Earthquake surface wave suppression method, device, equipment and medium

Abstract

The invention provides a seismic surface wave suppression method and device, equipment and a medium. Comprising the steps of obtaining a to-be-processed seismic signal; the seismic signal to be processed is input to a preset pulse neural network, the pulse neural network comprises an input layer, a hidden layer and an output layer, the input layer is connected with the hidden layer, and the hidden layer is connected with the output layer; performing conversion processing on the seismic signal to be processed through the input layer to obtain a corresponding pulse sequence; performing feature extraction processing on the pulse sequence through the hidden layer to obtain a plurality of features, and performing surface wave filtering processing according to the plurality of features to obtain a target feature sequence; and performing signal reconstruction processing on the target feature sequence through an output layer to obtain a to-be-processed seismic signal of the target. The surface wave interference is effectively reduced, and the surface wave suppression efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of exploration seismic data processing technology, and in particular to a seismic surface wave suppression method, apparatus, equipment and medium. Background Technology

[0002] In seismic exploration, surface waves have coherent noise characteristics of low frequency, low velocity, and high amplitude, which reduces the signal-to-noise ratio of seismic data and interferes with the extraction of effective signals. Therefore, surface waves need to be suppressed when processing seismic exploration data.

[0003] In existing technologies, deep learning methods based on convolutional neural networks are commonly used to process surface waves. However, deep learning methods based on convolutional neural networks have high computational energy consumption, poor processing performance in areas such as time-frequency aliasing of surface waves and effective waves, nonlinear interference caused by complex near-surface structures, and are prone to damaging effective signals, as well as insufficient ability to capture time-varying features.

[0004] Therefore, existing technologies suffer from poor suppression of seismic surface waves and low processing efficiency. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for suppressing seismic surface waves, which can achieve the effect of accurately suppressing seismic surface waves.

[0006] In a first aspect, embodiments of this application provide a seismic surface wave suppression method, including:

[0007] Acquire the seismic signal to be processed;

[0008] The seismic signal to be processed is input into a preset spiking neural network, which includes an input layer, a hidden layer, and an output layer. The input layer is connected to the hidden layer, and the hidden layer is connected to the output layer.

[0009] The input layer transforms the seismic signal to be processed to obtain the corresponding pulse sequence.

[0010] The pulse sequence is processed by a hidden layer to extract multiple features, and then surface wave filtering is performed based on these features to obtain the target feature sequence.

[0011] The target feature sequence is reconstructed through the output layer to obtain the target seismic signal.

[0012] In one possible implementation, the hidden layer includes a multi-scale convolutional kernel and a leakage integral firing model, wherein the multi-scale convolutional kernel and the leakage integral firing model are connected.

[0013] Accordingly, feature extraction processing is performed on the pulse sequence through a hidden layer to obtain multiple features, and surface wave filtering is performed based on these features to obtain the target feature sequence, including:

[0014] By using multi-scale convolution kernels, feature extraction processing is performed on the pulse sequence to obtain multiple features;

[0015] By utilizing the membrane potential accumulation characteristics of neurons in the leakage current integral firing model, surface wave energy accumulation regions are detected. During the detection process of surface wave energy accumulation regions, the neuron threshold is adaptively adjusted to obtain the target feature sequence.

[0016] In one possible implementation, the seismic signal to be processed is transformed through the input layer to obtain a corresponding pulse sequence, including:

[0017] The input layer normalizes the seismic signal to be processed to obtain a normalized seismic signal.

[0018] Based on the first preset dynamic threshold, the normalized seismic signal is encoded to obtain the corresponding pulse sequence.

[0019] In one possible implementation, the target feature sequence is reconstructed through an output layer to obtain the target seismic signal, including:

[0020] Through the output layer, a preset time-frequency joint loss function is used to perform signal reconstruction processing on the target feature sequence to obtain the target seismic signal. The preset time-frequency joint loss function incorporates a preset impulse sparsity constraint.

[0021] In one possible implementation, before inputting the seismic signal to be processed into a preset spiking neural network, the method further includes:

[0022] An input layer is established based on the preset normalization processing module and the preset pulse sequence conversion module;

[0023] A hidden layer is established based on a pre-defined multi-scale convolutional kernel and a pre-defined leakage integral distribution model;

[0024] The output layer is established based on a preset time-frequency joint loss function that incorporates preset impulse sparsity constraints;

[0025] A predefined spiking neural network is generated based on the input layer, hidden layer, and output layer.

[0026] In one possible implementation, after generating a preset spiking neural network based on the input layer, hidden layer, and output layer, the method further includes:

[0027] Seismic surface wave training samples are obtained, which include multiple seismic signal samples with seismic surface waves.

[0028] Based on the seismic surface wave training samples, a preset pulse neural network is trained to optimize the preset pulse neural network.

[0029] In one possible implementation, the pre-defined spiking neural network includes multiple synaptic matrices;

[0030] Accordingly, based on the seismic surface wave training samples, a pre-defined pulse neural network is trained to optimize the pre-defined pulse neural network, including:

[0031] Based on seismic surface wave training samples, multiple synaptic matrices are trained to optimize a pre-defined spiking neural network.

[0032] Secondly, embodiments of this application provide a seismic surface wave suppression device, including: a memory and a processor;

[0033] The memory stores instructions that the computer executes;

[0034] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0035] Thirdly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0036] Fourthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0037] This application provides a seismic surface wave suppression method, apparatus, device, and medium. The method involves acquiring a seismic signal to be processed; inputting the seismic signal to be processed into a preset spiking neural network, wherein the spiking neural network includes an input layer, a hidden layer, and an output layer, with the input layer connected to the hidden layer and the hidden layer connected to the output layer; transforming the seismic signal to be processed through the input layer to obtain a corresponding pulse sequence; extracting features from the pulse sequence through the hidden layer to obtain multiple features, and performing surface wave filtering based on these features to obtain a target feature sequence; and reconstructing the target feature sequence through the output layer to obtain the target seismic signal. Compared to existing technologies, this application uses a preset spiking neural network to process the seismic signal in a hierarchical manner, achieving feature extraction, surface wave filtering, and signal reconstruction of the seismic signal, thus improving the accuracy and efficiency of seismic signal processing and enhancing the suppression effect of seismic surface waves. Attached Figure Description

[0038] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0039] Figure 1 A schematic diagram of a seismic surface wave suppression system architecture is provided in this application;

[0040] Figure 2 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 1 ;

[0041] Figure 3 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 2 ;

[0042] Figure 4 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 3 ;

[0043] Figure 5 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 4 ;

[0044] Figure 6 Comparison diagram of output target seismic signals provided for this application;

[0045] Figure 7 Comparison diagram of output target seismic signal profiles provided for this application;

[0046] Figure 8 A schematic diagram of the seismic surface wave suppression device provided in this application;

[0047] Figure 9This is a structural schematic diagram of the seismic surface wave suppression device provided in this application.

[0048] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0049] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0050] It should be noted that all data involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0051] Seismic exploration is a method that uses seismic waves to detect underground geological structures. It mainly involves generating artificial seismic waves on or below the surface and recording data on the propagation and reflection of these artificial seismic waves underground. By analyzing this data, the structure and properties of underground rock strata can be inferred.

[0052] Seismic surface waves are seismic waves that propagate along the Earth's surface or interface. Because surface waves have strong energy, they can mask effective signals such as deep reflection waves, affecting data quality. Furthermore, seismic surface waves have the characteristics of low frequency and slow propagation speed, which leads to a reduction in the resolution of seismic data. Therefore, it is necessary to suppress seismic surface waves to improve the accuracy of seismic data.

[0053] In existing technologies, seismic surface waves can be filtered using methods such as frequency filtering, wavefield separation, or wavelet transform; or deep learning methods based on convolutional neural networks can be used to process surface waves. However, methods such as frequency filtering, wavefield separation, or wavelet transform suffer from high-frequency information loss and insufficient ability to capture time-varying features; while deep learning methods based on convolutional neural networks have high computational energy consumption, poor processing effects on time-frequency aliasing between seismic surface waves and effective waves, and nonlinear interference caused by complex near-surface structures, and are prone to damaging effective signals.

[0054] As a result, existing technologies suffer from poor suppression of seismic surface waves, leading to a large signal-to-noise ratio in seismic data and inaccurate data.

[0055] To address the aforementioned issues, the core concept of this application is as follows: a pre-defined spiking neural network comprising an input layer, a hidden layer, and an output layer is established, wherein the input layer of the spiking neural network is connected to the hidden layer, and the hidden layer is connected to the output layer; the seismic signal to be processed is input into the pre-defined spiking neural network, and the input layer performs transformation processing on the seismic signal to obtain a corresponding pulse sequence; the hidden layer performs feature extraction processing on the pulse sequence to obtain multiple features, and surface wave filtering processing is performed based on the multiple features to obtain a target feature sequence; the output layer performs signal reconstruction processing on the target feature sequence to obtain the target seismic signal, thereby improving the accuracy of seismic signal processing and the suppression effect of seismic surface waves.

[0056] Optionally, Figure 1 This is a schematic diagram of a seismic surface wave suppression system architecture provided in this application. Figure 1 As shown, the seismic surface wave suppression system architecture includes at least one of a data acquisition device 101, a processing device 102, and a display device 103.

[0057] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the above architecture. In other feasible embodiments of this application, the above architecture may include more or fewer components than illustrated, or combine some components, or split some components, or arrange different components, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown can be implemented in hardware, software, or a combination of both.

[0058] In the specific implementation process, the data acquisition device 101 may include an input / output interface or a communication interface, and the data acquisition device 101 can be connected to the processing device through the input / output interface or the communication interface.

[0059] The processing device 102 can input the seismic signal to be processed into a preset pulse neural network, wherein the pulse neural network includes an input layer, a hidden layer, and an output layer, the input layer is connected to the hidden layer, and the hidden layer is connected to the output layer; through the input layer, the seismic signal to be processed is transformed to obtain the corresponding pulse sequence; through the hidden layer, the pulse sequence is subjected to feature extraction processing to obtain multiple features, and surface wave filtering processing is performed based on the multiple features to obtain the target feature sequence; through the output layer, the target feature sequence is subjected to signal reconstruction processing to obtain the target seismic signal.

[0060] The display device 103 can also be a touch screen or the screen of a terminal device, used to receive user commands while displaying the above-mentioned content, so as to realize interaction with the user.

[0061] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0062] Figure 2 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 1 ,like Figure 2 As shown, the method includes:

[0063] S201. Obtain the seismic signal to be processed.

[0064] In this embodiment, the seismic signal to be processed is an artificially excited seismic wave that propagates underground and returns to the surface. The seismic signal to be processed includes seismic waveform, amplitude, frequency, phase, time information, polarization, energy, and source information. Among them, the seismic waveform includes transverse waves, longitudinal waves, reflected waves, refracted waves, direct waves, and surface waves.

[0065] S202. Input the seismic signal to be processed into a preset spiking neural network, wherein the spiking neural network includes an input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer and the hidden layer is connected to the output layer.

[0066] In this embodiment, the preset spiking neural network is a neural network model that simulates the dynamic characteristics of biological neurons, and transmits and processes information through pulses.

[0067] Furthermore, the spiking neural network adopts a hierarchical structure of input layer, hidden layer and output layer, which improves the real-time processing efficiency of the seismic signals to be processed.

[0068] S203. The seismic signal to be processed is transformed through the input layer to obtain the corresponding pulse sequence.

[0069] In this embodiment, the seismic signal to be processed is converted into a corresponding pulse sequence through the input layer of a preset spiking neural network, which effectively reduces data redundancy and preserves the temporal characteristics of the seismic signal, enabling real-time processing of the seismic signal and facilitating the extraction of multi-scale features.

[0070] S204. Through the hidden layer, feature extraction processing is performed on the pulse sequence to obtain multiple features, and surface wave filtering processing is performed based on the multiple features to obtain the target feature sequence.

[0071] Optionally, the hidden layer includes a multi-scale convolutional kernel and a leakage integral firing model, with the multi-scale convolutional kernel and the leakage integral firing model connected together.

[0072] In this embodiment, the hidden layer extracts multiple features from the pulse sequence through a multi-scale convolution kernel and a leakage current integral firing model, thereby obtaining multiple features; and detects the surface wave energy accumulation region based on the membrane potential accumulation characteristics, dynamically adjusts the neuron threshold, thereby realizing surface wave filtering processing, generating the target feature sequence, and preserving the effective signal.

[0073] S205. Through the output layer, the target feature sequence is processed for signal reconstruction to obtain the target seismic signal.

[0074] Optionally, the target feature sequence is reconstructed through the output layer to obtain the target seismic signal, including:

[0075] Through the output layer, a preset time-frequency joint loss function is used to perform signal reconstruction processing on the target feature sequence to obtain the target seismic signal. The preset time-frequency joint loss function incorporates a preset impulse sparsity constraint.

[0076] In this embodiment, the preset time-frequency joint loss function is shown in the following formula:

[0077]

[0078] Where L is the total loss value, used to measure the comprehensive loss function value of signal reconstruction processing; S is the seismic signal to be processed; The target seismic signal is represented by STFT, which is used to analyze the time-frequency characteristics of the signal. λ1 and λ2 are the trigger rates of the pulse sparsity constraint, which are used to reduce the loss of high-frequency information. In this embodiment, the trigger rate of the pulse sparsity constraint ranges from 0.2 to 0.8. For example, λ1 and λ2 can be 0.7 and 0.3, respectively.

[0079] The seismic surface wave suppression method provided in this application acquires the seismic signal to be processed, and sequentially converts the seismic signal into a pulse sequence through the input layer, hidden layer, and output layer of a preset pulse neural network. Multi-scale feature extraction is performed on the pulse sequence, and surface wave filtering is applied to obtain target sequence features. Signal reconstruction is then performed on the target sequence features to generate the target seismic signal. This improves the accuracy of feature extraction, effectively suppresses the interference of seismic surface waves on the seismic signal, preserves the effective signal, and enhances the efficiency and accuracy of seismic signal processing.

[0080] Figure 3 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 2 ,like Figure 3 As shown, in this embodiment... Figure 2Based on the embodiments, the method of converting the seismic signal to be processed through the input layer to obtain the corresponding pulse sequence in step S203 above will be described in detail. The method includes:

[0081] S301. The seismic signal to be processed is normalized through the input layer to obtain a normalized seismic signal.

[0082] In this embodiment, the seismic signal to be processed is normalized through the input layer, which effectively eliminates the differences in the dimensions and amplitude of the seismic signal, thereby improving the noise resistance of the seismic signal; the normalized seismic signal is easier to extract effective features, thereby improving the accuracy of seismic signal analysis.

[0083] Furthermore, the seismic signal to be processed is normalized using the following formula:

[0084]

[0085] Where x is the spatial location, t is the time, S(x,t) is the seismic signal to be processed at spatial location x and time t, and σ x Let S be the standard deviation of the seismic signal at spatial location x. norm (x,t) represents the normalized seismic signal.

[0086] S302. Based on the first preset dynamic threshold, the normalized seismic signal is encoded to obtain the corresponding pulse sequence.

[0087] In this embodiment, the first preset dynamic threshold V t h(x,t) is a threshold that is periodically adjusted with time t, used to control the pulse triggering condition, where the first preset dynamic threshold V t The setting of h(x,t) is shown in the following formula:

[0088] V t h(x,t)=V0·[1+0.2sin(2πf b t)]

[0089] Among them, f b V0 is the dominant frequency of the surface wave and V0 is the basic threshold. The dominant frequency of the surface wave refers to the main frequency components of the surface wave, which is used to adjust the frequency characteristics of the threshold. The basic threshold refers to the static threshold reference value, which is used to balance the sensitivity of pulse triggering.

[0090] Then according to the first preset dynamic threshold V t h(x,t) is used to encode the normalized seismic signal to obtain the corresponding pulse sequence P(x,t) as shown in the following formula:

[0091]

[0092] Where 1 represents a trigger pulse and 0 represents a no-trigger pulse.

[0093] This application provides a seismic surface wave suppression method that normalizes the seismic signal to be processed through a preset pulse neural network input layer to obtain a normalized seismic signal. Based on a first preset dynamic threshold, the normalized seismic signal is converted into a corresponding pulse sequence, which effectively improves the stability and accuracy of the seismic signal, highlights the characteristics of the seismic signal, and improves the real-time processing efficiency of the seismic signal.

[0094] Figure 4 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 3 ,like Figure 4 As shown, in this embodiment... Figure 2 Based on the embodiments, the method described in detail in step S204 above, which involves extracting features from the pulse sequence through a hidden layer to obtain multiple features, and then performing surface wave filtering based on these features to obtain the target feature sequence, includes:

[0095] S401. By using multi-scale convolution kernels, feature extraction processing is performed on the pulse sequence to obtain multiple features.

[0096] In this embodiment, for example, the multi-scale convolution kernel can be a 3×3×t convolution kernel, where t is the time dimension; by using the multi-scale convolution kernel to perform feature extraction processing on the pulse sequence, the features of different frequency components in the pulse sequence can be effectively captured.

[0097] S402. By utilizing the membrane potential accumulation characteristics of neurons in the leakage current integral firing model, the surface wave energy accumulation region is detected. During the detection of the surface wave energy accumulation region, the neuron threshold is adaptively adjusted to obtain the target feature sequence.

[0098] In this embodiment, the membrane potential accumulation characteristic of the leaky integral firing model neuron refers to simulating the dynamic characteristics of biological neurons through the accumulation of membrane potential, leakage, and firing processes; the mathematical expression of the leaky integral firing model is as follows:

[0099]

[0100] Where x is the position; t is the time; u(x,t) is the membrane potential at position x and time t; I(x,t) is the input current at position x and time t, and the input current I(x,t) is determined by the corresponding pulse sequence P(x,t); T m This is the membrane time constant, i.e., the decay time constant of the membrane potential.

[0101] In this embodiment, the membrane potential is calculated using the mathematical expression of the leakage current integral discharge model. The surface wave energy accumulation region is detected by accumulating the membrane potential; wherein, the surface wave energy accumulation region is the area where the membrane potential is concentratedly increased. The local pulse energy accumulation value U(x,t) is then calculated; the formula for calculating the local pulse energy accumulation value U(x,t) is as follows:

[0102]

[0103] Where i is the offset identifier of the current position; j is the offset identifier of the current time; k controls the coverage of the convolution kernel; W(i,j) is the weight value of the Gaussian weight kernel, that is, the weight of the Gaussian kernel function at position (i,j); P(x+i,t+j) is the pulse sequence value at position x+i and time t+j.

[0104] After calculating the local pulse energy accumulation value U(x,t), the neuron threshold is dynamically adjusted based on the local pulse energy accumulation value U(x,t). The adjustment method of the neuron threshold θ(x,t) is shown in the following formula:

[0105]

[0106] Where θ0 is the initial neuron threshold, i.e., the baseline threshold before adjustment, used to balance surface wave suppression and signal preservation; U max This is the maximum local pulse energy, used to limit the range of adjustment of the neuron threshold.

[0107] By dynamically adjusting the neuron threshold, when the membrane potential reaches the neuron threshold, the neuron fires a pulse and resets the membrane potential, thereby filtering the wavelet to obtain the target feature sequence.

[0108] The seismic surface wave suppression method proposed in this application extracts multiple features from the pulse sequence through multi-scale convolution kernels, combines the membrane potential accumulation characteristics of the neurons in the leakage current integral firing model, detects the surface wave energy accumulation area, and adaptively adjusts the neuron threshold to obtain the target feature sequence, thereby enhancing the accuracy and real-time performance of feature extraction.

[0109] Figure 5 A flowchart illustrating a seismic surface wave suppression method provided in this application. Figure 4 ,like Figure 5 As shown, in this embodiment... Figure 2 Based on the embodiment, before inputting the seismic signal to be processed into the preset pulse neural network in step S202 above, the method further includes:

[0110] S501. Establish the input layer based on the preset normalization processing module and the preset pulse sequence conversion module.

[0111] In this embodiment, the preset normalization processing module is used to normalize the seismic signal to be processed, so as to reduce the amplitude difference of the seismic signal to be processed and improve the stability of the pulse neural network.

[0112] The preset pulse sequence conversion module is used to convert the normalized seismic signal into a pulse sequence. The pulse coding used by the preset pulse sequence conversion module includes time coding and frequency coding.

[0113] S502. Based on the preset multi-scale convolution kernel and the preset leakage integral distribution model, establish the hidden layer.

[0114] In this embodiment, a preset multi-scale convolutional kernel and a preset leakage current integral firing model are connected to obtain the hidden layer of the spiking neural network, so that the spiking neural network can suppress the surface wave, thereby preserving the effective signal and improving the accuracy of the spiking neural network.

[0115] S503. Establish the output layer based on the preset time-frequency joint loss function that incorporates preset impulse sparsity constraints.

[0116] In this embodiment, a preset time-frequency joint loss function with preset pulse sparsity constraints is used to establish the output layer of the spiking neural network. The decoder then converts the pulse sequence into a continuous signal, effectively reducing the loss of high-frequency information, optimizing the fidelity of the seismic signal, and thus improving the accuracy of the spiking neural network.

[0117] S504. Generate a preset spiking neural network based on the input layer, hidden layer, and output layer.

[0118] In this embodiment, by constructing a pre-defined spiking neural network in layers, the ability of the pre-defined spiking neural network to extract spatiotemporal features is improved, and the efficiency and accuracy of the pre-defined spiking neural network in processing seismic signals are enhanced; and sparse connections and event-driven mechanisms are adopted to reduce the computational complexity of the pre-defined spiking neural network.

[0119] After generating a predefined spiking neural network based on the input layer, hidden layer, and output layer, the process also includes:

[0120] S505. Obtain earthquake surface wave training samples, wherein the earthquake surface wave training samples include multiple earthquake signal samples with earthquake surface waves.

[0121] In this embodiment, the seismic surface wave training samples may also include corresponding target seismic signal samples, which are used to improve the generalization ability of the preset pulse neural network.

[0122] S506. Based on the seismic surface wave training samples, train the preset pulse neural network to optimize the preset pulse neural network.

[0123] Optionally, the preset spiking neural network includes multiple synaptic matrices.

[0124] In this embodiment, the synapse matrix represents the connection strength and weights between neurons.

[0125] Accordingly, based on the seismic surface wave training samples, a pre-defined pulse neural network is trained to optimize the pre-defined pulse neural network, including:

[0126] Based on seismic surface wave training samples, multiple synaptic matrices are trained to optimize a pre-defined spiking neural network.

[0127] In this embodiment, by setting multiple synaptic matrices, the transmission and integration of pulse signals are completed, and the adaptive learning of the preset spiking neural network is realized.

[0128] The seismic surface wave suppression method provided in this application embodiment obtains a preset pulse neural network by pre-setting a pulse neural network and training the pulse neural network accordingly, so as to achieve accurate suppression of seismic surface waves.

[0129] Alternatively, for example, such as Figures 6 to 7 As shown, where, Figure 6 include Figure 6 a and Figure 6 b, Figure 6 'a' represents the target seismic signal output by the preset spiking neural network provided in this application embodiment. Figure 6 b represents the target seismic signal output by the surface wave suppression model in the existing technology; Figure 7 include Figure 7 a and Figure 7 b, Figure 7 a is a cross-sectional view of the target seismic signal output by the preset pulse neural network provided in the embodiments of this application. Figure 7 b is a profile of the target seismic signal output by the surface wave suppression model in the existing technology; through Figure 6 , Figure 7 By comparing the target seismic signals, it can be clearly seen that the preset pulse neural network provided in this application embodiment can effectively suppress the interference of seismic surface waves, improve the visibility of reflected signals, and thus improve the signal-to-noise ratio and resolution of the effective signal.

[0130] Figure 8 A schematic diagram of the seismic surface wave suppression device provided in this application is shown below. Figure 8 As shown, the seismic surface wave suppression device provided in this embodiment includes:

[0131] The acquisition module 801 is used to acquire the seismic signal to be processed.

[0132] The input module 802 is used to input the seismic signal to be processed into a preset spiking neural network, wherein the spiking neural network includes an input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer and the hidden layer is connected to the output layer.

[0133] The conversion module 803 is used to convert and process the seismic signal through the input layer to obtain the corresponding pulse sequence.

[0134] The processing module 804 is used to perform feature extraction processing on the pulse sequence through the hidden layer to obtain multiple features, and to perform surface wave filtering processing based on the multiple features to obtain the target feature sequence.

[0135] Optionally, the hidden layer includes a multi-scale convolutional kernel and a leakage integral firing model, with the multi-scale convolutional kernel and the leakage integral firing model connected together.

[0136] Module 805 is used to perform signal reconstruction processing on the target feature sequence through the output layer to obtain the target seismic signal.

[0137] In one possible implementation, the conversion module 803 is further used for:

[0138] The seismic signal is normalized through the input layer to obtain a normalized seismic signal.

[0139] Based on the first preset dynamic threshold, the normalized seismic signal is encoded to obtain the corresponding pulse sequence.

[0140] In one possible implementation, the processing module 804 is further used for:

[0141] By using multi-scale convolution kernels, feature extraction processing is performed on the pulse sequence to obtain multiple features;

[0142] By utilizing the membrane potential accumulation characteristics of neurons in the leakage current integral firing model, surface wave energy accumulation regions are detected. During the detection process of surface wave energy accumulation regions, the neuron threshold is adaptively adjusted to obtain the target feature sequence.

[0143] In one possible implementation, module 805 is further used for:

[0144] Through the output layer, a preset time-frequency joint loss function is used to perform signal reconstruction processing on the target feature sequence to obtain the target seismic signal. The preset time-frequency joint loss function incorporates a preset impulse sparsity constraint.

[0145] In one possible implementation, before inputting the seismic signal to be processed into a preset spiking neural network, the following is also included:

[0146] The module is used to establish the input layer based on the preset normalization processing module and the preset pulse sequence conversion module; to establish the hidden layer based on the preset multi-scale convolution kernel and the preset leakage current integral firing model; and to establish the output layer based on the preset time-frequency joint loss function that incorporates the preset pulse sparsity constraint.

[0147] The generation module is used to generate a preset spiking neural network based on the input layer, hidden layer, and output layer.

[0148] In one possible implementation, after generating a predefined spiking neural network based on the input layer, hidden layer, and output layer, the following is also included:

[0149] The acquisition module 801 can also be used to acquire seismic surface wave training samples, wherein the seismic surface wave training samples include multiple seismic signal samples with seismic surface waves.

[0150] The training module is used to train a preset pulse neural network based on seismic surface wave training samples in order to optimize the preset pulse neural network.

[0151] Optionally, the preset spiking neural network includes multiple synaptic matrices;

[0152] The training module is specifically used to train multiple synaptic matrices based on seismic surface wave training samples in order to optimize a preset spiking neural network.

[0153] The seismic surface wave suppression device provided in this embodiment can perform the method provided in the above-described method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0154] Figure 9 This is a structural schematic diagram of the seismic surface wave suppression device provided in this application. Figure 9 As shown, the seismic surface wave suppression device provided in this embodiment includes at least one processor 901 and a memory 902. Optionally, the seismic surface wave suppression device further includes a communication component 903. The processor 901, memory 902, and communication component 903 are connected via a bus 904.

[0155] In a specific implementation, at least one processor 901 executes computer execution instructions stored in memory 902, causing at least one processor 901 to perform the above-described method.

[0156] The specific implementation process of processor 901 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0157] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0158] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0159] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0160] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0161] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0162] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0163] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0164] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0165] The units described as separate components may or may not be physically separate. 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 the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0166] In addition, the functional units 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.

[0167] If a function is implemented as 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 technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0168] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0169] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for suppressing seismic surface waves, characterized in that, include: Acquire the seismic signal to be processed; The seismic signal to be processed is input into a preset pulse neural network, wherein the pulse neural network includes an input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer and the hidden layer is connected to the output layer; The input layer is used to convert and process the seismic signal to be processed to obtain the corresponding pulse sequence. The pulse sequence is processed by feature extraction through the hidden layer to obtain multiple features, and surface wave filtering is performed based on the multiple features to obtain the target feature sequence. The target feature sequence is reconstructed through the output layer to obtain the target seismic signal.

2. The method according to claim 1, characterized in that, The hidden layer includes a multi-scale convolutional kernel and a leakage current integral distribution model, and the multi-scale convolutional kernel and the leakage current integral distribution model are connected; Accordingly, the step of performing feature extraction processing on the pulse sequence through the hidden layer to obtain multiple features, and performing surface wave filtering processing based on the multiple features to obtain the target feature sequence includes: The pulse sequence is processed by the multi-scale convolution kernel to obtain multiple features; By utilizing the membrane potential accumulation characteristics of the neurons in the leakage current integral firing model, the surface wave energy accumulation region is detected, and during the detection process of the surface wave energy accumulation region, the neuron threshold is adaptively adjusted to obtain the target feature sequence.

3. The method according to claim 1, characterized in that, The process of converting the seismic signal to be processed through the input layer to obtain the corresponding pulse sequence includes: The input layer is used to normalize the seismic signal to be processed, so as to obtain a normalized seismic signal. The normalized seismic signal is encoded according to a first preset dynamic threshold to obtain a corresponding pulse sequence.

4. The method according to claim 1, characterized in that, The step of performing signal reconstruction processing on the target feature sequence through the output layer to obtain the target seismic signal includes: Through the output layer, a preset time-frequency joint loss function is used to perform signal reconstruction processing on the target feature sequence to obtain the target seismic signal. The preset time-frequency joint loss function incorporates a preset pulse sparsity constraint.

5. The method according to any one of claims 1 to 4, characterized in that, Before inputting the seismic signal to be processed into the preset pulse neural network, the method further includes: An input layer is established based on the preset normalization processing module and the preset pulse sequence conversion module; A hidden layer is established based on a pre-defined multi-scale convolutional kernel and a pre-defined leakage integral distribution model; The output layer is established based on a preset time-frequency joint loss function that incorporates preset impulse sparsity constraints; A preset spiking neural network is generated based on the input layer, the hidden layer, and the output layer.

6. The method according to claim 5, characterized in that, After generating the preset spiking neural network based on the input layer, the hidden layer, and the output layer, the method further includes: Obtain seismic surface wave training samples, wherein the seismic surface wave training samples include multiple seismic signal samples with seismic surface waves; The preset pulse neural network is trained based on the earthquake surface wave training samples to optimize the preset pulse neural network.

7. The method according to claim 6, characterized in that, The preset spiking neural network includes multiple synaptic matrices; Accordingly, the step of training the preset pulse neural network based on the seismic surface wave training samples to optimize the preset pulse neural network includes: The multiple synaptic matrices are trained based on the seismic surface wave training samples to optimize the preset spiking neural network.

8. A seismic surface wave suppression device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the seismic surface wave suppression method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the seismic surface wave suppression method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the seismic surface wave suppression method according to any one of claims 1 to 7.

Images