A method, system, medium and device for improving the time-frequency resolution of seismic data

Abstract

The present invention discloses a method, system, medium and device for improving the time-frequency resolution of seismic data, which constructs a large number of synthetic chirp signals; obtains the energy spectrum and Wigner-Ville distribution of the synthetic signals as label pairs; combines simple label pairs to obtain complex label pairs; builds an SCNN network model for improving time-frequency resolution; trains the model using the constructed label pairs; and maps the signals using the trained model to obtain the time-frequency spectrum with high time-frequency resolution of the signals. By constructing the time-frequency spectrum of synthetic signals as the training set, the present invention greatly reduces the workload of using the processing results of seismic signals as the training set; trains the model using a GPU, and calculates the time-frequency spectrum of the signals using the trained model, which greatly reduces the time and labor costs compared with traditional time-frequency processing methods.

Description

Technical Field

[0001] The present invention belongs to the technical field of exploration geophysics, and particularly relates to a method, a system, a medium and a device for improving the time-frequency resolution of seismic data. Background Art

[0002] Time-frequency analysis technology is a key technology in signal processing and analysis and plays a very important role in signal processing and analysis. The time-frequency analysis method combines the time domain and the frequency domain and depicts the signal in the form of a joint-domain time-frequency distribution, overcoming the defects of single time-frequency domain signal analysis, and can maximize the acquisition of the time-frequency characteristics of the signal and display them in the time-frequency plane. As a typical non-stationary signal, seismic signals can be processed and analyzed through time-frequency analysis to quickly and efficiently obtain relevant information about underground reservoirs, providing a theoretical basis for the subsequent analysis and interpretation of seismic data.

[0003] The existing main time-frequency methods currently include:

[0004] Existing Technology 1: Linear time-frequency methods: There are short-time Fourier transform, continuous wavelet transform, S transform, etc., which can provide good time-frequency representations for multi-component signals and have the characteristic of being able to reconstruct the original signal. Therefore, linear time-frequency analysis is one of the most widely used time-frequency analysis methods in the actual industry; the disadvantage is that it cannot express the time-frequency local characteristics of the signal, cannot accurately describe the relationship between frequency and time, and the time-frequency analysis effect for non-stationary signals needs to be improved.

[0005] Existing Technology 2: Cohen class distributions: Cohen class distributions are obtained by two-dimensional convolution of the Wigner-Ville distribution with different kernel functions, and are typical non-linear quadratic time-frequency distributions. Common Cohen class distributions include smoothed Wigner-Ville distribution, pseudo Wigner-Ville distribution, and smoothed pseudo Wigner-Ville distribution; the disadvantage is that the time-frequency resolution is relatively low and it does not have the characteristic of reconstructing the original signal.

[0006] Existing Technology 3: Squeezed time-frequency analysis: Squeezed time-frequency analysis is a compromise means for time and frequency resolution, applying the rearrangement technique to the scale spectrum. The main squeezed time-frequency analysis methods include second-order squeezed wavelet transform, high-order squeezed wavelet transform, second-order squeezed time-frequency transform, n-order squeezed time-frequency transform, etc. The disadvantage is that it cannot guarantee high resolution in both time and frequency, and only significantly improves the frequency resolution, while the time resolution is not ideal. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a method, system, medium and device for improving the time-frequency resolution of seismic data in view of the deficiencies in the above-mentioned prior art, so as to solve the technical problem that the time-frequency resolution of the spectrogram is low and the cross interference cannot be suppressed when processing seismic signals.

[0008] The present invention adopts the following technical solutions:

[0009] A method for improving the time-frequency resolution of seismic data includes the following steps:

[0010] S1. Construct a synthetic chirp signal x(t), obtain the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and construct simple label pairs;

[0011] S2. Combine the simple label pairs obtained in step S1 to obtain complex label pairs;

[0012] S3. Build an SCNN network model;

[0013] S4. Use the complex label pairs obtained in step S2 to train the SCNN network model obtained in step S3;

[0014] S5. Use the SCNN network model trained in step S4 to map the actual signal to obtain the time-frequency spectrum with high time-frequency resolution of the signal.

[0015] Specifically, in step S1, perform a short-time Fourier transform on the constructed chirp signal x(t), and then perform a modulus square calculation to obtain the corresponding energy spectrum SPEC x (t, f); use the obtained short energy as the input of the training set, and then calculate the Wigner-Ville distribution of the signal to obtain the Wigner-Ville distribution WVD x (t, f) corresponding one-to-one to the energy spectrum SPEC x (t, f) as simple label pairs.

[0016] Further, the Wigner-Ville distribution WVD x (t, f) is:

[0017]

[0018] The energy spectrum SPEC x (t, f) is:

[0019]

[0020] Where x is the chirp signal, t is time, τ is the integration variable, g(t) is the window function, f is the frequency of the signal, and x(τ) is the chirp signal after time shift.

[0021] Specifically, in step S2, the input of the complex label pair is composed of the energy spectra of more than two chirp signals, and the label of the complex label pair is composed of the corresponding Wigner-Ville distributions of more than two chirp signals.

[0022] Specifically, in step S3, the convolution kernel size in the SCNN network model is N*N, the modules of the SCNN network model are the convolutional layer and the ReLU activation function layer, and the convolutional layer contains K filters of size N*N*M.

[0023] Specifically, step S4 is as follows: First, preprocess the training set data by performing data augmentation and normalization operations; then set the parameters batch_size, learning rate, and epoch of the SCNN network model; finally, start the end-to-end training of the SCNN network model on the server using the GPU.

[0024] Specifically, step S5 is as follows: Construct a synthetic signal obtained by adding three chirp signal components, then perform a short-time Fourier transform on the constructed synthetic signal to obtain the corresponding energy spectrum, and then use the energy spectrum as the input to the trained SCNN network model. The trained network model can perform the function of a deconvolution sum, perform a convolution operation with the time-frequency spectrum to obtain a time-frequency spectrum with high time-frequency resolution, and obtain a mapping result with improved time-frequency resolution.

[0025] In a second aspect, an embodiment of the present invention provides a system for improving the time-frequency resolution of seismic data, including:

[0026] A construction module that constructs a synthetic chirp signal x(t), obtains the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and constructs a simple label pair;

[0027] A combination module that combines the simple label pairs obtained by the construction module to obtain complex label pairs;

[0028] A building module that builds an SCNN network model;

[0029] A training module that uses the complex label pairs obtained by the combination module to train the SCNN network model obtained by the building module;

[0030] An output module that uses the trained SCNN network model by the training module to map the actual signal to obtain a time-frequency spectrum with high time-frequency resolution of the signal.

[0031] In a third aspect, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method for improving the time-frequency resolution of seismic data are implemented.

[0032] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, including a computer program, where when the computer program is executed by a processor, the steps of the method for improving the time-frequency resolution of seismic data are implemented.

[0033] Compared with the prior art, the present invention has at least the following beneficial effects:

[0034] A method for improving the time-frequency resolution of seismic data in the present invention constructs a large number of chirp signal label pairs, where the input is the energy spectrum of the signal and the label is the Wigner-Ville distribution of the signal. Then, simple label pairs are combined to construct complex label pairs to expand the data set. Then, an SCNN network model for improving the time-frequency resolution is built. Then, using the convolution relationship between the Wigner-Ville distribution and the energy spectrum obtained by the short-time Fourier transform, the SCNN network model is trained using the data in the training set. The deconvolution relationship between the energy spectrum and the Wigner-Ville distribution of the signal is realized using the SCNN model, and then the mapping between the low-resolution time-frequency spectrogram and the high-resolution time-frequency spectrogram is realized using the trained SCNN model; improving the time-frequency resolution of seismic data using the deep learning method can combine the advantages of the two time-frequency analysis methods of the short-time Fourier transform and the Wigner-Ville distribution, improve the time-frequency resolution when processing the time-frequency spectrogram of seismic signals, and effectively suppress cross interference.

[0035] Furthermore, since there is a linear convolution relationship between the Wigner-Ville distribution and the energy spectrum of the signal, that is, the energy spectrum SPEC x (t,f) of the signal can be obtained by convolving the Wigner-Ville distribution of the signal with the Wigner-Ville distribution of a window function. Therefore, the short-time Fourier transform is performed on the constructed chirp signal x(t), and then the modulus square calculation is performed to obtain the corresponding energy spectrum SPEC x (t,f); the obtained short energy is used as the input of the training set, and then the Wigner-Ville distribution of the signal is calculated to obtain the Wigner-Ville distribution WVD x (t,f) corresponding one-to-one to the energy spectrum SPEC x (t,f), as a simple label pair, which can enable the network model to learn the deconvolution relationship between the input and the label, and realize the mapping between the time-frequency spectrogram with low time-frequency resolution and the time-frequency spectrogram with high time-frequency resolution.

[0036] Furthermore, for a single-component linear frequency modulation signal, the time-frequency spectrum obtained by calculating its Wigner-Ville distribution will not introduce cross-interference and has the highest time-frequency resolution. Therefore, the low-resolution time-frequency spectrum SPEC x (t,f) obtained by performing short-time Fourier transform on the signal and taking the modulus square is used as the input data, and the Wigner-Ville distribution is used as the label, enabling the network model to learn the mapping relationship that can improve the time-frequency resolution and avoid cross-interference terms.

[0037] Furthermore, to achieve the diversity and completeness of the training set data, the model is made to learn not only the mapping relationship between the low-time-frequency-resolution time-frequency spectrograms and the high-time-frequency-resolution time-frequency spectrograms of signals with simple signal components. The dataset is augmented by constructing complex label pairs through the combination of simple label pairs. The input of the complex label pair consists of the energy spectra of two or more linear frequency modulation signals, and the label of the complex label pair is composed of the corresponding Wigner-Ville distributions of two or more linear frequency modulation signals.

[0038] Furthermore, for the task of realizing the mapping between the low-time-frequency-resolution time-frequency spectrogram and the high-time-frequency-resolution time-frequency spectrogram, we expect the model to achieve a linear deconvolution relationship from the input data to the label data. Therefore, it is necessary to build an SCNN network model to meet the requirements of the task. The convolution kernel size in the SCNN network model is N*N, and the selection of the convolution kernel size can ensure that as many features of the input data as possible are extracted. The module of the SCNN network model is a convolution layer and a ReLU activation function layer. The convolution layer contains K filters of size N*N*M. These filters are used to perform convolution operations with the data to extract the features of the data, and the activation function is used to enable the network to learn these extracted features, making the network model play a deconvolution relationship from the input data to the label data..

[0039] Furthermore, to improve the convergence speed and accuracy of the model, the training set data is first preprocessed by performing data augmentation and normalization operations; then the parameters batch_size, learning rate, and epoch of the SCNN network model are set to ensure that the network model iterates regularly; the training process of the network model has a huge computational amount, and the GPU has a higher computational power than the CPU, which can greatly reduce the time consumed in the model training process. Therefore, the SCNN network model is trained end-to-end using the GPU on the server.

[0040] Further, in order to verify the mapping effect of the trained model on the time-frequency spectrum with low time-frequency resolution and further illustrate that the model also has a very effective time-frequency resolution improvement effect on complex signals, a synthetic signal obtained by adding three chirp signal components is constructed. Then, the constructed synthetic signal is subjected to short-time Fourier transform to obtain the corresponding energy spectrum. Then, the energy spectrum is used as the input and fed into the trained SCNN network model. After calculation, the mapping result with improved time-frequency resolution is obtained. By comparing the input data with the output data, it can be proved that the model effectively improves the time-frequency resolution of the time-frequency spectrum of the signal.

[0041] It can be understood that the beneficial effects of the above second aspect to the fourth aspect can be referred to the relevant descriptions in the above first aspect, and will not be elaborated here.

[0042] In summary, the present invention greatly reduces the workload of using the processing results of seismic signals as the training set by constructing the time-frequency spectrum of the synthetic signal as the training set; builds an SCNN model to combine the deep learning method with the traditional time-frequency processing means, which has quite an innovation; uses GPU to train the model, and calculates the time-frequency spectrum of the signal through the trained model, which greatly reduces the time and labor costs compared with the traditional time-frequency processing method.

[0043] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0044] Figure 1 is the flow chart of the present invention;

[0045] Figure 2 is the structure diagram of the SCNN network model.

[0046] Figure 3 is the schematic diagram of multiple synthesized chirp signals;

[0047] Figure 4 is Figure 2 the energy spectrum diagram of the signal;

[0048] Figure 5 is Figure 2 the network mapping result diagram of the signal with high time-frequency resolution. Detailed Embodiments

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] In the description of the present invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0051] It should also be understood that the terms used in the specification of the present invention are for the purpose of describing particular embodiments only and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly dictates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0052] It should be further understood that the term "and / or" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the character " / " herein generally indicates an "or" relationship between the contextually related objects.

[0053] It should be understood that although terms such as first, second, and third may be used in the embodiments of the present invention to describe preset ranges, etc., these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range can also be referred to as the second preset range, and similarly, the second preset range can also be referred to as the first preset range.

[0054] Depending on the context, the word "if" as used herein can be interpreted as "when" or "while" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)".

[0055] Structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the drawings. These figures are not drawn to scale, where for the purpose of clear expression, some details are enlarged and some details may be omitted. The shapes of the various regions and layers shown in the figures and their relative sizes and positional relationships are merely exemplary, and may actually deviate due to manufacturing tolerances or technical limitations, and those skilled in the art can additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0056] The present invention provides a method for improving the time-frequency resolution of seismic data, which can improve the time-frequency resolution when processing the spectrogram of seismic signals and effectively suppress cross interference. First, a large number of label pairs of chirp signals are constructed, where the input is the energy spectrum of the signal and the label is the Wigner-Ville distribution of the signal. Then, the simple label pairs are combined to construct complex label pairs to expand the data set. Then, an SCNN network model for improving the time-frequency resolution is built. Then, using the convolution relationship between the Wigner-Ville distribution and the energy spectrum obtained by the short-time Fourier transform, the SCNN network model is trained using the data in the training set. The deconvolution relationship between the energy spectrum and the Wigner-Ville distribution of the signal is realized by using the SCNN model, and then the mapping from the low-resolution time-frequency spectrogram to the high-resolution time-frequency spectrogram is realized by using the trained SCNN model. This technology can provide seismic data with high time-frequency resolution for work such as seismic attribute extraction.

[0057] Please refer to Figure 1 , a method for improving the time-frequency resolution of seismic data according to the present invention, comprising the following steps:

[0058] S1. Construct a synthesized chirp signal, obtain the energy spectrum and the Wigner-Ville distribution of the synthesized chirp signal, and construct simple label pairs;

[0059] The frequency of the chirp signal increases linearly with time within the sampling time. For the chirp signal x(t), it is:

[0060]

[0061] Taking the derivative of its phase can obtain the linear variation relationship of frequency with time:

[0062] f = f0 + μ0t

[0063] where f0 is called the starting frequency, μ0 is the frequency modulation rate, t is the time, f is the frequency of the signal, and j is the imaginary unit.

[0064] Since the Wigner-Ville distribution of the chirp signal has extremely high time-frequency resolution and does not generate cross interference, a large number of chirp signals are constructed, and their energy spectra and Wigner-Ville distributions are used as input and labels, so that the model can learn the deconvolution relationship between the input and the labels.

[0065] The specific method for obtaining the simple label pairs of the synthesized chirp signal is as follows:

[0066] After performing the short-time Fourier transform on the constructed chirp signal x(t) and then performing the modulus square calculation on it, its energy spectrum can be obtained:

[0067]

[0068] Among them, τ is the integration variable, and g(t) is the window function.

[0069] The time-frequency resolutions of different energy spectra of the window function are different. Generally, the Gaussian window function is adopted for the window function. At this time, the time-frequency resolution of the energy spectrum obtained by the short-time Fourier transform is the highest.

[0070] The expression of the window function is:

[0071]

[0072] Among them, σ is the variance of the window function, and μ is the mean value of the window function.

[0073] Take the obtained short energy as the input of the training set, and then calculate the Wigner-Ville distribution of the signal:

[0074]

[0075] Obtain the Wigner-Ville distribution corresponding one-to-one to the energy spectrum; the Wigner-Ville distribution has the highest time-frequency resolution and will not produce cross interference for the chirp signal. Take it as the label in the training set. Here, SPEC x (t, f) and WVD x (t, f) form a pair of label pairs.

[0076] S2. Combine the simple label pairs obtained in step S1 to obtain complex label pairs;

[0077] Randomly combine the simple label pairs constructed according to the chirp signal to obtain a large number of complex one-to-one corresponding label pairs. The input of the complex label pair is composed of the energy spectra of two or more chirp signals, such as SPEC x1 (t, f) + PEC x2 (t, f) + PEC x3 (t, f), and the label is composed of the corresponding Wigner-Ville distributions of two or more chirp signals, such as WVD x1 (t, f) + VD x2 (t, f) + VD x3 (t, f). Doing so can expand the data set and also enable the model to learn more complex deconvolution mapping relationships.

[0078] S3. Build an SCNN network model for improving the time-frequency resolution;

[0079] The SCNN network model is proposed based on the traditional CNN model. The basic framework of the CNN model consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. Since the SCNN network model needs to establish the connection between the short-time Fourier transform energy spectrum and the Wigner-Ville distribution; the pooling layer and the fully connected layer in the traditional convolutional network model are deleted, and the convolutional kernel size in the network is uniformly designed as N*N. The modules of the SCNN network model are the convolutional layer and the rectified linear unit (ReLU activation function layer). The convolutional layer contains K filters of size N*N*M, which are mainly used to extract the local features of the input layer. The ReLU activation function is used to avoid the vanishing gradient in network training. Finally, through the training of a large amount of data, the deconvolution operation between the input and output can be realized.

[0080] S4. Use the complex labels obtained in step S2 to train the SCNN network model obtained in step S3;

[0081] First, preprocess the training set data by performing data augmentation and normalization operations;

[0082] Then, set the batch_size, learning rate, and epoch of the SCNN network model;

[0083] Then, start the end-to-end training of the network on the server using the GPU.

[0084] S5. Use the SCNN network model trained in step S4 to map the actual signal to obtain the high time-frequency resolution time-frequency spectrum of the signal.

[0085] Construct a synthetic signal obtained by adding three chirp signal components, then perform the short-time Fourier transform on the constructed synthetic signal to obtain its energy spectrum, and then use the energy spectrum as the input of the model to the trained model. After calculation, the mapping result with improved time-frequency resolution can be obtained.

[0086] In another embodiment of the present invention, a system for improving the time-frequency resolution of seismic data is provided. The system can be used to implement the method for improving the time-frequency resolution of seismic data. Specifically, the system for improving the time-frequency resolution of seismic data includes a construction module, a combination module, a building module, a training module, and an output module.

[0087] Among them, the construction module constructs the synthetic chirp signal x(t), obtains the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and constructs simple label pairs;

[0088] The combination module combines the simple label pairs obtained by the construction module to obtain complex label pairs;

[0089] Building module, building an SCNN network model;

[0090] Training module, using the complex labels obtained by the combination module to train the SCNN network model built by the building module;

[0091] Output module, using the trained SCNN network model of the training module to map the actual signal to obtain the high time-frequency resolution time-frequency spectrum of the signal.

[0092] In another embodiment of the present invention, a terminal device is provided. The terminal device includes a processor and a memory. The memory is used to store a computer program. The computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function. The processor described in the embodiment of the present invention can be used for the operation of the method for improving the time-frequency resolution of seismic data, including:

[0093] Construct a synthetic chirp signal x(t), obtain the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and construct simple label pairs; combine the simple label pairs to obtain complex label pairs; build an SCNN network model; use the complex label pairs to train the SCNN network model; use the trained SCNN network model to map the actual signal to obtain the high time-frequency resolution time-frequency spectrum of the signal.

[0094] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a terminal device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and, of course, the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, and this storage space stores the operating system of the terminal. And, in this storage space, there is also stored one or more instructions suitable for being loaded and executed by a processor. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory (Non-Volatile Memory), such as at least one disk memory.

[0095] One or more instructions stored in the computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the method for improving the time-frequency resolution of seismic data in the above embodiments; the one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0096] Construct a synthetic chirp signal x(t), obtain the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and construct simple label pairs; combine the simple label pairs to obtain complex label pairs; build an SCNN network model; use the complex label pairs to train the SCNN network model; use the trained SCNN network model to map the actual signal to obtain the time-frequency spectrum with high time-frequency resolution of the signal.

[0097] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0098] The present invention provides an intelligent method for improving the time-frequency resolution of seismic data using a deep learning method. Now, a synthetic example a(t) with a sampling number of k and a sampling interval of 1 / k s is used for verification:

[0099] The signal a(t) is distributed in the time domain between [t1, t2]. The value of t1 is 1 / 512 s, and the value of t2 is 1 s. The time-domain waveform of the signal is as Figure 3 shown. The short-time Fourier transform is performed on the signal a(t), and the magnitude square of it is taken to obtain the energy spectrum of the signal. The energy spectrum of the signal a(t) is as Figure 4 shown, and the size of the energy spectrum of the signal is 256 * 512. Then, 3000 pairs of tags with a size of 256 * 512 are constructed for the training of the network model.

[0100] The size of the convolution kernel of the model is set to 3 * 3, and the convolutional layer contains 600 convolution kernels; then the network sets the parameter batch_size to 2, the learning rate to 0.001, and epoch to 30. An L1 regularization term is added to the loss function to prevent overfitting during the network training process; then, on the server side, an NVIDIA GTX1080Ti GPU is used to start the end-to-end training of the network; then, the energy spectrum of the signal a(t) is used as the input, and the trained network model is used to map it; then, the energy spectrum of the signal a(t) is fed into the model, and a high-resolution mapping result as shown in Figure 5 is obtained. The time-frequency resolution of the time-frequency spectrum mapped by the network model is effectively improved, and the cross-interference is also effectively suppressed.

[0101] In summary, for a method, system, medium, and device for improving the time-frequency resolution of seismic data according to the present invention, by constructing the time-frequency spectrum of the synthetic signal as the training set, the workload of using the processing results of seismic signals as the training set is greatly reduced; building an SCNN model combines the deep learning method with traditional time-frequency processing means, which has considerable innovation; using a GPU to train the model and calculating the time-frequency spectrum of the signal through the trained model greatly reduces the time and labor costs compared with traditional time-frequency processing methods.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiments can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0103] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0104] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in the form of hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0105] In the embodiments provided by the present invention, it should be understood that the disclosed device / terminal and method can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For example, the division of the module or unit is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0106] The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0107] In addition, in each embodiment of the present invention, the functional units may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above integrated units may be implemented in the form of hardware or in the form of software functional units.

[0108] If the above integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such understanding, to implement all or part of the processes in the above method embodiments of the present invention, it may also be completed by a computer program instructing relevant hardware. The computer program may be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0109] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0110] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the process in Figure 1One or more processes and / or boxes Figure 1 The functions specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 One or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes.

[0112] The above content is only to illustrate the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for improving the time-frequency resolution of seismic data, characterized in that, Including the following steps: S1. Construct a synthetic chirp signal x(t), obtain the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and construct simple label pairs; S2. Combine the simple label pairs obtained in step S1 to obtain complex label pairs; S3. Build an SCNN network model; S4. Use the complex label pairs obtained in step S2 to train the SCNN network model obtained in step S3; S5. Use the SCNN network model trained in step S4 to map the actual signal to obtain the high time-frequency resolution time-frequency spectrum of the signal.

2. The method for improving the time-frequency resolution of seismic data according to claim 1, characterized in that, In step S1, perform a short-time Fourier transform on the constructed linear frequency modulation signal x(t), and then perform modulus square calculation to obtain the corresponding energy spectrum SPEC x (t, f); use the obtained short energy as the input of the training set, and then calculate the Wigner-Ville distribution of the signal to obtain the Wigner-Ville distribution WVD x (t, f) that corresponds one-to-one to the energy spectrum SPEC x (t, f), as a simple label pair.

3. The method for improving the time-frequency resolution of seismic data according to claim 2, characterized in that, Wigner-Ville distribution WVD x (t, f) is as follows: Energy spectrum SpEC x (t,f) is as follows: Wherein, x is the chirp signal, t is the time, τ is the integration variable, g(t) is the window function, f is the frequency of the signal, and x(τ) is the chirp signal after time shift.

4. The method for improving the time-frequency resolution of seismic data according to claim 1, characterized in that, In step S2, the input of the complex label pair is composed of the energy spectra of two or more chirp signals, and the label of the complex label pair is composed of the corresponding Wigner-Ville distributions of two or more chirp signals.

5. The method for improving the time-frequency resolution of seismic data according to claim 1, characterized in that, In step S3, the convolution kernel size in the SCNN network model is N*N, the module of the SCNN network model is the convolution layer and the ReLU activation function layer, and the convolution layer contains K filters with a size of N*N*M.

6. The method for improving the time-frequency resolution of seismic data according to claim 1, characterized in that, Step S4 is specifically as follows: First, preprocess the training set data, perform data augmentation and normalization operations; then set the parameters batch_size, learning rate, and epoch of the SCNN network model; finally, start the end-to-end training of the SCNN network model on the server using the GPU.

7. The method for improving the time-frequency resolution of seismic data according to claim 1, characterized in that, Step S5 is specifically as follows: Construct a synthetic signal obtained by adding three chirp signal components, then perform a short-time Fourier transform on the constructed synthetic signal to obtain the corresponding energy spectrum, and then use the energy spectrum as the input to the trained SCNN network model. The trained network model can achieve the function of a deconvolution sum, perform a convolution operation with the time-frequency spectrum to obtain a time-frequency spectrum with high time-frequency resolution, and obtain the mapping result with improved time-frequency resolution.

8. A system for improving the time-frequency resolution of seismic data, characterized in that, Including: A construction module that constructs a synthetic chirp signal x(t), obtains the energy spectrum and Wigner-Ville distribution of the synthetic chirp signal x(t), and constructs simple label pairs; A combination module that combines the simple label pairs obtained by the construction module to obtain complex label pairs; A building module that builds an SCNN network model; A training module that uses the complex label pairs obtained by the combination module to train the SCNN network model obtained by the building module; An output module that uses the SCNN network model trained by the training module to map the actual signal to obtain the high time-frequency resolution time-frequency spectrum of the signal.

9. A computer-readable storage medium storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to execute any one of the methods according to claims 1 to 7.

10. A computing device, characterized in that, Including: One or more processors, a memory, and one or more programs, where the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods according to claims 1 to 7.

Images