Geomagnetic reference map construction method based on super-resolution network

Abstract

The invention belongs to the technical field of geomagnetic navigation, and particularly relates to a geomagnetic reference map construction method based on a super-resolution network, and the method comprises the following steps: S1, collecting geomagnetic data points in a rectangular region by using geomagnetic collection equipment, forming sparse collection data, and constructing a sparse matrix as a geomagnetic test set; s2, obtaining geomagnetic library data from a geological exploration bureau website to obtain a dense matrix, performing down-sampling on the dense matrix to obtain a sparse matrix, and combining the dense matrix and the sparse matrix to construct a geomagnetic training set; s3, inputting the geomagnetic training set into a super-resolution network composed of a generator and a discriminator for training; and S4, inputting the geomagnetic test set into the generator to obtain an output result, and further constructing a high-precision geomagnetic reference map. The geomagnetic reference map construction system based on the super-resolution network can more flexibly adapt to data of different scales and complexities, improves the accuracy of geomagnetic reference map construction, and has smaller mean square error and higher precision compared with a traditional Kriging interpolation method.

Description

Technical Field

[0001] The present invention relates to the technical field of geomagnetic navigation, and particularly to a method for constructing a geomagnetic reference map based on a super-resolution network. Background Art

[0002] In the past few decades, significant progress has been made in positioning technologies, among which inertial navigation technology and satellite positioning technology are the main research directions. However, the gyroscopes used in inertial navigation have the problem of cumulative drift error over time, and satellite signals are easily affected by external factors such as terrain and climate, making it difficult to further improve the navigation positioning accuracy. Considering that the geomagnetic field is relatively stable and less affected by objective environmental factors, geomagnetic navigation technology has gradually become an effective auxiliary positioning method, and constructing a high-precision geomagnetic reference map is a key step to achieve accurate and stable navigation.

[0003] In recent years, interpolation methods have been mainly used to construct geomagnetic reference maps. Common methods include inverse distance weighted interpolation, radial basis function method, and Kriging interpolation based on particle swarm optimization algorithm. Although these algorithms have improved in terms of the root mean square error evaluation index in the process of constructing geomagnetic reference maps, it is still difficult to accurately construct the detailed information between magnetic measurement points. Therefore, we propose a method for constructing a geomagnetic reference map based on a super-resolution network to solve the above problems. Summary of the Invention

[0004] (1) Technical Problems to be Solved

[0005] In view of the deficiencies of the prior art, the present invention provides a method for constructing a geomagnetic reference map based on a super-resolution network, which solves the problems raised in the above background art.

[0006] (2) Technical Solutions

[0007] In order to achieve the above object, the present invention specifically adopts the following technical solutions:

[0008] A method for constructing a geomagnetic reference map based on a super-resolution network includes the following steps:

[0009] S1: Use a geomagnetic acquisition device to collect geomagnetic data points in a rectangular area, form sparse acquisition data, and construct a sparse matrix as a geomagnetic test set;

[0010] S2: Obtain geomagnetic library data from the website of the Geological Survey, extract geomagnetic data points in several rectangular areas of the same size to form a dense matrix, and perform bicubic interpolation downsampling on these matrices to obtain a sparse matrix. Combine the above dense matrix and sparse matrix to construct a geomagnetic training set;

[0011] S3: The sparse matrix I in the geomagnetic training set LRInput into the generator composed of feature extraction module, feature enhancement module and upsampling module, and get the dense matrix I SR , and then I SR Compared with the dense matrix I in the original geomagnetic training set HR The input is sent to the discriminator composed of convolutional layers, deep convolution modules, and activation functions for judgment. If the dense matrix generated by the generator is I HR , then update the weights to continue training; otherwise, the training ends. During this period, the generator and the discriminator are trained alternately to minimize the loss function, so that the generator generates a dense matrix;

[0012] S4: Input the sparse matrix composed of geomagnetic data points in the geomagnetic test set into the trained generator to obtain a dense matrix, and prepare a high-precision geomagnetic reference map corresponding to the acquisition area.

[0013] Furthermore, the geomagnetic data in the geomagnetic test set in S1 is derived from a geomagnetic data acquisition device, using a Waite WT10F Gauss meter with a sensitivity of 0.0001Gs, which is suitable for measuring the magnetic field strength of the earth environment; a certain open rectangular area is selected to measure the magnetic field strength, and the abnormal values ​​in the measurement process are eliminated and re-measured, and finally the average value of three repeated measurements is selected as the sample point; the sample points collected include three components of the magnetic field B x , B y , B z , the total magnetic field strength expression is: And extract geomagnetic data points of size 64×64 to form a sparse matrix as the geomagnetic test set.

[0014] Furthermore, the geomagnetic data in the geomagnetic training set in S2 comes from the website of the Geological Survey Bureau, and a number of rectangular area magnetic field value points of size 128×128 are extracted to form a matrix; the matrix is ​​downsampled using bicubic interpolation to obtain a matrix of size 64×64, and the two are combined to construct the geomagnetic training set.

[0015] Furthermore, the bicubic interpolation formula is: In the formula, I(x+i,y+j) represents the magnetic field value matrix of the original data at different longitudes and latitudes; ω i,j is the weight calculated based on the bicubic interpolation kernel function; P(x,y) is the matrix of magnetic field values ​​at different longitudes and latitudes after downsampling.

[0016] Furthermore, the specific content of inputting the geomagnetic training set into the generator and the discriminator for training in S3 is: the sparse matrix I in the geomagnetic training set LR Input to the Conv layer in the feature extraction module of the generator G for feature extraction to obtain the output feature matrix F 1 , the expression is: F1 =Conv(I LR ,ω 1 )+b 1 ,ω 1 is the convolution kernel weight, b 1 is the bias term, setting the convolution kernel size to 3×3 and the step size to 1; for the convolution layer output F 1 Use the PReLU activation function in the feature extraction module for nonlinear mapping to obtain the activated feature matrix F 2 , the expression is: F 2 =PReLU(F 1 );

[0017] The obtained feature matrix F 2 The feature enhancement module consisting of two Conv layers, two PReLU layers, one BN layer, and one sum layer is sent to the output, and the number of times is set to 4. The process is repeated to efficiently learn complex features and stabilize the learning process. The feature matrix F is obtained. 3 , where the convolution kernel size is set to 3×3 and the step size is 1; Conv layer, PReLU layer, and Sum layer are added between the feature enhancement module and the upsampling module to the feature matrix F 3 Processing to obtain the feature matrix F 6 ; The expression is: F 6 =F 5 +I LR , where F 4 =Conv(F 3 ,ω 1 )+b 1 , F 5 =PReLU(F 4 ), set the convolution kernel size to 1×1 to preserve the input feature information, increase the nonlinear expression of the feature matrix, and accelerate the learning of the network through residual connections to ensure that information is not lost during the upsampling process;

[0018] F 6 The input is sent to the upsampling module consisting of the Conv layer, the PixelShuffer layer, and the PReLU layer, and this process is repeated. Finally, the output result I is obtained through the 9×9 convolution kernel. SR ;

[0019] The output result I in the generator SR and the matrix I composed of the original dense geomagnetic data points HR The data are input into the Conv layer of the discriminator D for convolution processing with a convolution kernel of 1×1. Then the obtained feature matrix is ​​activated by the LeakyReLU function to output a nonlinear feature matrix.

[0020] The obtained nonlinear feature matrix is ​​input into the deep convolution module composed of Conv layer, BN layer and LeakyReLU layer for processing, and the number of times is set to 7, and the process is repeated, where the convolution kernel is 3×3;

[0021] Construct a denselayer fully connected layer, pass it through a LeakyReLU activation function, and then pass it through a denselayer fully connected layer. The result is nonlinearly activated by the sigmoid function; the output result is 0 or 1, which is used to determine whether the dense matrix generated by the generator is I HR Or I SR ;

[0022] The loss function expression of the generator is: Where L G represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generative mapping function, and E represents the expectation. Representative I LR It obeys P distribution.

[0023] The loss function of the discriminator is the adversarial loss function, expressed as:

[0024]

[0025] Where L D represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generative mapping function, and E represents the expectation. represents the distribution of training data, I HR ~P train (I HR ) represents I HR It follows P distribution, represents the data distribution after the generator, Representative I LR It obeys P distribution.

[0026] Furthermore, the parameter correction linear unit of the PReLU activation function is an improved version of ReLU, and the mathematical expression of the PReLU activation function is:

[0027]

[0028] Among them, a i is a constant, x i is the input, y i is the output.

[0029] Furthermore, the LeakyReLU activation function is an improved version of Relu, which introduces a negative non-zero gradient, and its mathematical expression is:

[0030]

[0031] Among them, a i is a fixed parameter in the interval (1, +∞), x i is the input, y i is the output result.

[0032] Furthermore, the mathematical expression of the Sigmoid function is:

[0033]

[0034] Where x represents input and y represents output.

[0035] (III) Beneficial effects

[0036] Compared with the prior art, the present invention provides a method for constructing a geomagnetic reference map based on a super-resolution network, which has the following beneficial effects:

[0037] The traditional Kriging interpolation method has parameter selection problems when processing data from different regions. The geomagnetic reference map construction system based on the super-resolution network can more flexibly adapt to data of different scales and complexities, improving the accuracy of constructing the geomagnetic reference map. In addition, the super-resolution network can better capture the detailed information between magnetic measurement points, including local features and small changes, thereby achieving a more accurate geomagnetic reference map construction, with a smaller mean square error and higher accuracy than the traditional Kriging interpolation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the method steps of the present invention;

[0039] Figure 2 It is a block diagram of the geomagnetic data acquisition system of the present invention;

[0040] Figure 3 It is a schematic diagram of the generator structure of the present invention;

[0041] Figure 4 Schematic diagram of the discriminator structure of the present invention;

[0042] Figure 5 It is a structural schematic diagram of a feature extraction module in the generator of the present invention;

[0043] Figure 6 It is a structural schematic diagram of a feature enhancement module in the generator of the present invention;

[0044] Figure 7 Schematic diagram of the structure of the upsampling module in the generator of the present invention;

[0045] Figure 8Schematic diagram of the structure of the deep convolution module in the discriminator of the present invention;

[0046] Fig. 9 It is a schematic diagram comparing the method proposed in the present invention with the Kriging interpolation method. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example

[0049] like Figure 1-9 As shown, a system for constructing a geomagnetic reference map based on a super-resolution network includes the following steps:

[0050] S1: Use geomagnetic acquisition equipment to collect geomagnetic data points in a rectangular area to form sparse acquisition data and construct a sparse matrix as a geomagnetic test set.

[0051] The geomagnetic data in the geomagnetic test set used are from Figure 2 The geomagnetic data acquisition device shown uses a Waite WT10F Gaussmeter with a sensitivity of 0.0001Gs (approximately equal to 10nT), which is suitable for measuring the magnetic field strength of the earth environment. Figure 2 The magnetic sensor and conditioning circuit shown in the figure are used for the detection and conditioning of geomagnetic component signals; the A / D conversion module is responsible for real-time acquisition of analog signals and digital conversion; the memory module is used for real-time storage of signals; the FPGA controls the working mode of the acquisition, storage and communication modules and coordinates the operation of each module; the communication interface module realizes data feedback and transmission with the host computer; the power supply module converts the input power into the voltage required by each module and provides power support. Select an open rectangular area to measure the magnetic field strength, eliminate the abnormal values ​​in the measurement process and re-measure, and finally select the average value of three repeated measurements as the sample point. The sample points collected include three components of the magnetic field B x , B y , B z , the total magnetic field strength expression is: Furthermore, geomagnetic data points of size 64×64 are extracted to form a sparse matrix as a geomagnetic test set.

[0052] S2: Obtain geomagnetic database data from the website of the Geological Survey Bureau, extract geomagnetic data points in several rectangular areas of the same size, form dense matrices, perform bicubic interpolation downsampling on these matrices to obtain sparse matrices. The above dense matrices and sparse matrices are combined to construct a geomagnetic training set.

[0053] The geomagnetic data in the geomagnetic training set used in the present invention comes from the Geological Survey Bureau, and a number of 128×128 rectangular area magnetic field value points are extracted to form a matrix. The matrix is ​​downsampled using bicubic interpolation to obtain a 64×64 matrix, and the two are combined to form a training set. The bicubic interpolation formula is: In the formula, I(x+i,y+j) represents the magnetic field value matrix of the original data at different longitudes and latitudes; ω i,j is the weight calculated based on the bicubic interpolation kernel function; P(x,y) is the matrix of magnetic field values ​​at different longitudes and latitudes after downsampling.

[0054] S3: The sparse matrix I in the geomagnetic training set LR Input into the generator composed of feature extraction module, feature enhancement module and upsampling module, and get the dense matrix I SR , and then I SR Compared with the dense matrix I in the original geomagnetic training set HR The input is sent to the discriminator composed of convolutional layers, deep convolution modules and activation functions for judgment. If the dense matrix generated by the generator is I HR , then update the weights to continue training; otherwise, the training ends. During this period, the generator and the discriminator are trained alternately to minimize the loss function, so that the generator generates a dense matrix.

[0055] The sparse matrix I in the geomagnetic training set LR Input Figure 3 The Conv layer in the feature extraction module of the generator G shown in the figure extracts features to obtain the output feature matrix F 1 , the expression is: F 1 =Conv(I LR ,ω 1 )+b 1 ,ω 1 is the convolution kernel weight, b 1 is the bias term, setting the convolution kernel size to 3×3 and the step size to 1. 1 Use as Figure 5 The PReLU activation function in the feature extraction module shown in the figure performs nonlinear mapping to obtain the activated feature matrix F 2 , the expression is: F 2 =PReLU(F 1). The parameter correction linear unit of the PReLU activation function is an improved version of ReLU. The mathematical expression of the PReLU activation function is:

[0056]

[0057] Among them, a i is a small constant, x i is the input, y i is the output.

[0058] The obtained feature matrix F 2 It is fed into a layer consisting of two Conv layers, two PReLU layers, a BN layer, and a sum layer. Figure 6 The feature enhancement module shown in the figure is set to 4 times and the process is repeated to efficiently learn complex features and stabilize the learning process. The feature matrix F is output. 3 , where the convolution kernel size is set to 3×3 and the step size is 1.

[0059] Add Conv layer, PReLU layer and Sum layer between the feature enhancement module and the upsampling module to the feature matrix F 3 Processing to obtain the feature matrix F 6 , the expression is: F 6 =F 5 +I LR , where F 4 =Conv(F 3 ,ω 1 )+b 1 , F 5 =PReLU(F 4 ), where the convolution kernel size is set to 1×1 to preserve the input feature information, increase the nonlinear expression of the feature matrix, and accelerate the learning of the network through residual connections to ensure that information is not lost during the upsampling process.

[0060] F 6 Input to the Conv layer, PixelShuffer layer, PReLU layer, etc. Figure 7 The upsampling module shown in the figure is repeated, and the output result I is finally obtained through a 9×9 convolution kernel. SR .

[0061] The output result I in the generator SR and the matrix I consisting of dense geomagnetic data points in the original geomagnetic training set HR Enter together into Figure 4The Conv layer in the discriminator D shown performs convolution processing with a convolution kernel of 1×1. Then, the obtained feature matrix is passed through the LeakyReLU activation function to output a non-linear feature matrix. The LeakyReLU activation function is an improved version of Relu, introducing a negative non-zero gradient, and its mathematical expression is:

[0062]

[0063] where a i is a fixed parameter in the interval (1, +∞), x i is the input, and y i is the output result.

[0064] The obtained non-linear feature matrix is input into the deep convolution module composed of a Conv layer, a BN layer, and a LeakyReLU layer as Figure 8 shown for processing, and this process is repeated 7 times. The convolution kernels are all 3×3.

[0065] The output result of the previous step is input into a denselayer fully connected layer and a LeakyReLU layer, and then through another denselayer fully connected layer; introducing non-linearity and complexity to help the discriminator learn more complex and abstract features and improve the ability to distinguish between real data and generated data. The obtained result is non-linearly activated through the sigmoid function; the output result is 0 or 1, used to determine whether the input matrix comes from the generation network or the geomagnetic training set. The mathematical expression of the sigmoid function is:

[0066]

[0067] where x represents the input and y represents the output.

[0068] The expression of the loss function of the generator is: where L G represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generation mapping function, E represents the expectation, represents I LR obeys the P distribution.

[0069] The loss function of the discriminator is the adversarial loss function, and the expression is:

[0070]

[0071] where L D represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generation mapping function, E represents the expectation, represents the distribution of the training data, I HR ~Ptrain (I HR ) represents I HR obeys the P distribution, represents the data distribution after passing through the generator, represents I LR obeys the P distribution.

[0072] The discriminator performs 1000 iterations, and the generator also performs 1000 iterations. The generator and the discriminator are alternately trained to minimize the loss function, so that the generator generates a dense matrix.

[0073] S4: Input the sparse matrix composed of geomagnetic data points in the geomagnetic test set into the trained generator to obtain a dense matrix, and prepare a high-precision geomagnetic reference map corresponding to the acquisition area.

[0074] The mean square error results of comparing the method in the present invention with the Kriging interpolation method are shown in the following table, and the comparison graph is as Fig. 9 shown.

[0075]

[0076] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing a geomagnetic reference map based on a super-resolution network, characterized in that: The steps include: S1: Use geomagnetic acquisition equipment to collect geomagnetic data points in a rectangular area to form sparse acquisition data and construct a sparse matrix as a geomagnetic test set; S2: Obtain geomagnetic database data from the website of the Geological Survey Bureau, extract geomagnetic data points in several rectangular areas of the same size to form a dense matrix, perform bicubic interpolation downsampling on these matrices to obtain sparse matrices; combine the above dense matrix and sparse matrix to construct a geomagnetic training set; S3: The sparse matrix I in the geomagnetic training set LR Input into the generator composed of feature extraction module, feature enhancement module and upsampling module, and get the dense matrix I SR , and then I SR Compared with the dense matrix I in the original geomagnetic training set HR The input is sent to the discriminator composed of convolutional layers, deep convolution modules, and activation functions for judgment. If the dense matrix generated by the generator is I HR , then update the weights to continue training; otherwise, the training ends; during this period, the generator and the discriminator are trained alternately to minimize the loss function so that the generator generates a dense matrix; S4: Input the sparse matrix composed of geomagnetic data points in the geomagnetic test set into the trained generator to obtain a dense matrix, and prepare a high-precision geomagnetic reference map corresponding to the acquisition area.

2. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 1, characterized in that: The geomagnetic data in the geomagnetic test set in S1 comes from the geomagnetic data acquisition equipment, using the Waite WT10F Gauss meter with a sensitivity of 0.0001Gs, which is suitable for measuring the magnetic field strength of the earth environment; Select an open rectangular area to measure the magnetic field strength, remove the abnormal values ​​in the measurement process and re-measure, and finally select the average value of three repeated measurements as the sample point; the sample points collected include the three components of the magnetic field B x , B y , B z , the total magnetic field strength expression is: And extract geomagnetic data points of size 64×64 to form a sparse matrix as the geomagnetic test set.

3. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 1, characterized in that: The geomagnetic data in the geomagnetic training set in S2 comes from the website of the Geological Survey Bureau. A number of 128×128 rectangular area magnetic field value points are extracted to form a matrix. The matrix is ​​downsampled using bicubic interpolation to obtain a 64×64 matrix. The two are combined to form a geomagnetic training set.

4. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 3, characterized in that: The bicubic interpolation formula is: In the formula, I(x+i,y+j) represents the magnetic field value matrix of the original data at different longitudes and latitudes; ω i,j is the weight calculated based on the bicubic interpolation kernel function; P(x,y) is the matrix of magnetic field values ​​at different longitudes and latitudes after downsampling.

5. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 1, characterized in that: The specific content of inputting the geomagnetic training set into the generator and the discriminator for training in S3 is: the sparse matrix I in the geomagnetic training set LR The Conv layer in the feature extraction module of the generator G is input to extract features to obtain the output feature matrix F1, which is expressed as: F1 = Conv(I LR ,ω1)+b1, ω1 is the convolution kernel weight, b1 is the bias term, the convolution kernel size is set to 3×3, and the step size is 1; the PReLU activation function in the feature extraction module is used to perform nonlinear mapping on the convolution layer output F1 to obtain the activated feature matrix F2, the expression is: F2=PReLU(F1); The obtained feature matrix F2 is sent to a feature enhancement module consisting of two Conv layers, two PReLU layers, a BN layer, and a sum layer, and the number of times is set to 4. The process is repeated to efficiently learn complex features and stabilize the learning process. The output is the feature matrix F3, where the convolution kernel size is set to 3×3 and the step size is 1; Add Conv layer, PReLU layer and Sum layer between the feature enhancement module and the upsampling module to process the feature matrix F3 to obtain the feature matrix F6; the expression is: F6 = F5 + I LR , where F4 = Conv(F3,ω1)+b1, F5 = PReLU(F4), and the convolution kernel size is set to 1×1 to maintain the input feature information, increase the nonlinear expression of the feature matrix, and accelerate the learning of the network through residual connections to ensure that information is not lost during the upsampling process; F6 is input to the upsampling module consisting of the Conv layer, the PixelShuffer layer, and the PReLU layer, and this process is repeated. Finally, the output result I is obtained through the 9×9 convolution kernel. SR ; The output result I in the generator SR and the matrix I composed of the original dense geomagnetic data points HR The data are input into the Conv layer of the discriminator D for convolution processing with a convolution kernel of 1×1. Then the obtained feature matrix is ​​activated by the LeakyReLU function to output a nonlinear feature matrix. The obtained nonlinear feature matrix is ​​input into the deep convolution module composed of Conv layer, BN layer and LeakyReLU layer for processing, and the number of times is set to 7, and the process is repeated, where the convolution kernel is 3×3; Construct a denselayer fully connected layer, pass it through a LeakyReLU activation function, and then pass it through a denselayer fully connected layer, and the result is nonlinearly activated by the sigmoid function; The output result is 0 or 1, which is used to determine whether the dense matrix generated by the generator is I HR Or I SR ; The loss function expression of the generator is: Where L G represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generative mapping function, and E represents the expectation. Representative I LR Obey P distribution; The loss function of the discriminator is the adversarial loss function, expressed as: Where L D represents the adversarial loss function, D(·) represents the discriminant mapping function, G(·) represents the generative mapping function, and E represents the expectation. represents the distribution of training data, I HR ~P train (I HR ) represents I HR It follows P distribution, represents the data distribution after the generator, Representative I LR It obeys P distribution.

6. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 5, characterized in that: The parameter correction linear unit of the PReLU activation function is an improved version of ReLU. The mathematical expression of the PReLU activation function is: Among them, a i is a constant, x i is the input, y i is the output.

7. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 5, characterized in that: The LeakyReLU activation function is an improved version of Relu, which introduces negative non-zero gradients. Its mathematical expression is: Among them, a i is a fixed parameter in the interval (1, +∞), x i is the input, y i is the output result.

8. The method for constructing a geomagnetic reference map based on a super-resolution network according to claim 5, characterized in that: The mathematical expression of the Sigmoid function is: Where x represents input and y represents output.

Images