P-wave first sample construction method and device based on joint of compressive sensing and unsupervised learning

Abstract

This invention discloses a method and apparatus for constructing first-arrival samples based on compressed sensing and unsupervised learning. The method includes: preprocessing raw seismic data; preprocessing includes normalization, flattening, and / or denoising; clustering the preprocessed data using an unsupervised clustering algorithm and principal component analysis to obtain multiple categories of first-arrival sample clusters; for each category of first-arrival sample cluster, generating a random sampling matrix using compressed sensing theory, randomly projecting each first-arrival sample in the cluster to obtain a corresponding low-dimensional compressed representation, and generating random sample numbers based on the low-dimensional compressed representation; randomly selecting first-arrival samples according to a preset ratio based on the random sample numbers, and labeling the selected first-arrival samples to obtain a first-arrival sample set. By integrating clustering and selection, sample construction is achieved. Principal component analysis clustering classification makes the clustering results more stable, and selection after compressed projection ensures uniform coverage and reduces reliance on manual intervention.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of seismic data processing and interpretation technology, specifically to a method and apparatus for constructing first arrival samples based on compressed sensing and unsupervised learning. Background Technology

[0002] As seismic exploration advances towards higher precision and efficiency, first-arrival picking, a crucial step in seismic data processing, directly impacts subsequent velocity modeling and imaging quality. Deep learning-based first-arrival picking methods have garnered significant attention due to their high degree of automation and adaptability. However, the performance of deep learning models is highly dependent on the quality and representativeness of the training samples.

[0003] In practical applications, the sheer volume of seismic data, the diverse morphologies of first arrivals, and the high cost of manual annotation make it difficult to obtain high-quality, representative training samples. Traditional random sampling or manual selection methods are inefficient, highly subjective, and fail to capture the diverse characteristics of first arrivals. Existing first arrival picking sample construction mainly relies on the following methods: 1. Manual experience method: Interpreters visually select "typical traces" on the shot gather or superimposed profile. This method is time-consuming and greatly affected by subjective experience and display gain. When the surface structure of the work area is complex (such as desert, piedmont zone, loess plateau, etc.), rare waveforms such as low velocity zone, scattering, and rotation are often missed, resulting in an imbalance of training set categories and a decrease in the model's generalization ability. 2. Random / Equal Interval Sampling: This method involves sampling massive amounts of seismic data using fixed trace spacing or pseudo-random numbers. It is simple to implement but lacks physical constraints. 3. Simple clustering method: K-means and spectral clustering are directly applied to the original waveform or short time window amplitude attributes, but high-dimensional noise redundancy is not considered, and the clustering results are sensitive to the initial center; furthermore, uniform sampling is still used after clustering, and sparse clusters are not protected, and rare features are still insufficient.

[0004] In summary, existing technologies generally lack an integrated framework for feature extraction, clustering, and balanced sampling, making it difficult to achieve a balance between massive amounts of data and carefully selected samples. This has become a bottleneck restricting the initial picking accuracy and engineering efficiency of deep learning. Summary of the Invention

[0005] In view of the above problems, embodiments of the present invention are proposed to provide a method and apparatus for constructing first arrival samples based on compressed sensing and unsupervised learning to overcome or at least partially solve the above problems.

[0006] According to one aspect of the present invention, a method for constructing first-arrival samples based on a combination of compressed sensing and unsupervised learning is provided, the method comprising: The raw seismic data is preprocessed; preprocessing includes normalization, flattening and / or denoising. The preprocessed data is clustered and classified based on an unsupervised clustering algorithm and principal component analysis to obtain initial sample clusters of multiple categories. For each category of initial arrival sample cluster, a random sampling matrix is ​​generated using compressed sensing theory. Each initial arrival sample in the initial arrival sample cluster is randomly projected to obtain the corresponding low-dimensional compressed representation. Random sample numbers are generated based on the low-dimensional compressed representation. The random sample numbers are then randomly selected according to a preset ratio, and the selected initial arrival samples are labeled to obtain the initial arrival sample set.

[0007] According to another aspect of the present invention, an apparatus for constructing first-arrival samples based on a combination of compressed sensing and unsupervised learning is provided, comprising: The preprocessing module is suitable for preprocessing raw seismic data; preprocessing includes normalization, flattening, and / or denoising. The clustering module is suitable for performing clustering and classification on preprocessed data based on unsupervised clustering algorithms and principal component analysis to obtain initial sample clusters of multiple categories. The selection module is suitable for generating a random sampling matrix using compressed sensing theory for each category of initial arrival sample clusters, performing random projection on each initial arrival sample in the initial arrival sample cluster to obtain the corresponding low-dimensional compressed representation, and generating random sample numbers based on the low-dimensional compressed representation; randomly selecting initial arrival samples according to a preset ratio based on the random sample numbers, and labeling the selected initial arrival samples to obtain the initial arrival sample set.

[0008] According to another aspect of the present invention, a computing device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described method for constructing first arrival samples based on compressed sensing and unsupervised learning.

[0009] According to another aspect of the present invention, a computer storage medium is provided, the storage medium storing at least one executable instruction, the executable instruction causing a processor to perform operations corresponding to the above-described method for constructing initial arrival samples based on compressed sensing and unsupervised learning.

[0010] According to another aspect of the present invention, a computer program product is provided, comprising at least one executable instruction that causes a processor to perform operations corresponding to the above-described method for constructing initial arrival samples based on compressed sensing and unsupervised learning.

[0011] The method and apparatus for constructing first arrival samples based on compressed sensing and unsupervised learning provided in this invention integrates the following: clustering is performed on the original seismic data after dimensionality reduction and principal component analysis; random sampling matrix generated by compressed sensing is randomly projected and selected based on random sample numbers; the clustering results are more stable through principal component analysis; and selection after compressed projection ensures uniform coverage. The entire process reduces reliance on manual intervention and fundamentally solves the problems of low sample selection efficiency and poor representativeness in existing technologies, providing a reliable data foundation for the engineering application of deep learning first arrival picking models.

[0012] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific implementation methods of the embodiments of the present invention are described below. Attached Figure Description

[0013] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a method for constructing first-arrival samples based on a joint compressed sensing and unsupervised learning according to an embodiment of the present invention is shown; Figure 2 This diagram illustrates the unsupervised classification of data samples. Figure 3 The sample plane distribution and a typical single shot schematic diagram are shown; Figure 4 This diagram illustrates the spatial distribution of samples before and after compressed sensing selection. Figure 5 A comparative schematic diagram of various sample pickup lines is shown; Figure 6 A schematic diagram of a device for constructing first-arrival samples based on a joint compression sensing and unsupervised learning according to an embodiment of the present invention is shown. Figure 7 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. Detailed Implementation

[0014] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0015] Figure 1 A flowchart of a method for constructing first-arrival samples based on a joint compressed sensing and unsupervised learning according to an embodiment of the present invention is shown, as follows: Figure 1 As shown, the method includes the following steps: Step S101: Preprocess the raw seismic data.

[0016] Based on the raw seismic data, there may be issues such as large gain differences between different shots and traces, leading to inconsistent amplitude magnitudes; or, due to shot-receiver distances and surface undulations, the first arrival time may drift across different traces, causing inconsistent first arrival times; or, it may contain noise interference such as environmental noise, surface waves, and multiple waves; or, the data formats acquired under different operational conditions may be inconsistent. Considering the subsequent unsupervised clustering and compressed sensing sampling, preprocessing is performed first. Preprocessing includes normalization, flattening, and denoising to unify the format and scale of the seismic data, eliminate noise interference, and enhance the identifiability of first arrival features.

[0017] Specifically, normalization can eliminate differences in amplitude magnitude between different traces, ensuring that seismic data are on the same scale. Normalization can be achieved through methods such as intra-trace normalization, which calculates the ratio of the original seismic data within each trace to the maximum absolute value within that trace, thus ensuring the amplitude range of the processed seismic data is within a preset range, such as [0, 1]. Alternatively, global normalization or mean-variance normalization can be used depending on the implementation; this is not limited here. Flattening can eliminate time drift. Methods such as picking, peak alignment, and cross-correlation alignment can be used to align the first arrival times of the seismic data, facilitating subsequent clustering that focuses on waveform shape rather than time position. Denoising removes noise from the original seismic data that is not the first arrival wave, enhancing the recognizability of the first arrival feature. One or more combinations of methods such as bandpass filtering, median filtering, and wavelet transform denoising can be used, depending on the implementation; this is not limited here.

[0018] After preprocessing, the resulting seismic data has fixed dimensions, consistent amplitude range, first arrival alignment, and noise suppression, which facilitates subsequent clustering processing, avoids data with large variance dominating the principal component direction, provides stable input for clustering, and makes it easier for waveforms with similar shapes to be clustered into the same category.

[0019] Step S102: Based on the unsupervised clustering algorithm, the preprocessed data is clustered and classified according to principal component analysis to obtain initial sample clusters of multiple categories.

[0020] The preprocessed data can be represented as:

[0021] Where n is the number of sampling points for each data point (high dimension), and N is the total number of data points.

[0022] To achieve zero mean, subtract the mean from each data point. The covariance matrix is ​​calculated as follows:

[0023] Eigenvalue decomposition based on the covariance matrix Where V is the eigenvector matrix and Λ is the eigenvalue diagonal matrix. The obtained eigenvalues ​​are sorted in descending order (from largest to smallest). A predetermined number of features at the top of the sorted list are combined to obtain the projection matrix. For example, if the predetermined number is k, the resulting projection matrix would be as follows:

[0024] The low-dimensional principal components of the preprocessed data are determined based on the projection matrix. The low-dimensional principal components of the data are as follows:

[0025]

[0026] Clustering is performed based on low-dimensional principal component analysis using unsupervised clustering algorithms, such as K-means, PCA, DBSCAN, or autoencoder clustering. First arrivals with similar morphologies are grouped into the same category, ensuring that subsequent sampling can process different morphologies separately, avoiding the obscuring of rare features. This clustering results in multiple discrete clusters of first arrival samples, such as... , where C i That is, the i-th initial arrival sample cluster, each initial arrival sample cluster contains different initial arrival samples x1…x j .

[0027] Step S103: For each category of initial arrival sample cluster, a random sampling matrix is ​​generated using compressed sensing theory. Each initial arrival sample in the initial arrival sample cluster is randomly projected to obtain the corresponding low-dimensional compressed representation. Random sample numbers are generated based on the low-dimensional compressed representation. Random sample numbers are randomly selected according to a preset ratio, and the selected initial arrival samples are labeled to obtain the initial arrival sample set.

[0028] After obtaining the initial sample clusters, the initial samples (preprocessed data) in each initial sample cluster are renumbered, such as the initial sample cluster... ,Include The first arriving sample, i.e. When renumbering, a random sampling matrix is ​​generated using compressed sensing theory. This random sampling matrix satisfies the RIP constraint of isometry, where RIP guarantees that compressed sampling can preserve the key information of the original signal as losslessly as possible.

[0029] For each category of initial arrival sample cluster, a random sampling matrix is ​​used to perform a random linear projection on each initial arrival sample in the cluster, resulting in a low-dimensional compressed representation of each initial arrival sample:

[0030] The hash score of each initial sample is determined based on the random vector w and the low-dimensional compressed representation, and is used as the random sample number. For example, the uniform random hash score is calculated. As a random sample number:

[0031] The random sample numbers are arranged in ascending order, and the initial arrival samples corresponding to the first few random sample numbers in the sorting are selected according to a preset ratio. The preset ratio can be: The value ranges from 1% to 10%, as seen in the initial sample cluster. Select according to the preset ratio First arrival samples are selected from each cluster. The first arrival samples are merged, and each merged sample is labeled. This labeling can be done manually or semi-automatically, depending on the implementation. For a small number of samples with extremely high accuracy requirements, manual labeling is preferable; for a large number of samples, automatic labeling algorithms such as automatic picking can be used. For complex waveforms, manual verification can be performed, but there are no specific limitations here. The labeled first arrival samples from each cluster are then merged to obtain a first arrival sample set. This first arrival sample set can be input into a preset first arrival picking model to improve the model's generalization ability and picking accuracy on real-world data, thereby obtaining the trained preset first arrival picking model.

[0032] Furthermore, after selecting the initial arrival samples, class imbalance can be checked by calculating the ratio between the number of each initial arrival sample cluster and the number of preprocessed data, such as the number of a certain initial arrival sample cluster. If the number of preprocessed data is N, then the proportion is... If the proportion of / N is lower than the rare threshold, such as 1%, then the initial sample cluster can be determined to be a rare cluster, with a small number of clusters. The preset proportion of the determined rare clusters can be adjusted based on the verification results. Increase the number of rare clusters in the final initial sample set. Alternatively, if the proportion of rare clusters exceeds the oversized threshold, the initial sample cluster can be identified as an oversized cluster. If the number of oversized clusters is excessive, the preset proportion of the identified oversized clusters can be adjusted based on the verification results. To reduce and avoid an excessive number of super-large clusters in the final initial sample set.

[0033] In one alternative embodiment, Figure 2 Unsupervised classification of data samples from a certain district, including three categories. Figure 2 The corresponding sample principal component feature map is as follows: Class 1: high signal-to-noise ratio, easy to pick; Class 2: medium signal-to-noise ratio, relatively easy to pick; Class 3: low signal-to-noise ratio, difficult to pick. Figure 2 b is the main component feature vector classification, with blue representing class 1, green representing class 2, and yellow representing class 3; Figure 2 c represents the sample size statistics for each category. Figure 3 The diagram shows the spatial location of the classified samples and a typical single shot, where green corresponds to the first class, red to the second class, and purple to the third class. Figure 4 middle Figure 4 'a' represents the spatial location of the sample before selection. Figure 4 b represents the planar distribution of samples after compressed sensing selection, where green corresponds to the first class, red to the second class, and purple to the third class. Figure 5 This is a comparison chart showing the results of using the traditional method (green picking line), manually selecting samples for deep learning initial arrival picking (red picking line), and the method of this embodiment (blue picking line). Figure 5 It can be seen that the sample features obtained in this embodiment are more diverse and its accuracy is higher.

[0034] The first arrival sample construction method based on compressed sensing and unsupervised learning provided by the present invention performs clustering based on the original seismic data through dimensionality reduction and principal component analysis. The random sampling matrix generated by compressed sensing is randomly projected and selected based on random sample numbers in an integrated manner. The clustering results are more stable through principal component analysis, and the selection after compressed projection ensures uniform coverage. The whole process reduces the dependence on manual intervention and fundamentally solves the problems of low sample selection efficiency and poor representativeness in existing technologies. It provides a reliable data foundation for the engineering application of deep learning first arrival picking models.

[0035] Figure 6 A schematic diagram of the structure of the first-arrival sample construction device based on compressed sensing and unsupervised learning provided in an embodiment of the present invention is shown. Figure 6 As shown, the device includes: The preprocessing module 610 is suitable for preprocessing raw seismic data; the preprocessing includes normalization, flattening and / or denoising. Clustering module 620 is suitable for performing clustering and classification on preprocessed data based on unsupervised clustering algorithms and principal component analysis to obtain initial sample clusters of multiple categories; The selection module 630 is suitable for generating a random sampling matrix using compressed sensing theory for each category of initial arrival sample cluster, performing random projection on each initial arrival sample in the initial arrival sample cluster to obtain the corresponding low-dimensional compressed representation, and generating random sample numbers based on the low-dimensional compressed representation; randomly selecting initial arrival samples according to a preset ratio based on the random sample numbers, and labeling the selected initial arrival samples to obtain the initial arrival sample set.

[0036] Alternatively, the clustering module 620 is further adapted to: The covariance matrix is ​​calculated by zero-mean normalizing the preprocessed data. The eigenvalues ​​are decomposed based on the covariance matrix, sorted in descending order, and the first set of features are combined to obtain the projection matrix. The low-dimensional principal components of each preprocessed data are determined based on the projection matrix. Then, clustering and classification are performed based on the low-dimensional principal components using an unsupervised clustering algorithm to obtain initial sample clusters of multiple categories.

[0037] Optionally, module 630 is further adapted to: A random sampling matrix is ​​generated using compressed sensing theory, and the random sampling matrix satisfies the constraint isometry condition. For each category of initial arrival sample cluster, a random sampling matrix is ​​used to perform random linear projection on each initial arrival sample in the initial arrival sample cluster to obtain a low-dimensional compressed representation of each initial arrival sample. The hash score of each initial sample is determined based on the random vector and low-dimensional compressed representation, and used as the random sample number.

[0038] Optionally, module 630 is further adapted to: Arrange the random sample numbers in ascending order, and select the initial arrival samples corresponding to the first few random sample numbers according to a preset ratio; The first arrival samples selected from each first arrival sample cluster are merged and labeled to obtain the first arrival sample set.

[0039] Optionally, the device further includes: a rare adjustment module 640, adapted to perform class imbalance verification based on the ratio of the number of each initial sample cluster to the preprocessed data, determine rare clusters based on the verification results, and adjust the preset ratio of rare clusters; the rare clusters are determined based on the ratio of the number of clusters.

[0040] Optionally, the device further includes a training module 650, adapted to input the obtained initial arrival sample set into a preset initial arrival picking model to obtain a trained preset initial arrival picking model.

[0041] The descriptions of the above modules refer to the corresponding descriptions in the method embodiments, and will not be repeated here.

[0042] This invention also provides a non-volatile computer storage medium storing at least one executable instruction that can perform the operation corresponding to the method for constructing initial arrival samples based on compressed sensing and unsupervised learning in any of the above method embodiments.

[0043] This application provides a computer program product, which includes at least one executable instruction or computer program that enables a processor to perform the operation corresponding to the method for constructing initial arrival samples based on compressed sensing and unsupervised learning in any of the above method embodiments.

[0044] Figure 7 The diagram illustrates the structure of a computing device according to an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computing device.

[0045] like Figure 7 As shown, the computing device may include: a processor 702, a communication interface 704, a memory 706, and a communication bus 708.

[0046] in: The processor 702, communication interface 704, and memory 706 communicate with each other via communication bus 708.

[0047] The communication interface 704 is used to communicate with other network elements such as clients or other servers.

[0048] The processor 702 is used to execute program 710, which can specifically execute the relevant steps in the above embodiment of the method for constructing first arrival samples based on compressed sensing and unsupervised learning.

[0049] Specifically, program 710 may include program code that includes computer operation instructions.

[0050] The processor 702 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0051] Memory 706 is used to store program 710. Memory 706 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0052] Specifically, program 710 can be used to cause processor 702 to execute the method for constructing first arrival samples based on compressed sensing and unsupervised learning in any of the above method embodiments. The specific implementation of each step in program 710 can be found in the corresponding descriptions of the steps and units in the above embodiments for constructing first arrival samples based on compressed sensing and unsupervised learning, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0053] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the embodiments of the present invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing preferred embodiments of the present invention.

[0054] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0055] Similarly, it should be understood that, in order to streamline the embodiments of the invention and aid in understanding one or more of the various inventive aspects, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed embodiments of the invention require more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0056] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0057] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0058] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention can also be implemented as device or apparatus programs (e.g., computer programs and computer program products) for performing part or all of the methods described herein. Such programs implementing the embodiments of the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0059] It should be noted that the above embodiments are illustrative of the present invention and not restrictive of the invention, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the present invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for constructing first-arrival samples based on a combination of compressed sensing and unsupervised learning, characterized in that, The methods include: The raw seismic data is preprocessed; the preprocessing includes normalization, flattening and / or denoising. The preprocessed data is clustered and classified based on an unsupervised clustering algorithm and principal component analysis to obtain initial sample clusters of multiple categories. For each category of initial arrival sample cluster, a random sampling matrix is ​​generated using compressed sensing theory. Each initial arrival sample in the initial arrival sample cluster is randomly projected to obtain a corresponding low-dimensional compressed representation. Random sample numbers are generated based on the low-dimensional compressed representation. The random sample numbers are then randomly selected according to a preset ratio, and the selected initial arrival samples are labeled to obtain the initial arrival sample set.

2. The method according to claim 1, characterized in that, The process of clustering the preprocessed data using an unsupervised clustering algorithm and principal component analysis to obtain multiple initial sample clusters further includes: The covariance matrix is ​​calculated by zero-mean normalizing the preprocessed data. Based on the covariance matrix, eigenvalues ​​are decomposed and sorted in descending order. A predetermined number of features that are sorted first are combined to obtain the projection matrix. The low-dimensional principal components of each preprocessed data are determined based on the projection matrix. Then, clustering and classification are performed based on the low-dimensional principal components using an unsupervised clustering algorithm to obtain initial sample clusters of multiple categories.

3. The method according to claim 1, characterized in that, For each category of initial arrival sample cluster, a random sampling matrix is ​​generated using compressed sensing theory. Random projection is then performed on each initial arrival sample in the initial arrival sample cluster to obtain a corresponding low-dimensional compressed representation. The generation of random sample numbers based on the low-dimensional compressed representation further includes: A random sampling matrix is ​​generated using compressed sensing theory, and the random sampling matrix satisfies the constraint isometry condition. For each category of initial arrival sample cluster, the random sampling matrix is ​​used to perform random linear projection on each initial arrival sample in the initial arrival sample cluster to obtain a low-dimensional compressed representation corresponding to each initial arrival sample. The hash score of each initial sample is determined based on the random vector and the low-dimensional compressed representation, and used as the random sample number.

4. The method according to claim 1, characterized in that, The step of randomly selecting samples according to the random sample numbers at a preset ratio and labeling the selected initial arrival samples to obtain the initial arrival sample set further includes: Arrange the random sample numbers in ascending order, and select the initial arrival samples corresponding to the first multiple random sample numbers according to a preset ratio; The first arrival samples selected from each first arrival sample cluster are merged and labeled to obtain the first arrival sample set.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: A class imbalance check is performed based on the ratio of the number of each initial sample cluster to the preprocessed data. Rare clusters are determined based on the check results, and the preset ratio of rare clusters is adjusted. The rare clusters are determined based on the ratio.

6. The method according to any one of claims 1-4, characterized in that, The method further includes: The obtained initial arrival sample set is input into the preset initial arrival picking model to obtain the trained preset initial arrival picking model.

7. A device for constructing first-arrival samples based on compressed sensing and unsupervised learning, characterized in that, The device includes: The preprocessing module is suitable for preprocessing raw seismic data; the preprocessing includes normalization, flattening and / or denoising. The clustering module is suitable for performing clustering and classification on preprocessed data based on unsupervised clustering algorithms and principal component analysis to obtain initial sample clusters of multiple categories. The selection module is suitable for generating a random sampling matrix using compressed sensing theory for each category of initial arrival sample clusters, performing random projection on each initial arrival sample in the initial arrival sample cluster to obtain the corresponding low-dimensional compressed representation, and generating random sample numbers based on the low-dimensional compressed representation; randomly selecting initial arrival samples according to a preset ratio based on the random sample numbers, and labeling the selected initial arrival samples to obtain the initial arrival sample set.

8. A computing device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the method for constructing first arrival samples based on the combined compressed sensing and unsupervised learning as described in any one of claims 1-6.

9. A computer storage medium, characterized in that, The storage medium stores at least one executable instruction that causes the processor to perform the operation corresponding to the method for constructing first arrival samples based on the combined compressed sensing and unsupervised learning as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes at least one executable instruction that causes the processor to perform the operation corresponding to the method for constructing first arrival samples based on the joint compression sensing and unsupervised learning as described in any one of claims 1-6.

Images