DAS-VSP data coherent noise suppression method and system based on adaptive texture analysis

Abstract

The invention discloses a DAS-VSP data coherent noise suppression method and system based on adaptive texture analysis, and the method comprises the steps: respectively calculating vertical and horizontal derivatives and frequency values of effective signals and noise in synthesized DAS-VSP data, calculating the correlation between pixels, drawing a spatial correlation graph, and guiding the selection of a training strategy; expanding a traditional blind spot network into a neighborhood shielding network, rotating input data to construct a plurality of sub-graphs, and performing translation cutting to form a blind neighborhood; a corresponding blind neighborhood shielding strategy is designed, so that the network meets the condition that blind spot network noise pixels are independent; new asymmetric expansion convolution is proposed to ignore the influence of coherent noise; ordinary total variation regularization is expanded into texture total variation regularization, and texture weight factors are calculated according to different texture characteristics of signals and noise to smooth noise reserved signals; and performing a denoising test on the synthetic seismic data and the actual DAS-VSP seismic data, and analyzing a denoising effect through a quantitative index signal-to-noise ratio, structural similarity and a visualization effect.

Description

Background Art

[0002] Seismic exploration is a key technology for the discovery and development of oil and gas resources, relying primarily on seismic methods for the exploration and production of natural energy. In a typical seismic exploration operation, an artificial source is used to generate elastic waves. These waves propagate through the underground medium under the excitation of the source, and seismic detectors are placed on the surface or in wells to record them, capturing the ground vibrations and converting them into digital data. Seismic data is often affected by multiple noise sources, which makes identifying useful seismic signals a challenge and seriously affects subsequent seismic data processing and interpretation. Therefore, improving the signal-to-noise ratio is an important goal in seismic data processing.

[0003] With the growing global demand for oil and gas resources, the exploration and development of oil wells is increasingly demanding high-precision reservoir characterization to guide further production. Traditional seismic sensors, due to limitations such as low spatial sampling density, high cost, and difficult deployment, are no longer able to meet the demands of modern oil and gas exploration. Therefore, developing new sensing technologies to improve exploration efficiency has become an urgent need in the industry. Compared to traditional geophones, distributed acoustic sensing (DAS) systems offer a novel solution. DAS systems offer the advantages of full coverage, wide bandwidth, high sensitivity, and high density. They are also able to withstand high temperatures and pressures and are immune to electromagnetic interference, ensuring system stability. This technology utilizes short optical pulses to achieve high spatial resolution monitoring and has been successfully applied to vertical seismic data acquisition. However, DAS technology also faces challenges in practical application, particularly the significant noise interference during data acquisition, resulting in low signal-to-noise ratios (SNRs).

[0004] Improving the signal-to-noise ratio (SNR) of DAS-VSP (Distributed Fiber Vertical Seismic Profiling) seismic data requires removing multiple noise sources, including random noise, instrument interference, coupling noise, checkerboard noise, and horizontal noise. This is a challenging task. Traditional denoising methods often fail to effectively remove noise without damaging the valid signal. Therefore, developing methods that effectively denoise and preserve the seismic signal has become a research hotspot. Recently, deep learning techniques have demonstrated great potential in seismic data denoising. These methods can more accurately identify and separate noise from signal, thereby improving data quality and providing more reliable information for oil and gas exploration. However, supervised deep learning faces many challenges in practical applications, including the quality and quantity of training data and the generalization ability of the model. First, high-quality labeled data is crucial for the success of supervised learning, but obtaining precisely labeled data for seismic data is often costly and time-consuming, especially in complex geological structures. Furthermore, the inherent complexity of seismic data complicates model training, resulting in the potential for poor performance on new or unseen data. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies in the above-mentioned prior art and provide a DAS-VSP data coherent noise suppression method and system based on adaptive texture analysis, which is used to solve the technical problem that existing seismic data is affected by multiple noise sources and cannot be efficiently removed without damaging the effective signal, effectively circumventing the limitations of traditional methods and supervised deep learning methods.

[0006] The present invention adopts the following technical solutions: The DAS-VSP data coherent noise suppression method based on adaptive texture analysis includes the following steps: S1. Calculate the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data, draw a spatial correlation map based on the correlation between pixels, and select a training strategy; S2. Expand the blind spot network into a neighborhood masking network according to the training strategy selected in step S1, rotate the input data to construct multiple subgraphs, and perform translation and cropping to form a blind neighborhood; S3, setting a corresponding masking strategy for the blind neighborhood obtained in step S2 according to different noise correlation characteristics, so that the neighborhood masking network satisfies the condition that the noise pixels of the blind spot network are independent; S4, using asymmetric dilated convolution to ignore the influence of coherent noise in the effective signal obtained in step S1; S5. Extending the ordinary total variation regularization to texture total variation regularization, calculating the texture weight factor according to the different characteristics of the signal and noise, further smoothing the noise obtained in step S4 and retaining the signal, and combining the masking strategy obtained in step S3 to construct a self-supervised method; S6. Process the synthetic seismic data and the actual DAS-VSP seismic data based on the self-supervision method constructed in step S5 to suppress the coherent noise of the DAS-VSP data.

[0007] Preferably, step S1 is specifically: The derivative values ​​of the effective signal and noise at each point in the vertical and horizontal directions are calculated by convolving each pixel with the convolution kernels of [[-1], [0], [1]] and [[-1, 0, 1]], and the derivative frequency graph is drawn; Divide all data into 20*20 rectangular blocks and calculate the mean of the correlation coefficient of each block relative to the fixed position block; Finally, the spatial correlation map between pixels at different positions is obtained.

[0008] Preferably, in step S2, based on the U-net, for vertical noise, the input data is first rotated 90° and 270° respectively to create two branches with different directional information to form feature maps. After the convolution processing of each branch is completed, each feature map is spatially offset; For horizontal noise, two feature maps are constructed by rotating 0° and 180° respectively. After the convolution processing of each branch is completed, each feature map is spatially offset.

[0009] Preferably, k Row zero elements are added to the top of each feature map before it is fed into the network and finally cropped from the bottom of the output. k The row size is Finally, all feature maps are derotated and input into a 1×1 convolutional layer for fusion of the two branches.

[0010] Preferably, in step S3, the objective function of the neighborhood shielding network is expressed as:

[0011] in, represents the loss function, represents the denoising network, and are the subscripts in the vertical and horizontal directions, represents the noise input, are network parameters, Remove the receptive field of vertical coherent noise for the network, Remove the receptive field of horizontally coherent noise for the network.

[0012] Preferably, in step S4, for vertical noisy data, the asymmetric dilated convolution expands the original 3*3 convolution kernel to a 3*5 convolution kernel, and sets the dilation factor in the horizontal direction with step sizes of 2, 3, and 4, respectively, while the vertical direction does not expand; for horizontal noisy data, the convolution kernel size is expanded to 5*3, and the dilation factor is set in the vertical direction.

[0013] Preferably, in step S5, the texture total variation regularization of different noise forms is as follows:

[0014] in, represents the network input, Represents the network output result, represents the scaling factor, and represent the horizontal and vertical gradients respectively, represents the proposed vertical regularization strategy for removing vertical coherent noise, Denotes the proposed horizontal regularization strategy for removing horizontal coherent noise.

[0015] Preferably, the overall objective function of the neighborhood shielding network is:

[0016] in, represents the weight parameter, represents the loss function, represents the denoising network, and are the subscripts in the vertical and horizontal directions, represents the noise input, are network parameters, Remove the receptive field of vertical coherent noise for the network, Remove the receptive field of horizontally coherent noise for the network.

[0017] Preferably, in step S6, the signal-to-noise ratio is selected as a quantitative indicator to measure the denoising performance of the network, the visualization effect is presented by the denoised data and the removed noise data, and the denoising performance is analyzed using the time domain diagram of the single-channel signal.

[0018] In a second aspect, an embodiment of the present invention provides a DAS-VSP data coherent noise suppression system based on adaptive texture analysis, comprising: The calculation module calculates the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data, draws a spatial correlation map based on the correlation between pixels, and selects a training strategy; The extension module expands the blind spot network into a neighborhood masking network according to the selected training strategy, rotates the input data to construct multiple subgraphs, and performs translation and cropping to form a blind neighborhood; The masking module sets the corresponding masking strategy for the obtained blind neighborhood according to the different correlation characteristics of the noise, so that the neighborhood masking network satisfies the condition that the noise pixels of the blind spot network are independent; The convolution module uses asymmetric dilated convolution to ignore the influence of coherent noise in the obtained effective signal; The regularization module extends the conventional total variation regularization to texture total variation regularization. It calculates the texture weight factor based on the different characteristics of the signal and noise, further smoothes the noise and preserves the signal, and combines the masking strategy to build a self-supervised method. The noise suppression module processes the synthetic seismic data and the actual DAS-VSP seismic data based on the constructed self-supervision method to achieve coherent noise suppression of DAS-VSP data.

[0019] In a third aspect, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the DAS-VSP data coherent noise suppression method based on adaptive texture analysis are implemented.

[0020] 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 DAS-VSP data coherent noise suppression method based on adaptive texture analysis.

[0021] Compared with the prior art, the present invention has at least the following beneficial effects: The DAS-VSP data coherent noise suppression method based on adaptive texture analysis aims to address the limitations of existing traditional methods and supervised deep learning approaches, and improves existing self-supervised deep learning methods. First, the method thoroughly analyzes the vertical and horizontal correlations of different types of coherent noise. Furthermore, it uses a novel blind neighbor masking strategy to carefully design the receptive field, extending the traditional blind spot network (BSN) into a neighbor masking network (NMN). This improvement ensures that the NMN still meets the assumption of pixel independence in the presence of significant coherent noise, thereby enhancing the network's adaptability in complex noisy environments. Furthermore, the method introduces a texture total variation (TTV) regularization technique, which effectively eliminates correlated noise by constraining the texture characteristics of the noise and improves signal fidelity while preserving the local smooth structure of the effective signal. To further enhance the texture difference between the signal and noise, the method proposes an asymmetric dilated convolution (ADConv) technique, which has a larger receptive field in specific directions and acts as an information filter in the network, further optimizing the noise removal effect. Our method relies solely on observed noisy DAS-VSP data to remove correlated noise in a self-supervised manner, without requiring any additional annotated data or prior information. Finally, extensive experimental validation on both synthetic and real DAS-VSP data demonstrates the superior performance and broad applicability of our method for coherent noise suppression.

[0022] Furthermore, convolution kernels of [[-1], [0], [1]] and [[-1, 0, 1]] can effectively calculate the derivative frequency values ​​of the target data. At the same time, the mean of the correlation coefficient of each block relative to a fixed position block can better reflect the coherence between pixels. The larger the mean, the stronger the coherence of the pixel values ​​at the two positions. This lays the data foundation for the construction of subsequent strategies; after determining the correlation relationship between pixels, different strategies can be flexibly selected for denoising.

[0023] Furthermore, based on the U-net, for vertical and horizontal noise, the input data is first rotated 90° and 270°, and 0° and 180°, respectively, to create two branches with different directional information to form feature maps. After the convolution processing of each branch is completed, each feature map is spatially offset. The purpose of this processing is to build a blind spot network, so that the network learns to use the pixels around the target pixel to predict the denoising result of the target pixel instead of learning a simple pixel mapping relationship. In addition, this blind spot processing can also effectively ignore the information of some coherent noise around the target pixel, so that the network more closely meets the assumption of pixel-by-pixel noise independence in blind spot strategy training.

[0024] Furthermore, since step S1 proves that the coherence distance of the coherent noise is much larger than 1, the blind spot strategy in step S2 alone cannot effectively remove the coherent noise. Therefore, step S3 sets a corresponding masking strategy for the blind neighborhood obtained in step S2, masking the blind neighborhood of the target pixel's receptive field. The masking range should be greater than or equal to the coherence distance of the noise, so that it covers all the effective information of the coherent noise. At this time, the receptive field of the target pixel will exclude the noise information, satisfying the assumption of the independence of the noise pixels, and denoising can be performed.

[0025] Furthermore, based on the coherent noise texture features extracted in step S1, the design logic of the asymmetric dilated convolution is to allow the network to ignore the coherent noise information as much as possible and highlight the effective signal information. Specifically, this convolution can be extended and expanded in one direction based on the form and characteristics of the noise while remaining unchanged in other directions. This allows the asymmetric dilated convolution to have a larger receptive field in the direction of the effective signal while maintaining the receptive field in the direction of the noise. The hollow design further achieves the goal of ignoring noise information, which enables the feature map extracted by the network to make trade-offs and choices between signal and noise information to a certain extent.

[0026] Furthermore, according to the coherent noise texture features extracted in step S1, it can be seen that horizontal and vertical noise have strong coherence in the horizontal and vertical directions, respectively, and the signal strength variation in the corresponding directions is small, with the derivative value close to 0. The valid signal exhibits strong coherence in both the horizontal and vertical directions, but has large signal strength variation in both directions, and the derivative value has a large absolute value. This characteristic allows the noise characteristics to be ignored by the variation in one direction while being enhanced by the variation in another direction, while the information extracted from the signal characteristics by the variation in each direction is relatively small. This difference in the characteristics of the signal and noise allows us to improve traditional variational regularization. Specifically, we retain the variation terms that enhance the noise characteristics to enable the network to smooth the detected noise as much as possible. For the variation terms that weaken the noise characteristics, we can use them to calculate a weighting factor. This factor effectively detects the valid signal information and assigns it a smaller weight, while assigning a larger weight to the noise information. This makes texture total variation regularization more effective at smoothing noise-preserving signals than traditional variational regularization.

[0027] Furthermore, the organic combination of neighborhood masking network (NMN) technology, texture total variation (TTV) regularization technology and asymmetric dilated convolution (ADConv) technology fully utilizes the differences in texture features of signal and noise for denoising from three different perspectives. The signal-to-noise ratio effectively quantitatively measures the network's powerful denoising performance, and the visualization effect and time domain diagram of the single-channel signal effectively and qualitatively demonstrate the effectiveness of the network in removing noise and retaining the signal.

[0028] 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.

[0029] In summary, this paper, by studying the spatial correlation within noise, extends the blind spot network (BSN) into a neighborhood masking network (NMN) with a specialized neighborhood masking architecture. This design strategically configures the receptive field to filter out noise-related information around the predicted blind spot while preserving the integrity of valid signal data. Furthermore, it integrates texture total variation regularization (TTV) technology to intelligently adjust weights based on the different texture characteristics of noise and signal. This results in excellent noise smoothing while preserving valid signal. Furthermore, an asymmetric dilated convolution (ADConv) technique is employed to reduce the impact of noise on blind spot prediction accuracy. This paper excels in extracting valid signals obscured by correlated noise and preserving fine details of weak geological structures.

[0030] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings to be used in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 This is a network structure diagram of the present invention; Figure 2 The derivative frequency graph and spatial correlation graph of noise and effective signal are calculated respectively for the present invention; FIG3 is a diagram of the first synthetic seismic test data of the present invention, wherein (a) is the noisy data of the first synthetic seismic test data of the present invention; (b) is the data of the first synthetic seismic test data of the present invention after denoising by a neighborhood masking network (NMN) with texture total variation regularization (TTV) and asymmetric dilated convolution (ADConv); (c) is a diagram showing the difference between the noisy data of FIG3(a) of the first synthetic seismic test data of the present invention and the denoising result of FIG3(b), i.e., the removed noise data; FIG4 is a diagram of the second synthetic seismic test data of the present invention, wherein (a) is the noisy data of the second synthetic seismic test data of the present invention; (b) is the data of the second synthetic seismic test data of the present invention after denoising by a neighborhood masking network (NMN) with texture total variation regularization (TTV) and asymmetric dilated convolution (ADConv); (c) is a diagram showing the difference between the noisy data of FIG4(a) and the denoising result of FIG4(b) of the second synthetic seismic test data of the present invention, i.e., the removed noise data; FIG5 is a diagram of the first actual DAS-VSP seismic data of the present invention, wherein (a) is the noisy data of the first actual DAS-VSP seismic data of the present invention; (b) is the data of the first actual DAS-VSP seismic data of the present invention after denoising by a neighborhood masking network (NMN) with texture total variation regularization (TTV) and asymmetric dilated convolution (ADConv); (c) is the difference data between the noisy data of the first actual DAS-VSP seismic data of the present invention, FIG5(a), and the denoising result, FIG5(b), which is the removed noise data; (d) is a local self-similarity diagram of the difference data of the first actual DAS-VSP seismic data of the present invention, FIG5(c), and the denoising result, FIG5(b); (e) is a time domain diagram comparing the 150th track of the first actual DAS-VSP seismic data of the present invention before denoising (gray line) and after denoising (black line); FIG6 is a diagram of the second actual DAS-VSP seismic data of the present invention, wherein (a) is the noisy data of the second actual DAS-VSP seismic data of the present invention; (b) is the data of the second actual DAS-VSP seismic data of the present invention after denoising by a neighborhood masking network (NMN) with texture total variation regularization (TTV) and asymmetric dilated convolution (ADConv); (c) is the difference data between the noisy data of FIG6(a) of the second actual DAS-VSP seismic data of the present invention and the denoising result FIG6(b), that is, the removed noise data; (d) is a local self-similarity diagram of the difference data of FIG6(c) of the second actual DAS-VSP seismic data of the present invention and the denoising result FIG6(b); (e) is a time domain diagram comparing the 100th track of the second actual DAS-VSP seismic data of the present invention before denoising (gray line) and after denoising (black line); FIG7 is a diagram of the third actual DAS-VSP seismic data of the present invention, wherein (a) is the noisy data of the third actual DAS-VSP seismic data of the present invention; (b) is the data of the third actual DAS-VSP seismic data of the present invention after denoising by a neighborhood masking network (NMN) with texture total variation regularization (TTV) and asymmetric dilated convolution (ADConv); (c) is the difference data between the noisy data of the third actual DAS-VSP seismic data of the present invention in FIG7(a) and the denoised result in FIG7(b); (d) is a local self-similarity diagram of the difference data of the third actual DAS-VSP seismic data of the present invention in FIG7(c) and the denoised result in FIG7(b); (e) is a time domain diagram comparing the 10th track of the third actual DAS-VSP seismic data of the present invention before denoising (gray line) and after denoising (black line); Figure 8 A schematic diagram of a computer device provided in accordance with an embodiment of the present invention; Figure 9 The present invention is a block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0034] In the description of the present invention, it is to be understood that the terms “include” and “comprise” indicate the presence of the 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.

[0035] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0036] It should be further understood that the term "and / or" as used in the present specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0037] It should be understood that although the terms "first," "second," and "third" may be used to describe preset ranges in embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are merely used to distinguish one preset range from another. For example, without departing from the scope of 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.

[0038] 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 the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0039] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.

[0040] This paper provides a DAS-VSP data coherent noise suppression method based on adaptive texture analysis. It fully analyzes the vertical and horizontal correlations of different types of coherent noise. Based on this, a novel blind neighborhood masking strategy is used to carefully design the receptive field, extending the traditional blind spot network (BSN) into a neighborhood masking network (NMN). This improvement enables the NMN to still meet the assumption of noise pixel independence even in the presence of significant coherent noise, thereby enhancing the network's adaptability in complex noisy environments. Furthermore, the paper introduces the texture total variation regularization (TTV) technique, which effectively eliminates correlated noise by constraining the texture characteristics of the noise and improves signal fidelity while maintaining the local smooth structure of the effective signal. To further enhance the texture difference between the signal and noise, the paper proposes an asymmetric dilated convolution (ADConv) technique, which has a larger receptive field in specific directions and acts as an information filter in the network, further optimizing the noise removal effect. The proposed method relies solely on the observed noisy DAS-VSP data to remove correlated noise in a self-supervised manner without the need for any additional labeled data or prior information. Finally, through extensive experimental verification on synthetic and real DAS-VSP data, the proposed method demonstrates its superior performance and wide applicability in coherent noise suppression.

[0041] The present invention discloses a DAS-VSP data coherent noise suppression method based on adaptive texture analysis. The method utilizes a self-supervised deep learning method to train only original noisy data on distributed fiber vertical seismic (DAS-VSP) data to suppress coherent noise with spatial correlation, thereby providing basic data for subsequent seismic data processing. The method specifically includes the following steps: S1. Calculate the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data respectively, calculate the correlation between pixels and draw a spatial correlation map to guide the selection of training strategies; The derivative values ​​of the effective signal and noise at each point in the vertical and horizontal directions are calculated by convolving each pixel with the convolution kernels of [[-1], [0], [1]] and [[-1, 0, 1]], and then the derivative frequency graph is drawn; All data are then divided into 20*20 rectangular blocks, and for each block, the mean of the correlation coefficient relative to the fixed position block is calculated; Finally, a spatial correlation map between pixels at different positions is obtained.

[0042] The derivative frequency plot and spatial correlation plot reveal that the characteristics of DAS-VSP effective signals and noise have obvious differences in the vertical and horizontal directions.

[0043] S2. Expand the traditional blind spot network (BSN) into a neighborhood masking network (NMN), rotate the input data to construct multiple subgraphs, and perform translation and cropping to form a blind neighborhood, thereby expanding the receptive field of the blind spot; Based on U-net, for vertical noise, we first rotate the input data by 90° and 270° to create two branches with different directional information to form feature maps. After the convolution processing of each branch is completed, each feature map is spatially offset to ensure that the receptive field of each branch does not include the center pixel; Specifically, k Row zero elements are added to the top of each feature map before it is fed into the network and finally cropped from the bottom of the output. k row, that is, the size is Blind neighborhood mask of ; Finally, all feature maps are derotated and input into a 1×1 convolutional layer to fuse the two branches.

[0044] This small convolution kernel does not change the spatial dimension and is mainly used to adjust and synthesize feature channels from different branches. This not only maintains the spatial structure of the output, but also effectively integrates information from all directions. In addition, it allows the same convolution kernel to be used (parameter sharing) to process information entering from different directions, thereby improving training efficiency and reducing model complexity.

[0045] For horizontal noise, two feature maps are constructed by rotating 0° and 180° respectively, and the subsequent processing is exactly the same.

[0046] S3. According to the different correlation characteristics of noise, design the corresponding blind neighborhood masking strategy so that the network meets the condition that the noise pixels of the blind spot network are independent; The objective function of a common blind spot network is:

[0047] in, represents the loss function, represents the denoising network, and are the subscripts in the vertical and horizontal directions, Indicates the receptive field around the blind spot excluding itself, represents the noise input.

[0048] For vertical noisy data, increase the blind spot size from 1 to k , create a size of 2 k-1, and modify the receptive field of the middle pixel to:

[0049] Occlusion 2 k After -1 adjacent seismic trace, the receptive field of the middle pixel only includes several trace sets on the left and right sides. These trace sets already contain most of the information for predicting the effective signal components of the target pixel and the noise components contained in them are statistically irrelevant, which meets the training conditions of the blind spot network.

[0050] For horizontal noise data, select Mask k The adjacent time sampling periods are taken as the central pixel, and the receptive field of the middle pixel is:

[0051] Therefore, the objective function of the network is expressed as:

[0052] in, are network parameters, Remove the receptive field of vertical coherent noise for the network, Remove the receptive field of horizontally coherent noise for the network.

[0053] S4, using asymmetric dilated convolution (ADConv), which has a larger receptive field in a specific direction and further ignores the influence of coherent noise; For vertically noisy data, asymmetric dilated convolution expands the original 3*3 convolution kernel to a 3*5 convolution kernel and sets dilation factors in the horizontal direction with step sizes of 2, 3, and 4, respectively, while maintaining no dilation in the vertical direction. Testing has shown that compared to the traditional 3*3 convolution kernel, the receptive field in the horizontal direction is significantly increased, while the receptive field in the vertical direction remains unchanged.

[0054] For horizontally noisy data, the convolution kernel size is increased to 5*3, and an expansion factor is set in the vertical direction. This convolution design excludes noise information from the blind zone receptive field, greatly filtering out the influence of correlated noise. Without significantly increasing the number of parameters, it effectively expands the model's receptive field in specific directions, thereby improving the network's ability to recognize long-range spatially correlated features.

[0055] S5. Extending the ordinary total variation regularization to texture total variation regularization (TTV), the texture weight factor can be calculated according to the different characteristics of the signal and noise to further smooth the noise and preserve the signal; The traditional TV regularization formula is:

[0056] in, represents the data to be denoised, represents the TV regularization term, and represent the horizontal and vertical gradients respectively.

[0057] For level-correlated noisy data, only select Smooth noise, that is:

[0058] in, Represents the weight that determines the smoothing degree of noise and valid signal areas.

[0059] because It has a strong ability to separate noise and signal. For the noise area, The larger it is, the smoother the noise will be. For the signal area, The smaller it is, the less signal loss can be avoided.

[0060] The weight that determines the smoothing degree of noise and valid signal areas pass Calculation, that is:

[0061] in, represents the network input, Indicates the network output result, 0.01 means to prevent division by zero, Represents the scaling factor, whose value is determined by the coherence characteristics of the noise.

[0062] For different noise forms, the total variation regularization formula of texture is written as:

[0063] The corresponding overall objective function of the network is:

[0064] in, Represents the weight parameter. These two objective functions correspond to the vertical and horizontal noise denoising tasks of DAS-VSP data, respectively.

[0065] S6. This self-supervised method is applied to synthetic seismic data and real DAS-VSP seismic data, and the denoising effect is analyzed through quantitative indicators such as signal-to-noise ratio, structural similarity, and visualization effect.

[0066] The signal-to-noise ratio (SNR) and structural similarity (SSIM) were selected as quantitative indicators to measure the denoising performance of the network. The visualization effect was presented through the denoised data and the noise-removed data (the difference between the noisy data and the denoised data). The denoising performance was also analyzed using the local self-similarity graph of the denoised data and the time domain graph of the single-channel signal. The two were combined to analyze whether the network can well preserve the seismic signal while removing correlated noise.

[0067] Those skilled in the art will appreciate that various aspects of the present invention may be implemented as systems, methods, or program products. Accordingly, various aspects of the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, microcode, etc.), or in a combination of hardware and software, collectively referred to herein as "circuits," "modules," or "platforms."

[0068] In another embodiment of the present invention, a DAS-VSP data coherent noise suppression system based on adaptive texture analysis is provided. The system can be used to implement the above-mentioned DAS-VSP data coherent noise suppression method based on adaptive texture analysis. Specifically, the DAS-VSP data coherent noise suppression system based on adaptive texture analysis includes a calculation module, an expansion module, a masking module, a convolution module, a regularization module and a noise suppression module.

[0069] The calculation module calculates the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data, draws a spatial correlation map based on the correlation between pixels, and selects a training strategy. The extension module expands the blind spot network into a neighborhood masking network according to the selected training strategy, rotates the input data to construct multiple subgraphs, and performs translation and cropping to form a blind neighborhood; The masking module sets the corresponding masking strategy for the obtained blind neighborhood according to the different correlation characteristics of the noise, so that the neighborhood masking network satisfies the condition that the noise pixels of the blind spot network are independent; The convolution module uses asymmetric dilated convolution to ignore the influence of coherent noise in the obtained effective signal; The regularization module extends the conventional total variation regularization to texture total variation regularization. It calculates the texture weight factor based on the different characteristics of the signal and noise, further smoothes the noise and preserves the signal, and combines the masking strategy to build a self-supervised method. The noise suppression module processes the synthetic seismic data and the actual DAS-VSP seismic data based on the constructed self-supervision method to achieve coherent noise suppression of DAS-VSP data.

[0070] 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 may be 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, 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 DAS-VSP data coherent noise suppression method based on adaptive texture analysis, including: The vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data are calculated respectively, and a spatial correlation diagram is drawn according to the correlation between pixels, and a training strategy is selected; according to the selected training strategy, the blind spot network is expanded into a neighborhood masking network, and the input data is rotated to construct multiple subgraphs and then translated and cropped to form a blind neighborhood; according to the different correlation characteristics of the noise, corresponding masking strategies are set for the obtained blind neighborhoods, so that the neighborhood masking network satisfies the condition of pixel independence of the blind spot network noise; asymmetric dilated convolution is used to ignore the influence of coherent noise in the obtained effective signal; ordinary total variation regularization is extended to texture total variation regularization, and the texture weight factor is calculated according to the different characteristics of the signal and noise. The noise is further smoothed and the signal is retained. A self-supervised method is constructed in combination with the masking strategy; based on the constructed self-supervised method, the synthetic seismic data and actual DAS-VSP seismic data are processed to achieve coherent noise suppression in DAS-VSP data.

[0071] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a terminal device, used to store programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the terminal device and, of course, extended storage media supported by the terminal device. It may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by a processor. These instructions may be one or more computer programs (including program code). It should be noted that more specific examples (a non-exhaustive list) of computer-readable storage media herein include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk-read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0072] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, which carry readable program code. Such propagated data signals can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than a readable storage medium, which can send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0073] The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, and the like, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a stand-alone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0074] One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the DAS-VSP data coherence noise suppression method based on adaptive texture analysis in the above embodiment. One or more instructions stored in the computer-readable storage medium may be loaded and executed by the processor to perform the following steps: The vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data are calculated respectively, and a spatial correlation diagram is drawn according to the correlation between pixels, and a training strategy is selected; according to the selected training strategy, the blind spot network is expanded into a neighborhood masking network, and the input data is rotated to construct multiple subgraphs and then translated and cropped to form a blind neighborhood; according to the different correlation characteristics of the noise, corresponding masking strategies are set for the obtained blind neighborhoods, so that the neighborhood masking network satisfies the condition of pixel independence of the blind spot network noise; asymmetric dilated convolution is used to ignore the influence of coherent noise in the obtained effective signal; ordinary total variation regularization is extended to texture total variation regularization, and the texture weight factor is calculated according to the different characteristics of the signal and noise. The noise is further smoothed and the signal is retained. A self-supervised method is constructed in combination with the masking strategy; based on the constructed self-supervised method, the synthetic seismic data and actual DAS-VSP seismic data are processed to achieve coherent noise suppression in DAS-VSP data.

[0075] See also Figure 8 The terminal device is a computer device. 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 by the processor 61. When executed by the processor 61, the computer program 63 implements the DAS-VSP data coherent noise suppression method based on adaptive texture analysis in the embodiment. To avoid repetition, the details are not described here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the DAS-VSP data coherent noise suppression system based on adaptive texture analysis in the embodiment. To avoid repetition, the details are not described here.

[0076] The computer device 60 may be a desktop computer, a notebook computer, a PDA, a cloud server, or other computing devices. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. It will be understood by those skilled in the art that Figure 8 This is merely an example of the computer device 60 and does not constitute a limitation of the computer device 60 . The computer device 60 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.

[0077] The processor 61 may be a central processing unit (CPU), other general-purpose processors, central processing units (CPUs), graphics processors (GPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), other programmable logic devices, discrete gate or transistor logic devices, quantum computing-based data processing logic, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0078] 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.

[0079] 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 about to be output.

[0080] Any reference to memory, database, or other media used in the various embodiments provided herein 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 storage, 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. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0081] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0082] See also Figure 9 The terminal device is an electronic device 600, which is a general-purpose computing device. Components of the electronic device 600 may include, but are not limited to, at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), and a display unit 640.

[0083] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present invention described in the above method section of this specification. For example, the processing unit 610 can perform the following steps: Figure 1 Follow the steps shown in .

[0084] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0085] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0086] Bus 630 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0087] The electronic device 600 can also communicate with one or more external devices 700 (e.g., a keyboard, a pointing device, a Bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 600, and / or any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., a router, a modem, etc.). Such communication can occur via an input / output (I / O) interface 650. Furthermore, the electronic device 600 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 660. The network adapter 660 can communicate with other modules of the electronic device 600 via the bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the electronic device 600, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0088] 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, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can 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 of 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.

[0089] See also Figure 2 The spatial characteristics of two noise data and a synthetic clean signal data in the vertical and horizontal directions are analyzed, and derivative frequency plots and spatial correlation plots are given. The synthetic clean signal data is simulated using the complex-valued adaptive coefficient finite-difference frequency domain (CVAC-FDFD) method.

[0090] Specifically, for vertical noise, its vertical derivative is closer to zero, indicating that it is smoother in the vertical direction, while horizontal noise is smoother in the horizontal direction.

[0091] However, unlike the correlated noise, the horizontal and vertical derivatives of the synthetic data are very close, with no significant difference. The spatial correlation plot also shows that the correlated noise exhibits strong correlation in the corresponding direction, but very weak correlation in other directions. The signal, on the other hand, exhibits strong correlation in all directions.

[0092] Based on this observation, we design a neighborhood masking network (NMN), texture total variation regularization (TTV), and asymmetric dilated convolution (ADConv) strategy to fully utilize the different data characteristics of signals and noise in the horizontal and vertical directions.

[0093] Please refer to Figures 3(a) to 4(c), which respectively show the denoising effects of two synthetic seismic test data. Figures 3(a) and 4(a) show the noisy seismic data. It is obvious that the effective signal is contaminated by strong energy-coherent noise; Figures 3(b) and 4(b) show the results after denoising using the Neighborhood Masking Network (NMN) with Texture Total Variation Regularization (TTV) and Asymmetric Dilated Convolution (ADConv). It can be seen that the noise is largely removed and the seismic signal is clearer. Figure 3(c) shows the difference between the noisy data (Figure 3(a)) and the denoised data (Figure 3(b), and Figure 4(c) shows the difference between the noisy data (Figure 4(a)) and the denoised data (Figure 4(b). It can be seen that the seismic signal is basically intact and a large amount of noise is suppressed. From the quantitative indicator analysis, the signal-to-noise ratio (SNR) of the first synthetic data increased from the initial -2.63dB to 21.32dB, and the structural similarity (SSIM) increased from the initial 0.7251 to 0.9985; The signal-to-noise ratio of the second synthetic data increased from -1.05dB to 21.26dB, and the structural similarity (SSIM) increased from the initial 0.7915 to 0.9981. Both the signal-to-noise ratio and structural similarity were greatly improved.

[0094] The above analysis shows that the noise is well suppressed and the signal is well preserved in the synthetic seismic data.

[0095] Please refer to Figures 5(a) to 7(e) for further testing of the network's denoising performance using three actual DAS-VSP seismic data. Figures 5(a), 6(a), and 7(a) are actual noisy DAS-VSP seismic data. The first two data contain 192 channels, each with 6400 samples, and a sampling rate of 0.001s. The last data contains 512 channels, each with 2400 samples, and a sampling rate of 0.002s. It can be seen that a large amount of strong energy coherent noise and some random noise overwhelm the effective seismic signal, making it impossible to clearly identify. Figures 5(b), 6(b), and 7(b) show the data after denoising by the present invention. The quality of the seismic data is significantly improved, and the effective signals that are relatively weak due to noise interference become more continuous and clear. Correspondingly, Figures 5(c), 6(c), and 7(c) are the difference data between the noisy data and the denoised data. No visible signal leakage is shown in the removed noise part, and the noise is suppressed to a great extent; Furthermore, Figures 5(d), 6(d), and 7(d) analyze the local self-similarity plots of the denoised data and the noise-removed data. It can be seen that the local self-similarity of the two data sets is very weak, indicating almost no signal leakage. Subsequently, single-channel signal analysis was performed on the 150th, 100th, and 10th channels of three actual DAS-VSP seismic data sets, respectively. Figures 5(e), 6(e), and 7(e) show the results of the time-domain analysis of the single-channel signals, demonstrating that the noise is significantly suppressed. Actual DAS-VSP seismic data demonstrate that the present invention is suitable for effectively removing complex correlated noise while preserving effective seismic signals.

[0096] In summary, this paper presents a method and system for suppressing coherent noise in DAS-VSP data based on adaptive texture analysis. This method leverages the unique texture and data characteristics of various noise types associated with DAS-VSP data. This innovative approach is based entirely on existing observational data and requires no external training datasets. Specifically, by studying the spatial correlation within the noise, the blind spot network (BSN) is extended to a neighborhood masking network (NMN), which features a specialized neighborhood masking structure. This design strategically configures the receptive field to filter out noise-related information around the prediction blind spot while maintaining the integrity of the valid signal data. Furthermore, the method integrates texture total variation regularization (TTV) technology to intelligently adjust weights based on the different texture characteristics of the noise and signal. This method effectively smooths the noise while preserving the valid signal. Furthermore, an asymmetric dilated convolution (ADConv) technique is employed to reduce the impact of noise on blind spot prediction accuracy. This method excels in extracting valid signals obscured by correlated noise and preserving fine details of weak geological structures.

[0097] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and 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 into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into 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, and will not be repeated here.

[0098] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0099] 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. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0100] 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 merely illustrative. For example, the division of the modules or units is merely a logical functional division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, and can be electrical, mechanical, or other forms.

[0101] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0102] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0103] 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 process 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. When the computer program is executed by a processor, it can implement the steps of the above-mentioned various method embodiments. 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, USB flash drive, mobile hard disk, magnetic 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.

[0104] 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 block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0105] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0107] The above content is 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 DAS-VSP data coherent noise suppression method based on adaptive texture analysis, characterized in that: The following steps are involved: S1. Calculate the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data, draw a spatial correlation diagram based on the correlation between pixels, and select a training strategy; S2, according to the training strategy selected in step S1, the blind spot network is expanded into a neighborhood masking network, the input data is rotated to construct multiple sub-graphs, and the sub-graphs are translated and cropped to form a blind neighborhood; S3, setting a corresponding masking strategy for the blind neighborhood obtained in step S2 according to different noise correlation characteristics, so that the neighborhood masking network satisfies the condition that the noise pixels of the blind spot network are independent; S4, using asymmetric dilated convolution to ignore the influence of coherent noise in the effective signal obtained in step S1; S5, extending the ordinary total variation regularization to texture total variation regularization, calculating the texture weight factor according to the different characteristics of the signal and the noise, further smoothing the noise obtained in step S4 and retaining the signal, and constructing a self-supervised method in combination with the masking strategy obtained in step S3; S6. Process the synthetic seismic data and the actual DAS-VSP seismic data based on the self-supervision method constructed in step S5 to suppress the coherent noise of the DAS-VSP data.

2. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1 is characterized in that: Step S1 is specifically as follows: The derivative values ​​of the effective signal and noise at each point in the vertical and horizontal directions are calculated by convolving each pixel with the convolution kernels of [[-1], [0], [1]] and [[-1, 0, 1]], and the derivative frequency graph is drawn; Divide all data into 20*20 rectangular blocks and calculate the mean of the correlation coefficient of each block relative to the fixed position block; Finally, a spatial correlation map between pixels at different positions is obtained.

3. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1 is characterized in that: In step S2, based on U-net, for vertical noise, the input data is first rotated 90° and 270° to create two branches with different directional information to form feature maps. After the convolution processing of each branch is completed, each feature map is spatially offset; For horizontal noise, two feature maps are formed by rotating 0° and 180° respectively. After the convolution processing of each branch is completed, each feature map is spatially offset.

4. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 3 is characterized in that: Will k Row zero elements are added to the top of each feature map before it is fed into the network and finally cropped from the bottom of the output k The row size is Finally, all feature maps are derotated and input into a 1×1 convolutional layer for fusion of the two branches.

5. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1, characterized in that: In step S3, the objective function of the neighborhood shielding network is expressed as: in, represents the loss function, represents the denoising network, and are the subscripts in the vertical and horizontal directions, represents the noise input, are network parameters, Remove the receptive field of vertical coherent noise for the network, Remove the receptive field of horizontally coherent noise from the network.

6. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1, characterized in that: In step S4, for vertical noisy data, asymmetric dilated convolution expands the original 3*3 convolution kernel to a 3*5 convolution kernel, and sets the dilation factor in the horizontal direction with step sizes of 2, 3, and 4, respectively, while there is no dilation in the vertical direction; for horizontal noisy data, the convolution kernel size is expanded to 5*3, and the dilation factor is set in the vertical direction.

7. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1, characterized in that: In step S5, the texture total variation regularization of different noise forms is as follows: in, represents the network input, Represents the network output result, represents the scaling factor, and represent the horizontal and vertical gradients respectively, represents the proposed vertical regularization strategy for removing vertical coherent noise, Denotes the proposed horizontal regularization strategy for removing horizontal coherent noise.

8. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 7 is characterized in that: The overall objective function of the neighborhood shielding network is: in, represents the weight parameter, represents the loss function, represents the denoising network, and are the subscripts in the vertical and horizontal directions, represents the noise input, are network parameters, Remove the receptive field of vertical coherent noise for the network, Remove the receptive field of horizontally coherent noise from the network.

9. The DAS-VSP data coherent noise suppression method based on adaptive texture analysis according to claim 1, characterized in that: In step S6, the signal-to-noise ratio is selected as a quantitative indicator to measure the denoising performance of the network. The visualization effect is presented through the denoised data and the removed noise data, and the denoising performance is analyzed using the time domain diagram of the single-channel signal.

10. A DAS-VSP data coherent noise suppression system based on adaptive texture analysis, characterized in that: include: The calculation module calculates the vertical and horizontal derivatives and their frequency values ​​of the effective signal and noise in the synthetic DAS-VSP data, draws a spatial correlation map according to the correlation between pixels, and selects a training strategy; The extension module expands the blind spot network into a neighborhood masking network according to the selected training strategy, rotates the input data to construct multiple sub-graphs, and performs translation and cropping to form a blind neighborhood; The masking module sets the corresponding masking strategy for the obtained blind neighborhood according to the different correlation characteristics of the noise, so that the neighborhood masking network satisfies the condition that the noise pixels of the blind spot network are independent; The convolution module uses asymmetric dilated convolution to ignore the influence of coherent noise in the obtained effective signal; Regularization module, which expands the ordinary total variation regularization into texture total variation regularization, calculates the texture weight factor according to the different characteristics of signal and noise, further smoothes the obtained noise and retains the signal, and builds a self-supervised method in combination with the masking strategy; The noise suppression module processes the synthetic seismic data and the actual DAS-VSP seismic data based on the constructed self-supervision method to achieve coherent noise suppression of DAS-VSP data.

Images