Geomagnetic gradient tensor depth representation learning method and system

Abstract

The invention discloses a geomagnetic gradient tensor depth representation learning method and system, and belongs to the technical field of information processing, and the method comprises the steps: obtaining a geomagnetic three-component data matrix of a magnetic object, and building an original gradient tensor data set; generating a gradient matrix based on the original gradient tensor data set, and performing orthogonal decomposition on the gradient matrix to obtain mutually independent rotation isovariant features and direction sensitive features; inputting the rotation isovariant features and the direction sensitive features into a rotation isovariant convolutional neural network for feature extraction; constructing a geometric constraint loss function including an angle constraint loss function and a direction constraint loss function, and optimizing the training network; the geomagnetic gradient tensor data is subjected to deep representation based on the trained network, the innovative gradient matrix orthogonal decomposition technology solves the problem that rotation invariance and direction sensitivity are difficult to consider at the same time, the feature representation accuracy is improved by about 30-40%, the processing speed is improved by 2-3 times, and the anti-interference capability is remarkably enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information processing, in particular to a geomagnetic gradient tensor deep representation learning method and system, and more particularly to a method and system for feature extraction and representation of geomagnetic gradient data using deep learning and tensor analysis techniques. BACKGROUND

[0002] Geomagnetic gradient data has wide application value in geological exploration, underwater target detection, intelligent traffic management and other fields. Traditional geomagnetic gradient data processing methods are mainly based on signal processing and statistical analysis, which are difficult to effectively extract deep features in complex geomagnetic fields, and have obvious shortcomings in processing rotation transformation and direction sensitive information.

[0003] With the development of deep learning technology, it is possible to use neural networks to process geomagnetic gradient data. However, existing deep learning methods still have the following problems when processing geomagnetic gradient tensor data: on the one hand, it is difficult to maintain rotation invariance and direction sensitivity at the same time, often resulting in a trade-off situation; on the other hand, the feature representation ability of geomagnetic gradient tensor data is limited, making it difficult to capture complex patterns and internal structures in the data. In addition, the existing method has high computational complexity, limited real-time processing capability, and insufficient stability in high noise environment.

[0004] Therefore, there is an urgent need for a method and system that can effectively process geomagnetic gradient tensor data while maintaining rotation invariance and direction sensitivity, and having efficient feature representation capability. SUMMARY

[0005] The purpose of the present application is to provide a geomagnetic gradient tensor deep representation learning method and system, which aims to solve the problems of existing technology that cannot maintain rotation invariance and direction sensitivity at the same time, limited feature representation capability, low computational efficiency and insufficient anti-interference ability.

[0006] The present application provides a geomagnetic gradient tensor deep representation learning method, comprising: Obtaining a geomagnetic three-component data matrix of a magnetic object, and establishing an original gradient tensor data set according to the geomagnetic three-component data matrix; Based on the original gradient tensor data set, a gradient matrix is generated, and the gradient matrix is orthogonally decomposed to obtain rotation invariant features and direction sensitive features, wherein the rotation invariant features and the direction sensitive features are independent of each other; The rotation invariant features and the direction sensitive features are input into a rotation invariant convolutional neural network for feature extraction to obtain feature representations; construct a geometric constraint loss function comprising an angle constraint loss function and a direction constraint loss function, and train the rotation equivariant convolutional neural network by optimizing the geometric constraint loss function; based on the trained rotation equivariant convolutional neural network, deep feature of input magnetic gradient tensor data is obtained.

[0007] As a preferred, the magnetic three-component data matrix of the magnetic object is obtained, and the original gradient tensor data set is established according to the magnetic three-component data matrix. Specifically, it comprises: obtaining a plurality of magnetic three-component data matrices of magnetic objects and establishing a magnetic three-component data set of magnetic objects, the magnetic three-component data set of magnetic objects comprises a plurality of elements, and each element contains a magnetic three-component data matrix of a magnetic object; extracting the first magnetic three-component data matrix, the second magnetic three-component data matrix and the third magnetic three-component data matrix contained in the magnetic three-component data set of magnetic objects; establishing an original spatial three-component data set according to the first magnetic three-component data matrix, the second magnetic three-component data matrix and the third magnetic three-component data matrix; performing matrix difference operation on the original spatial three-component data set to extract the original gradient tensor data set.

[0008] As a preferred, the gradient matrix is generated based on the original gradient tensor data set, and the rotation equivariant feature and the direction sensitive feature are obtained by orthogonal decomposition of the gradient matrix. Specifically, it comprises: preprocessing the original gradient tensor data set, including zero mean and unit amplitude normalization; converting the preprocessed original gradient tensor data set into a gradient matrix form; constructing an orthogonal decomposition matrix wherein R is the orthogonal decomposition matrix, G is the gradient matrix, W is the rotation equivariant feature, the transpose of the gradient matrix; by eigenvalue decomposition, the gradient matrix is decomposed into two-channel information independent of the rotation equivariant feature and the direction sensitive feature, wherein the rotation equivariant feature dimension is 4 times the original gradient tensor dimension, and the direction sensitive feature dimension is 2 times the original gradient tensor dimension.

[0009] As a preferred, the rotation equivariant convolutional neural network comprises: a gradient matrix orthogonal decomposition structure composed of a plurality of convolution modules and a plurality of pooling modules, wherein the convolution module comprises a convolution layer, a ReLU layer, a batch normalization layer and a maximum pooling layer, and the pooling module is a maximum pooling layer; a rotation-invariant feature and a direction-sensitive feature, respectively. a feature fusion module configured to fuse the rotation-invariant feature and the direction-sensitive feature to generate a target gradient tensor feature.

[0010] Preferably, the angle constraint loss function is configured to constrain the stability of the rotation-invariant feature under rotation transformation, including: introducing an angle constraint in the feature space to limit the change of the included angle between the feature vectors; quantifying the stability of the rotation-invariant feature under rotation transformation; ensuring the orthogonality of the feature vectors to avoid information redundancy.

[0011] Preferably, the direction constraint loss function is configured to ensure the orthogonality and effectiveness of the direction-sensitive feature, including: constructing a direction constraint loss function to evaluate the quality of the direction-sensitive feature; introducing a loss metric to quantify the deviation of the direction-sensitive feature from the ideal state; establishing a feature orthogonality constraint to ensure the capture of complete directional information.

[0012] Preferably, the optimization process of the geometric constraint loss function includes: initializing network parameters to construct an initial feature representation; calculating the current feature representation through forward propagation; calculating the angle constraint loss function and the direction constraint loss function; synthesizing a total loss function for backpropagation; updating network parameters to optimize feature extraction capability; monitoring the trend of the loss function, and saving the optimal network parameters when the convergence condition is reached.

[0013] Preferably, based on the trained rotation-invariant convolutional neural network, the input geomagnetic gradient tensor data is deeply characterized to obtain a target gradient tensor feature, including: converting the input geomagnetic gradient tensor data into a 3×3×3 tensor structure; converting the 3×3×3 tensor structure into a 1×27 tensor through a first convolutional layer; obtaining a 1×9 first feature tensor through a first pooling layer; extracting features through a second convolutional layer to generate a feature matrix; separating the rotation-invariant feature and the direction-sensitive feature through the orthogonal decomposition module; The target gradient tensor feature representation is generated through the feature fusion module.

[0014] As preferred, it further comprises: The target gradient tensor feature is applied to application scenarios such as geological resource exploration, underwater target detection or intelligent traffic management. For geological resource exploration, regional grid point collection, real-time gradient tensor feature extraction, intelligent identification and marking of abnormal points, three-dimensional modeling and evaluation of resource distribution are performed. For underwater target detection, regional scanning collection of magnetic field data, real-time extraction of gradient tensor features, target detection, classification and positioning, trajectory tracking and behavior analysis are performed. For intelligent traffic management, multi-point synchronous collection of vehicle magnetic field data, real-time extraction of gradient tensor features, vehicle type identification and flow statistics, traffic state evaluation and early warning are performed.

[0015] The geomagnetic gradient tensor deep representation learning system comprises: The geomagnetic gradient data acquisition module is used to obtain a geomagnetic three-component data matrix of a magnetic object, and to establish an original gradient tensor data set according to the geomagnetic three-component data matrix. The gradient tensor construction module is used to generate a gradient matrix based on the original gradient tensor data set, and to perform orthogonal decomposition on the gradient matrix to obtain rotationally invariant features and direction-sensitive features, wherein the rotationally invariant features and the direction-sensitive features are independent of each other. The rotationally invariant feature extraction module is used to input the rotationally invariant features and the direction-sensitive features into a rotationally invariant convolutional neural network for feature extraction to obtain a feature representation. The geometric constraint optimization module is used to construct a geometric constraint loss function comprising an angle constraint loss function and a direction constraint loss function, and to train the rotationally invariant convolutional neural network by optimizing the geometric constraint loss function. The feature application module is used to perform deep representation on input geomagnetic gradient tensor data based on the trained rotationally invariant convolutional neural network to obtain a target gradient tensor feature, and to apply the target gradient tensor feature to a related application scenario.

[0016] The beneficial effects of the present application include: 1. Through the innovative gradient matrix orthogonal decomposition technology, the effective separation of rotationally invariant features and direction-sensitive features is realized, the problem of difficult to balance rotation invariance and direction sensitivity is solved, and the feature representation accuracy is improved by about 30%-40%.

[0017] 2. A special rotation equivariant convolutional neural network is designed, which significantly improves the feature extraction capability and processing speed is about 2-3 times higher than that of traditional methods, and is more suitable for real-time application scenarios.

[0018] 3. A geometric constraint loss function is introduced to ensure the quality and reliability of feature extraction, and stable feature extraction performance can be maintained in a high noise environment, and the system signal-to-noise ratio requirement is reduced from ≥15dB of traditional methods to ≥8dB, which is improved by about 7dB.

[0019] 4. The whole process integration from data acquisition to feature application is realized, which has high engineering practicability and scalability, and is suitable for many application fields such as geological exploration, underwater target detection and intelligent traffic management. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The flowchart of the geomagnetic gradient tensor deep representation learning method provided by the embodiment of the application.

[0021] Figure 2 The gradient matrix orthogonal decomposition schematic diagram provided by the embodiment of the application.

[0022] Figure 3 The rotation equivariant convolutional neural network structure schematic diagram provided by the embodiment of the application.

[0023] Figure 4 The geometric constraint loss function optimization flowchart provided by the embodiment of the application.

[0024] Figure 5 The geomagnetic gradient tensor deep representation learning system structure schematic diagram provided by the embodiment of the application. DETAILED DESCRIPTION

[0025] Please refer to Figure 1 - Figure 5 , the application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to illustrate and explain the application, and are not used to limit the application.

[0026] Referring to Figure 1 , the application provides a geomagnetic gradient tensor deep representation learning method, comprising the following steps: S1: Obtain a geomagnetic three-component data matrix of a magnetic object, and establish an original gradient tensor data set according to the geomagnetic three-component data matrix.

[0027] S2: Based on the original gradient tensor data set, a gradient matrix is generated, and the gradient matrix is orthogonally decomposed to obtain rotation equivariant features and direction sensitive features, wherein the rotation equivariant features and the direction sensitive features are independent of each other.

[0028] S3: inputting the rotation equivariant feature and the direction sensitive feature into a rotation equivariant convolutional neural network to perform feature extraction, to obtain a feature representation.

[0029] S4: constructing a geometric constraint loss function comprising an angle constraint loss function and a direction constraint loss function, and training the rotation equivariant convolutional neural network by optimizing the geometric constraint loss function.

[0030] S5: based on the trained rotation equivariant convolutional neural network, performing deep characterization on input magnetic gradient tensor data to obtain a target gradient tensor feature.

[0031] The following will be described in detail: S1: obtaining a magnetic three-component data matrix of a magnetic object and establishing an original gradient tensor data set, in an embodiment of the present application, a magnetic three-component data matrix of a magnetic object is obtained, and an original gradient tensor data set is established according to the magnetic three-component data matrix, which specifically comprises: First, a plurality of magnetic three-component data matrices of magnetic objects are obtained and established as a magnetic three-component data set of magnetic objects. Preferably, data acquisition is performed by a magnetometer acquisition host and a geomagnetic gradient magnetometer, and the sampling frequency is set to 100 Hz to ensure that the time resolution of the data meets the subsequent processing requirements. The magnetic three-component data set of the magnetic object includes a plurality of elements, and each element contains a magnetic three-component data matrix of a magnetic object.

[0032] Next, the first magnetic three-component data matrix, the second magnetic three-component data matrix and the third magnetic three-component data matrix contained in the magnetic three-component data set of the magnetic object are extracted. The three data matrices correspond to the magnetic field component measurement values in the x, y and z orthogonal directions, respectively.

[0033] Subsequently, an original spatial three-component data set is established according to the first magnetic three-component data matrix, the second magnetic three-component data matrix and the third magnetic three-component data matrix. Here, the elements in the original spatial three-component data set are magnetic spatial three-component data matrices, denoted as 、 and , representing the first, second and third magnetic spatial three-component data matrices, respectively.

[0034] Finally, the original spatial three-component data set is subjected to matrix difference operation to extract the original gradient tensor data set. Matrix difference operation refers to calculating the magnetic field difference between adjacent spatial points to obtain spatial gradient information. The original gradient tensor data set can be represented as: , wherein is the original gradient tensor data set, For the The original gradient tensor data matrix, n is the number of samples in the data set. is a The third-order tensor contains the gradient information of the magnetic field components in the x, y, and z directions in three-dimensional space.

[0035] During the actual implementation, to improve data quality, the collected data underwent preprocessing, including denoising, filtering, and outlier detection. For noise suppression, a bandpass filter was used with cutoff frequencies set at 0.1 Hz and 40 Hz, effectively filtering out ambient noise and high-frequency interference. For outlier detection, a threshold was set at the mean ±3 standard deviations. Data points outside this range were marked as outliers and interpolated.

[0036] S2: Generate a gradient matrix and perform orthogonal decomposition. In one embodiment of the present invention, a gradient matrix is ​​generated based on the original gradient tensor data set, and the gradient matrix is ​​orthogonally decomposed to obtain rotational equivariant features and direction-sensitive features, specifically including: First, the original gradient tensor dataset is preprocessed, including zero-meaning and unit amplitude normalization. Zero-meaning refers to subtracting the mean from the data so that the center of the data distribution is at the origin, which can be expressed as: , in, is the zero-mean gradient tensor, and N is the total number of samples in the dataset.

[0037] Unit amplitude normalization means dividing the data by its modulus to make the data amplitude uniform, which can be expressed as: , in, is the normalized gradient tensor, for The modulus of , calculated as the square root of the sum of the squares of its elements.

[0038] Next, the pre-processed original gradient tensor data set is converted into a gradient matrix form. This step is to reconstruct the 3×3×3 third-order tensor into a matrix representation for subsequent processing. The converted gradient matrix is ​​denoted as G, and its dimension is , where m is the number of samples and n is the feature dimension (here it is 27, which is the dimension of the 3×3×3 tensor after flattening).

[0039] Then, construct the orthogonal decomposition matrix , where R is the orthogonal decomposition matrix, G is the gradient matrix, and W is the rotational equivariant feature. is the transpose of the gradient matrix. This form of decomposition has important mathematical significance, which decomposes the gradient matrix into two independent parts of rotationally invariant features and direction sensitive features.

[0040] Finally, the gradient matrix is decomposed into two-channel information independent of the rotationally invariant features and the direction sensitive features by eigenvalue decomposition. Specifically, the matrix G is singular value decomposed (SVD): , where U and V are orthogonal matrices, is a diagonal matrix containing singular values.

[0041] Based on the SVD decomposition result, the rotationally invariant features W and the direction sensitive features D can be obtained: , , where, and are the singular value diagonal matrices related to the rotationally invariant features and the direction sensitive features extracted from .

[0042] In the present application, the rotationally invariant feature dimension is 4 times the original gradient tensor dimension, and the direction sensitive feature dimension is 2 times the original gradient tensor dimension. This dimension setting is verified by a large number of experiments, which can maintain the integrity of information while providing sufficient representation ability.

[0043] The core innovation of this orthogonal decomposition method is that it can separate the information in the gradient tensor into rotationally invariant part and direction sensitive part, solving the problem that traditional methods are difficult to maintain both characteristics at the same time.

[0044] S3: rotationally invariant convolutional neural network feature extraction, in an embodiment of the present application, the rotationally invariant features and the direction sensitive features are input into a rotationally invariant convolutional neural network for feature extraction to obtain feature representation. The rotationally invariant convolutional neural network has a special structure design, including: a gradient matrix orthogonal decomposition structure composed of multiple convolution modules and multiple pooling modules. The convolution module includes a convolution layer, a ReLU layer, a batch normalization layer and a max pooling layer, and the pooling module is a max pooling layer. Specifically, the convolution layer is responsible for feature extraction, the ReLU layer introduces nonlinear transformation, the batch normalization layer stabilizes the training process, and the max pooling layer reduces dimension and extracts significant features.

[0045] The orthogonal decomposition module is composed of two 3x3 convolution layers and a 1x1 convolution layer. The two 3x3 convolution layers are used to capture local feature correlation, and the 1x1 convolution layer is used to fuse inter-channel information. The role of this module is to decompose the gradient matrix into two-channel information independent of the rotation-invariant feature and the direction-sensitive feature.

[0046] The feature fusion module is used to fuse the rotation-invariant feature and the direction-sensitive feature to generate a target gradient tensor feature. This module forms the final feature representation by weighted fusion of the two types of features.

[0047] In the network implementation, the first convolution layer uses 32 3x3 convolution kernels with a step size of 1 and a padding of 1; the second convolution layer uses 64 3x3 convolution kernels with a step size of 1 and a padding of 1; the two 3x3 convolution layers in the orthogonal decomposition module each use 128 convolution kernels, and the 1x1 convolution layer uses 256 convolution kernels. These parameter settings are optimized to achieve a good balance between computational efficiency and feature extraction capability.

[0048] The convolution operation can be represented as: , wherein, is the input feature map, is the output feature map, W is the convolution kernel weight, b is the bias term, is the activation function (ReLU), and * represents the convolution operation.

[0049] The ReLU activation function is defined as: , The batch normalization operation is defined as: , wherein, X is the input, y is the output, is the batch mean, is the batch variance, and are learnable parameters, is a small constant (usually set to ) to prevent division by zero.

[0050] The max-pooling operation is defined as: , wherein, is the value of position in the input feature map, is the pooling window region centered at , and is the value of position in the output feature map.

[0051] In the implementation process, a deep network structure is formed by stacking multiple convolution modules and pooling modules, and features from low to high levels can be extracted layer by layer. Preferably, the network depth is set to 5 layers, which is a compromise between model complexity and feature extraction capability.

[0052] S4: Geometric constraint loss function construction and optimization. In an embodiment of the present application, a geometric constraint loss function including an angle constraint loss function and a direction constraint loss function is constructed, and the rotation equivariant convolutional neural network is trained by optimizing the geometric constraint loss function.

[0053] First, the angle constraint loss function is used to constrain the stability of the rotation equivariant feature under rotation transformation, including: introducing an angle constraint in the feature space to limit the change of the included angle between feature vectors; quantifying the stability of the rotation equivariant feature under rotation transformation; ensuring the orthogonality of the feature vectors to avoid information redundancy.

[0054] The angle constraint loss function can be expressed as: , wherein, is the included angle between the feature vectors and , calculated as: , N is the number of feature vectors, and are feature vectors in the rotation equivariant feature.

[0055] Next, the direction constraint loss function is used to ensure the orthogonality and effectiveness of the direction sensitive feature, including: constructing a direction constraint loss function to evaluate the quality of the direction sensitive feature; introducing a loss metric to quantify the deviation of the direction sensitive feature from the ideal state; establishing a feature orthogonality constraint to ensure that complete directional information is captured.

[0056] The direction constraint loss function can be expressed as: , wherein, is the current direction sensitive feature, is the ideal direction sensitive feature (obtained by prior knowledge or supervision signal), is the Frobenius norm, is a weight coefficient (set to 0.5 in the experiment), M is the dimension of the direction sensitive feature, , represents the inner product of two direction sensitive features.

[0057] Then, the angle constraint loss function and the direction constraint loss function are combined to form a total loss function: , wherein, is a task-related loss (such as a classification loss, a regression loss, etc.), 、 and are loss function weight coefficients for balancing the importance of different loss terms. In experiments, 、 and are set to 0.3, 0.3 and 0.4 respectively, which are the better configurations verified by a large number of experiments.

[0058] Preferably, the optimization process of the geometric constraint loss function comprises: 1. Initialize network parameters and construct initial feature representations. The network parameter initialization adopts the He initialization method, which can effectively prevent the gradient vanishing or explosion problem in deep network.

[0059] 2. Calculate the current feature representation by forward propagation. The input data is processed by each layer of the network to generate the current feature representation.

[0060] 3. Calculate the angle constraint loss function and the direction constraint loss function. According to the current feature representation, the values of the two types of loss functions are calculated.

[0061] 4. Synthesize the total loss function and perform back propagation. The loss terms are combined by weighting to form the total loss function, and then the gradient is calculated by the back propagation algorithm.

[0062] 5. Update the network parameters and optimize the feature extraction capability. The network parameters are updated using the Adam optimizer, and the learning rate is initially set to 0.001 and adopts the learning rate decay strategy, which is decayed to 0.9 times of the original every 10 epochs.

[0063] 6. Monitor the trend of the loss function, and save the optimal network parameters when the convergence condition is reached. The convergence condition is set to the total loss function change being less than 10 -4 for 5 consecutive epochs, or reaching the maximum number of epochs (set to 100).

[0064] The optimization process adopts a batch processing mode, and the number of samples per batch is set to 64, which achieves a good balance between computational efficiency and optimization stability. Early stopping strategy is adopted in the optimization process to avoid overfitting problem.

[0065] S5: Deep representation based on the trained network, in an embodiment of the present application, based on the trained rotation equivariant convolutional neural network, the input geomagnetic gradient tensor data is deeply represented to obtain the target gradient tensor feature. Specifically, it includes: Firstly, the input geomagnetic gradient tensor data is converted into a 3x3x3 tensor structure. This step is to organize the original geomagnetic data into a standard three-order tensor form for subsequent processing.

[0066] Then, the 3x3x3 tensor structure is converted into a 1x27 tensor through the first convolutional layer. The first convolutional layer uses 32 3x3 convolutional kernels with a step size of 1 and a padding of 1 to convert the input tensor into a feature map, and then the feature map is flattened into a 1x27 vector.

[0067] Then, the first feature tensor of 1x9 is obtained after the first pooling layer. The first pooling layer uses a 3x3 pooling window with a step size of 3 to reduce the dimension of the 1x27 vector to a 1x9 feature tensor.

[0068] Subsequently, the second convolutional layer is used to extract features and generate a feature matrix. The second convolutional layer uses 64 3x3 convolutional kernels with a step size of 1 and a padding of 1 to further extract feature information and form a feature matrix.

[0069] Then, the rotation equivariant feature and the direction sensitive feature are separated through the orthogonal decomposition module. The orthogonal decomposition module is composed of two 3x3 convolutional layers and one 1x1 convolutional layer, which decomposes the feature matrix into rotation equivariant feature and direction sensitive feature.

[0070] Finally, the target gradient tensor feature representation is generated through the feature fusion module. The feature fusion module fuses the rotation equivariant feature and the direction sensitive feature into the final target gradient tensor feature through weighted combination.

[0071] The target gradient tensor feature can be represented as: , wherein, is the target gradient tensor feature, is the rotation equivariant feature, is the direction sensitive feature, and is the feature fusion weighting coefficient, indicating the importance of the two types of features. In actual application, according to the specific task requirements, the values of and can be adjusted to balance the rotation invariance and direction sensitivity.

[0072] Preferably, in the test phase, model compression techniques, including pruning and quantization, are adopted to improve the processing efficiency. Pruning refers to removing network connections that have less impact on the output. In the experiment, the pruning threshold is set to the 90th percentile of the absolute weight values. Quantization refers to converting 32-bit floating-point parameters to 8-bit integers to reduce computational and storage overhead. After compression, the model size is reduced by about 75%, the inference speed is increased by about 2.5 times, and the accuracy is only decreased by about 1%-2%, which is an acceptable compromise in practical applications.

[0073] In an embodiment of the present application, the target gradient tensor feature can be applied to geological resource exploration, underwater target detection, or intelligent traffic management, etc.

[0074] For geological resource exploration, the method of the present application performs regional grid point collection, real-time gradient tensor feature extraction, abnormal point intelligent identification and marking, resource distribution three-dimensional modeling and evaluation. Specifically, first, grid points are set in the exploration area, usually with a grid spacing of 5-10 meters to ensure the spatial resolution of the collected data; then real-time magnetic gradient data are collected and features are extracted; then abnormal points are identified by setting a threshold (usually ±3 times the standard deviation of the regional average); finally, a three-dimensional resource model is constructed based on the distribution of abnormal points. Experiments show that the positioning accuracy of the method of the present application in iron ore resource exploration is improved by about 35%, and the exploration efficiency is improved by about 40%.

[0075] For underwater target detection, the method of the present application performs regional scanning of magnetic field data, real-time extraction of gradient tensor features, target detection, classification and positioning, trajectory tracking and behavior analysis. Specifically, first, a underwater magnetic gradient sensor array is used to scan the target area; then the gradient tensor features are extracted in real time; then the target is detected and classified through feature matching, with a classification accuracy of up to 92%; finally, the trajectory is analyzed through continuous tracking. In complex underwater environments, the target detection distance of the method of the present application is improved by about 25% compared with traditional methods, and it can still maintain an identification rate of more than 85% under the condition of a signal-to-noise ratio as low as 8dB.

[0076] For intelligent traffic management, the method of the present application performs multi-point synchronous collection of vehicle magnetic field data, real-time extraction of gradient tensor features, vehicle type identification and flow statistics, traffic state evaluation and warning issuance. Specifically, first, a magnetic gradient sensor array is buried at key nodes on the road; then the magnetic field disturbance generated by passing vehicles is collected in real time; then the gradient tensor features are extracted for vehicle identification and statistics; finally, the traffic state is evaluated based on the flow data and a warning is issued. In practical applications, the vehicle identification accuracy is more than 95%, the flow statistics error is controlled within ±3%, and the warning response time is less than 2 seconds, meeting the real-time traffic management requirements.

[0077] Reference Figure 5The application provides a geomagnetic gradient tensor deep characterization learning system, comprising a geomagnetic gradient data acquisition module 1, a gradient tensor construction module 2, a rotationally invariant feature extraction module 3, a geometric constraint optimization module 4 and a feature application module 5.

[0078] The geomagnetic gradient data acquisition module 1 is used for acquiring a geomagnetic three-component data matrix of a magnetic object and establishing an original gradient tensor data set according to the geomagnetic three-component data matrix. The module comprises a magnetometer support, a magnetometer acquisition host and a geomagnetic gradient magnetometer, wherein the geomagnetic gradient magnetometer comprises three triaxial magnetometers which are arranged at equal intervals in space. Preferably, the sensitivity of the triaxial magnetometer is not less than 0.1 nT, and the sampling rate is not less than 100 Hz, so as to meet the high-precision data acquisition requirement.

[0079] The gradient tensor construction module 2 is used for generating a gradient matrix based on the original gradient tensor data set and performing orthogonal decomposition on the gradient matrix to obtain rotationally invariant features and direction-sensitive features, wherein the rotationally invariant features and the direction-sensitive features are independent of each other. The module is responsible for data preprocessing, gradient calculation and orthogonal decomposition, so as to ensure the data quality for subsequent feature extraction.

[0080] The rotationally invariant feature extraction module 3 is used for inputting the rotationally invariant features and the direction-sensitive features into a rotationally invariant convolutional neural network to perform feature extraction and obtain feature representation. The module realizes the forward propagation process of the network, including convolution, activation, normalization and pooling operations and the like, so as to form effective feature representation.

[0081] The geometric constraint optimization module 4 is used for constructing a geometric constraint loss function comprising an angle constraint loss function and a direction constraint loss function and training the rotationally invariant convolutional neural network by optimizing the geometric constraint loss function. The module is responsible for training and optimizing the network, and improves the feature extraction capability through loss calculation, back propagation and parameter updating.

[0082] The feature application module 5 is used for performing deep characterization on input geomagnetic gradient tensor data based on the trained rotationally invariant convolutional neural network, obtaining target gradient tensor features and applying the target gradient tensor features to related application scenarios. The module is the output end of the system and is responsible for applying the extracted features to actual tasks, such as geological exploration, underwater target detection or intelligent traffic management and the like.

[0083] The modules are connected through standardized interfaces to form a complete data processing flow. Preferably, high-speed bus structure is adopted for communication between the modules, and the data transmission rate is not less than 100 MB / s, so as to ensure the real-time performance of the system. The system adopts modular design, so that it is convenient to upgrade and expand, and each module can be individually optimized or replaced according to application requirements.

[0084] In the system implementation, the hardware configuration comprises: a multi-core processor (main frequency >=2.5GHz) is used for a processing platform, memory >=8GB, storage >=256GB SSD; the software architecture comprises an operating system layer, a middleware layer and an application layer, and supports multi-task parallel processing and remote access.

[0085] The system of the application realizes full-process integration from data acquisition to feature application, has high engineering practicability and scalability, can stably run in various complex environments, and provides a complete solution for geomagnetic gradient data processing.

[0086] The above-mentioned embodiments only express the specific implementation of the application, and the description is more specific and detailed, but it cannot be understood as a limitation on the scope of the patent of the application. It should be pointed out that, for ordinary skilled persons in the art, without departing from the concept of the application, a number of modifications and improvements can be made, which belong to the protection scope of the application.

Claims

1. A deep representation learning method for geomagnetic gradient tensor, characterized by: include: Acquire a geomagnetic three-component data matrix of a magnetic object, and establish an original gradient tensor data set according to the geomagnetic three-component data matrix; Based on the original gradient tensor data set, a gradient matrix is ​​generated, and the gradient matrix is ​​orthogonally decomposed to obtain rotational equivariant features and direction-sensitive features, wherein the rotational equivariant features and the direction-sensitive features are independent of each other; Inputting the rotation equivariant feature and the direction-sensitive feature into a rotation equivariant convolutional neural network for feature extraction to obtain a feature representation; Constructing a geometric constraint loss function including an angle constraint loss function and a direction constraint loss function, and training the rotation equivariant convolutional neural network by optimizing the geometric constraint loss function; Based on the trained rotational equivariant convolutional neural network, the input geomagnetic gradient tensor data is deeply characterized to obtain the target gradient tensor features.

2. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The step of obtaining a geomagnetic three-component data matrix of a magnetic object and establishing an original gradient tensor data set according to the geomagnetic three-component data matrix specifically includes: Acquire geomagnetic three-component data matrices of a plurality of magnetic objects and establish a geomagnetic three-component data set of a magnetic object, wherein the geomagnetic three-component data set of a magnetic object includes a plurality of elements, and each element includes a geomagnetic three-component data matrix of a magnetic object; Extracting a first geomagnetic three-component data matrix, a second geomagnetic three-component data matrix, and a third geomagnetic three-component data matrix contained in the geomagnetic three-component data set of the magnetic object; Establishing an original spatial three-component data set according to the first geomagnetic three-component data matrix, the second geomagnetic three-component data matrix and the third geomagnetic three-component data matrix; A matrix difference operation is performed on the original spatial three-component data set to extract the original gradient tensor data set.

3. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The method generates a gradient matrix based on the original gradient tensor data set, and performs orthogonal decomposition on the gradient matrix to obtain rotational equivariant features and direction-sensitive features, specifically including: Preprocessing the raw gradient tensor dataset, including zero-meaning and unit amplitude normalization; Converting the preprocessed raw gradient tensor data set into a gradient matrix form; Constructing an orthogonal decomposition matrix , where R is the orthogonal decomposition matrix, G is the gradient matrix, and W is the rotational equivariant feature. is the transpose of the gradient matrix; Through eigenvalue decomposition, the gradient matrix is ​​decomposed into two-channel information consisting of the rotation equivariant feature and the direction-sensitive feature, which are independent of each other, wherein the dimension of the rotation equivariant feature is 4 times the dimension of the original gradient tensor, and the dimension of the direction-sensitive feature is 2 times the dimension of the original gradient tensor.

4. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The rotation equivariant convolutional neural network includes: A gradient matrix orthogonal decomposition structure composed of multiple convolution modules and multiple pooling modules, wherein the convolution module includes a convolution layer, a ReLU layer, a batch normalization layer and a maximum pooling layer, and the pooling module is a maximum pooling layer; An orthogonal decomposition module, consisting of two 3×3 convolutional layers and one 1×1 convolutional layer, for decomposing the gradient matrix into two independent channels of information, namely, the rotation-equivariant feature and the direction-sensitive feature; The feature fusion module is used to fuse the rotation equivariant feature with the direction-sensitive feature to generate a target gradient tensor feature.

5. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The angle constraint loss function is used to constrain the stability of the rotational equivariant feature under rotation transformation, including: Introduce angle constraints in the feature space to limit the angle changes between feature vectors; quantifying the stability of the rotationally equivariant feature under rotational transformations; Ensure the orthogonality of feature vectors to avoid information redundancy.

6. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The direction constraint loss function is used to ensure the orthogonality and effectiveness of the direction-sensitive features, including: Constructing a direction-constrained loss function to evaluate the quality of the direction-sensitive features; Introducing a loss metric to quantify the deviation of the direction-sensitive feature from the ideal state; Establish feature orthogonality constraints to ensure that complete directional information is captured.

7. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: The optimization process of the geometric constraint loss function includes: Initialize network parameters and construct initial feature representation; Calculate the current feature representation through forward propagation; Calculating the angle constraint loss function and the direction constraint loss function; Synthesize the total loss function and perform backpropagation; Update network parameters and optimize feature extraction capabilities; Monitor the trend of the loss function and save the optimal network parameters after reaching the convergence condition.

8. The geomagnetic gradient tensor deep representation learning method according to claim 4, characterized in that: The rotational equivariant convolutional neural network after training is used to deeply characterize the input geomagnetic gradient tensor data to obtain target gradient tensor features, specifically including: Convert the input geomagnetic gradient tensor data into a 3×3×3 tensor structure; The 3×3×3 tensor structure is converted into a 1×27 tensor through the first convolutional layer; After the first pooling layer, the first feature tensor of 1×9 is obtained; Extract features through the second convolutional layer to generate a feature matrix; Separating the rotational equivariant features and the direction-sensitive features through the orthogonal decomposition module; The target gradient tensor feature representation is generated through the feature fusion module.

9. The geomagnetic gradient tensor deep representation learning method according to claim 1, characterized in that: Also includes: Applying the target gradient tensor feature to application scenarios such as geological resource exploration, underwater target detection or intelligent traffic management; Among them, for geological resource exploration, regional grid point collection, real-time gradient tensor feature extraction, intelligent identification and marking of abnormal points, and three-dimensional modeling and evaluation of resource distribution are performed; For underwater target detection, it performs area scanning to collect magnetic field data, extract gradient tensor features in real time, target detection, classification and positioning, trajectory tracking and behavior analysis; For intelligent traffic management, it performs multi-point synchronous collection of vehicle magnetic field data, real-time extraction of gradient tensor features, vehicle type identification and flow statistics, traffic status assessment and warning issuance.

10. Geomagnetic gradient tensor deep representation learning system, characterized by: include: A geomagnetic gradient data acquisition module is used to obtain a geomagnetic three-component data matrix of a magnetic object and to establish an original gradient tensor data set based on the geomagnetic three-component data matrix; a gradient tensor construction module, configured to generate a gradient matrix based on the original gradient tensor dataset, and perform orthogonal decomposition on the gradient matrix to obtain rotational equivariant features and direction-sensitive features, wherein the rotational equivariant features and the direction-sensitive features are independent of each other; a rotation equivariant feature extraction module, configured to input the rotation equivariant feature and the direction-sensitive feature into a rotation equivariant convolutional neural network for feature extraction to obtain a feature representation; A geometric constraint optimization module is used to construct a geometric constraint loss function including an angle constraint loss function and an orientation constraint loss function, and train the rotation equivariant convolutional neural network by optimizing the geometric constraint loss function; The feature application module is used to perform in-depth characterization of the input geomagnetic gradient tensor data based on the trained rotational equivariant convolutional neural network, obtain target gradient tensor features, and apply the target gradient tensor features to relevant application scenarios.

Images