Magnetic interference compensation method and system cooperatively driven by model data

Abstract

The invention discloses a magnetic interference compensation method and system cooperatively driven by model data, and relates to the field of magnetic field compensation, and the method comprises the steps: obtaining a magnetic anomaly measurement value, and measuring a direction cosine and an analog signal of a magnetic field; the analog signal is generated by an electronic device in the aircraft; on the basis of the BP neural network, combining a T-L model to construct an interference magnetic field determination model; inputting the magnetic anomaly measurement value, the direction cosine of the measurement magnetic field and the analog signal into an interference magnetic field determination model to obtain an interference magnetic field; and performing interference compensation on the magnetic anomaly measurement value based on the interference magnetic field. The method breaks through the normal form barrier of a traditional calibration method and a pure data driving model, can achieve the deep coupling of the intrinsic feature extraction of the magnetic field and the physical constraint of the interference field, and improves the compensation precision and generalization capability of the interference magnetic field.

Description

Technical Field

[0001] The present application relates to the field of magnetic field compensation, and in particular to a magnetic interference compensation method and system driven by collaborative model data. Background Art

[0002] In recent years, autonomous navigation technology in satellite-denied environments has become a research focus. Geomagnetic navigation, with its passive nature, 24 / 7 operation, and low cost, has demonstrated significant application value. However, magnetometer data not only contains the actual geomagnetic field but also magnetic interference caused by other magnetic materials. This significantly reduces the accuracy of geomagnetic field measurements and severely limits navigation reliability.

[0003] Currently, research on aeromagnetic compensation focuses on two main approaches: hard compensation and soft compensation. Hard compensation methods are time-consuming, labor-intensive, costly, and ineffective. For example, the commonly used telescopic rod technique struggles to achieve sufficient distance on an aircraft to protect the magnetometer mounted on the rod from interference from the cabin's magnetic field. Currently, the most commonly used soft compensation method is the Tolles-Lawson model and its optimization.

[0004] The Tolles-Lawson model, abbreviated as the TL model, divides the carrier's interfering magnetic field into three parts: a constant magnetic field, an induced magnetic field, and an eddy current magnetic field. The model uses the angle between the measured magnetic field vector and the coordinate axes of the carrier coordinate system to calculate the direction cosines, and uses the first-order terms, derivative terms, and cross-coupling terms of the three direction cosines to model the above three magnetic fields respectively, so as to estimate the strength of the carrier's interfering magnetic field. The TL model is an idealized model, and its interfering magnetic field only considers the constant magnetic field, the induced magnetic field, and the eddy current magnetic field, and its performance is poor in complex environments. In addition, due to the correlation of aeromagnetic data, the product of the information matrix and its transpose may degenerate into a singular matrix, resulting in the compensation parameter vector having no unique solution, which in turn leads to poor compensation accuracy and generalization ability of the interfering magnetic field. Summary of the Invention

[0005] In order to solve the above-mentioned problems existing in the prior art, the present application provides a magnetic interference compensation method and system driven by model data collaboration.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a magnetic interference compensation method driven by model data collaboration, comprising:

[0008] Obtaining magnetic anomaly measurements, direction cosines of the measured magnetic field, and analog signals; the analog signals are generated by electronic equipment in the aircraft;

[0009] On the basis of the BP neural network, a model for determining the interference magnetic field is constructed in combination with the TL model;

[0010] Inputting the magnetic anomaly measurement value, the direction cosine of the measured magnetic field and the analog signal into the interference magnetic field determination model to obtain the interference magnetic field;

[0011] Interference compensation is performed on the magnetic anomaly measurement value based on the interfering magnetic field.

[0012] Optionally, based on the BP neural network, a TL model is combined to construct an interference magnetic field determination model, including:

[0013] A first network and a second network are constructed based on the BP neural network; the first network is used to obtain a predicted value of a compensation parameter vector of a TL model based on input data; the second network is used to obtain a random interference factor term based on the input data; the random interference factor term is used to optimize an interference magnetic field determined by the TL model;

[0014] The predicted value of the compensation parameter vector output by the first network and the random interference factor term output by the second network are coupled to complete the construction of the interference magnetic field determination model.

[0015] Optionally, the model data collaboratively driven magnetic interference compensation method further includes: in the process of training the first network, using a ridge regression method to determine initial compensation parameters of the TL model to initialize the first network.

[0016] Alternatively, the formula y=Ac is used i +r i coupling the predicted value of the compensation parameter vector output by the first network and the random interference factor term output by the second network;

[0017] Where y is the interference magnetic field, A is the information matrix, c i is the predicted value of the compensation parameter vector, r i is the random interference factor term.

[0018] Optionally, drawing on the characteristics of the elastic network, an L1 regularization term and an L2 regularization term are simultaneously added to the loss function of the interference magnetic field determination model;

[0019] The loss function of the interference magnetic field determination model is expressed as:

[0020]

[0021] In the formula, loss is the loss function value, H m is the magnetic anomaly measurement value, y is the interference magnetic field, H labelis the magnetic anomaly label value, λ1 is the L1 regularization coefficient, λ2 is the L2 regularization coefficient, w is the weight parameter of the interference magnetic field determination model, ||w||1 is the L1 regularization term, is the L2 regularization term.

[0022] Optionally, the first network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer;

[0023] In the first network, the feature ratio between the input layer, the first hidden layer, the second hidden layer, the third hidden layer and the output layer is 26:200:120:70:18.

[0024] Optionally, the second network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and an output layer;

[0025] In the second network, the feature ratio among the input layer, the first hidden layer, the second hidden layer, the third hidden layer, the fourth hidden layer and the output layer is 26:120:60:40:20:1.

[0026] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the magnetic interference compensation method collaboratively driven by model data provided above.

[0027] In the third aspect, the present application provides a magnetic interference compensation system driven by collaborative model data, comprising: a processor and a memory; the processor is connected to the memory; a computer program is stored on the memory; the processor executes the computer program to implement the magnetic interference compensation method driven by collaborative model data as described in any one of claims 1-7.

[0028] Optionally, the memory is a computer-readable storage medium.

[0029] According to the specific embodiments provided in this application, this application has the following technical effects:

[0030] The present application provides a magnetic interference compensation method and system driven by model data collaboration. For the problem of carrier interference magnetic field compensation, by combining the TL model with the BP neural network to construct an interference magnetic field determination model, this magnetic interference collaborative compensation architecture based on model data hybrid drive realizes deep coupling of magnetic field intrinsic feature extraction and interference field physical constraints, and can break through the paradigm barriers of traditional calibration methods and pure data-driven models. Moreover, by adopting the constructed magnetic interference collaborative compensation architecture based on model data hybrid drive to determine the interference magnetic field compensation, so as to complete the compensation of the interference magnetic field, the compensation accuracy and generalization ability of the interference magnetic field compensation can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 A flowchart of a magnetic interference compensation method driven by model data collaboration provided in one embodiment of the present application;

[0033] Figure 2 A schematic diagram of a carrier coordinate system provided in one embodiment of the present application;

[0034] Figure 3 A schematic diagram of the training process of the interference magnetic field determination model provided in one embodiment of the present application;

[0035] Figure 4 A schematic diagram of a data processing process for an interference magnetic field determination model provided in one embodiment of the present application;

[0036] Figure 5 A schematic structural diagram of a magnetic interference compensation system driven collaboratively by model data provided in one embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0038] This application adopts a soft compensation approach. First, a neural network is used to solve the compensation parameter vector of the TL model. On the one hand, this breaks the paradigm of traditional model solution and improves the expressiveness of linear models in complex environments. On the other hand, it can prevent the complex collinearity problem caused by the strong correlation of aeromagnetic data. Second, the strong nonlinear characteristics of the neural network are used to introduce a random interference factor term to improve the shortcomings of the traditional TL model, enabling it to have better compensation capabilities in complex cabin environments.

[0039] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0040] In an exemplary embodiment, the present application provides a magnetic interference compensation method driven by model data collaboration, which is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to a server as an example for explanation. Figure 1 As shown, the method includes:

[0041] Step 100: Obtain magnetic anomaly measurement values, direction cosines of the measured magnetic field, and analog signals. The analog signals are generated by electronic equipment in the aircraft.

[0042] Step 101: Based on the BP neural network and combined with the TL model, an interference magnetic field determination model is constructed.

[0043] Step 102: Input the magnetic anomaly measurement value, the direction cosine of the measured magnetic field, and the analog signal into an interference magnetic field determination model to obtain an interference magnetic field.

[0044] Step 103: Perform interference compensation on the magnetic anomaly measurement value based on the interfering magnetic field.

[0045] By implementing the above steps 100 to 103, the present application can achieve deep coupling of magnetic field intrinsic feature extraction and interference field physical constraints, break through the paradigm barriers of traditional calibration methods and pure data-driven models, and improve the compensation accuracy and generalization ability of interference magnetic field compensation.

[0046] In another exemplary embodiment of the present application, considering that the TL model is an idealized physical model, it only considers the constant magnetic field, induced magnetic field and eddy current magnetic field, and does not consider nonlinear components such as random magnetic fields without physical laws generated in airborne equipment. This may cause the magnetometer installed in the cabin to be unable to use the TL model to compensate for the measurement data normally. In order to solve this problem, the present application uses the BP neural network output to obtain a random interference factor term to simulate the random magnetic field generated in the cabin, and introduces it into the modeling of the interference magnetic field. In addition, the use of neural network training to obtain the compensation parameter vector can, on the one hand, break the barrier of the TL model to solve the compensation parameters and improve the expression ability of the linear model in complex environments. On the other hand, it can prevent possible complex collinearity problems. Based on this, in this embodiment, the implementation process of step 101 provided above in the present application can be replaced by the following steps 1 and 2.

[0047] Step 1: Based on BP neural network, construct two sub-networks, the first network and the second network, with the same input. Figure 4 The network 1 in FIG is used to obtain the predicted value of the compensation parameter vector of the TL model based on the input data. The first network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer. The second network (i.e. Figure 4 Network 2) in [ 1 ] is used to obtain a random interference factor term based on the input data. The random interference factor term is used to optimize the interfering magnetic field determined by the TL model. The second network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and an output layer.

[0048] Step 2: Couple the predicted value of the compensation parameter vector output by the first network and the random interference factor term output by the second network to complete the construction of the interference magnetic field determination model.

[0049] In another exemplary embodiment of the present application, in order to improve the performance of the model, the present application uses the USAF-MIT (Magnetic Signal Enhancement Challenge) open source data as the data basis to construct a data set. This data set is used to train the first network and the second network provided above in this application. Since the two sub-networks are trained in the same way, they are collectively referred to as training the neural network below. Among them, the USAF-MIT open source data was published in 2020, which includes the flight information of the carrier, magnetometer data and various sensor data.

[0050] Based on the above description, if Figure 3As shown in the figure, during the neural network training process, feature data from five time points (time T, time T-1, time T-2, time T-3, and time T-4) was selected to fit the output at the final time point. Nine features were selected at each time point, corresponding to the magnetic anomaly measurement, the direction cosines of the measured magnetic field, and the sensor signal. The magnetic anomaly measurement is calculated by subtracting the main magnetic field and diurnal magnetic field from the uncompensated total magnetic field of scalar magnetometer No. 4 (installed on the aircraft's rear cabin floor). The direction cosines, represented by flux_b_x, flux_b_y, and flux_b_z in the dataset, are the three-component magnetic field strength of the vector magnetometer and can provide richer geometric information. The sensor signals, cur_strb, cur_ac_hi, cur_heat, vol_bat_1, and vol_block in the dataset, are analog signals generated by electronic equipment in the aircraft and are used to simulate the complex airborne equipment environment. Five sensor signals were selected based on the voltages and currents generated by common cabin activities, such as the inertial navigation system, air conditioning, and strobe lights. By introducing changes in these voltages and currents, the complex interference sources in the cabin are mapped into the expression of the TL model parameters.

[0051] The feature data from the five selected moments was first normalized and then dimensionality reduced using principal component analysis. The number of principal components that achieved a cumulative explained variance of 99.99% was selected. This reduced the number of features from 45 to 26, allowing for neural network training. The neural network training process used features from the current moment (i.e., moment T) and the previous moments (i.e., moments T-1, T-2, T-3, and T-4) to fit the label value at the current moment.

[0052] During the training process, the TL model compensation parameter vector containing 18 coefficients is obtained through the first network. After the compensation parameter vector is multiplied by the information matrix of known data, the following is obtained: Figure 2The carrier coordinate system shown includes an interfering magnetic field comprising a constant magnetic field, an induced magnetic field, and an eddy current magnetic field. At the same time, in order to increase the training speed of the network and improve the overall compensation effect, the compensation parameter vector of the TL model obtained by the ridge regression method is used in the first network to initialize the network, that is, the compensation parameter vector is used as the initial output in the first network training iteration process. In this embodiment, the structure of the first network is 26:200:120:70:18; the second network obtains a random interference factor term to further optimize the composition of the interfering magnetic field in the TL model, making it perform better in complex environments. The structure of the second network is 26:120:60:40:20:1; the final output of the neural network is an interfering magnetic field, which includes a constant magnetic field, an induced magnetic field, an eddy current magnetic field, and a random magnetic field.

[0053] Furthermore, during training, the first network and the second network set common network parameters, wherein the learning rate is 0.0001, the batch size is 128, the number of iterations is 300, the optimizer is the Adam optimizer, and the activation function is a smooth and non-monotonic Swish function.

[0054] In another exemplary embodiment of the present application, Figure 2 As shown, H m is the magnetic anomaly measurement value (i.e., measured magnetic field), H e is the Earth's magnetic field, H i is the interference magnetic field, X, Y, and Z are the angles between the measured magnetic field and the three axes of the carrier coordinate system. Based on this, the relationships shown in equations (1) to (3) can be obtained.

[0055] H m =H e +H i (1)

[0056]

[0057] Because |H i | 2 / |H m | 2 is very small and can be ignored, and according to Taylor expansion Further simplifying formula (3) yields formula (4), where x is an unknown number.

[0058]

[0059] Where, To measure the direction cosines of the magnetic field, is the interfering magnetic field in the carrier coordinate system, ε=|H m |-|H e |.

[0060] Based on the ideal assumption, the composition of the interfering magnetic field in the TL model is shown in Equation (5).

[0061] H i =H per +H inc +H edd (5)

[0062] Where H per is a constant magnetic field, H inc is the induced magnetic field, H edd is the eddy current magnetic field, where:

[0063]

[0064]

[0065] Where, is the first-order derivative of the transposed direction cosine of the measured magnetic field, and a*, b*, and c* correspond to the 18 coefficients of the TL model compensation parameter vector.

[0066] Substituting equation (5) into equation (4), we can obtain equation (9).

[0067]

[0068] Simplifying formula (9) can give the form of formula (10).

[0069] ε=Ac (10)

[0070]

[0071] c=[a1,a2,a3,b 11 ,(b 12 +b 21 ),(b 13 +b 31 ),b 22 ,(b 23 +b 32 ),b 33 , c 11 , c 12 ,c 13 ,c 21 ,c 22 ,c 23 ,c 31 ,c 32 ,c 33 ] T (12)

[0072] Where A is the information matrix, which is all known information. c is the compensation parameter vector, which is all unknown data. , cos′X is the first-order derivative of cosX. b12 +b 21 The term can be abbreviated as b 12 , the remaining terms can be simplified in the same way, and a vector containing 18 unknowns is obtained, as shown in formula (13).

[0073] c=[a1,a2,a3,b 11 ,b 12 ,b 13 ,b 22 ,b 23 ,b 33 ,c 11 , c 12 ,c 13 ,c 21 ,c 22 ,c 23 ,c 31 ,c 32 ,c 33 ] T (13)

[0074] As shown in formula (14), when using the TL model for compensation, it is necessary to solve the compensation parameter vector c and then calculate the interference magnetic field to achieve the purpose of compensation.

[0075] c=(A T A) -1 A T ε (14)

[0076] Based on the above description, the first network can obtain the compensation parameter vector. The interference magnetic field including the constant magnetic field, induced magnetic field, and eddy current magnetic field in the carrier coordinate system can be calculated through the above formula (10). The second network can obtain the random interference factor term in addition to the interference magnetic field in the TL model. The outputs of the first and second networks are combined to obtain the output interference magnetic field y, as shown in formula (15), and the loss function shown in formula (16) is constructed.

[0077] y=Ac i +r i (15)

[0078]

[0079] In the formula, loss is the loss function value, H label is the magnetic anomaly label value, which is calculated by subtracting the main magnetic field and diurnal magnetic field from the compensated total magnetic field of scalar magnetometer No. 1 (the scalar magnetometer installed on the tail pin), λ1 is the L1 regularization coefficient, and λ2 is the L2 regularization coefficient. i is the predicted value of the compensation parameter vector, r i is the random interference factor, w is the weight parameter of the interference magnetic field determination model, ||w||1 is the L1 regularization term, is the L2 regularization term.

[0080] This application draws on the characteristics of the elastic network when constructing the loss function, and incorporates both L1 and L2 regularization terms. L1 regularization penalizes the absolute value of the weights, tending to produce sparse solutions, which is suitable for feature selection; L2 regularization penalizes the square of the weights, tending to make the weights uniformly approach small values, which is suitable for preventing overfitting.

[0081] Based on the above description, the data processing flow of the interference magnetic field determination model is as follows: Figure 4 As shown in the figure, the final output result is the interference magnetic field. After obtaining the interference magnetic field, the interference magnetic field compensation can be performed on the magnetic field measurement value to obtain high-precision geomagnetic field data. On the one hand, it can be used to prepare high-precision geomagnetic maps, and on the other hand, it can better assist positioning in geomagnetic matching.

[0082] In addition, the interference sources in the cabin under complex environments can be analyzed and modeled separately, and then added to the TL model to make the TL model more performant in complex environments.

[0083] In summary, the present application uses a neural network to optimize and solve the process of solving the compensation parameter vector of the TL model, uses the strong nonlinearity of the neural network to obtain a random interference factor term, adds it to the TL model, and completes the improvement of the TL model. It can be seen that the present application analyzes the modeling mechanism and model defects of the TL model, uses the strong nonlinear characteristics of the neural network to introduce a random interference factor term to improve the shortcomings of the traditional physical model, and uses the neural network to solve the compensation parameter vector of the TL model, breaking the paradigm of traditional model solving. Through the organic combination of model-driven and data-driven, the present application improves the compensation ability and generalization ability of the traditional method.

[0084] In an exemplary embodiment, a magnetic interference compensation system driven by model data collaboration is provided. The system may be a computer device, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store interference magnetic field compensation data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a magnetic interference compensation method driven by model data collaboration is realized.

[0085] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0086] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0087] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0088] In an exemplary embodiment, a computer program product may also be provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0089] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0090] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (RRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0091] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0092] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A magnetic interference compensation method driven by model data collaboration, characterized in that: include: Obtaining magnetic anomaly measurements, direction cosines of the measured magnetic field, and analog signals; the analog signals are generated by electronic equipment in the aircraft; On the basis of BP neural network and combined with TL model, the interference magnetic field determination model is constructed; Inputting the magnetic anomaly measurement value, the direction cosine of the measured magnetic field and the analog signal into the interference magnetic field determination model to obtain the interference magnetic field; Interference compensation is performed on the magnetic anomaly measurement value based on the interfering magnetic field.

2. The magnetic interference compensation method driven by model data collaboration according to claim 1, characterized in that: Based on the BP neural network, the TL model is combined to construct an interference magnetic field determination model, including: A first network and a second network are constructed based on the BP neural network; the first network is used to obtain a predicted value of a compensation parameter vector of a TL model based on input data; the second network is used to obtain a random interference factor term based on the input data; the random interference factor term is used to optimize an interference magnetic field determined by the TL model; The predicted value of the compensation parameter vector output by the first network and the random interference factor term output by the second network are coupled to complete the construction of the interference magnetic field determination model.

3. The magnetic interference compensation method driven by model data collaboration according to claim 2, characterized in that: The model data collaboratively driven magnetic interference compensation method further includes: in the process of training the first network, using a ridge regression method to determine the initial compensation parameters of the TL model to initialize the first network.

4. The magnetic interference compensation method driven by model data collaboration according to claim 2, characterized in that: Using the formula y = Ac i +r i coupling the predicted value of the compensation parameter vector output by the first network and the random interference factor term output by the second network; Where y is the interference magnetic field, A is the information matrix, c i is the predicted value of the compensation parameter vector, r i is the random interference factor term.

5. The magnetic interference compensation method driven by model data collaboration according to claim 2, characterized in that: Drawing on the characteristics of the elastic network, the L1 regularization term and the L2 regularization term are simultaneously added to the loss function of the interference magnetic field determination model; The loss function of the interference magnetic field determination model is expressed as: In the formula, loss is the loss function value, H m is the magnetic anomaly measurement value, y is the interfering magnetic field, H label is the magnetic anomaly label value, λ1 is the L1 regularization coefficient, λ2 is the L2 regularization coefficient, w is the weight parameter of the interference magnetic field determination model, ||w||1 is the L1 regularization term, is the L2 regularization term.

6. The magnetic interference compensation method driven by model data collaboration according to claim 2, characterized in that: The first network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer; In the first network, the feature ratio between the input layer, the first hidden layer, the second hidden layer, the third hidden layer and the output layer is 26:200:120:70:

18.

7. The magnetic interference compensation method driven by model data collaboration according to claim 2, characterized in that: The second network includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and an output layer; In the second network, the feature ratio among the input layer, the first hidden layer, the second hidden layer, the third hidden layer, the fourth hidden layer and the output layer is 26:120:60:40:20:

1.

8. A magnetic interference compensation system driven by model data collaboration, characterized in that: include: processor and memory; The processor is connected to the memory; the memory stores a computer program; The processor executes the computer program to implement the magnetic interference compensation method driven by model data collaboration according to any one of claims 1 to 7.

9. The magnetic interference compensation system driven by model data collaboration according to claim 8, characterized in that: The memory is a computer-readable storage medium.

Images