A method, device, equipment and medium for reconstructing magnetotelluric impedance tensor

Abstract

The present invention provides a method, device, equipment and medium for reconstructing magnetotelluric impedance tensor, which relates to the technical field of magnetotelluric data processing, so as to alleviate the technical problems existing in the prior art that when the signal-to-noise ratio is low, signal features are difficult to extract and the denoising effect is limited. Based on the initial noisy impedance tensor data, the impedance tensor reconstruction model is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to perform preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; de-entangle the initial noisy impedance tensor data based on the phase winding points to obtain de-entangled noisy impedance tensor data; perform final reconstruction on the de-entangled noisy impedance tensor data to obtain target impedance tensor data, so as to improve the reconstruction quality of the impedance tensor and reduce noise interference.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetotelluric data processing, and in particular to a method, device, equipment and medium for reconstructing magnetotelluric impedance tensor. Background Art

[0002] Magnetotelluric sounding is an important physical means of earth exploration that uses natural electromagnetic fields as field sources to study the electrical structure of the earth's interior. Due to human production activities, electromagnetic signals measured on the surface are often interfered by various noises, which affects the accurate estimation of the impedance tensor and further affects the reliability of geological interpretation results.

[0003] Existing noise suppression technologies are mainly divided into two types: time domain and frequency domain, including empirical mode decomposition, variational mode decomposition, deep learning and other methods, as well as robust estimation and power spectrum data screening methods. However, most of these methods are based on the characteristic differences between signals and noise in time series for denoising. When the signal-to-noise ratio is low, signal features are difficult to extract and the denoising effect is limited. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method, device, equipment and medium for reconstructing a magnetotelluric impedance tensor, so as to improve the reconstruction quality of the impedance tensor and reduce noise interference and phase error.

[0005] In a first aspect, the present invention provides a method for reconstructing a magnetotelluric impedance tensor, comprising:

[0006] Obtaining initial noisy impedance tensor data;

[0007] Based on the initial noisy impedance tensor data, an impedance tensor reconstruction model is used to denoise and reconstruct the initial noisy impedance tensor data to obtain target impedance tensor data; wherein, the impedance tensor reconstruction model is to preliminarily reconstruct the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; to identify winding information on the preliminary reconstructed impedance tensor data to obtain phase winding points; to unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and to finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data.

[0008] Optionally, the impedance tensor reconstruction model includes: a reconstruction sub-model, a phase winding information identification sub-model and a final reconstruction sub-model, and performs denoising and reconstruction processing on the initial noisy impedance tensor data, including:

[0009] Based on the reconstruction sub-model, the phase winding information identification sub-model and the final reconstruction sub-model, a denoising and reconstruction operation is iteratively performed on the initial noisy impedance tensor data until it is determined that the iteration termination condition is met, and the impedance tensor data output by the last denoising and reconstruction operation is determined as the target impedance tensor data;

[0010] Among them, the denoising and reconstruction operation includes: inputting the initial noisy impedance tensor data into the reconstruction sub-model to obtain preliminary reconstructed impedance tensor data; inputting the preliminary reconstructed impedance tensor data into the phase winding information identification sub-model to obtain the phase winding point; obtaining the unwrapped noisy impedance tensor data based on the phase winding point and the initial noisy impedance tensor data; inputting the unwrapped noisy impedance tensor data into the final reconstruction sub-model to obtain the reconstructed impedance tensor data; obtaining the marked strong noise based on the initial noisy impedance tensor data and the reconstructed impedance tensor data; replacing the marked strong noise in the initial noisy impedance tensor data with the reconstructed impedance tensor data to obtain new initial noisy impedance tensor data.

[0011] Optionally, the method further comprises:

[0012] Acquire a training data set, wherein the training data set includes a plurality of training sample data; each training sample data includes initial noisy impedance tensor data, target winding impedance tensor data, target phase winding point and target unwinding impedance tensor data;

[0013] Based on the training data set, an iterative training operation is performed on the initial impedance tensor reconstruction model until it is determined that the iterative training termination condition is met, and the impedance tensor reconstruction model is obtained based on the weights and thresholds of the initial impedance tensor reconstruction model updated when the iterative training operation is performed for the last time; wherein the iterative training operation includes:

[0014] Select target training sample data from the training data set;

[0015] Input the initial noisy impedance tensor data in the target training sample data into the initial reconstruction sub-model to obtain the training value of the winding impedance tensor data; input the training value of the winding impedance tensor data into the initial phase winding information identification sub-model to obtain the training value of the phase winding point; based on the phase winding point and the initial noisy impedance tensor data, obtain the unwinding noisy impedance tensor data, and input the unwinding noisy impedance tensor data into the final reconstruction sub-model to obtain the training value of the unwinding impedance tensor data;

[0016] Based on the error between the training value of the winding impedance tensor data and the target winding impedance tensor data in the target training sample data, the weights and thresholds of the initial reconstruction sub-model are updated; based on the error between the training value of the phase winding point and the target phase winding point in the target training sample data, the weights and thresholds of the initial phase winding information identification sub-model are updated; based on the error between the training value of the unwinding impedance tensor data and the target unwinding impedance tensor data in the target training sample data, the weights and thresholds of the initial final reconstruction sub-model are updated.

[0017] Optionally, both the reconstruction sub-model and the final reconstruction sub-model are machine learning models:

[0018] Both the reconstruction sub-model and the final reconstruction sub-model are convolutional neural network models that receive the initial noisy impedance tensor data in the target training sample data through the input layer, obtain the reconstructed impedance tensor data based on the initial noisy impedance tensor data through the hidden layer, and output the reconstructed impedance tensor data through the output layer; wherein the hidden layer includes multiple convolution units, residual units and channel attention units, the convolution unit is used to extract the features of the initial noisy impedance tensor data, the residual unit is used for the connection between two convolution units, and the channel attention unit is used based on the optimized feature channel.

[0019] Optionally, the phase winding information identification sub-model is a machine learning model:

[0020] The phase winding information identification sub-model is a convolutional neural network model that receives the initial noisy impedance tensor data in the target training sample data through the input layer, obtains the phase winding point based on the initial noisy impedance tensor data through the hidden layer, and outputs the phase winding point through the output layer; wherein the hidden layer includes multiple convolution units and residual units, the convolution unit is used to extract the features of the initial noisy impedance tensor data, and the residual unit is used for the connection between two convolution units.

[0021] Optionally, obtain a training data set, including:

[0022] Obtain standard impedance data;

[0023] Add noise that meets actual observation conditions to the standard impedance data to obtain initial noisy impedance tensor data;

[0024] Convert the standard impedance data into apparent resistivity sequence and phase sequence to obtain target winding impedance tensor data;

[0025] Based on the phase sequence of standard impedance data, a target phase winding point is obtained;

[0026] Based on the target phase winding point, the phase sequence in the standard impedance data is unwound to obtain the target unwound impedance tensor data.

[0027] A training data set is obtained based on the initial noisy impedance tensor data, the target winding impedance tensor data, the target phase winding point and the target unwinding impedance tensor data.

[0028] In a second aspect, the present invention provides a device for reconstructing a magnetotelluric impedance tensor, comprising:

[0029] An acquisition unit, used for acquiring initial noisy impedance tensor data;

[0030] A reconstruction unit is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data using an impedance tensor reconstruction model based on the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to perform preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data.

[0031] In a third aspect, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method of reconstructing the earth battery impedance tensor when executing the computer program.

[0032] In a fourth aspect, the present invention further provides a computer-readable storage medium storing computer instructions, which implement the above-mentioned method for reconstructing the magnetotelluric impedance tensor when executed by a processor.

[0033] The embodiments of the present invention provide a method, device, equipment and medium for reconstructing a magnetotelluric impedance tensor, which obtains initial noisy impedance tensor data; based on the initial noisy impedance tensor data, uses an impedance tensor reconstruction model to denoise and reconstruct the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to preliminarily reconstruct the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data, so as to improve the reconstruction quality of the impedance tensor and reduce noise interference.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 A flow chart of a method for reconstructing a magnetotelluric impedance tensor provided by an embodiment of the present invention is shown;

[0037] Figure 2 A flow chart showing the impedance tensor reconstruction model provided in an embodiment of the present invention performs denoising and reconstruction processing on initial noisy impedance tensor data to obtain target impedance tensor data;

[0038] Figure 3 A schematic diagram of a process flow of an impedance tensor reconstruction model training process provided by an embodiment of the present invention is shown;

[0039] Figure 4 A schematic structural diagram of a device for reconstructing a magnetotelluric impedance tensor provided by an embodiment of the present invention is shown;

[0040] Figure 5 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0042] In order to facilitate those skilled in the art to better understand the present application, the technical terms involved in the present application are briefly introduced below.

[0043] The impedance tensor is a key parameter that describes the relationship between the electric field and the magnetic field, while the apparent resistivity and phase are important characterization parameters of the impedance tensor, which are used to simplify and explain the physical meaning of the impedance tensor.

[0044] The impedance tensor reconstruction model is a convolutional neural network model that realizes denoising and reconstruction of the impedance tensor by learning the mapping relationship between the initial noisy impedance tensor and the target impedance tensor; in the present application, the impedance tensor reconstruction model includes a reconstruction sub-model, a phase winding information identification sub-model and a final reconstruction sub-model, the reconstruction sub-model is used to perform preliminary reconstruction of the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; the phase winding information identification sub-model is used to perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; the initial noisy impedance tensor data is unwound based on the phase winding points to obtain unwound noisy impedance tensor data; the final reconstruction sub-model is used to perform final reconstruction of the unwound noisy impedance tensor data to obtain target impedance tensor data;

[0045] Among them, the reconstruction sub-model and the final reconstruction sub-model are the same deep residual convolutional neural network model with convolutional units, residual units and channel attention units, but the model weight parameters of the two are different; the phase winding information recognition sub-model is a residual convolutional neural network model with convolutional units and residual units;

[0046] Among them, the convolution unit can gradually extract the features of the initial noisy impedance tensor sequence from low-level to high-level through multi-layer convolution operations, and enhance the recognition ability of complex impedance patterns; the residual unit allows features to flow directly in the network through residual connections, maintains the transmission of important features, and better captures and reconstructs complex impedance tensor features; the channel attention unit can automatically identify and enhance the most critical feature channels for reconstructing the impedance tensor through the attention mechanism, suppress irrelevant or noisy feature channels, thereby improving the denoising effect and reconstruction quality.

[0047] After introducing the technical terms involved in this application, the technical solutions provided by this application are described in detail.

[0048] Figure 1 A schematic diagram of a method for reconstructing the magnetotelluric impedance tensor provided in an embodiment of the present application. Figure 1 As shown, the method comprises at least the following steps:

[0049] Step S110: Obtain initial noisy impedance tensor data.

[0050] In the embodiment of the present application, when the magnetotelluric impedance tensor is reconstructed, initial noisy impedance tensor data can be obtained.

[0051] Since each component of the impedance tensor is a complex number and each component contains amplitude and phase information, the impedance tensor varies greatly in the reconstruction calculation and the calculation is unstable. Therefore, it is necessary to process the initial noisy impedance tensor data to obtain the apparent resistivity and phase sequence corresponding to the initial noisy impedance tensor data, and reconstruct the impedance tensor data based on the apparent resistivity and phase sequence.

[0052] Step S120: Based on the initial noisy impedance tensor data, an impedance tensor reconstruction model is used to denoise and reconstruct the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to preliminarily reconstruct the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; to identify winding information on the preliminary reconstructed impedance tensor data to obtain phase winding points; to unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and to finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data.

[0053] In an embodiment of the present application, the initial noisy impedance tensor data is input into an impedance tensor reconstruction model, and the impedance tensor reconstruction model performs preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; winding information is identified on the preliminary reconstructed impedance tensor data to obtain phase winding points; the initial noisy impedance tensor data is unwound based on the phase winding points to obtain unwound noisy impedance tensor data; the unwound noisy impedance tensor data is finally reconstructed to obtain target impedance tensor data, thereby improving the reconstruction quality of the impedance tensor and reducing noise interference, thereby providing more accurate and reliable data support for geophysical exploration.

[0054] The following is a detailed description of the process of denoising and reconstructing the initial noisy impedance tensor data to obtain the target impedance tensor data using the impedance tensor reconstruction model provided in the embodiment of the present application. Figure 2 The impedance tensor reconstruction model of the embodiment of the present application performs denoising and reconstruction processing on the initial noisy impedance tensor data to obtain the target impedance tensor data in the following process:

[0055] Step S210: inputting the initial noisy impedance tensor data into the reconstruction sub-model to obtain preliminary reconstructed impedance tensor data;

[0056] Step S220: inputting the preliminary reconstructed impedance tensor data into the phase winding information identification sub-model to obtain the phase winding point;

[0057] Step S230: unwrapping the initial noisy impedance tensor data based on the phase winding point to obtain unwrapped noisy impedance tensor data;

[0058] Step S240: inputting the unwrapped noisy impedance tensor data into the final reconstruction sub-model to obtain reconstructed impedance tensor data;

[0059] Step S250: obtaining a marked strong noise based on the initial noisy impedance tensor data and the reconstructed impedance tensor data;

[0060] Step S260: replacing the marked strong noise in the initial noisy impedance tensor data with the reconstructed impedance tensor data to obtain new initial noisy impedance tensor data;

[0061] Step S270: determine whether the iteration termination condition is met; if so, determine the impedance tensor data output by the last denoising and reconstruction operation as the target impedance tensor data; if not, return to step S210; wherein the iteration termination condition is that the number of iterations is not less than a preset iteration number threshold.

[0062] The training process of the impedance tensor reconstruction model provided in the embodiment of the present application is described in detail below. Figure 3 As shown, the impedance tensor reconstruction model training process provided in the embodiment of the present application is as follows:

[0063] Step S310: Acquire a training data set, wherein the training data set includes a plurality of training sample data; each training sample data includes initial noisy impedance tensor data, target winding impedance tensor data, target phase winding point and target unwinding impedance tensor data.

[0064] In the embodiment of the present application, when obtaining the training data set, the following methods may be used but are not limited to:

[0065] First, obtain standard impedance data;

[0066] Then, noise that meets the actual observation conditions is added to the standard impedance data to obtain the initial noisy impedance tensor data; the standard impedance data is converted into an apparent resistivity sequence and a phase sequence to obtain the target winding impedance tensor data; based on the phase sequence of the standard impedance data, the target phase winding point is obtained; based on the target phase winding point, the phase sequence in the standard impedance data is unwound to obtain the target unwound impedance tensor data;

[0067] Finally, a training data set is obtained based on the initial noisy impedance tensor data, the target winding impedance tensor data, the target phase winding point and the target unwinding impedance tensor data.

[0068] Step S320: Select target training sample data from the training data set.

[0069] Step S330: inputting the initial noisy impedance tensor data in the target training sample data into the initial reconstruction sub-model to obtain the training value of the winding impedance tensor data;

[0070] Inputting the training value of the winding impedance tensor data into the initial phase winding information identification sub-model to obtain the training value of the phase winding point;

[0071] Based on the phase winding point and the initial noisy impedance tensor data, the unwinding noisy impedance tensor data is obtained, and the unwinding noisy impedance tensor data is input into the final reconstruction sub-model to obtain the training value of the unwinding impedance tensor data.

[0072] Step S340: updating the weights and thresholds of the initial reconstruction sub-model based on the error between the training value of the winding impedance tensor data and the target winding impedance tensor data in the target training sample data;

[0073] Based on the error between the training value of the phase winding point and the target phase winding point in the target training sample data, the weights and thresholds of the initial phase winding information identification sub-model are updated;

[0074] Based on the error between the training value of the entanglement-free impedance tensor data and the target entanglement-free impedance tensor data in the target training sample data, the weights and thresholds of the initial and final reconstruction sub-models are updated.

[0075] In an embodiment of the present application, the reconstruction sub-model and the final reconstruction sub-model are both machine learning models; wherein, the reconstruction sub-model and the final reconstruction sub-model are both convolutional neural network models that receive initial noisy impedance tensor data in target training sample data through an input layer, obtain reconstructed impedance tensor data based on the initial noisy impedance tensor data through a hidden layer, and then output the reconstructed impedance tensor data through an output layer; wherein the hidden layer includes multiple convolution units, residual units and channel attention units, the convolution unit is used to extract features of the initial noisy impedance tensor data, the residual unit is used for connection between two convolution units, and the channel attention unit is used to optimize feature channels.

[0076] Step S350: determine whether the iterative training termination condition is met; if so, execute step S360; if not, return to step S320; wherein, the iterative training termination condition is that the number of iterations is not less than a preset iteration number threshold, or the prediction error is not higher than the error threshold.

[0077] Step S360: Based on the weights and thresholds of the initial reconstruction sub-model, the initial final reconstruction sub-model and the initial phase winding information identification sub-model updated during the last iterative training operation, the reconstruction sub-model, the final reconstruction sub-model and the phase winding information identification sub-model are obtained.

[0078] In an embodiment of the present application, the phase winding information identification sub-model is a machine learning model; the phase winding information identification sub-model is a convolutional neural network model that receives the initial noisy impedance tensor data in the target training sample data through the input layer, obtains the phase winding point based on the initial noisy impedance tensor data through the hidden layer, and outputs the phase winding point through the output layer; wherein the hidden layer includes multiple convolution units and residual units, the convolution unit is used to extract the features of the initial noisy impedance tensor data, and the residual unit is used for the connection between two convolution units.

[0079] Furthermore, in the implementation of this application, the reconstruction sub-model, the final reconstruction sub-model and the phase winding information identification sub-model all use the mean square error as the main loss function to measure the difference between the prediction results of each sub-model and the true value, and use the Adam optimizer to optimize each sub-model, wherein the initial learning rate of the Adam optimizer is set to 0.0001, and the batch size is set to: the learning rate is halved every 20 batches, and the training is performed for a total of 1000 rounds to ensure stable convergence, and finally three trained sub-models are obtained.

[0080] Furthermore, during the training process, the reconstruction sub-model and the final reconstruction sub-model are trained independently. After the reconstruction sub-model is trained, the training data is re-input into the reconstruction sub-model for reconstruction, and the output data obtained is used as the input of the phase entanglement information identification sub-model for training. This can ensure the training effect of the phase entanglement information identification sub-model and make the phase entanglement information identification sub-model more suitable for the phase unwrapping task in impedance tensor reconstruction.

[0081] In a method for reconstructing the magnetotelluric impedance tensor provided in the present application, an efficient reconstruction of the noisy impedance tensor is achieved through an impedance tensor reconstruction model. Specifically, the impedance tensor reconstruction model includes a reconstruction sub-model, a phase winding information identification sub-model and a final reconstruction sub-model, wherein the reconstruction sub-model can effectively remove part of the noise and extract key features; the phase winding information identification sub-model can handle the phase unwrapping problem to ensure the continuity and accuracy of the phase data; the final reconstruction sub-model performs a final reconstruction of the unwrapped noisy impedance tensor, further improving the accuracy and detail retention of the impedance tensor. Compared with traditional data processing methods, the method for reconstructing the magnetotelluric impedance tensor of the present application improves the reconstruction quality of the impedance tensor, reduces noise interference, and thus provides more accurate and reliable data support for geophysical exploration.

[0082] Based on the above embodiments, the present application provides a device for reconstructing the magnetotelluric impedance tensor. Figure 4 As shown, the device for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application at least includes:

[0083] An acquisition unit 410, configured to acquire initial noisy impedance tensor data;

[0084] The reconstruction unit 420 is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data using an impedance tensor reconstruction model based on the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to perform preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data.

[0085] In an optional embodiment, the reconstruction unit 420 is further configured to: the impedance tensor reconstruction model includes: a reconstruction sub-model, a phase winding information identification sub-model and a final reconstruction sub-model, and perform denoising and reconstruction processing on the initial noisy impedance tensor data, including:

[0086] Based on the reconstruction sub-model, the phase winding information identification sub-model and the final reconstruction sub-model, a denoising and reconstruction operation is iteratively performed on the initial noisy impedance tensor data until it is determined that the iteration termination condition is met, and the impedance tensor data output by the last denoising and reconstruction operation is determined as the target impedance tensor data;

[0087] Among them, the denoising and reconstruction operation includes: inputting the initial noisy impedance tensor data into the reconstruction sub-model to obtain preliminary reconstructed impedance tensor data; inputting the preliminary reconstructed impedance tensor data into the phase winding information identification sub-model to obtain the phase winding point; obtaining the unwrapped noisy impedance tensor data based on the phase winding point and the initial noisy impedance tensor data; inputting the unwrapped noisy impedance tensor data into the final reconstruction sub-model to obtain the reconstructed impedance tensor data; based on the initial noisy impedance tensor data, obtaining the marked strong noise; replacing the marked strong noise in the initial noisy impedance tensor data with the reconstructed impedance tensor data to obtain new initial noisy impedance tensor data.

[0088] In an optional embodiment, the apparatus further includes a training unit 430, which is used to obtain a training data set, wherein the training data set includes a plurality of training sample data; each training sample data includes initial noisy impedance tensor data, target winding impedance tensor data, target phase winding point and target unwinding impedance tensor data;

[0089] Based on the training data set, an iterative training operation is performed on the initial impedance tensor reconstruction model until it is determined that the iterative training termination condition is met, and the impedance tensor reconstruction model is obtained based on the weights and thresholds of the initial impedance tensor reconstruction model updated when the iterative training operation is performed for the last time; wherein the iterative training operation includes:

[0090] Select target training sample data from the training data set;

[0091] Input the initial noisy impedance tensor data in the target training sample data into the initial reconstruction sub-model to obtain the training value of the winding impedance tensor data; input the training value of the winding impedance tensor data into the initial phase winding information identification sub-model to obtain the training value of the phase winding point; based on the phase winding point and the initial noisy impedance tensor data, obtain the unwinding noisy impedance tensor data, and input the unwinding noisy impedance tensor data into the final reconstruction sub-model to obtain the training value of the unwinding impedance tensor data;

[0092] Based on the error between the training value of the winding impedance tensor data and the target winding impedance tensor data in the target training sample data, the weights and thresholds of the initial reconstruction sub-model are updated; based on the error between the training value of the phase winding point and the target phase winding point in the target training sample data, the weights and thresholds of the initial phase winding information identification sub-model are updated; based on the error between the training value of the unwinding impedance tensor data and the target unwinding impedance tensor data in the target training sample data, the weights and thresholds of the initial final reconstruction sub-model are updated.

[0093] In an optional embodiment, the reconstruction sub-model and the final reconstruction sub-model are both machine learning models: the reconstruction sub-model and the final reconstruction sub-model are both convolutional neural network models that receive initial noisy impedance tensor data in target training sample data through an input layer, obtain reconstructed impedance tensor data based on the initial noisy impedance tensor data through a hidden layer, and then output the reconstructed impedance tensor data through an output layer; wherein the hidden layer includes multiple convolution units, residual units and channel attention units, the convolution unit is used to extract features of the initial noisy impedance tensor data, the residual unit is used for the connection between two convolution units, and the channel attention unit is used to optimize the feature channel.

[0094] In an optional embodiment, the phase winding information identification sub-model is a machine learning model: the phase winding information identification sub-model is a convolutional neural network model that receives initial noisy impedance tensor data in target training sample data through an input layer, obtains phase winding points based on the initial noisy impedance tensor data through a hidden layer, and outputs phase winding points through an output layer; wherein the hidden layer includes multiple convolution units and residual units, the convolution unit is used to extract features of the initial noisy impedance tensor data, and the residual unit is used for the connection between two convolution units.

[0095] In an optional embodiment, the apparatus further includes a training unit 430 further configured to:

[0096] Obtain standard impedance data;

[0097] Add noise that meets actual observation conditions to the standard impedance data to obtain initial noisy impedance tensor data;

[0098] Convert the standard impedance data into apparent resistivity sequence and phase sequence to obtain target winding impedance tensor data;

[0099] Based on the phase sequence of standard impedance data, a target phase winding point is obtained;

[0100] Based on the target phase winding point, the phase sequence in the standard impedance data is unwound to obtain the target unwound impedance tensor data;

[0101] A training data set is obtained based on the initial noisy impedance tensor data, the target winding impedance tensor data, the target phase winding point and the target unwinding impedance tensor data.

[0102] After introducing the method and device for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application, the electronic device provided in the embodiment of the present application is briefly introduced next.

[0103] See also Figure 5 As shown, the electronic device 500 provided in the embodiment of the present application includes at least a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program, the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application is implemented.

[0104] The electronic device 500 provided in the embodiment of the present application may further include a bus 503 connecting different components (including the processor 501 and the memory 502). The bus 503 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, and the like.

[0105] The memory 502 may include a readable storage medium in the form of a volatile memory, such as a random access memory 5021 and / or a cache memory 5022, and may further include a read-only memory 5023. The memory 502 may also include a program tool 5025 having a group (at least one) of program modules 5024, the program modules 5024 including but not limited to an operating subsystem, one or more application programs, other program modules, and program data, each of which or some combination thereof may include the implementation of a network environment.

[0106] The processor 501 may be a processing element or a collective term for multiple processing elements. For example, the processor 501 may be a central processing unit or one or more integrated circuits configured to implement the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application. Specifically, the processor 501 may be a general-purpose processor, including but not limited to a CPU, a dedicated integrated circuit, a readily available programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc.

[0107] The electronic device 500 can communicate with one or more external devices 504 (e.g., keyboards, remote controls, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 500 (e.g., mobile phones, computers, etc.), and / or communicate with devices that enable the electronic device 500 to communicate with one or more other electronic devices 500 (e.g., routers, modems, etc.). Such communication can be performed through an input / output (I / O) interface 505. In addition, the electronic device 500 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 506. Figure 5 As shown, the network adapter 506 communicates with other modules of the electronic device 500 via the bus 503. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 500, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, disk arrays (Redundant Arrays of Independent Disks, RAID) subsystems, tape drives, and data backup storage subsystems.

[0108] It should be noted that Figure 5 The electronic device 500 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0109] The computer-readable storage medium provided in the embodiment of the present application is introduced below. The computer-readable storage medium provided in the embodiment of the present application stores computer instructions, and when the computer instructions are executed by the processor, the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application is implemented. Specifically, the computer instructions can be built-in or installed in the processor, so that the processor can implement the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application by executing the built-in or installed computer instructions.

[0110] In addition, the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application can also be implemented as a computer program product, which includes a program code. When the program code is run on a processor, it implements the method for reconstructing the magnetotelluric impedance tensor provided in the embodiment of the present application.

[0111] The computer program product provided in the embodiments of the present application may adopt one or more computer-readable storage media, and the computer-readable storage medium may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any suitable combination of the above. Specifically, more specific examples of computer-readable storage media (a non-exhaustive list) include an electrical connection with one or more wires, a portable disk, a hard disk, RAM, ROM, Erasable Programmable Read Only Memory (EPROM), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0112] The computer program product provided in the embodiment of the present application may adopt a CD-ROM and include program code, and may also be run on an electronic device such as a computer. However, the computer program product provided in the embodiment of the present application is not limited thereto. In the embodiment of the present application, the computer-readable storage medium may be any tangible medium containing or storing program code, and the program code may be used by or in combination with an instruction execution system, apparatus, or device.

[0113] It should be noted that, although several units or subunits of the device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided into multiple units to be embodied.

[0114] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0115] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0116] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for reconstructing a magnetotelluric impedance tensor, characterized in that: include: Obtaining initial noisy impedance tensor data; Based on the initial noisy impedance tensor data, an impedance tensor reconstruction model is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to perform preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; and finally reconstruct the unwinding noisy impedance tensor data to obtain target impedance tensor data; Wherein, the impedance tensor reconstruction model includes: a reconstruction sub-model, a phase winding information identification sub-model and a final reconstruction sub-model; based on the initial noisy impedance tensor data, the impedance tensor reconstruction model is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data to obtain target impedance tensor data, including: based on the reconstruction sub-model, the phase winding information identification sub-model and the final reconstruction sub-model, iteratively performing denoising and reconstruction operations on the initial noisy impedance tensor data until it is determined that the iteration termination condition is met, and the impedance tensor data output by the last execution of the denoising and reconstruction operation is determined as the target impedance tensor data; wherein, the denoising and reconstruction operation includes: Initial noisy impedance tensor data is input into the reconstruction sub-model to obtain preliminary reconstructed impedance tensor data; the preliminary reconstructed impedance tensor data is input into the phase winding information identification sub-model to obtain phase winding points; unwrapped noisy impedance tensor data is obtained based on the phase winding points and the initial noisy impedance tensor data; the unwrapped noisy impedance tensor data is input into the final reconstruction sub-model to obtain reconstructed impedance tensor data; based on the initial noisy impedance tensor data and the reconstructed impedance tensor data, marked strong noise is obtained; the marked strong noise in the initial noisy impedance tensor data is replaced by the reconstructed impedance tensor data to obtain new initial noisy impedance tensor data; The method also includes: obtaining a training data set, wherein the training data set includes a plurality of training sample data; each of the training sample data includes initial noisy impedance tensor data, target winding impedance tensor data, target phase winding point and target unwinding impedance tensor data; based on the training data set, performing an iterative training operation on the initial impedance tensor reconstruction model until it is determined that an iterative training termination condition is met, and obtaining the impedance tensor reconstruction model based on the weights and thresholds of the initial impedance tensor reconstruction model updated when the iterative training operation was last performed; wherein the iterative training operation includes: selecting target training sample data from the training data set; inputting the initial noisy impedance tensor data in the target training sample data into the initial reconstruction sub-model to obtain the training value of the winding impedance tensor data; inputting the training value of the winding impedance tensor data into the initial phase winding The information identification submodel obtains the training value of the phase winding point; based on the phase winding point and the initial noisy impedance tensor data, the unwrapped noisy impedance tensor data is obtained, and the unwrapped noisy impedance tensor data is input into the final reconstruction submodel to obtain the training value of the unwrapped impedance tensor data; based on the error between the training value of the winding impedance tensor data and the target winding impedance tensor data in the target training sample data, the weights and thresholds of the initial reconstruction submodel are updated; based on the error between the training value of the phase winding point and the target phase winding point in the target training sample data, the weights and thresholds of the initial phase winding information identification submodel are updated; based on the error between the training value of the unwrapped impedance tensor data and the target unwrapped impedance tensor data in the target training sample data, the weights and thresholds of the initial and final reconstruction submodels are updated.

2. The method for reconstructing the magnetotelluric impedance tensor according to claim 1, characterized in that: The reconstruction sub-model and the final reconstruction sub-model are both machine learning models: The reconstruction sub-model and the final reconstruction sub-model are both convolutional neural network models that receive initial noisy impedance tensor data in target training sample data through an input layer, obtain reconstructed impedance tensor data based on the initial noisy impedance tensor data through a hidden layer, and then output the reconstructed impedance tensor data through an output layer; wherein the hidden layer includes multiple convolution units, residual units and channel attention units, the convolution unit is used to extract features of the initial noisy impedance tensor data, the residual unit is used for connection between two of the convolution units, and the channel attention unit is used to optimize feature channels.

3. The method for reconstructing the magnetotelluric impedance tensor according to claim 1, characterized in that: The phase winding information identification sub-model is a machine learning model: The phase winding information identification sub-model is a convolutional neural network model that receives initial noisy impedance tensor data in target training sample data through an input layer, obtains phase winding points based on the initial noisy impedance tensor data through a hidden layer, and outputs the phase winding points through an output layer; wherein the hidden layer includes multiple convolution units and residual units, the convolution unit is used to extract the features of the initial noisy impedance tensor data, and the residual unit is used for the connection between two of the convolution units.

4. The method for reconstructing the magnetotelluric impedance tensor according to claim 1, characterized in that: Get the training data set, including: Obtain standard impedance data; Adding noise that meets actual observation conditions to the standard impedance data to obtain initial noisy impedance tensor data; Converting the standard impedance data into an apparent resistivity sequence and a phase sequence to obtain target winding impedance tensor data; Based on the phase sequence of the standard impedance data, obtaining a target phase winding point; Based on the target phase winding point, unwinding the phase sequence in the standard impedance data to obtain target unwinding impedance tensor data; The training data set is obtained based on the initial noisy impedance tensor data, the target winding impedance tensor data, the target phase winding point and the target unwinding impedance tensor data.

5. A device for reconstructing a magnetotelluric impedance tensor, characterized in that: include: An acquisition unit, used for acquiring initial noisy impedance tensor data; A reconstruction unit is used to perform denoising and reconstruction processing on the initial noisy impedance tensor data using an impedance tensor reconstruction model based on the initial noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model is to perform preliminary reconstruction on the initial noisy impedance tensor data to obtain preliminary reconstructed impedance tensor data; perform winding information identification on the preliminary reconstructed impedance tensor data to obtain phase winding points; unwind the initial noisy impedance tensor data based on the phase winding points to obtain unwinding noisy impedance tensor data; perform final reconstruction on the unwinding noisy impedance tensor data to obtain target impedance tensor data; wherein the impedance tensor reconstruction model includes: a reconstruction sub-model, a phase winding information identification sub-model model and a final reconstruction sub-model; based on the initial noisy impedance tensor data, using the impedance tensor reconstruction model to perform denoising and reconstruction processing on the initial noisy impedance tensor data to obtain target impedance tensor data, including: based on the reconstruction sub-model, the phase winding information identification sub-model and the final reconstruction sub-model, iteratively performing a denoising and reconstruction operation on the initial noisy impedance tensor data until it is determined that the iteration termination condition is met, and the impedance tensor data output by the last execution of the denoising and reconstruction operation is determined as the target impedance tensor data; wherein the denoising and reconstruction operation includes: inputting the initial noisy impedance tensor data into the reconstruction sub-model to obtain preliminary reconstructed impedance tensor data; inputting the preliminary reconstructed impedance tensor data into the A phase winding information identification submodel is used to obtain a phase winding point; based on the phase winding point and the initial noisy impedance tensor data, unwrapped noisy impedance tensor data is obtained; the unwrapped noisy impedance tensor data is input into the final reconstruction submodel to obtain reconstructed impedance tensor data; based on the initial noisy impedance tensor data and the reconstructed impedance tensor data, a marked strong noise is obtained; the marked strong noise in the initial noisy impedance tensor data is replaced by the reconstructed impedance tensor data to obtain new initial noisy impedance tensor data; and a training data set is obtained, wherein the training data set includes a plurality of training sample data; each of the training sample data includes initial noisy impedance tensor data, target winding impedance tensor data, Target phase winding point and target unwinding impedance tensor data; based on the training data set, perform an iterative training operation on the initial impedance tensor reconstruction model until it is determined that the iterative training termination condition is met, and obtain the impedance tensor reconstruction model based on the weights and thresholds of the initial impedance tensor reconstruction model updated when the iterative training operation was last performed; wherein the iterative training operation includes: selecting target training sample data from the training data set; inputting the initial noisy impedance tensor data in the target training sample data into the initial reconstruction sub-model to obtain the training value of the winding impedance tensor data; inputting the training value of the winding impedance tensor data into the initial phase winding information identification sub-model to obtain the training value of the phase winding point;Based on the phase winding point and the initial noisy impedance tensor data, the unwrapped noisy impedance tensor data is obtained, and the unwrapped noisy impedance tensor data is input into the final reconstruction sub-model to obtain the training value of the unwrapped impedance tensor data; based on the error between the training value of the winding impedance tensor data and the target winding impedance tensor data in the target training sample data, the weights and thresholds of the initial reconstruction sub-model are updated; based on the error between the training value of the phase winding point and the target phase winding point in the target training sample data, the weights and thresholds of the initial phase winding information identification sub-model are updated; based on the error between the training value of the unwrapped impedance tensor data and the target unwrapped impedance tensor data in the target training sample data, the weights and thresholds of the initial and final reconstruction sub-model are updated. ; 6. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for reconstructing the magnetotelluric impedance tensor according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method for reconstructing the magnetotelluric impedance tensor according to any one of claims 1 to 4 is implemented.

Images