Training method of fault recognition model based on orthogonal annotation, fault recognition method, electronic device and storage medium

Abstract

The present invention provides a training method for a fault recognition model based on orthogonal labeling, a fault recognition method, an electronic device, and a storage medium, which can be applied to the field of fault recognition technology. The method includes: slicing a seismic data sample set in multiple orthogonal dimensions to obtain a multi-dimensional slice sample set; obtaining initial pseudo-labels for the multi-dimensional slice sample set by alternating multiple rounds of pseudo-label generation operations and parameter update operations of a segmentation network; performing multiple rounds of cross-optimization between different dimensions on the initial pseudo-labels of the multi-dimensional slice sample set to fuse pseudo-labels of multi-dimensional information; and performing multiple rounds of semi-supervised training on the fault recognition model using the seismic data sample set with pseudo-labels and an unlabeled seismic data training set to obtain a trained fault recognition model.

Description

Technical Field

[0001] The present invention relates to the technical field of fault recognition, and in particular to a training method for a fault recognition model based on orthogonal annotation, a fault recognition method, an electronic device, and a storage medium. Background Art

[0002] Fault identification is a crucial step in seismic data interpretation. Identifying faults can reveal the movement history of strata, fracture characteristics, and the distribution of oil and gas reservoirs, which is crucial for oil and gas exploration and geological hazard assessment. However, current fault identification technology faces numerous challenges, including the complexity and noise of seismic data, the diversity of fault morphology, limitations in data quality and resolution, and a lack of high-quality labels. These factors complicate the extraction and accurate identification of fault features. Current fault identification methods are primarily categorized into supervised learning, weakly supervised learning, and semi-supervised learning.

[0003] Existing fault identification methods based on supervised learning rely on a large number of high-quality labels, resulting in high labeling costs. While weakly supervised and semi-supervised learning methods require fewer labels than supervised learning methods, they still typically require a certain number of high-quality labels. While the introduction of extremely weakly supervised learning methods based on orthogonal labeling can address some of these technical issues, existing pseudo-label generation strategies struggle to generate high-quality fault pseudo-labels without pre-training on real labels. Summary of the Invention

[0004] In view of the above problems, the present invention provides a training method for a fault recognition model based on orthogonal annotation, a fault recognition method, an electronic device, and a storage medium that improve the accuracy and efficiency of fault recognition.

[0005] According to a first aspect of the present invention, a method for training a fault recognition model based on orthogonal annotation is provided, comprising:

[0006] Slicing the seismic data sample set in multiple orthogonal dimensions to obtain a multi-dimensional slice sample set, wherein the multi-dimensional slice sample set includes a time dimension slice sample set, an inline dimension slice sample set, and a cross dimension slice sample set;

[0007] The segmentation network is used to generate pseudo labels for the multi-dimensional slice sample set, and the loss value in the pseudo label generation process is used to update the parameters of the segmentation network. The initial pseudo labels of the multi-dimensional slice sample set are obtained by alternating multiple rounds of pseudo label generation and parameter update operations.

[0008] The initial pseudo labels of the multi-dimensional slice sample set are cross-optimized in multiple rounds across different dimensions using an optimized loss function. The three-dimensional information of the seismic data volume is introduced into the cross-optimization process to obtain pseudo labels that integrate the multi-dimensional information.

[0009] A semi-supervised learning framework based on contextual prototype-aware learning technology uses a pseudo-labeled seismic data sample set and an unlabeled seismic data training set to perform multiple rounds of training on the fault recognition model. The training process of the fault recognition model is supervised by a training loss function to obtain a trained fault recognition model.

[0010] According to an embodiment of the present invention, the above-mentioned method of generating pseudo labels for a multi-dimensional slice sample set using a segmentation network, and updating parameters of the segmentation network using the loss value in the pseudo label generation process, and obtaining initial pseudo labels for the multi-dimensional slice sample set by alternating multiple rounds of pseudo label generation operations and parameter update operations includes:

[0011] Using the first segmentation network, downsampling and upsampling operations are performed on the first time slice in the time dimension slice sample set to generate the first round of iterative pseudo labels for the first time slice;

[0012] The pseudo-label loss function is used to process the first-round iteration pseudo-label of the first time slice and the true label corresponding to the first time slice, and the pseudo-label loss value of the first round iteration is calculated;

[0013] Through the back-propagation operation, the first segmentation network is updated with the pseudo-label loss value of the first iteration to obtain the first segmentation network after the first iteration;

[0014] Iteratively performing a pseudo-label generation operation, a pseudo-label loss value calculation operation, and a parameter update operation until the number of iterations meets a first preset iteration value, thereby obtaining an initial pseudo-label for the first time slice;

[0015] The first segmentation network after the previous iteration and the initial pseudo label of the previous time slice are used to perform the same processing operation as the first time slice on each time slice in the time dimension slice sample set to obtain the initial pseudo label of the time dimension slice sample set.

[0016] According to an embodiment of the present invention, the above-mentioned method of generating pseudo labels for the multi-dimensional slice sample set using a segmentation network, and updating the parameters of the segmentation network using the loss value in the pseudo label generation process, and obtaining the initial pseudo labels for the multi-dimensional slice sample set by alternating multiple rounds of pseudo label generation operations and parameter update operations further includes:

[0017] Using the second segmentation network, the inline dimension slice sample set is processed in the same way as the time dimension slice sample set to obtain the initial pseudo label of the inline dimension slice sample set;

[0018] The third segmentation network is used to perform the same processing operation on the cross-dimensional slice sample set as that on the time dimension slice sample set to obtain the initial pseudo label of the cross-dimensional slice sample set.

[0019] According to an embodiment of the present invention, the above-mentioned use of the optimization loss function to perform multiple rounds of cross-optimization between different dimensions on the initial pseudo labels of the multi-dimensional slice sample set, and by introducing the three-dimensional information of the seismic data volume in the cross-optimization process, the pseudo labels that integrate the multi-dimensional information are obtained, including:

[0020] Perform mask generation operation on pixels in the seismic data sample set that are lower than a preset pixel threshold to obtain a pixel mask, and perform pairwise combination of the multi-dimensional slice sample set on the dimensions to obtain a cross-dimensional slice sample set;

[0021] The pixel mask and the initial pseudo-label of the cross-dimensional slice sample set corresponding to the pixel mask are processed by optimizing the loss function to obtain the cross-optimization loss value;

[0022] Use the cross-optimization loss value to update the parameters of the segmentation network obtained during the initial pseudo-label generation process in the cross-optimization phase;

[0023] The segmentation network obtained in the cross optimization stage is used to regenerate the initial pseudo labels for the multi-dimensional slice samples.

[0024] The calculation operation of the cross-optimization loss value, the segmentation network parameter update operation in the cross-optimization stage, and the regeneration operation of the initial pseudo-label are iteratively performed until the cross-optimization loss value is lower than the preset threshold, and a pseudo-label that integrates multi-dimensional information is obtained.

[0025] According to an embodiment of the present invention, the semi-supervised learning framework based on contextual prototype-aware learning technology uses a seismic data sample set with pseudo labels and a seismic data training set with unlabeled data to perform multiple rounds of training on the fault recognition model, and uses a training loss function to supervise the training process of the fault recognition model. The trained fault recognition model includes:

[0026] A semi-supervised learning framework is constructed based on the teacher neural network and the student neural network, and the student neural network is used as a fault recognition model;

[0027] Introducing contextual prototype-aware learning technology into the semi-supervised learning framework and constructing a training loss function based on class activation maps;

[0028] The teacher neural network is used to process the unlabeled earthquake data training set to obtain a first output result, and the student neural network is used to process the earthquake data sample set with pseudo labels to obtain a second output result;

[0029] The training loss value is calculated by processing the first output result and the second output result through the training loss function, and the parameters of the student neural network are updated using the training loss value based on the back propagation mechanism;

[0030] Based on the student neural network after parameter update, the teacher neural network parameters are updated through the exponential average moving operation;

[0031] The training loss value calculation operation, output result acquisition operation, and neural network parameter update operation are iteratively performed until the preset training conditions are met to obtain a trained fault recognition model.

[0032] According to an embodiment of the present invention, the above-mentioned introduction of contextual prototype-aware learning technology into the semi-supervised learning framework and construction of a training loss function based on a class activation map include:

[0033] The output of the teacher neural network is thresholded to generate a binary mask, and masked average pooling is performed on the unlabeled seismic data training set based on the binary mask to obtain the instance prototype;

[0034] The instance prototypes are stored in a dynamically updateable support library, and the instance prototypes are clustered using a preset clustering algorithm;

[0035] The clustered instance prototypes are used to construct a candidate context prototype set, and the cosine similarity between each candidate context prototype and the instance prototype in the candidate context prototype set is calculated;

[0036] Based on cosine similarity, multiple candidate context prototypes that are most similar to the instance prototype are selected from the candidate context prototype set, and the context-aware prototype set is constructed using the selected multiple candidate context prototypes.

[0037] According to an embodiment of the present invention, the above-mentioned introduction of contextual prototype-aware learning technology into the semi-supervised learning framework and construction of a training loss function based on a class activation map further include:

[0038] Calculate the positive correlation weight of each context-aware prototype and the instance prototype in the context-aware prototype set based on the parameter-free identity mapping layer and activation function;

[0039] Performing weighted calculation on each context-aware prototype in the context-aware prototype set using the positive correlation weight to obtain a weighted context-aware prototype set;

[0040] Calculate the distribution gap between the instance prototype and each context-aware prototype in the context-aware prototype set, and obtain an offset term that represents the distribution gap;

[0041] The instance prototype is replaced by the result of the operation of the instance prototype and the offset term to obtain a revised instance prototype, and the weighted context-aware prototype set is revised by using the revised instance prototype to obtain a revised context-aware prototype set;

[0042] The class activation map is obtained using the revised context-aware prototype set and the unlabeled seismic data training set, and the training loss function is constructed using the class activation map.

[0043] A second aspect of the present invention provides a fault identification method, comprising:

[0044] The trained fault recognition model is used to process the fault data to be identified to obtain a fault recognition result, wherein the trained fault recognition model is obtained by training the above-mentioned fault recognition model based on orthogonal annotation, and the fault data to be identified includes an earthquake fault image.

[0045] A third aspect of the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0046] The fourth aspect of the present invention further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the computer program or instructions are executed by a processor.

[0047] The training method of the above-mentioned fault recognition model based on orthogonal annotation provided by the present invention significantly reduces the dependence on high-quality full labels by introducing the orthogonal annotation method, thereby reducing the manual annotation cost in the fault identification process; at the same time, a special pseudo-label generation strategy is designed for the complexity of seismic data, which is more adaptable to the nonlinear and high-dimensional characteristics of seismic data; in addition, the introduction of the semi-supervised learning framework enables the present invention to only generate pseudo-labels for a small amount of data, while combining a large amount of unlabeled data to complete model training. At the same time, context-aware learning technology is introduced into the semi-supervised learning framework, which further enhances the ability of the present invention to extract seismic data features. Therefore, the fault recognition model trained by the training method of the fault recognition model based on orthogonal annotation provided by the present invention has wide applicability, is not only suitable for fault recognition, but can also be extended to other segmentation tasks that are difficult to label with high quality, and has high application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0049] Figure 12 is an application scenario diagram of a fault identification model training method and a fault identification method based on orthogonal annotation according to an embodiment of the present invention;

[0050] Figure 2 is a flow chart of a method for training a fault recognition model based on orthogonal annotation according to an embodiment of the present invention;

[0051] Figure 3 2 is a schematic diagram of a pseudo-label generation process according to an embodiment of the present invention;

[0052] Figure 4 is a schematic diagram of training a fault recognition model according to a semi-supervised learning framework according to an embodiment of the present invention;

[0053] Figure 5 is a structural block diagram of a training device for a fault recognition model based on orthogonal annotation according to an embodiment of the present invention;

[0054] Figure 6 4 is a block diagram of an electronic device suitable for implementing a training method for a fault recognition model based on orthogonal annotation and a fault recognition method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concept of the present invention.

[0056] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the presence of the features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0057] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0058] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0059] Existing fault identification methods include supervised fault identification methods, weakly supervised fault identification methods and semi-supervised fault identification methods.

[0060] Supervised fault identification methods rely on high-quality, manually annotated labels, using deep learning models to learn fault characteristics from data. These methods offer the advantages of high accuracy and model stability, but they are highly dependent on labels. Seismic data is complex, nonlinear, and highly noisy, making it difficult to obtain true fault labels, requiring significant time and effort from experts. Furthermore, while synthetic data can compensate for label deficiencies to some extent, its distribution differs from real data, limiting its generalization capabilities.

[0061] Given the aforementioned technical challenges with supervised fault identification methods, exploring fault identification techniques using weakly supervised and semi-supervised learning is of great value. These methods can utilize partially labeled data and a large amount of unlabeled data for training when labels are insufficient, reducing the reliance on labels to some extent. However, these methods still typically require a certain number of high-quality labels as initial guidance, and balancing the quality of pseudo-labels with model robustness is often difficult.

[0062] In the task of medical image segmentation, the orthogonal labeling method has been proposed and applied. This method only labels some orthogonal vertical slices, effectively reducing the need for full labeling, alleviating the labeling burden, and achieving extremely weak supervised learning. Its advantage is that it can use a very small amount of annotation to generate pseudo-labels to guide model training. However, this method cannot be directly applied to the field of earthquake fault identification. The current pseudo-label generation strategies for orthogonal labeling mainly include image registration and pre-trained neural network-based methods: the former requires a strict registration relationship between images, which is difficult to meet for seismic data due to the complexity of the acquisition method; the latter requires real labels to pre-train the neural network, and the quality of the pseudo-labels is directly related to it, which still cannot completely get rid of the dependence on real labels.

[0063] In order to overcome at least one of the problems of the prior art, an embodiment of the present invention provides a training method and a fault recognition method for a fault recognition model based on orthogonal labeling, which relies on an extremely small number of labeled slices to generate a large number of reliable pseudo-labels, and then uses the pseudo-labels to guide the training of the fault recognition model according to a semi-supervised learning framework, so that the fault recognition model obtained by training can improve the recognition efficiency and accuracy of fault data.

[0064] Figure 1 2 is an application scenario diagram of a fault recognition model training method and a fault recognition method based on orthogonal annotation according to an embodiment of the present invention.

[0065] like Figure 1 As shown, the application scenario 100 according to this embodiment may include the field of fault identification technology. A network 104 is used as a medium for providing a communication link between a first terminal device 101, a second terminal device 102, a third terminal device 103, and a server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0066] A user may use a first terminal device 101, a second terminal device 102, or a third terminal device 103 to interact with a server 105 via a network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, or the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (for example only).

[0067] The first terminal device 101 , the second terminal device 102 , and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0068] The server 105 may be a server that provides various services, such as a background management server (for example only) that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process received data such as user requests, and feed back processing results (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal devices.

[0069] It should be noted that the training method and fault identification method of the orthogonal annotation-based fault identification model provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the training device of the orthogonal annotation-based fault identification model provided in the embodiment of the present invention can generally be set in the server 105. The training method and fault identification method of the orthogonal annotation-based fault identification model provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the training device of the orthogonal annotation-based fault identification model provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0070] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0071] The following will be based on Figure 1 The scene described by Figures 2 to 4 The training method of the fault recognition model based on orthogonal annotation and the fault recognition method of the disclosed embodiment are described in detail.

[0072] Figure 2 4 is a flow chart of a method for training a fault recognition model based on orthogonal annotation according to an embodiment of the present invention.

[0073] like Figure 2 As shown, the above-mentioned training method of the fault recognition model based on orthogonal annotation includes operations S210 to S240.

[0074] In operation S210 , the seismic data sample set is sliced ​​in multiple orthogonal dimensions to obtain a multi-dimensional slice sample set, wherein the multi-dimensional slice sample set includes a time dimension slice sample set, an inline dimension slice sample set, and a cross dimension slice sample set.

[0075] Operation S210 is used to construct the data sample set used for model training, that is, to cut the seismic data sample set into a number of three-dimensional volumes of specific dimensions. The present invention selects the data dimensions as (128, 128, 128), which represent the vertical (time dimension, the same below), inline (inline dimension, the same below) and crossline (cross dimension, the same below) directions respectively. Three-dimensional seismic volume , the vertical direction of the 3D seismic volume Slices are recorded as , in the inline direction Slices are recorded as , in the crossline direction Slices are recorded as . No. The labels of the 3D seismic volumes are denoted by , a label slice is required in each direction to guide the generation of pseudo labels, that is 、 and .

[0076] The three dimensions mentioned above: vertical, inline, and crossline are orthogonal to each other. Based on the technical insights of this disclosure, those skilled in the art may employ other orthogonal dimensions, not limited to the three mentioned above, or convert non-orthogonal dimensions into orthogonal dimensions through data preprocessing.

[0077] In operation S220, the segmentation network is used to generate pseudo labels for the multidimensional slice sample set, and the loss value in the pseudo label generation process is used to update the parameters of the segmentation network. The initial pseudo labels of the multidimensional slice sample set are obtained by alternating multiple rounds of pseudo label generation operations and parameter update operations.

[0078] Operation S220 is used for the initial generation of pseudo labels. In the initial generation process of pseudo labels, a segmentation network, such as a U-Net network, is used to predict pseudo labels. U-Net is a convolutional neural network with a symmetrical structure of encoder and decoder. The encoder extracts multi-level features through convolution and pooling, and the decoder gradually restores the spatial resolution through convolution and upsampling. At the same time, it uses jump connections to fuse the high-resolution features of the encoder to achieve end-to-end segmentation from input to output. Initialize three U-Net networks for the expansion of the three directions of orthogonal annotation. The U-Net network used for pseudo label expansion in the vertical direction is recorded as , the U-Net network used for pseudo-label expansion in the inline direction is recorded as , the U-Net network used for pseudo-label expansion in the crossline direction is recorded as .

[0079] The above operation S220 involves two processes, namely, the pseudo-label generation process and the U-Net network parameter update process. The above two processes are performed alternately. After the parameters of the U-Net network are iteratively updated to a certain extent, the latest updated U-Net is used to generate the initial pseudo-label.

[0080] Based on the technical inspiration provided by the present invention, those skilled in the art can adopt other reasonable image segmentation networks according to actual needs.

[0081] In operation S230, the initial pseudo labels of the multi-dimensional slice sample set are cross-optimized in multiple rounds in different dimensions using an optimization loss function, and the three-dimensional information of the seismic data volume is introduced into the cross-optimization process to obtain pseudo labels that integrate the multi-dimensional information.

[0082] The above operation S230 mainly optimizes the loss value calculation and updates the parameters of the U-Net network again, realizes the fusion of information between different dimensions through cross-optimization, and uses the pseudo-label that integrates multi-dimensional information as the label of the seismic data sample set.

[0083] In operation S240, a semi-supervised learning framework based on contextual prototype-aware learning technology is used to perform multiple rounds of training on the fault recognition model using a pseudo-labeled seismic data sample set and an unlabeled seismic data training set, and a training loss function is used to supervise the training process of the fault recognition model to obtain a trained fault recognition model.

[0084] The above operations S210 to S240 schematically illustrate a training method for a fault recognition model for earthquake fault data identification. Under the technical inspiration of the present invention, the above operations S210 to S240 can also be applied to the training of other tomographic image recognition models, such as the recognition of computed tomography images (CT images), that is, a model or method for other tomographic image recognition can be obtained through the above operations S210 to S240.

[0085] The training method of the above-mentioned fault recognition model based on orthogonal annotation provided by the present invention significantly reduces the dependence on high-quality full labels by introducing the orthogonal annotation method, thereby reducing the manual annotation cost in the fault identification process; at the same time, a special pseudo-label generation strategy is designed for the complexity of seismic data, which is more adaptable to the nonlinear and high-dimensional characteristics of seismic data; in addition, the introduction of the semi-supervised learning framework enables the present invention to only generate pseudo-labels for a small amount of data, while combining a large amount of unlabeled data to complete model training. At the same time, context-aware learning technology is introduced into the semi-supervised learning framework, which further enhances the ability of the present invention to extract seismic data features. Therefore, the fault recognition model trained by the training method of the fault recognition model based on orthogonal annotation provided by the present invention has wide applicability, is not only suitable for fault recognition, but can also be extended to other segmentation tasks that are difficult to label with high quality, and has high application value.

[0086] According to an embodiment of the present invention, the above-mentioned method of generating pseudo labels for a multi-dimensional slice sample set using a segmentation network and updating parameters of the segmentation network using the loss value in the pseudo label generation process, and obtaining initial pseudo labels for the multi-dimensional slice sample set by alternating multiple rounds of pseudo label generation operations and parameter update operations includes: using a first segmentation network to perform downsampling operations and upsampling operations on a first time slice in a time dimension slice sample set in succession to generate a first-round iterative pseudo label for the first time slice; processing the first-round iterative pseudo label of the first time slice and the true label corresponding to the first time slice using a pseudo label generation loss function to calculate a first-round iterative pseudo label loss value; updating parameters of the first segmentation network using the first-round iterative pseudo label loss value through a backpropagation operation to obtain a first segmentation network after the first round of iteration; iteratively performing pseudo label generation operations, pseudo label loss value calculation operations, and parameter update operations until the number of iterations meets a first preset iteration value to obtain an initial pseudo label for the first time slice; using the first segmentation network after the previous round of iteration and the initial pseudo label of the previous time slice to perform the same processing operation as the first time slice on each time slice in the time dimension slice sample set to obtain an initial pseudo label for the time dimension slice sample set.

[0087] According to an embodiment of the present invention, the above-mentioned method of using a segmentation network to generate pseudo labels for a multi-dimensional slice sample set, and using the loss value in the pseudo label generation process to update the parameters of the segmentation network, and obtaining the initial pseudo labels of the multi-dimensional slice sample set by alternating multiple rounds of pseudo label generation operations and parameter update operations also includes: using a second segmentation network to perform the same processing operation on the inline dimension slice sample set as the time dimension slice sample set to obtain the initial pseudo labels of the inline dimension slice sample set; using a third segmentation network to perform the same processing operation on the cross-dimensional slice sample set as the time dimension slice sample set to obtain the initial pseudo labels of the cross-dimensional slice sample set.

[0088] The above embodiment involves the generation process of initial pseudo labels in multiple dimensions (i.e., vertical, inline, and crossline three orthogonal dimensions). Figure 3 The above process is further explained in detail.

[0089] Figure 3 2 is a schematic diagram of a pseudo-label generation process according to an embodiment of the present invention.

[0090] The pseudo-label generation process includes the initial pseudo-label generation process (1) and the initial pseudo-label optimization process (2).

[0091] (1.1) Initial generation of pseudo labels: Generate the pseudo labels for the second slice based on the true labels of the first slice (the true labels of the seismic data sample set itself, and take out the true labels of the dimension through orthogonal annotation to perform the initial generation of pseudo labels): earthquake data samples The pseudo label generation process is as follows Figure 3 As shown, take the vertical direction as an example. First, The first slice in the vertical direction Enter the U-net network , and forward propagation is performed. According to the network Output and true label slices Calculating supervised loss The process is shown in formulas (1) to (3):

[0092] (1),

[0093] (2)

[0094] (3),

[0095] in, is the cross entropy loss function, is the dice loss function, No. Pixels are recorded as , No. Pixels are recorded as According to the supervision loss and cross loss Calculate the total loss function as shown in formula (4):

[0096] (4),

[0097] Among them, the cross loss in the first round of pseudo label generation process The value of is always 0. For hyperparameters that do not participate in the update. Back propagation calculates the loss function The gradient of the U-net network is updated according to the gradient descent Repeat the forward propagation, back propagation, and parameter update operations several times, and record the final network output, i.e., the pseudo label. .

[0098] Repeat the above steps using the second slice As a network Input, according to the network Output and the previous pseudo-label slice Calculating supervised loss .network After several training sessions, record the final output pseudo-label The pseudo-label generation process in the inline and crossline directions is synchronized with the vertical direction and is completely consistent. After training, pseudo labels are obtained and ,network After training, pseudo labels are obtained and .

[0099] (1.2) Generation of remaining pseudo labels: Repeat the steps in (1.1) in the vertical, inline and crossline directions simultaneously. Update the neural network parameters under the guidance of the pseudo label of the previous slice to obtain a new pseudo label. Finally, all the pseudo label slices obtained in the vertical direction constitute , all pseudo-label slices obtained in the inline direction constitute , all pseudo-label slices obtained in the crossline direction constitute .

[0100] According to an embodiment of the present invention, the above-mentioned use of the optimization loss function to perform multiple rounds of cross-optimization between different dimensions on the initial pseudo-labels of the multi-dimensional slice sample set, and introducing the three-dimensional information of the seismic data volume in the cross-optimization process to obtain pseudo-labels that fuse multi-dimensional information includes: performing a mask generation operation on the pixels in the seismic data sample set that are lower than a preset pixel threshold to obtain a pixel mask, and performing two-by-two dimensional combinations on the multi-dimensional slice sample set to obtain a cross-dimensional slice sample set; processing the pixel mask and the initial pseudo-labels of the cross-dimensional slice sample set corresponding to the pixel mask by using the optimization loss function to obtain a cross-optimization loss value; using the cross-optimization loss value to update the parameters of the segmentation network obtained in the cross-optimization stage in the initial pseudo-label generation process; using the segmentation network obtained in the cross-optimization stage to regenerate the initial pseudo-labels in the multi-dimensional slice sample set; iteratively performing the calculation operation of the cross-optimization loss value, the segmentation network parameter update operation in the cross-optimization stage, and the regeneration operation of the initial pseudo-label until the cross-optimization loss value is lower than the preset threshold to obtain a pseudo-label that fuses multi-dimensional information.

[0101] The optimization process of the initial pseudo-labels is further described in detail below through a specific implementation method.

[0102] The optimization process of the initial pseudo-label (2): Figure 3 As shown, repeat all the steps in the initial pseudo-label generation process (1) to generate new pseudo-label slices and then , and In the process of calculating the loss, it is necessary to select pixels whose uncertainty is lower than the threshold and generate the corresponding mask Cross Loss It is no longer 0, and its calculation is shown in formulas (5) to (8):

[0103] (5),

[0104] (6),

[0105] (7),

[0106] (8),

[0107] in, is the mask of the two selected pseudo-label bodies A single pixel value, for A single pixel value, for A single pixel value, for The process of generating pseudo labels is repeated until The value of is reduced to below the set threshold. Compared with the pseudo-labels initially generated in the initial pseudo-label generation process (1), the pseudo-labels obtained at this time introduce the three-dimensional information of the seismic data volume. , and The pseudo labels generated in these three directions gradually converge, and finally one of them is selected as the final pseudo label. Repeating the initial pseudo-label generation process (1) can obtain fault pseudo-labels for any number of samples.

[0108] According to an embodiment of the present invention, the above-mentioned semi-supervised learning framework based on contextual prototype-aware learning technology uses a seismic data sample set with pseudo labels and a seismic data training set with unlabeled data to perform multiple rounds of training on the fault recognition model, and uses a training loss function to supervise the training process of the fault recognition model to obtain a trained fault recognition model, which includes: constructing a semi-supervised learning framework based on a teacher neural network-student neural network, and using the student neural network as a fault recognition model; introducing contextual prototype-aware learning technology into the semi-supervised learning framework, and constructing a training loss function based on a class activation map; using the teacher neural network to process the unlabeled seismic data training set to obtain a first output result, and using the student neural network to process the seismic data sample set with pseudo labels to obtain a second output result; calculating the training loss value by processing the first output result and the second output result through the training loss function, and updating the parameters of the student neural network based on the back-propagation mechanism; based on the student neural network after the parameters are updated, updating the parameters of the teacher neural network through an exponential average moving operation; iteratively performing the training loss value calculation operation, the output result acquisition operation, and the neural network parameter update operation until the preset training conditions are met to obtain a trained fault recognition model.

[0109] Figure 4 2 is a schematic diagram of a fault recognition model trained within a semi-supervised learning framework according to an embodiment of the present invention.

[0110] The forward propagation of the fault recognition model uses the pseudo-label with fused multi-dimensional information as the label of the seismic data sample set, and records the above data as , and its corresponding pseudo label is recorded as The data without labels is denoted as .

[0111] The semi-supervised learning framework in this invention is based on the common teacher-student framework. The overall framework structure is as follows Figure 4 As shown in Figure 2, both the teacher network and the student network adopt the U-Net network structure constructed with three-dimensional convolutional layers. The teacher network is denoted as , student network notes . Pseudo-label the data Input the student network to obtain output . Input the teacher model and student model to obtain and . Loss function and The calculation of is shown in formulas (9) and (10):

[0112] (9),

[0113] (10),

[0114] in, is the cross entropy loss function, is the class activation graph.

[0115] Get class activation map The process will be expanded in subsequent embodiments.

[0116] According to an embodiment of the present invention, the above-mentioned context prototype-aware learning technology is introduced into the semi-supervised learning framework, and a training loss function based on the class activation map is constructed, including: thresholding the output results of the teacher neural network to generate a binary mask, and performing mask average pooling on the unlabeled seismic data training set according to the binary mask to obtain an instance prototype; storing the instance prototype in a dynamically updateable support library, and clustering the instance prototype using a preset clustering algorithm; using the clustered instance prototypes to construct a candidate context prototype set, and calculating the cosine similarity between each candidate context prototype in the candidate context prototype set and the instance prototype; selecting multiple candidate context prototypes that are most similar to the instance prototype from the candidate context prototype set based on the cosine similarity, and using the selected multiple candidate context prototypes to construct a context-aware prototype set.

[0117] According to an embodiment of the present invention, the above-mentioned introduction of context prototype-aware learning technology into the semi-supervised learning framework and construction of a training loss function based on a class activation map also include: calculating the positive correlation weight of each context-aware prototype and the instance prototype in the context-aware prototype set based on a parameter-free identity mapping layer and an activation function; using the positive correlation weight to perform weighted calculation on each context-aware prototype in the context-aware prototype set to obtain a weighted context-aware prototype set; calculating the distribution gap between the instance prototype and each context-aware prototype in the context-aware prototype set to obtain an offset term representing the distribution gap; replacing the instance prototype with the result after the instance prototype and the offset term operation to obtain a corrected instance prototype, and using the corrected instance prototype to correct the weighted context-aware prototype set to obtain a corrected context-aware prototype set; using the corrected context-aware prototype set and an unlabeled seismic data training set to obtain a class activation map, and using the class activation map to construct a training loss function.

[0118] The following is a specific implementation method and combined with the attached Figure 4 The process of training the fault recognition model using the semi-supervised learning framework provided by the present invention is further described in detail.

[0119] like Figure 4 As shown in Figure 2, contextual prototype-aware learning includes mask average pooling, k-means clustering, output class activation map, and calculation of loss function.

[0120] Mask average pooling: For the output of the teacher network Thresholding to generate a binary mask .according to For unlabeled data Perform mask average pooling to obtain , as shown in formula (11):

[0121] (11),

[0122] in, For three-dimensional seismic data, the indexes corresponding to the three dimensions are respectively , and express. The output of the teacher model has four dimensions. The first dimension is the identification type, i.e. fault and non-fault, and its index is Indicates that n is 0 or 1, and the other three dimensions are consistent. The dimensions and same.

[0123] k-means clustering: cluster each data instance prototype Stored in the support library, the support library will be dynamically updated with training. Apply the k-means clustering algorithm to divide similar prototypes into multiple cluster groups. Each cluster group represents a context feature pattern of the category. , extract all instance prototypes from the support library and build a candidate context prototype set .

[0124] Output class activation map: Calculate each context prototype With the current instance prototype The similarity of is measured using cosine similarity, as shown in formula (12):

[0125] (12).

[0126] Based on the calculated similarity, select the most similar instance from the candidate set context prototypes to form a context-aware prototype set, as shown in formula (13):

[0127] (13),

[0128] For each context prototype , calculated based on the parameter-free identity mapping layer and softmax function and the instance prototype Positive correlation weight , as shown in formula (14):

[0129] (14),

[0130] in and It is the parameter-free identity mapping layer in feature transformation, Is used to adjust the weight The scale factor of . The filtered context prototypes are weighted as shown in formula (15):

[0131] (15).

[0132] Calculate the distribution gap between the dense distribution center of the current instance feature and the context prototype. Introduce an offset term for the current instance prototype , as shown in formula (16):

[0133] (16),

[0134] Align instance features to the distribution center of context features, that is, use Replace the original .

[0135] Using context-aware prototype collections , for each pixel position of the image Calculate the enhanced class activation value as shown in formula (17):

[0136] (17),

[0137] in, For selected categories The number of most relevant context prototypes. The class activation map containing all fault and non-fault categories is denoted as .

[0138] Calculating the loss function: combining Figure 4 The teacher model output shown is and class activation maps , calculate the loss function, as shown in formula (18):

[0139] (18).

[0140] Loss function weight sum: The loss function of the semi-supervised learning framework based on the teacher-student framework that introduces contextual prototype-aware learning technology is , and The weighted sum of is calculated as described in formula (19):

[0141] (19).

[0142] Backpropagation: Gradients are backpropagated based on the loss function calculated in the loss function weight sum to update the network parameters in the framework. Gradient propagation is performed only within the student network and the context-prototype-aware learning model, while the teacher network weights are updated using the exponential moving average of the student network weights. Throughout the training process, the context prototype set and instance features are dynamically updated via the support library, continuously optimizing the output of the teacher model.

[0143] The second aspect of the present invention provides a fault identification method, including: using a trained fault identification model to process the fault data to be identified to obtain a fault identification result, wherein the trained fault identification model is trained by the above-mentioned training method of the fault identification model based on orthogonal annotation, and the fault data to be identified includes an earthquake fault image.

[0144] By utilizing the training method of the fault recognition model provided by the present invention, a fault recognition model is obtained based on the weights of the student neural network, and is used for recognizing fault data.

[0145] Those skilled in the art can identify other tomographic data, such as computed tomography images (CT images), etc., based on the teachings of the present invention.

[0146] Based on the above-mentioned training method of the fault recognition model based on orthogonal annotation, the present invention also provides a training device of the fault recognition model based on orthogonal annotation. Figure 5 The device is described in detail.

[0147] Figure 5 4 is a structural block diagram of a training device for a fault recognition model based on orthogonal annotation according to an embodiment of the present invention.

[0148] like Figure 5 As shown, the training device 500 of the above-mentioned fault recognition model based on orthogonal annotation includes an orthogonal slicing module 510, a pseudo-label generation module 520, a pseudo-label optimization module 530 and a semi-supervised training module 540.

[0149] The orthogonal slicing module 510 is used to slice the seismic data sample set in multiple orthogonal dimensions to obtain a multi-dimensional slice sample set, wherein the multi-dimensional slice sample set includes a time dimension slice sample set, an inline dimension slice sample set, and a cross-dimensional slice sample set; in one embodiment, the orthogonal slicing module 510 can be used to perform the operation S210 described above, which will not be repeated here.

[0150] The pseudo-label generation module 520 is used to generate pseudo-labels for the multi-dimensional slice sample set using the segmentation network, and to update the parameters of the segmentation network using the loss value in the pseudo-label generation process, thereby obtaining the initial pseudo-labels for the multi-dimensional slice sample set by alternating multiple rounds of pseudo-label generation operations and parameter update operations. In one embodiment, the pseudo-label generation module 520 can be used to perform the operation S220 described above, which will not be repeated here.

[0151] The pseudo-label optimization module 530 is used to perform multiple rounds of cross-optimization among different dimensions on the initial pseudo-labels of the multi-dimensional slice sample set using an optimized loss function, and to obtain pseudo-labels that integrate multi-dimensional information by introducing three-dimensional information of the seismic data volume during the cross-optimization process. In one embodiment, the pseudo-label optimization module 530 can be used to perform the operation S230 described above, which will not be repeated here.

[0152] The semi-supervised training module 540 is used in a semi-supervised learning framework based on contextual prototype-aware learning technology, which uses a pseudo-labeled seismic data sample set and an unlabeled seismic data training set to perform multiple rounds of training on the fault recognition model, and uses a training loss function to supervise the training process of the fault recognition model to obtain a trained fault recognition model; in one embodiment, the semi-supervised training module 540 can be used to perform the operation S240 described above, which will not be repeated here.

[0153] According to embodiments of the present invention, any multiple modules among the orthogonal slicing module 510, pseudo-label generation module 520, pseudo-label optimization module 530, and semi-supervised training module 540 may be combined into a single module, or any one of these modules may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in a single module. According to embodiments of the present invention, at least one of the orthogonal slicing module 510, pseudo-label generation module 520, pseudo-label optimization module 530, and semi-supervised training module 540 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or may be implemented in hardware or firmware through any other reasonable means of circuit integration or packaging, or implemented in any one of software, hardware, and firmware, or any suitable combination of any of these. Alternatively, at least one of the orthogonal slicing module 510, the pseudo-label generation module 520, the pseudo-label optimization module 530, and the semi-supervised training module 540 may be at least partially implemented as a computer program module, which may perform corresponding functions when executed.

[0154] Figure 6 4 is a block diagram of an electronic device suitable for implementing a training method for a fault recognition model based on orthogonal annotation and a fault recognition method according to an embodiment of the present invention.

[0155] like Figure 6 As shown, an electronic device 600 according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 602 or programs loaded from a storage unit 608 into a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or related chipsets and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0156] Various programs and data required for the operation of the electronic device 600 are stored in the RAM 603. The processor 601, ROM 602, and RAM 603 are connected to each other via a bus 604. The processor 601 executes the programs in the ROM 602 and / or RAM 603 to perform various operations according to the method flow of the embodiment of the present invention. It should be noted that the programs may also be stored in one or more memories other than the ROM 602 and RAM 603. The processor 601 may also execute the programs stored in the one or more memories to perform various operations according to the method flow of the embodiment of the present invention.

[0157] According to an embodiment of the present invention, electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to bus 604. Electronic device 600 may also include one or more of the following components connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or modem. Communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. Removable media 611, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 610 as needed, so that computer programs read from the removable media can be installed into storage section 608 as needed.

[0158] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0159] According to an embodiment of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present invention, a computer-readable storage medium may include the ROM 602 and / or RAM 603 described above, and / or one or more memories other than ROM 602 and RAM 603.

[0160] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0161] It will be understood by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or coupled in various ways, even if such combinations or couplings are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention may be combined and / or coupled in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or couplings fall within the scope of the present invention.

[0162] The above describes embodiments of the present invention. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A training method for a fault recognition model based on orthogonal annotation, characterized in that: The method comprises: Slicing the seismic data sample set in multiple orthogonal dimensions to obtain a multi-dimensional slice sample set, wherein the multi-dimensional slice sample set includes a time dimension slice sample set, an inline dimension slice sample set, and a cross dimension slice sample set; Generate pseudo labels for the multidimensional slice sample set using a segmentation network, and update parameters of the segmentation network using the loss value in the pseudo label generation process, and obtain initial pseudo labels for the multidimensional slice sample set by alternating multiple rounds of pseudo label generation and parameter update operations; Using an optimized loss function, the initial pseudo labels of the multi-dimensional slice sample set are cross-optimized in multiple rounds across different dimensions, and by introducing three-dimensional information of the seismic data volume into the cross-optimization process, pseudo labels that integrate the multi-dimensional information are obtained; A semi-supervised learning framework based on contextual prototype perception learning technology is used to perform multiple rounds of training on the fault recognition model using a seismic data sample set with the pseudo labels and an unlabeled seismic data training set, and a training loss function is used to supervise the training process of the fault recognition model to obtain a trained fault recognition model.

2. The method according to claim 1, characterized in that Generating pseudo labels for the multidimensional slice sample set using a segmentation network, and updating parameters of the segmentation network using the loss value in the pseudo label generation process, and obtaining initial pseudo labels for the multidimensional slice sample set by alternating multiple rounds of pseudo label generation and parameter update operations, including: Using a first segmentation network, successively performing downsampling operations and upsampling operations on a first time slice in the time dimension slice sample set to generate a first-round iterative pseudo label for the first time slice; Processing the first-round iterative pseudo-label of the first time slice and the true label corresponding to the first time slice through the pseudo-label generation loss function, and calculating the first-round iterative pseudo-label loss value; By back-propagation, the first segmentation network is updated with the pseudo-label loss value of the first iteration to obtain a first segmentation network after the first iteration. Iteratively performing a pseudo-label generation operation, a pseudo-label loss value calculation operation, and a parameter update operation until the number of iterations meets a first preset iteration value, thereby obtaining an initial pseudo-label for the first time slice; The first segmentation network after the previous iteration and the initial pseudo label of the previous time slice are used to perform the same processing operation as the first time slice on each time slice in the time dimension slice sample set to obtain the initial pseudo label of the time dimension slice sample set.

3. The method according to claim 2, characterized in that Also includes: Using a second segmentation network, the inline dimension slice sample set is subjected to the same processing operation as that of the time dimension slice sample set to obtain an initial pseudo label of the inline dimension slice sample set; The third segmentation network is used to perform the same processing operation on the cross-dimensional slice sample set as that on the time dimension slice sample set to obtain an initial pseudo label of the cross-dimensional slice sample set.

4. The method according to claim 1, wherein The initial pseudo labels of the multi-dimensional slice sample set are cross-optimized in multiple rounds in different dimensions using an optimized loss function. By introducing the three-dimensional information of the seismic data volume into the cross-optimization process, the pseudo labels that integrate the multi-dimensional information are obtained. Performing a mask generation operation on pixels in the seismic data sample set that are lower than a preset pixel threshold to obtain a pixel mask, and performing two-by-two dimensional combinations on the multi-dimensional slice sample set to obtain a cross-dimensional slice sample set; Processing the pixel mask and the initial pseudo labels of the cross-dimensional slice sample set corresponding to the pixel mask through the optimization loss function to obtain a cross-optimization loss value; Using the cross-optimization loss value to update the parameters of the segmentation network obtained in the initial pseudo-label generation process in the cross-optimization phase; Regenerating the initial pseudo labels for the multi-dimensional slice samples using the segmentation network obtained in the cross optimization stage; The calculation operation of the cross-optimization loss value, the segmentation network parameter update operation in the cross-optimization stage, and the regeneration operation of the initial pseudo-label are iteratively performed until the cross-optimization loss value is lower than a preset threshold, thereby obtaining the pseudo-label that integrates the multi-dimensional information.

5. The method according to claim 1, wherein A semi-supervised learning framework based on contextual prototype-aware learning technology is used to train the fault recognition model for multiple rounds using a seismic data sample set with pseudo labels and a seismic data training set with unlabeled data. The training process of the fault recognition model is supervised using a training loss function, and the trained fault recognition model includes: Constructing the semi-supervised learning framework based on the teacher neural network-student neural network, and using the student neural network as the fault recognition model; Introducing the contextual prototype-aware learning technology into the semi-supervised learning framework and constructing a training loss function based on class activation maps; Using the teacher neural network to process the unlabeled seismic data training set to obtain a first output result, and using the student neural network to process the seismic data sample set with the pseudo labels to obtain a second output result; Processing the first output result and the second output result by the training loss function to calculate a training loss value, and using the training loss value to update parameters of the student neural network based on a backpropagation mechanism; Based on the student neural network after parameter update, the teacher neural network parameters are updated by exponential moving average operation; The training loss value calculation operation, output result acquisition operation, and neural network parameter update operation are iteratively performed until the preset training conditions are met to obtain a trained fault recognition model.

6. The method according to claim 5, characterized in that Introducing the contextual prototype-aware learning technology into the semi-supervised learning framework and constructing a training loss function based on a class activation map includes: Thresholding the output result of the teacher neural network to generate a binary mask, and performing mask average pooling on the unlabeled seismic data training set according to the binary mask to obtain an instance prototype; Storing the instance prototypes in a dynamically updateable support library, and clustering the instance prototypes using a preset clustering algorithm; Constructing a candidate context prototype set using the instance prototypes after clustering, and calculating the cosine similarity between each candidate context prototype in the candidate context prototype set and the instance prototype; A plurality of candidate context prototypes that are most similar to the instance prototype are selected from the candidate context prototype set based on the cosine similarity, and a context-aware prototype set is constructed using the selected plurality of candidate context prototypes.

7. The method according to claim 6, characterized in that Also includes: Calculating a positive correlation weight between each context-aware prototype in the context-aware prototype set and the instance prototype based on a parameter-free identity mapping layer and an activation function; Performing weighted calculation on each context-aware prototype in the context-aware prototype set using the positive correlation weight to obtain a weighted context-aware prototype set; Calculating a distribution gap between the instance prototype and each context-aware prototype in the context-aware prototype set to obtain an offset term representing the distribution gap; Replacing the instance prototype with a result of an operation between the instance prototype and the offset term to obtain a revised instance prototype, and revising the weighted context-aware prototype set with the revised instance prototype to obtain a revised context-aware prototype set; The class activation map is obtained by using the modified context-aware prototype set and the unlabeled seismic data training set, and the training loss function is constructed by using the class activation map.

8. A fault identification method, characterized in that: The method comprises: The fault data to be identified is processed using the trained fault identification model to obtain a fault identification result, wherein the trained fault identification model is trained using the training method described in any one of claims 1 to 7, and the fault data to be identified includes an earthquake fault image.

9. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images