third filtering.

[0083] In some embodiments:

[0084] In S3, the spectral moment statistical parameters include a first spectral moment statistical parameter, a second spectral moment statistical parameter and a third spectral moment statistical parameter. The above three different spectral moment statistical parameters represent the characteristics of local seismic image energy changes; specifically, the first spectral moment statistical parameter is used to characterize the sum of energy changes in all directions of the seismic image; the second spectral moment statistical parameter is used to characterize the place where the seismic image energy changes most strongly; the third spectral moment statistical parameter is used to characterize the place where the seismic image energy changes the most after eliminating the background energy.

[0085] The expression of the spectral moment statistical parameter is:

[0086] M s =m′ 20 +m′ 02 +m′ 11 ;

[0087]

[0088] Among them, M s represents the first spectral moment statistical parameter; Δ 22 represents the second spectral moment statistical parameter; Λ2 represents the third spectral moment statistical parameter; m′ 20 represents the filtered spectral element in the first filtered seismic spectral element matrix; m′ 02 represents the filtered spectral element in the second filtered seismic spectral element matrix; m′ 11 Represents the filtered spectral moment element in the third filtered seismic spectral moment element matrix.

[0089] In some embodiments:

[0090] In S4, the formula for performing boundary extraction on the spectral moment statistical parameter is:

[0091]

[0092] G=G x +G y ;

[0093] Wherein, F(x,y) represents the spectral moment statistical parameter; G x represents the gradient result of the spectral moment statistical parameter in the lateral change direction of the seismic wave; G yrepresents the gradient result of the spectral moment statistical parameter in the longitudinal change direction of the seismic wave; G represents the boundary result of the spectral moment statistical parameter.

[0094] Since the spectral moment statistical parameters in the present invention are three, namely, the first spectral moment statistical parameter, the second spectral moment statistical parameter and the third spectral moment statistical parameter; therefore, the boundary extraction of the spectral moment statistical parameters is to extract the boundaries of the first spectral moment statistical parameter, the second spectral moment statistical parameter and the third spectral moment statistical parameter respectively. When F(x, y) represents the first spectral moment statistical parameter, G x It represents the gradient result of the first spectral moment statistical parameter in the lateral change direction of the seismic wave, G y represents the gradient result of the first spectral moment statistical parameter in the longitudinal direction of the seismic wave, and G represents the boundary result of the first spectral moment statistical parameter; when F(x,y) represents the second spectral moment statistical parameter, G x G represents the gradient result of the second spectral moment statistical parameter in the lateral change direction of the seismic wave. y represents the gradient result of the second spectral moment statistical parameter in the longitudinal direction of the seismic wave, and G represents the boundary result of the second spectral moment statistical parameter; when F(x,y) represents the third spectral moment statistical parameter, G x G represents the gradient result of the third spectral moment statistical parameter in the lateral change direction of the seismic wave. y It represents the gradient result of the third spectral moment statistical parameter in the longitudinal change direction of the seismic wave, and G represents the boundary result of the third spectral moment statistical parameter.

[0095] In order to verify the implementation effect of the present invention, this embodiment selects the following Figure 2 The actual 2D seismic profile image shown is a 2D pre-stack migration seismic image. Figure 2 It can be seen from the seismic image shown that due to the interference of random noise, the boundaries of the underground anomaly bodies are blurred, and it is difficult to identify the boundaries of the anomaly bodies in weak signal areas.

[0096] Through the actual two-dimensional seismic image, a 3×3 sliding window is selected to calculate the spectral moment element. Figure 3 to Figure 5 They are spectral moment elements m 20 、m 02 and m 11 The result is Figure 3 Describes the lateral anomaly boundaries of the seismic image. Figure 4 The vertical anomaly boundary of the seismic image is shown. Figure 5 The abnormal boundaries in all directions of the seismic image are depicted. The boundaries of some abnormal bodies in the three component images are blurred, and the abnormal body morphology of weak signals is difficult to identify due to noise interference.

[0097] The spectral element m 20 、m 02 and m 11Perform 3×3 sliding window calculations on the two-dimensional Gaussian filter to obtain the filtered spectral element m′ 20 , m′ 02 and m′ 11 like Figures 9 to 11 As shown in Figure 2, it can be seen that the two-dimensional Gaussian filtering effectively eliminates the random noise in the spectral element image, but the morphological boundary of the abnormal body with weak signal is relatively fuzzy.

[0098] By spectral moment element m 20 、m 02 and m 11 The constructed original spectral moment statistical parameter M s , Δ 22 and Λ2 as Figure 6 to Figure 8 shown. Figure 6 is the original spectral moment statistical parameter M s It can be seen that this parameter identifies the main abnormal body morphology, but due to noise interference, the morphology of the abnormal body with weak signals is not identified, and only the longitudinal boundary of the abnormal body is identified, and the horizontal boundary recognition is poor. Figure 7 The statistical parameter Δ2 of the original spectral moment is used to identify the morphology of the main anomaly, but it is difficult to distinguish the boundary features of noise and weak signals. Figure 8 is the statistical parameter Λ2 of the original spectral moment. Although more anomalies are identified, including some anomalies with weak signals, the identified boundary features are relatively fuzzy.

[0099] The new spectral moment statistical parameter M constructed by the filtered spectral moment element s , Δ 22 and Λ2 as Figure 12 to Figure 14 shown. Fig.12 is the new spectral moment statistical parameter M s It can be seen that this parameter identifies the main abnormal body morphology. While identifying the longitudinal boundary of the abnormal body, it also identifies the lateral boundary and effectively suppresses the interference of noise. It also identifies some abnormal body boundaries with weak signals. The abnormal boundary depicted is better than the original spectral moment statistical parameter M. s Closer to reality. Fig.13 is the new spectral moment statistical parameter Δ 22 , effectively identifying the contour boundary of the anomaly, and well identifying the boundary of the anomaly with weak signals. Fig.14 The new spectral moment statistical parameter Λ2 is used to more precisely identify the boundary of the anomaly, and the contour boundary of the anomaly is closer to the actual situation. Although the new spectral moment statistical parameter can well identify the boundary of the anomaly in the seismic image and suppress the noise, the contour boundary feature of the anomaly is not prominent, so the present invention combines the edge operator with the seismic image spectral moment method.

[0100] The new spectral moment statistical parameter (M s , Δ 22, Λ2) to extract the boundary and obtain the result image after the boundary enhancement of the seismic image anomaly body (G M , G Δ , G Δ )like Figure 15 to Figure 17 As shown in the figure, the new spectral moment statistical parameters can more clearly characterize the boundary features of the anomaly body in the seismic image while retaining the boundary of the weak signal anomaly body.

[0101] Embodiment 2:

[0102] Based on the above-mentioned seismic image recognition method based on random theory, the present invention also provides a seismic image recognition system based on random theory.

[0103] like Fig.18 As shown, a seismic image recognition system based on random theory includes:

[0104] A seismic spectrum moment element matrix calculation module, which is used to obtain a seismic image and calculate a seismic spectrum moment element matrix of a second-order energy spectrum moment of the seismic image;

[0105] A filtering module, which is used to filter the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix;

[0106] A spectral moment statistical parameter construction module, which is used to construct spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix;

[0107] The boundary extraction module is used to extract the boundary of the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image.

[0108] The specific functions of each module in the seismic image recognition system based on random theory of the present invention refer to the specific steps in the seismic image recognition method based on random theory of the present invention, which will not be repeated here.

[0109] Embodiment three:

[0110] Based on the above-mentioned seismic image recognition method based on random theory, the present invention also provides a seismic image recognition device based on random theory.

[0111] like Fig.19 As shown, a seismic image recognition device based on random theory includes a processor, a memory and a program code stored in the memory. When the program code is executed by the processor, the seismic image recognition method based on random theory as described above is implemented.

[0112] like Fig.19As shown, a seismic image recognition device based on random theory is presented in the form of a general computing device. The components of a seismic image recognition device based on random theory may include but are not limited to: a processor, a memory, and a bus connecting different system components. Among them, the processor, the interface, and the memory communicate with each other through the bus. The bus represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any bus structure in a variety of bus structures. The computing device for extracting geophysical fracture information typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computing device for extracting geophysical fracture information, including volatile and non-volatile media, removable and non-removable media. The processor executes various functional applications and data processing by running the program code stored in the memory. The processor may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs. The memory may include a computer system readable medium in the form of a volatile memory for storing the seismic image recognition method program, and the memory may include a high-speed RAM memory, or other removable / non-removable, volatile / non-volatile computer system storage medium. A seismic image recognition device based on random theory can communicate with external devices (such as keyboards, mice, displays, etc.), and can also communicate with any device that communicates with other computing devices (such as network cards, modems, network adapters, etc.).

[0113] In a seismic image recognition method, system and device based on random theory of the present invention:

[0114] 1. The present invention uses two-dimensional Gaussian filtering to perform denoising preprocessing on the spectral moment element matrix of the seismic image, which can improve the recognition ability of the seismic spectral moment image. For example, the existing seismic spectral moment method calculates the spectral moment element results of the seismic image (such as Figure 3 to Figure 5 There are false anomaly edges caused by interference waves and noise in the spectral moment element. False anomalies such as noise contained in the spectral moment element will also appear in the subsequent constructed seismic spectral moment statistical parameters (such as Figure 6 to Figure 8 Compared with the results of the above methods, the present invention suppresses interference waves and noise signals well (such as Figures 9 to 11 As shown in the figure), the actual boundary of the abnormal body can be effectively identified.

[0115] 2. The new spectral moment statistical parameters can more accurately characterize the actual boundaries of underground anomalies. For example, seismic images are obtained by seismic amplitude imaging. Since seismic waves propagate vertically and the underground is mainly distributed in horizontal layers, this often leads to strong longitudinal energy of seismic amplitude, which suppresses the lateral energy and affects the lateral recognition effect of seismic images. Figure 6 As shown in Figure 1, the spectral moment statistical parameters of seismic images are mainly horizontal, and lack vertical recognition. The new spectral moment statistical parameters proposed in this invention solve this problem, and the lateral boundaries of the anomaly body are also better displayed (such as Fig.12 As shown in Figure 2, the overall boundary contour of the underground anomaly is better identified. This is of great significance for accurately identifying underground anomalies and crustal structures.

[0116] 3. New spectral moment statistical parameters are constructed through the seismic spectral moment element matrix. The new spectral moment statistical parameters can better identify the boundaries of small anomalies in seismic images than existing statistical parameters. For example, existing seismic spectral moment parameters (such as Figure 7 and Figure 8 It is difficult to clearly identify the boundary features of small abnormal bodies. The new spectral moment statistical parameters (such as Fig.13 and Fig.14 The boundaries of small anomalies in seismic images can be clearly identified.

[0117] 4. Although the existing seismic spectral moment method can identify the shape of underground anomalies, it is not effective in describing the boundaries (e.g. Figure 7 , Figure 8 and Fig. 9 As shown in Figure 2), some weak boundaries are relatively fuzzy. The edge computing part (boundary extraction) proposed in the present invention can effectively enhance the abnormal body boundary (such as Figure 15 to Figure 17 The extracted boundary is more continuous and accurate, which is beneficial to the characterization of underground anomalies and the subsequent study of crustal structure and state.

[0118] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A seismic image recognition method based on random theory, characterized in that: include: S1, acquiring a seismic image, and calculating a seismic spectral moment element matrix of a second-order energy spectral moment of the seismic image; S2, filtering the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix; S3, constructing spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix; S4, performing boundary extraction on the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image; The S1 is specifically: Acquire a seismic image, and calculate a spectral moment element of a local second-order energy spectral moment of each point in the seismic image; The spectral moment element set of the local second-order energy spectral moment of all points in the seismic image is used to form the seismic spectral moment element matrix; Wherein, the seismic spectral moment element matrix includes a first seismic spectral moment element matrix, a second seismic spectral moment element matrix and a third seismic spectral moment element matrix; the spectral moment elements in the first seismic spectral moment element matrix represent the variance of the slope in the lateral change direction of the seismic wave; the spectral moment elements in the second seismic spectral moment element matrix represent the variance of the slope in the longitudinal change direction of the seismic wave; the spectral moment elements in the third seismic spectral moment element matrix represent the covariance of the slope in the lateral change direction and the longitudinal change direction of the seismic wave; In S2, a two-dimensional Gaussian smoothing filter method is specifically used to filter the seismic spectrum element matrix; In S3, the spectral moment statistical parameters include a first spectral moment statistical parameter, a second spectral moment statistical parameter and a third spectral moment statistical parameter; the first spectral moment statistical parameter is used to characterize the sum of energy changes in all directions of the seismic image; the second spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is the strongest; the third spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is the largest after eliminating the background energy; In S4, the formula for performing boundary extraction on the spectral moment statistical parameter is: G=G x +G y ; Wherein, F(x,y) represents the spectral moment statistical parameter; G x G represents the gradient result of the spectral moment statistical parameter in the lateral change direction of the seismic wave; y represents the gradient result of the spectral moment statistical parameter in the longitudinal change direction of the seismic wave; G represents the boundary result of the spectral moment statistical parameter.

2. The seismic image recognition method based on random theory according to claim 1, characterized in that: The discrete spatial domain convolution form of the spectral moment element in the seismic spectral moment element matrix is ​​expressed as: Among them, m 20 represents the spectral element in the first seismic spectral element matrix; m 02 represents the spectral element in the second seismic spectral element matrix; m 11 represents the spectral moment element in the third seismic spectral moment element matrix; S(x j ,y i ) represents the local facet of the S wave field intercepted by the sliding window in the seismic image; j=1, 2, ..., n; i=1, 2, ..., l; n and l are the numbers of calculation points in the x-direction and y-direction in the sliding window, respectively; specifically, the x-direction represents the lateral change direction of the seismic wave, and the y-direction represents the longitudinal change direction of the seismic wave.

3. The seismic image recognition method based on random theory according to claim 1, characterized in that: The formula for filtering the seismic spectrum element matrix using the two-dimensional Gaussian smoothing filter method is: M ′ (x,y)=M(x,y)*G(x,y,δ g ); Among them, M ′ (x, y) represents the filtered seismic spectrum matrix; M (x, y) represents the seismic spectrum matrix; G (x, y, δ g ) represents a two-dimensional Gaussian smoothing filter function, and δ g represents the standard deviation; x and y represent the number of points of the seismic image in the lateral change direction of the seismic wave and the longitudinal change direction of the seismic wave, respectively.

4. The seismic image recognition method based on random theory according to claim 1, characterized in that: The seismic spectral moment element matrix includes a first seismic spectral moment element matrix, a second seismic spectral moment element matrix and a third seismic spectral moment element matrix; the spectral moment elements in the first seismic spectral moment element matrix represent the variance of the slope in the lateral change direction of the seismic wave; the spectral moment elements in the second seismic spectral moment element matrix represent the variance of the slope in the longitudinal change direction of the seismic wave; the spectral moment elements in the third seismic spectral moment element matrix represent the covariance of the slope in the lateral change direction and the longitudinal change direction of the seismic wave; The filtered seismic spectrum moment element matrix includes a first filtered seismic spectrum moment element matrix, a second filtered seismic spectrum moment element matrix and a third filtered seismic spectrum moment element matrix; the first filtered seismic spectrum moment element matrix is ​​obtained by filtering the first seismic spectrum moment element matrix; The second filtered seismic spectrum moment element matrix is ​​obtained by filtering the second seismic spectrum moment element matrix; The third filtered seismic spectral moment element matrix is ​​obtained by filtering the third seismic spectral moment element matrix; The expression of the spectral moment statistical parameter is: M s =m′ 20 +m′ 02 +m′ 11 ; Among them, M s represents the first spectral moment statistical parameter; Δ 22 represents the second spectral moment statistical parameter; Λ2 represents the third spectral moment statistical parameter; m′ 20 represents the filtered spectral element in the first filtered seismic spectral element matrix; m′ 02 represents the filtered spectral element in the second filtered seismic spectral element matrix; m′ 11 Represents the filtered spectral moment element in the third filtered seismic spectral moment element matrix.

5. A seismic image recognition system based on random theory, characterized in that: include: A seismic spectrum moment element matrix calculation module, which is used to obtain a seismic image and calculate a seismic spectrum moment element matrix of a second-order energy spectrum moment of the seismic image; A filtering module, which is used to filter the seismic spectrum moment element matrix to obtain a filtered seismic spectrum moment element matrix; A spectral moment statistical parameter construction module, which is used to construct spectral moment statistical parameters for characterizing local energy in the seismic image using the filtered seismic spectral moment element matrix; A boundary extraction module, which is used to extract the boundary of the spectral moment statistical parameters to obtain the boundary of the abnormal body in the seismic image; The seismic spectrum moment element matrix calculation module is specifically used for: Acquire a seismic image, and calculate a spectral moment element of a local second-order energy spectral moment of each point in the seismic image; The spectral moment element set of the local second-order energy spectral moment of all points in the seismic image is used to form the seismic spectral moment element matrix; Wherein, the seismic spectral moment element matrix includes a first seismic spectral moment element matrix, a second seismic spectral moment element matrix and a third seismic spectral moment element matrix; the spectral moment elements in the first seismic spectral moment element matrix represent the variance of the slope in the lateral change direction of the seismic wave; the spectral moment elements in the second seismic spectral moment element matrix represent the variance of the slope in the longitudinal change direction of the seismic wave; the spectral moment elements in the third seismic spectral moment element matrix represent the covariance of the slope in the lateral change direction and the longitudinal change direction of the seismic wave; In the filtering module, a two-dimensional Gaussian smoothing filtering method is specifically used to filter the seismic spectrum element matrix; In the spectral moment statistical parameter construction module, the spectral moment statistical parameters include a first spectral moment statistical parameter, a second spectral moment statistical parameter and a third spectral moment statistical parameter; the first spectral moment statistical parameter is used to characterize the sum of energy changes in all directions of the seismic image; the second spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is the strongest; the third spectral moment statistical parameter is used to characterize the place where the energy change of the seismic image is the largest after the background energy is eliminated; In the boundary extraction module, the formula for performing boundary extraction on the spectral moment statistical parameters is: G=G x +G y ; Wherein, F(x,y) represents the spectral moment statistical parameter; G x G represents the gradient result of the spectral moment statistical parameter in the lateral change direction of the seismic wave; y represents the gradient result of the spectral moment statistical parameter in the longitudinal change direction of the seismic wave; G represents the boundary result of the spectral moment statistical parameter.

6. A seismic image recognition device based on random theory, characterized in that: The invention comprises a processor, a memory and a program code stored in the memory, wherein when the program code is executed by the processor, the seismic image recognition method based on random theory as claimed in any one of claims 1 to 4 is implemented.

Images