A method, apparatus, equipment, and medium for suppressing linear surface wave noise in seismic data.
By constructing a denoising model using an adaptive multi-scale dilated convolutional network, the problem of low efficiency and reliance on manual intervention in seismic data noise suppression in existing technologies is solved, achieving intelligent surface wave noise suppression and improving the quality and stability of seismic data.
Patent Information
- Application Number
- CN202211231888.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-30
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2042-09-30
AI Technical Summary
Existing methods for suppressing noise in seismic data rely on strong manual intervention, resulting in low denoising efficiency and poor denoising performance when the assumptions about surface waves and effective signals do not match.
An adaptive multi-scale dilated convolutional network is adopted, and a denoising model is constructed by combining multiple convolutional modules. Intelligent surface wave noise suppression is achieved through signal-to-noise ratio evaluation, data augmentation, and loss value optimization.
This improved the data quality of seismic data, enhanced the network's ability to identify data at different scales, and ensured the stability and fidelity of the denoised data.
Smart Images

Figure CN117849878B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of seismic data optimization technology, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for suppressing linear surface wave noise in seismic data. Background Technology
[0002] Noise suppression in seismic data is a critical process in seismic data processing. Surface waves in seismic data typically exhibit characteristics such as low frequency, low intensity, and high amplitude, which severely impact the signal-to-noise ratio and imaging quality of seismic data.
[0003] Common surface wave noise suppression methods are based on the differences between surface waves and effective waves in the transform domain, such as the FK domain, wavelet domain, or curve domain. Existing technologies utilize the following methods to achieve noise suppression:
[0004] 1. A one-dimensional prediction error filter is used to separate the surface wave and the effective signal in the frequency domain based on the waveform tilt angle;
[0005] 2. The signal-to-noise separation effect is improved by using a local orthogonalization algorithm after the bandpass filter;
[0006] 3. Local time-frequency transformation (LTFT) was used to achieve better separation between the signal and the surface wave.
[0007] However, the above methods often contain assumptions about noise and signals. When the assumptions are close to the actual situation, the denoising effect is more significant. When there are significant differences between the two, the denoising effect will be poor. In addition, the above methods rely heavily on the design of the filter. However, due to the complexity of the filter parameter design, the denoising effect is difficult to guarantee.
[0008] Currently, various surface wave noise suppression algorithms based on traditional methods have been developed and applied to some extent in actual production. However, traditional methods are mainly limited by strong human intervention, and require a large amount of manual parameter tuning during the denoising process, which greatly affects the denoising efficiency.
[0009] In summary, existing technologies suffer from problems such as excessive manual intervention and low noise reduction efficiency in the process of suppressing seismic data noise. Summary of the Invention
[0010] To address the above-mentioned problems, embodiments of the present invention provide a method, apparatus, electronic device, and computer-readable storage medium for suppressing linear surface wave noise in seismic data.
[0011] In a first aspect, embodiments of the present invention provide a method for suppressing linear surface wave noise in seismic data, comprising:
[0012] Acquire actual earthquake data, evaluate the signal-to-noise ratio based on the actual earthquake data, and obtain training earthquake data;
[0013] An initial sample set is prepared based on the training seismic data, and data augmentation is performed on the sample data in the initial sample set to obtain a standard sample set;
[0014] A two-dimensional convolutional layer is constructed. A feature extraction module and an output module are generated based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer, respectively. A multi-scale dilated convolutional module is generated based on the normalization layer and the activation layer. An initial denoising model is constructed based on the feature extraction module, the multi-scale dilated convolutional module, and the output module.
[0015] The standard sample set is input into the initial denoising model for training to obtain the training denoising result. The loss value is calculated based on the training denoising result and the standard sample set to obtain the training loss value. The initial denoising model is optimized using the training loss value to obtain the standard denoising model.
[0016] Obtain the seismic data to be denoised, and perform convolution calculation on the model to be denoised using the standard denoising model to obtain the denoised seismic data.
[0017] According to an embodiment of the present invention, the step of performing signal-to-noise ratio evaluation based on the actual seismic data to obtain training seismic data includes:
[0018] The row and column pixel noise standard deviations are calculated on the actual seismic data to obtain the row noise standard deviation and the column noise standard deviation.
[0019] Calculate the row equivalent noise standard deviation and the column equivalent noise standard deviation based on the row noise standard deviation and the column noise standard deviation;
[0020] A reference noise standard deviation is selected from the row equivalent noise standard deviation and the column equivalent noise standard deviation, and the signal-to-noise ratio of the actual seismic data is calculated based on the reference noise standard deviation.
[0021] The actual seismic data is filtered based on the signal-to-noise ratio to obtain training seismic data.
[0022] According to an embodiment of the present invention, calculating the row equivalent noise standard deviation and the column equivalent noise standard deviation based on the row noise standard deviation and the column noise standard deviation includes:
[0023] The standard deviation of row equivalent noise and the standard deviation of column equivalent noise are calculated using the following formula:
[0024]
[0025]
[0026] in, The standard deviation of the equivalent noise of the row; The equivalent noise standard deviation of the column is given. Let c be the standard deviation of the noise in the i-th row, i∈(1,n), n is the total number of pixels in the row, and c is the identifier of the row noise; Let r be the standard deviation of the noise in the j-th row, j∈(1,m), where m is the total number of pixels in the column and r is the identifier of the row noise.
[0027] According to an embodiment of the present invention, the step of performing data augmentation on the sample data in the initial sample set to obtain a standard sample set includes:
[0028] The sample data is segmented and cropped to obtain cropped data;
[0029] The cropped data is resampled to obtain sampled data;
[0030] The sampled data are then aggregated to obtain a standard sample set.
[0031] According to an embodiment of the present invention, the step of generating a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer respectively includes:
[0032] The feature extraction module is obtained by sequentially stacking the two-dimensional convolutional layer, the normalization layer, and the activation layer;
[0033] The normalization layer, the activation layer, the two-dimensional convolutional layer, and the two-dimensional convolutional layer with 1 channel are stacked sequentially to obtain the output module.
[0034] According to an embodiment of the present invention, the step of generating a multi-scale dilated convolution module based on the normalization layer and the activation layer includes:
[0035] The dilation rate is determined based on the width and depth of the preset dilated convolutional network block, and the dilated convolutional layer is generated based on the dilation rate and the preset convolutional kernel.
[0036] The expansion rate is determined using the following formula:
[0037] dilationrate=mod((i-1)*2+j,10)
[0038] Where, dilation rate is the dilation rate; i is the depth of the dilated convolutional network block, 1≤i≤8; j is the width of the dilated convolutional network block, 0≤j≤2; mod is the modulo operator;
[0039] Dilated convolutional data blocks are generated based on the normalization layer, the activation layer, and the dilated convolutional layer.
[0040] A multi-scale dilated convolution module is generated based on the depth and the dilated convolution data block.
[0041] According to an embodiment of the present invention, the step of inputting the standard sample set into the initial denoising model for training to obtain the training denoising result includes:
[0042] The feature extraction module in the initial denoising model is used to extract features from the samples in the standard sample set to obtain seismic data image features;
[0043] The seismic data image features are multi-scale convolutionally performed using the multi-scale dilated convolution module in the initial denoising model to obtain a multi-scale feature vector.
[0044] The multi-scale feature vector is normalized using the output module of the initial denoising model to obtain the training denoising result.
[0045] Secondly, embodiments of the present invention provide a linear surface wave noise suppression device for seismic data, characterized in that it comprises:
[0046] The training seismic data generation module is used to acquire actual seismic data, evaluate the signal-to-noise ratio based on the actual seismic data, and obtain training seismic data.
[0047] A standard sample set generation module is used to prepare an initial sample set based on the training seismic data, and to perform data augmentation on the sample data in the initial sample set to obtain a standard sample set;
[0048] An initial denoising model generation module is used to construct a two-dimensional convolutional layer, generate a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer and a preset activation layer, generate a multi-scale dilated convolutional module based on the normalization layer and the activation layer, and construct an initial denoising model based on the feature extraction module, the multi-scale dilated convolutional module and the output module.
[0049] The standard denoising model generation module is used to input the standard sample set into the initial denoising model for training, obtain the training denoising result, calculate the loss value based on the training denoising result and the standard sample set, obtain the training loss value, and optimize the initial denoising model using the training loss value to obtain the standard denoising model.
[0050] The standard denoising module is used to acquire seismic data to be denoised, and to perform convolution calculations on the model to be denoised using the standard denoising model to obtain denoised seismic data.
[0051] Thirdly, embodiments of the present invention provide an electronic device, which includes:
[0052] processor;
[0053] Memory used to store the processor's executable instructions;
[0054] The processor is configured to execute the instructions to implement a linear surface wave noise suppression method for seismic data as described in the first aspect above.
[0055] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for suppressing linear surface wave noise in seismic data as described in the first aspect above.
[0056] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial effects:
[0057] The embodiments of this invention improve data quality by evaluating the signal-to-noise ratio of actual seismic data; enhance sample diversity and network generalization ability by performing data augmentation processing; enhance the nonlinear characteristics of the network and further accelerate convergence by establishing a feature extraction model; significantly increase the network's receptive field by establishing a multi-scale dilated convolution model while reducing network depth, enhancing the network's ability to identify data at different scales and achieving noise identification to a greater extent, thereby realizing intelligent processing of surface wave noise suppression; and ensure that the relative relationship between energy and amplitude of spatiotemporal domain data remains unchanged before and after denoising by calculating and optimizing the initial denoising model, thereby improving the stability of the denoised seismic data. Attached Figure Description
[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0059] Figure 1 The flowchart of the linear surface wave noise suppression method for seismic data according to Embodiment 1 of the present invention is shown.
[0060] Figure 2 The image shows the seismic data to be denoised according to Embodiment 1 of the present invention;
[0061] Figure 3 The image shows the denoised seismic data according to Embodiment 1 of the present invention;
[0062] Figure 4 The image shows the noise removed from the seismic data to be denoised according to Embodiment 1 of the present invention relative to the denoised seismic data;
[0063] Figure 5 The diagram shows the functional modules of the seismic data linear surface wave noise suppression device according to Embodiment 3 of the present invention.
[0064] Figure 6 The diagram shows the composition of an electronic device for implementing the linear surface wave noise suppression method for seismic data according to Embodiment 4 of the present invention. Detailed Implementation
[0065] The present disclosure will be further described below with reference to the embodiments shown in the accompanying drawings.
[0066] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0067] This invention proposes a linear surface wave noise suppression method for seismic data. Based on an adaptive multi-scale dilated convolutional network, a denoising model is built by combining multiple convolutional modules. This model can significantly increase the network's receptive field while reducing the network depth, enhancing the network's ability to identify data at different scales and achieving noise identification to a greater extent, thus realizing intelligent processing of surface wave noise suppression. The denoising model is used to train a diverse sample set for denoising, and loss calculation and model optimization are performed based on the training results to improve the denoising effect and denoising capability of the model, achieving linear surface wave denoising of the seismic data to be denoised.
[0068] Example 1
[0069] like Figure 1 As shown, this invention proposes a method for suppressing linear surface wave noise in seismic data, comprising the following steps:
[0070] S1. Obtain actual earthquake data, evaluate the signal-to-noise ratio based on the actual earthquake data, and obtain training earthquake data.
[0071] In this embodiment of the invention, the actual seismic data can be pre-stack seismic data, represented as a surface wave image containing features such as low frequency, low intensity, and high amplitude.
[0072] In this embodiment of the invention, high-quality seismic data can be selected from actual seismic data by evaluating the signal-to-noise ratio.
[0073] In this embodiment of the invention, the step of evaluating the signal-to-noise ratio based on the actual seismic data to obtain training seismic data includes:
[0074] The row and column pixel noise standard deviations are calculated on the actual seismic data to obtain the row noise standard deviation and the column noise standard deviation.
[0075] Calculate the row equivalent noise standard deviation and the column equivalent noise standard deviation based on the row noise standard deviation and the column noise standard deviation;
[0076] A reference noise standard deviation is selected from the row equivalent noise standard deviation and the column equivalent noise standard deviation, and the signal-to-noise ratio of the actual seismic data is calculated based on the reference noise standard deviation.
[0077] The actual seismic data is filtered based on the signal-to-noise ratio to obtain training seismic data.
[0078] In this embodiment of the invention, the larger of the row equivalent noise standard deviation and the column equivalent noise standard deviation can be selected as the reference noise standard deviation.
[0079] In this embodiment of the invention, the row noise standard deviation matrix and the column noise standard deviation matrix corresponding to the actual seismic data are as follows:
[0080]
[0081]
[0082] Where, σ col σ is the row noise standard deviation matrix corresponding to the actual seismic data; row This is the column noise standard deviation matrix corresponding to the actual seismic data. Let c be the standard deviation of the noise in the nth row, where n is the total number of pixels in the row, and c is the identifier of the row noise. Let r be the standard deviation of the noise in the m-th row, where m is the total number of pixels in the column and r is the identifier for the row noise.
[0083] In detail, in this embodiment of the invention, the standard deviation of row equivalent noise and the standard deviation of column equivalent noise are calculated using the following formula:
[0084]
[0085]
[0086] in, The standard deviation of the equivalent noise of the row; The equivalent noise standard deviation of the column is given. Let c be the standard deviation of the noise in the i-th row, i∈(1,n), n is the total number of pixels in the row, and c is the identifier of the row noise; Let r be the standard deviation of the noise in the j-th row, j∈(1,m), where m is the total number of pixels in the column and r is the identifier of the row noise.
[0087] In this embodiment of the invention, the image signal-to-noise ratio (SNR) reflects the ratio of valid information to invalid information in an image. Therefore, a higher SNR indicates higher quality seismic data, and vice versa.
[0088] S2. Prepare an initial sample set based on the training seismic data, and perform data augmentation on the sample data in the initial sample set to obtain a standard sample set.
[0089] In this embodiment of the invention, the initial sample set includes multiple original seismic data images and denoised seismic data images; in this embodiment of the invention, data samples can be prepared using TFRecord.
[0090] In this embodiment of the invention, the step of performing data augmentation on the sample data in the initial sample set to obtain a standard sample set includes:
[0091] The sample data is segmented and cropped to obtain cropped data;
[0092] The cropped data is resampled to obtain sampled data;
[0093] The sampled data are then aggregated to obtain a standard sample set.
[0094] In this embodiment of the invention, the sample data can be divided into blocks of 128*512 to control the size of the data to be denoised and improve the denoising effect. By resampling the truncated data, the sample diversity can be improved and the generalization ability of the subsequent denoising network can be enhanced.
[0095] S3. Construct a two-dimensional convolutional layer, generate a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer and a preset activation layer, generate a multi-scale dilated convolutional module based on the normalization layer and the activation layer, and construct an initial denoising model based on the feature extraction module, the multi-scale dilated convolutional module and the output module.
[0096] In this embodiment of the invention, the two-dimensional convolutional layer can be composed of a convolutional layer with 16 channels and a 1x1 kernel.
[0097] In this embodiment of the invention, the step of generating a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer respectively includes:
[0098] The feature extraction module is obtained by sequentially stacking the two-dimensional convolutional layer, the normalization layer, and the activation layer;
[0099] The normalization layer, the activation layer, the two-dimensional convolutional layer, and the two-dimensional convolutional layer with 1 channel are stacked sequentially to obtain the output module.
[0100] In this embodiment of the invention, the convolution kernel in the two-dimensional convolutional layer with 1 channel can also be 1x1.
[0101] In this embodiment of the invention, the step of generating a multi-scale dilated convolution module based on the normalization layer and the activation layer includes:
[0102] The dilation rate is determined based on the width and depth of the preset dilated convolutional network block, and the dilated convolutional layer is generated based on the dilation rate and the preset convolutional kernel.
[0103] Dilated convolutional data blocks are generated based on the normalization layer, the activation layer, and the dilated convolutional layer.
[0104] A multi-scale dilated convolution module is generated based on the depth and the dilated convolution data block.
[0105] Furthermore, in this embodiment of the invention, the expansion rate can be determined using the following formula:
[0106] dilationrate=mod((i-1)*2+j,10)
[0107] Where, dilationrate is the dilation rate; i is the depth of the dilated convolutional network block, 1≤i≤8; j is the width of the dilated convolutional network block, 0≤j≤2; and mod is the modulo operator.
[0108] In this embodiment of the invention, the width can be set to 2, the size of the convolution kernel in the dilated convolutional layer can be set to 3x3, and the depth can be set to 8.
[0109] In this embodiment of the invention, an initial denoising model is constructed by sequentially using a feature extraction module, a multi-scale dilated convolution module, and an output module. The network layer structure of this model can be as follows: a two-dimensional convolutional layer with 16 channels and a 1x1 kernel, a batch normalization layer, an activation layer (with ReLU activation function), a dilated convolutional data block with a depth of 8 (the data block structure is: a batch normalization layer, an activation layer (with ReLU activation function), two dilated convolutional layers with a width of 2, wherein the kernel size of the dilated convolutional layer is 3x3), a batch normalization layer, an activation layer (with ReLU activation function), a two-dimensional convolutional layer with 16 channels and a 1x1 kernel, and a two-dimensional convolutional layer with 1 channel and a 1x1 kernel.
[0110] S4. Input the standard sample set into the initial denoising model for training to obtain the training denoising result. Calculate the loss value based on the training denoising result and the standard sample set to obtain the training loss value. Optimize the initial denoising model using the training loss value to obtain the standard denoising model.
[0111] In this embodiment of the invention, the step of inputting the standard sample set into the initial denoising model for training to obtain the trained denoising result includes:
[0112] The feature extraction module in the initial denoising model is used to extract features from the samples in the standard sample set to obtain seismic data image features;
[0113] The seismic data image features are multi-scale convolutionally performed using the multi-scale dilated convolution module in the initial denoising model to obtain a multi-scale feature vector.
[0114] The multi-scale feature vector is normalized using the output module of the initial denoising model to obtain the training denoising result.
[0115] In this embodiment of the invention, preliminary feature extraction of samples can be achieved through the two-dimensional convolutional layer in the feature extraction module. Subsequent normalization and activation layers in the feature extraction module accelerate feature convergence, thus achieving feature extraction of the samples. The multi-scale dilated convolution module, through dilated convolution blocks of a certain depth, uses a multi-scale dilation operator based on the dilation rate within the dilated convolution blocks to significantly increase the network's receptive field while reducing network depth, enhancing the network's ability to recognize data at different scales and achieving greater noise recognition. The output module maintains the relative relationship between the energy and amplitude of the spatiotemporal data corresponding to the samples after feature extraction and multi-scale dilation.
[0116] In this embodiment of the invention, the L2 norm (least squared error, LSE) loss function can be used to calculate the loss value between the training denoising result and the denoised seismic data in the standard sample set. Based on the loss value calculation result, the Adam optimizer is used to iteratively update the initial denoising model.
[0117] S5. Obtain the seismic data to be denoised, and perform convolution calculation on the model to be denoised using the standard denoising model to obtain the denoised seismic data.
[0118] In this embodiment of the invention, the standard denoising model is used to perform convolution calculation on the model to be denoised to obtain denoised seismic data. This process is similar to the process in step S4 above, where the standard sample set is input into the initial denoising model for training to obtain the training denoising result. Therefore, it will not be described in detail here.
[0119] In this embodiment of the invention, for Figure 2 The seismic data to be denoised shown (data acquisition time: 0–5000 s) can be obtained by performing convolution calculations using the standard denoising model described above. Figure 3 The denoised seismic data shown (data acquisition time: 0–5000 s) has had scattered surface wave noise removed from single shots, along with the seismic data to be denoised. Figure 3 The denoised seismic data in the middle is relatively Figure 2 The noise to be removed from the seismic data to be denoised is as follows: Figure 4 (Data acquisition time: 0-5000s) As shown. This embodiment of the invention can effectively suppress scattered surface wave noise in a single shot, and the removed noise does not contain effective waves, thus achieving fidelity in the noise removal process.
[0120] The embodiments of this invention improve data quality by evaluating the signal-to-noise ratio of actual seismic data; enhance sample diversity and network generalization ability by performing data augmentation processing; enhance the nonlinear characteristics of the network and further accelerate convergence by establishing a feature extraction model; significantly increase the network's receptive field by establishing a multi-scale dilated convolution model while reducing network depth, enhancing the network's ability to identify data at different scales and achieving noise identification to a greater extent, thereby realizing intelligent processing of surface wave noise suppression; and ensure that the relative relationship between energy and amplitude of spatiotemporal domain data remains unchanged before and after denoising by calculating and optimizing the initial denoising model, thereby improving the stability of the denoised seismic data.
[0121] Example 2
[0122] To better understand the present invention, a second embodiment is provided below to further explain the data augmentation of sample data in the initial sample set.
[0123] In this embodiment of the invention, the step of performing data augmentation on the sample data in the initial sample set to obtain a standard sample set includes:
[0124] The sample data is subjected to Gabor filtering transformation to obtain multi-directional, multi-scale depth images;
[0125] The region to be repaired in a multi-directional, multi-scale depth image is obtained. The target mask generation method is determined based on the edge position of the region to be repaired and the depth image. The mask of the region to be repaired in the multi-directional, multi-scale depth image is determined based on the target mask generation method.
[0126] By combining the mask of the area to be repaired with the fast traversal algorithm, holes in the area to be repaired in the multi-directional and multi-scale depth image are filled to obtain the repaired depth image.
[0127] The repaired depth image is subjected to median filtering to obtain an enhanced depth image. A standard sample set is then constructed based on the depth image and the sample data.
[0128] In this embodiment of the invention, the method for image enhancement after filling holes in the area to be repaired may further include performing histogram equalization, bilateral filter filtering, Sobel operator boundary extraction, watershed algorithm-based segmentation, average value filling of segmented blocks, and pixel quantization on the hole-repaired depth image in sequence.
[0129] Example 3
[0130] like Figure 5 As shown in the figure, this embodiment also provides a functional block diagram of a linear surface wave noise suppression device for seismic data.
[0131] The seismic data linear surface wave noise suppression device 100 described in this embodiment can be installed in an electronic device. Depending on the functions implemented, the seismic data linear surface wave noise suppression device 100 may include a training seismic data generation module 101, a standard sample set generation module 102, an initial denoising model generation module 103, a standard denoising model generation module 104, and a standard denoising module 105. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.
[0132] In this embodiment, the functions of each module / unit are as follows:
[0133] The training earthquake data generation module 101 is used to acquire actual earthquake data, perform signal-to-noise ratio evaluation based on the actual earthquake data, and obtain training earthquake data.
[0134] The standard sample set generation module 102 is used to prepare an initial sample set based on the training seismic data, and to perform data augmentation on the sample data in the initial sample set to obtain a standard sample set.
[0135] The initial denoising model generation module 103 is used to construct a two-dimensional convolutional layer, generate a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer and a preset activation layer, generate a multi-scale dilated convolutional module based on the normalization layer and the activation layer, and construct an initial denoising model based on the feature extraction module, the multi-scale dilated convolutional module and the output module.
[0136] The standard denoising model generation module 104 is used to input the standard sample set into the initial denoising model for training to obtain the training denoising result, calculate the loss value based on the training denoising result and the standard sample set to obtain the training loss value, and optimize the initial denoising model using the training loss value to obtain the standard denoising model.
[0137] The standard denoising module 105 is used to acquire seismic data to be denoised, and to perform convolution calculations on the model to be denoised using the standard denoising model to obtain denoised seismic data.
[0138] Example 4
[0139] like Figure 6 As shown, this embodiment also provides a computer electronic device, which may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as a seismic data linear surface wave noise suppression program.
[0140] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., executing a linear surface wave noise suppression program for seismic data) and calls data stored in the memory 11 to perform various functions of the electronic device and process data.
[0141] The memory 11 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device, such as a plug-in portable hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 11 can include both internal and external storage units of the electronic device. The memory 11 can be used not only to store application software and various types of data installed on the electronic device, such as the code for a linear surface wave noise suppression program for seismic data, but also to temporarily store data that has been output or will be output.
[0142] The communication bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.
[0143] The communication interface 13 is used for communication between the aforementioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, Bluetooth interface, etc.), typically used to establish communication connections between the electronic device and other electronic devices. The user interface may be a display, an input unit (such as a keyboard), or, optionally, a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device and to display a visual user interface.
[0144] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.
[0145] For example, although not shown, the electronic device may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.
[0146] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.
[0147] The seismic data linear surface wave noise suppression program stored in the memory 11 of the electronic device is a combination of multiple instructions. When run in the processor 10, it can achieve the following:
[0148] Acquire actual earthquake data, evaluate the signal-to-noise ratio based on the actual earthquake data, and obtain training earthquake data;
[0149] An initial sample set is prepared based on the training seismic data, and data augmentation is performed on the sample data in the initial sample set to obtain a standard sample set;
[0150] A two-dimensional convolutional layer is constructed. A feature extraction module and an output module are generated based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer, respectively. A multi-scale dilated convolutional module is generated based on the normalization layer and the activation layer. An initial denoising model is constructed based on the feature extraction module, the multi-scale dilated convolutional module, and the output module.
[0151] The standard sample set is input into the initial denoising model for training to obtain the training denoising result. The loss value is calculated based on the training denoising result and the standard sample set to obtain the training loss value. The initial denoising model is optimized using the training loss value to obtain the standard denoising model.
[0152] Obtain the seismic data to be denoised, and perform convolution calculation on the model to be denoised using the standard denoising model to obtain the denoised seismic data.
[0153] Specifically, the specific implementation method of the processor 10 for the above instructions can be referred to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, and will not be repeated here.
[0154] Furthermore, if the modules / units integrated into the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).
[0155] Example 5
[0156] This embodiment provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the linear surface wave noise suppression method for seismic data as described above.
[0157] This program code can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 Steps of a specified function in one or more processes.
[0158] Storage media include permanent and non-permanent, removable and non-removable media, and can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by computing devices.
[0159] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application. When the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.
[0160] It should be understood that the terms used in this way can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in a sequence other than those illustrated or described herein.
[0161] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.
[0162] The modules described as separate components may or may not be physically separate. The components shown as modules 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 the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0163] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.
[0164] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0165] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.
[0166] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.
[0167] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.
[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for suppressing linear surface wave noise in seismic data, characterized in that, The method includes: Acquire actual earthquake data, evaluate the signal-to-noise ratio based on the actual earthquake data, and obtain training earthquake data; An initial sample set is prepared based on the training seismic data, and data augmentation is performed on the sample data in the initial sample set to obtain a standard sample set; A two-dimensional convolutional layer is constructed. A feature extraction module and an output module are generated based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer, respectively. A multi-scale dilated convolutional module is generated based on the normalization layer and the activation layer. An initial denoising model is constructed based on the feature extraction module, the multi-scale dilated convolutional module, and the output module. The standard sample set is input into the initial denoising model for training to obtain the training denoising result. The loss value is calculated based on the training denoising result and the standard sample set to obtain the training loss value. The initial denoising model is optimized using the training loss value to obtain the standard denoising model. Obtain the seismic data to be denoised, and perform convolution calculation on the seismic data to be denoised using the standard denoising model to obtain denoised seismic data; The step of generating a multi-scale dilated convolution module based on the normalization layer and the activation layer includes: The dilation rate is determined based on the width and depth of the preset dilated convolutional network block, and the dilated convolutional layer is generated based on the dilation rate and the preset convolutional kernel. The expansion rate is determined using the following formula: in, For expansion rate; To increase the depth of convolutional network blocks, ; To expand the width of convolutional network blocks, ; For the remainder operator; Dilated convolutional data blocks are generated based on the normalization layer, the activation layer, and the dilated convolutional layer. A multi-scale dilated convolution module is generated based on the depth and the dilated convolution data block.
2. The method for suppressing linear surface wave noise in seismic data as described in claim 1, characterized in that, The step of evaluating the signal-to-noise ratio based on the actual seismic data to obtain training seismic data includes: The row and column pixel noise standard deviations are calculated on the actual seismic data to obtain the row noise standard deviation and the column noise standard deviation. Calculate the row equivalent noise standard deviation and the column equivalent noise standard deviation based on the row noise standard deviation and the column noise standard deviation; A reference noise standard deviation is selected from the row equivalent noise standard deviation and the column equivalent noise standard deviation, and the signal-to-noise ratio of the actual seismic data is calculated based on the reference noise standard deviation. The actual seismic data is filtered based on the signal-to-noise ratio to obtain training seismic data.
3. The method for suppressing linear surface wave noise in seismic data as described in claim 1, characterized in that, The step of performing data augmentation on the sample data in the initial sample set to obtain a standard sample set includes: The sample data is segmented and cropped to obtain cropped data; The cropped data is resampled to obtain sampled data; The sampled data are then aggregated to obtain a standard sample set.
4. The method for suppressing linear surface wave noise in seismic data as described in claim 1, characterized in that, The step of generating a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer, and a preset activation layer, respectively, includes: The feature extraction module is obtained by sequentially stacking the two-dimensional convolutional layer, the normalization layer, and the activation layer; The normalization layer, the activation layer, the two-dimensional convolutional layer, and the two-dimensional convolutional layer with 1 channel are stacked sequentially to obtain the output module.
5. The method for suppressing linear surface wave noise in seismic data as described in any one of claims 1 to 4, characterized in that, The step of inputting the standard sample set into the initial denoising model for training to obtain the trained denoising result includes: The feature extraction module in the initial denoising model is used to extract features from the samples in the standard sample set to obtain seismic data image features; The seismic data image features are multi-scale convolutionally performed using the multi-scale dilated convolution module in the initial denoising model to obtain a multi-scale feature vector. The multi-scale feature vector is normalized using the output module of the initial denoising model to obtain the training denoising result.
6. A device for suppressing linear surface wave noise in seismic data, characterized in that, The device includes: The training seismic data generation module is used to acquire actual seismic data, evaluate the signal-to-noise ratio based on the actual seismic data, and obtain training seismic data. A standard sample set generation module is used to prepare an initial sample set based on the training seismic data, and to perform data augmentation on the sample data in the initial sample set to obtain a standard sample set; An initial denoising model generation module is used to construct a two-dimensional convolutional layer, generate a feature extraction module and an output module based on the two-dimensional convolutional layer, a preset normalization layer and a preset activation layer, generate a multi-scale dilated convolutional module based on the normalization layer and the activation layer, and construct an initial denoising model based on the feature extraction module, the multi-scale dilated convolutional module and the output module. The standard denoising model generation module is used to input the standard sample set into the initial denoising model for training, obtain the training denoising result, calculate the loss value based on the training denoising result and the standard sample set, obtain the training loss value, and optimize the initial denoising model using the training loss value to obtain the standard denoising model. A standard denoising module is used to acquire seismic data to be denoised, and to perform convolution calculations on the seismic data to be denoised using the standard denoising model to obtain denoised seismic data. The step of generating a multi-scale dilated convolution module based on the normalization layer and the activation layer includes: The dilation rate is determined based on the width and depth of the preset dilated convolutional network block, and the dilated convolutional layer is generated based on the dilation rate and the preset convolutional kernel. The expansion rate is determined using the following formula: in, For expansion rate; To increase the depth of convolutional network blocks, ; To expand the width of convolutional network blocks, ; For the remainder operator; Dilated convolutional data blocks are generated based on the normalization layer, the activation layer, and the dilated convolutional layer. A multi-scale dilated convolution module is generated based on the depth and the dilated convolution data block.
7. An electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the seismic data linear surface wave noise suppression method as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method for suppressing linear surface wave noise in seismic data as described in any one of claims 1 to 5.
