Five-dimensional seismic data interpolation method and system based on coordinates
By adopting NeRF theory and nuclear norm regularization of the profile-by-section interpolation method in five-dimensional seismic data processing, the problem of missing irregular trajectories and additional labeled data requirements is solved, and efficient and accurate five-dimensional seismic data reconstruction and modeling is achieved. robustness.
Patent Information
- Application Number
- CN202410953182.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-16
AI Technical Summary
The prior art is difficult to effectively deal with the problem of missing irregular trajectories in five-dimensional seismic data, and additional labeling of data is required for interpolation.
Unsupervised seismic data point-by-point interpolation model based on NeRF theory is adopted, combined with additional convolutional layers to modify the MLP framework, and a profile-by-section interpolation model is constructed, and a kernel norm regularization is introduced into the objective function to realize the reconstruction of five-dimensional seismic data. .
It significantly improves data processing efficiency, enables the reconstruction of five-dimensional seismic data faster and more accurately, and improves the noise immunity and robustness of the model, eliminating the need for additional labeled data.
Smart Images

Figure CN118939953A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of seismic signal processing, and in particular relates to a coordinate-based five-dimensional seismic data interpolation method and system. Background Art
[0002] In exploration seismology, fully sampled seismic data lays a solid foundation for subsequent tasks, such as deconvolution, synchronous source separation, and fault detection. However, in actual seismic data acquisition, due to physical or economic limitations, there are often missing traces in the observed data. These missing traces carry valuable geological information and may have an adverse effect on subsequent data processing and interpretation. Therefore, the reconstruction of seismic data has attracted widespread attention from academia and industry.
[0003] At present, deep learning technology has demonstrated significant potential in various fields, and seismic data interpolation is no exception. Compared with traditional methods, deep learning methods alleviate the need for a large number of manual parameter adjustments. In addition, because deep learning utilizes nonlinear activation functions, neural network models can more effectively capture nonlinear features in data, thereby enhancing information extraction and representation capabilities.
[0004] Various types of neural networks have been applied to seismic data interpolation, including autoencoder-based methods, generative adversarial networks, deep internal networks, and model-driven networks. However, existing work has mainly focused on low-dimensional data, such as two-dimensional or three-dimensional seismic data. Compared with these data, five-dimensional seismic data can better represent the directionality of geological features and the changes in seismic response at different offsets and azimuths, thereby providing a more comprehensive understanding of the subsurface properties. However, using deep learning to represent complex geological environments in five-dimensional data faces unique challenges because five-dimensional convolutions require more computational resources than three-dimensional convolutions. Existing methods take advantage of the characteristics of three-dimensional convolution operations to construct a five-dimensional convolutional neural network. This method uses synthetic five-dimensional seismic data as labels to train the network to enable it to have the ability to interpolate data. However, in order to apply this method more widely, further improvements in computational efficiency are necessary.
[0005] In recent years, there has been a great interest in coordinate-based neural representations in the field of computer vision, and the Neural Radiance Field (NeRF) method has been proposed. This method requires building a model including a Fourier Feature Map (FFM) module and a Multilayer Perceptron (MLP), which is able to encode the wavefield in a point-wise manner. The FFM module can convert the spatial coordinates into Fourier domain encoding, while the MLP maps the Fourier encoding to pixel values, such as color, density, or occupancy. By training the model with observed data points, the complete and continuous wavefield information will be encoded in the weight parameters of the MLP. Therefore, the missing data can be reconstructed by querying the values corresponding to each coordinate.
[0006] There is a coordinate-based seismic shot point interpolation method that uses an MLP network to encode the seismic wave field and queries the network at unobserved shot point locations to recover the lost shot point set; however, this method still has shortcomings. It mainly solves the problem of incomplete shot inspection sets, but does not address the challenge of missing irregular trajectories in high-dimensional data. Summary of the invention
[0007] The technical problem to be solved by the present invention is to provide a coordinate-based five-dimensional seismic data interpolation method and system for addressing the deficiencies in the above-mentioned prior art, so as to solve the technical problem that the prior art method requires additional labeled data.
[0008] The present invention adopts the following technical solutions:
[0009] A coordinate-based five-dimensional seismic data interpolation method comprises the following steps:
[0010] S1. Construct an unsupervised point-by-point interpolation model for seismic data based on NeRF theory;
[0011] S2, modifying the MLP framework with additional convolutional layers to build a profile-by-profile interpolation model based on the constructed unsupervised seismic data point-by-point interpolation model;
[0012] S3, introduce nuclear norm regularization and construct the objective function of the profile-by-profile interpolation model;
[0013] S4. Read the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
[0014] Preferably, the objective function of the unsupervised seismic data point-by-point interpolation model is:
[0015]
[0016] Among them, v1, v2, v3, v4, v5 are the five-dimensional coordinates of seismic data, D obs is the five-dimensional seismic data to be solved, θ is the parameter weight of MLP, M θ is the function of MLP mapping coordinates to complete seismic data, Represents the mean square error of the matrix.
[0017] Preferably, the five-dimensional seismic data D obs for:
[0018]
[0019] in, is the Hadamard product, P is the sampling matrix, and Z is the complete seismic data.
[0020] Preferably, the elements in the sampling matrix P satisfy the following conditions:
[0021]
[0022] Among them, d v1,v2,v3,v4,v5 is the five-dimensional seismic data D obs The elements with index v1,v2,v3,v4,v5 in .
[0023] 5. The coordinate-based five-dimensional seismic data interpolation method according to claim 1, characterized in that the profile-by-profile interpolation model includes an encoder and a decoder, the encoder includes a predefined single FFM layer γ(v), and an encoder based on the MLP prototype, the encoder converts the coordinate v into a corresponding latent variable Then, the convolutional network-based decoder is transformed from the variable Reconstruct the complete seismic profile using N τ Represents a combination of an encoder and a decoder.
[0024] Preferably, a single FFM layer γ(v) is specifically:
[0025] γ K (v)=[cos(w1v1),sin(w1v1),cos(w1v2),sin(w1v2),…,cos(w K v n ),sin(w K v n )] T
[0026] Where K represents the total number of components, and the i-th frequency component is linearly sampled by w i =π2 i Indicates that v represents any coordinate normalized within [0, 1].
[0027] Preferably, for seismic data that does not use conventional time difference correction, the objective function is defined as:
[0028]
[0029] Among them, S x and S y are the total lengths of the inline axis and the cross-line axis in the source, R x is the total length of the receiver internal shaft, r x is the index of the receiver on the inline axis, P is the sampling matrix;
[0030] For the common midpoint seismic data using NMO technique, the objective function is defined as:
[0031]
[0032] Among them, M x and M y are the maximum lengths of the inline axis and the cross axis respectively, U is the tensor expansion method, is the corresponding learning goal, M θ is the function of MLP that maps coordinates to complete seismic data.
[0033] Preferably, the objective function of the profile-by-profile interpolation model is:
[0034]
[0035] Among them, λ is a hyperparameter used to balance the fidelity and regularization terms in the objective function, ‖‖ * is the nuclear norm, M x and M y are the maximum lengths of the inline axis and the cross axis, M θ is the function of MLP mapping coordinates to complete seismic data, U is the tensor expansion method, (P:,m x ,m y ,:,:) is a specific part selected from the sampling matrix, is the corresponding learning objective, and F is the Frobenius norm.
[0036] Preferably, the nuclear norm ‖X‖ * as follows:
[0037]
[0038] Among them, σ i is the i-th singular value of X, and k is the total number of singular values. In a second aspect, an embodiment of the present invention provides a coordinate-based five-dimensional seismic data interpolation system, comprising:
[0039] The construction module builds an unsupervised point-by-point interpolation model for seismic data based on NeRF theory;
[0040] The interpolation module, which modifies the MLP framework with additional convolutional layers, builds a profile-by-profile interpolation model based on the constructed unsupervised seismic data point-by-point interpolation model;
[0041] Function module, introduces nuclear norm regularization and constructs the objective function of the profile-by-profile interpolation model;
[0042] Reconstruction module, reads the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
[0043] In a third aspect, a computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned coordinate-based five-dimensional seismic data interpolation method when executing the computer program.
[0044] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned coordinate-based five-dimensional seismic data interpolation method.
[0045] Compared with the prior art, the present invention has at least the following beneficial effects:
[0046] A coordinate-based five-dimensional seismic data interpolation method is an unsupervised learning method that eliminates the need for additional labeled data. Based on NeRF's original point-by-point method, the unique characteristics of seismic data are utilized to improve the convolutional network decoder so that it outputs data on a profile-by-profile basis. Compared with the point-by-point method, this technology can significantly increase data processing efficiency by 40 times, thereby enabling faster and more accurate reconstruction of five-dimensional seismic data. Nuclear norm regularization is added to the objective function, which can effectively improve the noise resistance and robustness of the model.
[0047] Furthermore, in order to supplement the missing seismic traces and obtain the complete recorded data Z, the sampling results of the restored data Z are required to be as close as possible to the observed data. Based on the above conditions, an unsupervised seismic data point-by-point interpolation model is constructed to minimize the observed data D obs And the model predicts data M θ The mean square error (v) enables the model to accurately reconstruct the seismic data and ensure that the difference between the model output and the actual observed data is as small as possible.
[0048] Furthermore, in order to reduce the computational resource consumption of the point-by-point interpolation model, we introduce a new profile-by-profile interpolation model, which consists of an encoder and a decoder. The encoder uses the FFM layer and MLP structure to convert the high-dimensional input coordinates v into low-dimensional latent variables The decoder uses convolution operations to extract important features from the input data and provide a concise and informative representation for subsequent decoding. The feature information is extracted and fused to reconstruct high-quality seismic profiles. The combination of encoder and decoder N τ This constitutes an end-to-end learning model. By optimizing the parameters of the entire model, an efficient mapping from input coordinates to seismic profiles is achieved, ensuring that the difference between the model output and the actual observed data is as small as possible.
[0049] Furthermore, different objective functions are used for different seismic data. For seismic data that do not use conventional time difference correction, in order to accurately locate specific data points in the seismic data, s x ,s y ,r x Construct objective functions for the indexes. These indexes can effectively handle holes and discontinuities in the data, ensuring that the model can effectively interpolate and reconstruct on incomplete datasets.
[0050] Furthermore, for the common midpoint seismic data using the NMO technique, since the application of NMO correction will significantly enhance the event continuity of a specific gather, it helps our network to effectively capture their correlation. Therefore, an intuitive strategy is to fix the m in the 5D seismic dataset. x ,m y , and extract the 3D tensor Construct the objective function. The tensor includes multiple gathers at the same common midpoint, giving similar features to different gathers, thereby enhancing the continuity and correlation of the data and ensuring that the model can effectively interpolate and reconstruct on incomplete data sets.
[0051] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.
[0052] In summary, based on the original point-by-point method of NeRF, this paper utilizes the unique characteristics of seismic data to improve the convolutional network decoder so that it can output data on a profile-by-profile basis. Compared with the point-by-point method, this technology can significantly improve the data processing efficiency by 40 times, thereby enabling faster and more accurate reconstruction of five-dimensional seismic data; at the same time, nuclear norm regularization is added to the objective function, which can effectively improve the noise resistance and robustness of the model. .
[0053] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 A schematic diagram of a coordinate system for describing five-dimensional seismic data in three-dimensional seismic exploration;
[0055] Figure 2 Schematic diagram of the MLP network structure of the point-by-point interpolation method;
[0056] Figure 3 Schematic diagrams of two different tensor expansion strategies, where (a) is a profile generated by common offset azimuth gathers, and (b) is a profile generated by common azimuth gathers;
[0057] Figure 4Schematic diagrams of actual data examples of two different expansion strategies, where (a) is a profile generated by common offset azimuth gathers, and (b) is a profile generated by azimuth segmented gathers;
[0058] Figure 5 Schematic diagram of the network architecture of the interpolation method proposed in the present invention, where (a) takes the position index as input and uses MLP+ConvNets to extract the position from the compact latent space. (b) is the architecture of the NeRSI block, which aims to increase the feature map by the S factor;
[0059] Figure 6 Schematic diagram of the results obtained by reconstructing synthetic data using different methods, where (a) is 8 real seismic data gathers, (b) is seismic data with 80% of seismic traces irregularly lost, (c) is the recovery result using the PMF-TR method, (d) is the reconstruction error using the PMF-TR method, (e) is the recovery result using the point-by-point method, (f) is the reconstruction error using the point-by-point method, (g) is the recovery result using the profile-by-profile method, and (h) is the reconstruction error using the profile-by-profile method. The dark arrows indicate that the PMF-TR method and the point-by-point method do not fully recover the tilt structure in the complete data, while in the area marked by the light arrows, both methods have signal leakage;
[0060] Figure 7 Schematic diagram of seismic gathers obtained with fixed offset bin=6 and azimuth bin=4, where (a) is actual data, (b) is the reconstruction result using the PMF-TR method, and (c) is the reconstruction result using the method of the present invention;
[0061] Figure 8 Schematic diagram of seismic gathers obtained with fixed offset bin=6 and CMP x number=4, where (a) is the actual data, (b) is the reconstruction result using the PMF-TR method, and (c) is the reconstruction result using the method of the present invention;
[0062] Fig. 9 Schematic diagram of post-stack seismic data, where (a) is before reconstruction, (b) is after reconstruction using the PMF-TR method, and (c) is after reconstruction using the method of the present invention;
[0063] Fig.10 Schematic diagram of reconstruction results when the regularization factor takes different values, where (a) is incomplete seismic data, (b) is the reconstruction result without using nuclear norm regularization, and (c) is the reconstruction result using nuclear norm regularization;
[0064] Fig.11 It is a flowchart of the present invention.
[0065] Fig.12 A schematic diagram of a computer device provided by an embodiment of the present invention;
[0066] Fig.13 The block diagram of a chip provided according to an embodiment of the present invention. DETAILED DESCRIPTION
[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0068] In the description of the present invention, it should be understood that the terms “include” and “comprises” indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.
[0069] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0070] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.
[0071] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.
[0072] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.
[0073] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.
[0074] The present invention provides a coordinate-based five-dimensional seismic data interpolation method. Based on the NeRF theory, the intrinsic structural characteristics between profiles are effectively utilized, so that the convolutional network decoder outputs data profile by profile, and nuclear norm regularization is used in the objective function. It is a coordinate-based unsupervised internal learning method, which eliminates the need for additional labeled data and can effectively increase data processing efficiency and noise resistance of the model.
[0075] See also Fig.11 The present invention provides a coordinate-based five-dimensional seismic data interpolation method. Based on NeRF theory, a section-by-section seismic data reconstruction network is built and the objective function is updated, which further improves the seismic data processing efficiency and reconstruction accuracy. The specific implementation steps are as follows:
[0076] S1. Based on NeRF theory, an unsupervised point-by-point interpolation model for seismic data is constructed;
[0077] See also Figure 1 , the seismic data is structured as a five-dimensional tensor. For the source-receiver coordinates, the observation tensor is represented as:
[0078] D obs (t,s x ,s y ,r x ,r y )
[0079] Among them, (s x ,s y ,r x ,r y ) are the source and receiver location coordinates.
[0080] Using different data orderings, the observation tensor is represented as:
[0081] D obs (t,m x ,m y ,h x ,h y )
[0082] Among them, t represents time, m x and m y Indicates the coordinates of the midpoints of inline and cross lines, h x and h y Indicates inline and crossing line offsets.
[0083] Seismic data is represented as:
[0084] D obs (t,m x ,m y ,|h|,θ)
[0085] Where |h| represents the offset distance and θ represents the azimuth angle.
[0086] For ease of description, the above coordinate positions are expressed as:
[0087] v=[v1,v2,v3,v4,v5]
[0088] Assume that the observed data D obs Corresponding to the complete seismic data Z, the five-dimensional seismic data D obs for:
[0089]
[0090] in, is the Hadamard product, P is the sampling matrix, and Z is the complete seismic data.
[0091] The elements in the sampling matrix P satisfy the following conditions:
[0092]
[0093] Among them, d v1,,,, is the five-dimensional seismic data D obs The elements indexed as v1, v2, v3, v4, v5 in Z. Seismic data is conceptualized as a wave field, which is essentially a function that maps coordinates to corresponding values in Z:
[0094] z v1v2v3v4v5 =(v)
[0095] The present invention uses MLP to describe this mapping:
[0096] z v1v2v3v4v5 = θ(v)
[0097] Among them, θ is the parameter weight of MLP.
[0098] Considering that complete seismic data cannot be observed directly, the present invention uses the observed data D obs Speculation M θ , the objective function is:
[0099]
[0100] Among them, v1, v2, v3, v4, v5 are the five-dimensional coordinate representations of seismic data, D obs is the seismic data to be solved, θ is the parameter weight of MLP, M θ is the function of MLP mapping coordinates to complete seismic data, Represents the mean square error of the matrix.
[0101] See also Figure 2 ,The network architecture of the point-by-point interpolation model of seismic data consists of an encoder and a decoder. The encoder consists of 16 fully connected layers and the decoder consists of 2 fully connected layers.
[0102] S2, modifying the MLP framework with additional convolutional layers to build a profile-by-profile interpolation model based on the model built in step S1;
[0103] See also Figure 5 , the present invention develops an innovative network structure, combining additional convolutional layers to modify the MLP framework. The network architecture includes an encoder and a decoder; the encoder has two main components, the first is a predefined single FFM layer γ(v) to improve the position encoding ability, and the second is an encoder based on the MLP prototype, which can effectively transform the coordinate v into the corresponding latent variable Then, the convolutional network-based decoder is Reconstruct the complete seismic profile using N τ Represents the combination of encoder and decoder. The specific model is:
[0104] M θ (v)= τ (γ(v))
[0105] Furthermore, there is considerable evidence that standard MLPs often face challenges in capturing high-frequency variations. Specifically, directly feeding the original data locations v into the MLP produces over-smoothed results, leading to the loss of high-frequency details in the reconstructed data.
[0106] To address this issue, v is expanded to its Fourier spectrum before being fed into the MLP.
[0107] The present invention introduces FFM position coding, and the FFM position coding is:
[0108] γ K (v)=[cos(w1v1),sin(w1v1),cos(w1v2),sin(w1v2),
[0109] …,cos(w K v n ),sin(w K v n )] T
[0110] Where K represents the total number of components, and the i-th frequency component is linearly sampled by w i =2 i Indicates that V represents any coordinate normalized within [0, 1].
[0111] Although the point-by-point method has shown effectiveness in seismic data reconstruction, its single-point output strategy has obvious limitations. This method treats each point in the wave field as an independent entity, which results in a large amount of computing resources required for model training. In real data scenarios, especially when the signal-to-noise ratio is low, the model tends to overfit the noise in the data. Therefore, in order to overcome these challenges, the present invention introduces a profile-by-profile interpolation model based on the point-by-point interpolation model. The output of this model is no longer limited to a single pixel, but provides a complete seismic profile, thereby capturing the characteristics of seismic data more comprehensively.
[0112] Furthermore, since convolution operations are conducive to extracting observable and regular features, profiles with clear structural information and stable event continuity are very suitable for learning convolutional neural networks. In order to ensure that the profile has good predictability and regularity, it is crucial to choose an appropriate output profile size, especially considering the five-dimensional characteristics of seismic data.
[0113] Furthermore, the present invention explores two different scenarios, namely, seismic data without conventional moveout (NMO) correction and seismic data with common midpoint (CMP) processing using NMO.
[0114] Different objective functions are used for different types of seismic data.
[0115] For seismic data that are not corrected using conventional moveout (NMO), the present invention fixes s x ,s y and r x , we get the earthquake set D obs (:,s x ,s y ,r x,:); by giving V = (s x ,s y ,r x ) T , construct the data set {v,D obs (:,s x ,s y ,r x ,:)}, and define the objective function as:
[0116]
[0117] Among them, S x and S y are the total lengths of the inline axis and the cross-line axis in the source, R x Indicates the total length of the inline shaft in the receiver.
[0118] For CMP seismic data processed by NMO, the application of NMO correction significantly enhances the continuity of the gathers, which helps the neural network capture the correlation between gathers. Therefore, when choosing the profile size, a common strategy is to fix the m of the five-dimensional seismic data set. x and m y , extract the 3D tensor Since the tensor All traces that include the same common midpoint will give different events similar characteristics, so this slicing method helps to enhance The similarity of the data in , makes it more suitable for convolutional networks.
[0119] For the sake of For a more detailed comparison of different sections of the invention, the present invention adopts two different expansion methods, in which the time dimension remains unchanged. Figure 3 As shown, the tensor Expand to a matrix or Since the amplitude variation caused by azimuthal anisotropy is usually not as obvious as that caused by offset variation, Figure 4 As shown, Figure 4 Compared with b, Figure 4 The unfolding method in a produces a more structurally coherent section.
[0120] The present invention adopts Figure 3 The expansion strategy (marked as U) shown in a creates a dataset for training Where v = (m x ,m y ) T is the input of the network, is the corresponding learning goal, and the objective function is defined as:
[0121]
[0122] Among them, M x and M y are the maximum lengths of the inner and cross axes, respectively. U is a tensor expansion method. X is the m in the fixed five-dimensional seismic data. x and m y The extracted three-dimensional tensor is expanded into a two-dimensional matrix, v = (m x ,m y ) T is the input of the network.
[0123] S3, introduce nuclear norm regularization and construct the objective function;
[0124] In seismic exploration, real pre-stack seismic data often contain a lot of noise, which makes it quite complicated to accurately identify useful signals. Since neural networks tend to learn and simulate these noises, the output results may be full of noise. In order to meet this challenge, the present invention introduces nuclear norm regularization. As a convex relaxation form of low-rank regularization, nuclear norm regularization can effectively improve the generalization ability of the model while effectively controlling the complexity of the model. By introducing nuclear norm regularization, the present invention aims to better suppress the influence of noise on the model during the learning process, thereby improving the accuracy and robustness of seismic data processing.
[0125] Introducing nuclear norm regularization into the objective function is:
[0126]
[0127] Among them, λ is a hyperparameter used to balance the fidelity and regularization terms in the objective function, ‖‖ * is the nuclear norm,
[0128] The nuclear norm is defined as follows:
[0129]
[0130] Among them, σ i is the i-th singular value of X.
[0131] S4. Read the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the network to realize the reconstruction of five-dimensional seismic data.
[0132] After completing the model training, process each v=(m x ,m y ) T , obtain their respective profiles for seismic data reconstruction.
[0133] In another embodiment of the present invention, a coordinate-based five-dimensional seismic data interpolation system is provided, which can be used to implement the above-mentioned coordinate-based five-dimensional seismic data interpolation method. Specifically, the coordinate-based five-dimensional seismic data interpolation system includes a construction module, an interpolation module, a function module and a reconstruction module.
[0134] Among them, the construction module builds an unsupervised point-by-point interpolation model of seismic data based on NeRF theory;
[0135] The interpolation module, which modifies the MLP framework with additional convolutional layers, builds a profile-by-profile interpolation model based on the constructed unsupervised seismic data point-by-point interpolation model;
[0136] Function module, introduces nuclear norm regularization and constructs the objective function of the profile-by-profile interpolation model;
[0137] Reconstruction module, reads the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
[0138] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the coordinate-based five-dimensional seismic data interpolation method, including:
[0139] Based on NeRF theory, an unsupervised point-by-point interpolation model for seismic data is constructed; based on the constructed unsupervised point-by-point interpolation model for seismic data, an interpolation model for profiles is constructed by modifying the MLP framework with additional convolutional layers; nuclear norm regularization is introduced to construct the objective function of the interpolation model for profiles; the original five-dimensional seismic data D is read. obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
[0140] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (Non-Volatile Memory), such as at least one disk memory.
[0141] The processor may load and execute one or more instructions stored in a computer-readable storage medium to implement the corresponding steps of the coordinate-based five-dimensional seismic data interpolation method in the above embodiment; the processor may load and execute the following steps:
[0142] Based on NeRF theory, an unsupervised point-by-point interpolation model for seismic data is constructed; based on the constructed unsupervised point-by-point interpolation model for seismic data, an interpolation model for profiles is constructed by modifying the MLP framework with additional convolutional layers; nuclear norm regularization is introduced to construct the objective function of the interpolation model for profiles; the original five-dimensional seismic data D is read. obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
[0143] See also Fig.12 , the terminal device is a computer device, and the computer device 60 of this embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the computer program 63 is executed by the processor 61, the method for calculating the composition of fluid in the reservoir transformation wellbore in the embodiment is realized. To avoid repetition, it is not described one by one here. Alternatively, when the computer program 63 is executed by the processor 61, the functions of each model / unit in the calculation system for the composition of fluid in the reservoir transformation wellbore in the embodiment are realized. To avoid repetition, it is not described one by one here.
[0144] The computer device 60 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will appreciate that Fig.12This is only an example of the computer device 60 and does not constitute a limitation of the computer device 60. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.
[0145] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, central processing units, graphics processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, data processing logic based on quantum computing, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
[0146] The memory 62 may be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 may also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the computer device 60.
[0147] Furthermore, the memory 62 may include both an internal storage unit of the computer device 60 and an external storage device. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 may also be used to temporarily store data that has been output or is to be output.
[0148] Any reference to memory, database or other media used in the embodiments provided in the present application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.
[0149] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0150] See also Fig.13 , the terminal device is a chip, and the chip 600 of this embodiment includes a processor 622, which may be one or more, and a memory 632 for storing a computer program executable by the processor 622. The computer program stored in the memory 632 may include one or more modules, each corresponding to a set of instructions. In addition, the processor 622 may be configured to execute the computer program to execute the above-mentioned generalizable monocular absolute depth map estimation method.
[0151] In addition, the chip 600 may further include a power supply component 626 and a communication component 650, wherein the power supply component 626 may be configured to perform power management of the chip 600, and the communication component 650 may be configured to implement communication, for example, wired or wireless communication, of the chip 600. In addition, the chip 600 may further include an input / output interface 658. The chip 600 may operate based on an operating system stored in the memory 632.
[0152] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0153] The present invention selected an effective traditional PMF-TR method for further comparison, and the results are as follows Figure 6 shown. Figure 6 a shows 8 seismic gathers, s x =16, r y =32,s y The range is 1 to 16, r x The range is 1 to 8. As can be seen, the raw data includes both horizontal and tilted events, which presents a challenge for reconstruction. Figure 6 b plots the seismic data with 80% irregular missing traces. The recovery results of PMF-TR, point-by-point method and profile method are shown in Figure 6 c. Figure 6 e and Figure 6 g, the corresponding reconstruction error is Figure 6 d. Figure 6 B and Figure 6 h. All three methods are effective in data interpolation. Quantitative calculations show that the SNRs of the PMF-TR method, the point-by-point method, and the profile method are 12.37dB, 16.61dB, and 18.67dB, respectively. Both implicit representation methods outperform the PMF-TR method, proving the superior performance of the implicit network. In addition, as shown by the dark arrows, the tilted structure in the complete data is not fully restored by the PMF-TR method and the point-by-point method. In the area marked by the light arrows, it can be clearly seen that both the PMF-TR method and the point-by-point method have signal leakage. In contrast, the profile method proposed in the present invention successfully restores the tilted structure without obvious leakage of useful signals. This may be attributed to the fact that the convolutional network decoder based on the profile method has a larger receptive field and therefore has a stronger feature extraction capability in reconstructing data with complex features.
[0154] Compared with synthetic data, real data presents greater complexity and is significantly affected by noise. In order to further evaluate the effectiveness of the profile-based reconstruction method proposed in this invention, PMF-TR is selected as the benchmark method, and a land dataset acquired in Canada is used as the dataset for comparison.
[0155] Figure 7 The slice diagram of the result of reconstructing the actual data using PMF-TR and the method of the present invention is shown. Specifically, Figure 7 a shows the CMP y section of the five images observed with fixed CMP x (bin=4) and offset (bin=6). Obviously, the original data segment contains a large number of missing traces and noise, which poses a considerable challenge to reconstruction. Figure 7 b shows the CMP y profile obtained by PMF-TR, which effectively restores Figure 7 The missing channel in a. However, noise interference is still present. Figure 7 c shows the results of the present invention, wherein Figure 7 All missing traces in a are accurately restored. Compared with the PMF-TR method, the image restored by the proposed method has a higher signal-to-noise ratio and better effect. These results also prove the effectiveness of the nuclear norm in processing noisy data.
[0156] In order to further verify the effectiveness of the method of the present invention, the offset bin=6 and the azimuth bin=4 were fixed, and the results were observed from different angles. Figure 8 a shows the cross section of the original seismic record, with Figure 7 Compared to a, these sections show larger notches. Figure 8 b and Figure 8 c shows the reconstruction results of the PMF-TR method and the proposed method. The proposed method can effectively restore the missing track even at the location with significant gaps. Compared with the PMF-TR method, the two methods have Figure 8 The regions indicated by the light colored boxes all produced good results, and the method of the present invention showed excellent lateral consistency in the restored reflections. On the other hand, in the regions marked by the dark colored boxes, significant noise was observed in the results restored by the PMF-TR method. In contrast, the method of the present invention was able to effectively process these regions.
[0157] For a comprehensive comparison, Fig. 9 a shows the post-stack seismic data before interpolation. The original post-stack data has a significantly low SNR, and many reflections are masked by noise, especially in shallow areas. The reconstructed post-stack cubes obtained using PMF-TR and the method of the present invention are shown in Figure 2, respectively. Fig. 9 b and Fig. 9 c, and Fig. 9Compared with the original post-stack data in a, both effectively eliminate the pre-stack noise and significantly enhance the signal energy. Fig. 9 The profile shown in b still shows slight residual noise in the shallow area. Fig. 9 There is almost no random noise in the profile shown in c, which confirms the effectiveness of the proposed method in recovering seismic data containing noise and irregular sampling.
[0158] In order to further verify the effect of nuclear norm regularization, another five-dimensional seismic data collected in China was selected as a data set for comparison. Fig.10 In FIG. 4 , a set of original CMP gathers is shown, where CMP x=3 and CMP y=3. Fig.10 b is the reconstruction result without nuclear norm regularization (λ=0), Fig.10 c is the reconstruction result when λ=0.0001. Fig.10 b and Fig.10 c, it can be observed that when λ changes from 0 to 0.0001, the noise interference of the reconstruction result is significantly reduced. This shows the effectiveness of the nuclear norm regularization constraint. However, when λ exceeds a certain threshold, the reconstruction quality will decrease slightly, indicating that a careful balance needs to be struck between data fidelity and regularization.
[0159] In the above examples, the present invention can effectively interpolate five-dimensional seismic data, laying a foundation for subsequent data analysis.
[0160] In summary, the present invention provides a coordinate-based five-dimensional seismic data interpolation method and system, which eliminates the need for additional labeled data. This is the first attempt at this method in five-dimensional seismic data processing. Based on the original point-by-point method of NeRF, the unique characteristics of seismic data are utilized to improve the convolutional network decoder so that it outputs data profile by profile. Compared with the point-by-point method, this technology can significantly improve the data processing efficiency by 40 times, thereby enabling faster and more accurate reconstruction of five-dimensional seismic data. Nuclear norm regularization is added to the objective function, which can effectively improve the noise resistance and robustness of the model.
[0161] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.
[0162] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0163] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0164] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0165] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0166] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0167] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.
[0168] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0169] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0170] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0171] The above contents are only for explaining the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.
Claims
1. A coordinate-based five-dimensional seismic data interpolation method, characterized in that: The following steps are involved: S1. Construct an unsupervised point-by-point interpolation model for seismic data based on NeRF theory; S2, modifying the MLP framework with additional convolutional layers to build a profile-by-profile interpolation model based on the constructed unsupervised seismic data point-by-point interpolation model; S3, introduce nuclear norm regularization and construct the objective function of the profile-by-profile interpolation model; S4. Read the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
2. The coordinate-based five-dimensional seismic data interpolation method according to claim 1, characterized in that: The objective function of the unsupervised seismic data point-by-point interpolation model is: Among them, v1, v2, v3, v4, v5 are the five-dimensional coordinates of seismic data, D obs is the five-dimensional seismic data to be solved, θ is the parameter weight of MLP, M θ is the function of MLP mapping coordinates to complete seismic data, Represents the mean square error of the matrix.
3. The coordinate-based five-dimensional seismic data interpolation method according to claim 2, characterized in that: 5D seismic data obs for: in, is the Hadamard product, P is the sampling matrix, and Z is the complete seismic data.
4. The coordinate-based five-dimensional seismic data interpolation method according to claim 3, characterized in that: The elements in the sampling matrix P satisfy the following conditions: Among them, d v1,v2,v3,v4,v5 is the five-dimensional seismic data D obs The elements with index v1,v2,v3,v4,v5 in .
5. The coordinate-based five-dimensional seismic data interpolation method according to claim 1, characterized in that: The profile-by-profile interpolation model consists of an encoder and a decoder. The encoder includes a predefined single FFM layer γ(v) and an encoder based on the MLP prototype. The encoder transforms the coordinate v into the corresponding latent variable Then, the convolutional network-based decoder is transformed from the variable Reconstruct the complete seismic profile using N τ Represents a combination of an encoder and a decoder.
6. The coordinate-based five-dimensional seismic data interpolation method according to claim 5, characterized in that: A single FFM layer γ(v) is specifically: γ K (v)\[cos(w1v1),sin(w1v1),cos(w1v2),sin(w1v2), …,cos(w K v n ),sin(w K v n )] T Where K represents the total number of components, and the i-th frequency component is linearly sampled by w i =π2 i Indicates that v represents any coordinate normalized within [0, 1].
7. The coordinate-based five-dimensional seismic data interpolation method according to claim 5, characterized in that: For seismic data without conventional time difference correction, the objective function is defined as: Among them, S x and S y are the total lengths of the inline axis and the cross-line axis in the source, R x is the total length of the receiver internal shaft, r x is the index of the receiver on the inline axis, P is the sampling matrix; For the common midpoint seismic data using NMO technique, the objective function is defined as: Among them, M x and M y are the maximum lengths of the inline axis and the cross axis respectively, U is the tensor expansion method, is the corresponding learning goal, M θ is the function of MLP that maps coordinates to complete seismic data.
8. The coordinate-based five-dimensional seismic data interpolation method according to claim 1, characterized in that: The objective function of the profile-by-profile interpolation model is: Among them, λ is a hyperparameter used to balance the fidelity and regularization terms in the objective function, ‖‖ * is the nuclear norm, M x and M y are the maximum lengths of the inline axis and the cross axis, M θ is the function of MLP mapping coordinates to complete seismic data, U is the tensor expansion method, (P:,m x ,m y ,:,:) is a specific part selected from the sampling matrix, is the corresponding learning objective, and F is the Frobenius norm.
9. The coordinate-based five-dimensional seismic data interpolation method according to claim 8, characterized in that: Nuclear norm ‖X‖ * as follows: Among them, σ i is the i-th singular value of X, and k is the total number of singular values.
10. A coordinate-based five-dimensional seismic data interpolation system, characterized in that: include: The construction module builds an unsupervised point-by-point interpolation model for seismic data based on NeRF theory; The interpolation module, which modifies the MLP framework with additional convolutional layers, builds a profile-by-profile interpolation model based on the constructed unsupervised seismic data point-by-point interpolation model; Function module, introduces nuclear norm regularization and constructs the objective function of the profile-by-profile interpolation model; Reconstruction module, reads the original five-dimensional seismic data D obs , the five-dimensional seismic data with missing traces are input into the profile-by-profile interpolation model to achieve five-dimensional seismic data reconstruction.
Citation Information
Patent Citations
Exploration data first arrival pickup method based on U-MLP network
CN115311562A
Five-dimensional seismic data low-rank constraint reconstruction method, system, medium and equipment
CN115511004A
Dynamic scene reconstruction method based on explicit and implicit hybrid coding
CN117252987A
Volume video processing method and system based on dynamic NeRF
CN118075493A
Deformable neural radiance fields
WO2022104299A1
