A method and system for compensating aeromagnetic data of a multi-rotor UAV

Abstract

The present invention is applicable to the field of aeromagnetic data processing technology and provides a method and system for compensating aeromagnetic data for a multi-rotor unmanned aerial vehicle (UAV). The method includes the following steps: obtaining aeromagnetic data from the multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data; determining total geomagnetic field data based on the three-component fluxgate data; calculating a direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data; using a preset PINN neural network, the direction cosine matrix and the total geomagnetic field data are used as input values ​​to output a magnetic interference prediction value; and compensating the aeromagnetic data based on the magnetic interference prediction value. The PINN neural network used in the present invention, combined with the physical information contained in the data, can have a strong fitting and compensation capability, while not requiring reliance on a large amount of data support, resulting in high compensation accuracy for aeromagnetic data.

Description

Technical Field

[0001] The present invention belongs to the technical field of aeromagnetic data processing, and in particular relates to an aeromagnetic data compensation method and system for a multi-rotor unmanned aerial vehicle. Background Art

[0002] Multi-rotor drone aeromagnetic surveying involves mounting measuring instruments (optically pumped magnetometers, fluxgate magnetometers) and auxiliary equipment on a multi-rotor drone to measure magnetic data along fixed survey lines over a survey area. However, due to interference from the aerial platform and magnetic equipment, compensating for this interference field is a crucial step in aeromagnetic data processing. The effectiveness of this compensation determines the quality of subsequent data processing.

[0003] To address the problem of aeromagnetic interference compensation, Tolles et al. first classified magnetic interference based on its causes and properties into three types: a constant interference field generated by the magnetization of the aircraft's hard magnetic materials by the external magnetic field; an induced interference field generated by the magnetization of the aircraft's soft magnetic materials by the Earth's magnetic field during aircraft motion; and an eddy current interference field generated by the aeromagnetic system's loop conductors cutting through the external magnetic field during aircraft maneuvers. They then established the Tolles-Lawson (TL) equation, a classic mathematical model for magnetic interference fields. To solve the TL equation, Leliak et al. designed a set of reference maneuvers in north-east-south-west headings based on the TL equation. In the mathematical model established by Tolles and Lawson, the three interference fields were simply expressed as linear equations, which were then solved to compensate for the Earth's magnetic field. However, actual magnetic interference is complex, and the overall impact of the three interference fields cannot be simply described using linear equations. Therefore, a nonlinear neural network was introduced into the data processing to avoid multicollinearity between variables, which can lead to low model solution accuracy and thus affect the compensation results.

[0004] Traditional neural networks for solving nonlinear problems typically use BP neural networks, which take the product of 16 direction cosines and the measured total geomagnetic field strength as input and output the magnetic interference as the measured total geomagnetic field strength minus the regional geomagnetic background field information measured by geomagnetic stations. Although this interprets the data as nonlinear and compensates for it through backpropagation, the BP neural network's parameter optimization typically uses gradient descent, which can lead to local minima rather than global minima. Furthermore, because it replicates the relationship between given inputs and outputs, it is prone to overfitting.

[0005] The above compensation methods have high requirements on the quality of data collected by multi-rotor UAV aviation platforms, and they are all data-driven compensation. Since the accuracy of linear compensation is not high and nonlinear compensation may have problems such as gradient disappearance or gradient explosion, their work efficiency and compensation effect are not ideal. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for compensating aeromagnetic data of a multi-rotor UAV, aiming to solve the above technical problems.

[0007] The present invention is implemented as follows: a method for compensating aeromagnetic data of a multi-rotor UAV, comprising the following steps:

[0008] Acquire aeromagnetic data of a multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data;

[0009] Determining total geomagnetic field data based on the three-component fluxgate data;

[0010] Calculating a direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data;

[0011] Based on a preset PINN neural network, the direction cosine matrix and the total geomagnetic field data are used as input values ​​to output a magnetic interference prediction value;

[0012] The aeromagnetic data is compensated according to the magnetic interference prediction value.

[0013] Preferably, the three-component fluxgate data includes the components of the geomagnetic field in three directions, respectively Indicates; total geomagnetic field data is expressed as Total geomagnetic field data The calculation formula is as follows:

[0014] .

[0015] Preferably, the direction cosine matrix includes multiple direction cosines in the TL equation; the direction cosines are determined by the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field; the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field are respectively expressed as 、 、 express;

[0016] The TL equation is:

[0017]

[0018]

[0019] ;

[0020] Where, It is the preset 18 compensation coefficients; H p is the constant field interference value; H i is the induction field interference value; H ec is the eddy current field interference value; H t is the total magnetic interference value.

[0021] Preferably, the construction method of the PINN neural network is as follows: constructing a feedforward fully connected neural network, in which each neuron is connected to all neurons in the previous and next layers, and after receiving input, each neuron is activated according to its own judgment to generate an output signal and then aggregated to achieve training of the information source; selecting the product of the direction cosine matrix and the total geomagnetic field data as input, the magnetic interference value as output, and selecting the softmax function as the nonlinear activation function of the feedforward fully connected neural network, and using the feedforward fully connected neural network to calculate the output value; the total loss function of the feedforward fully connected neural network includes a physical information loss term and a data loss term, and the output value is constrained by the total loss function, and compared with the set threshold of the total loss function to determine whether re-iteration is needed, thereby achieving efficient convergence.

[0022] Preferably, the model parameters of the feedforward fully connected neural network are expressed as:

[0023] ;

[0024] in, represents the output of the lth layer, f represents the nonlinear activation function of the layer, is the weight matrix of this layer, is the output of the previous layer, is the bias of this layer.

[0025] Preferably, the expression of the nonlinear activation function is:

[0026] ;

[0027] Among them, X i It represents the i-th group of data of the input direction cosine matrix, j represents that the current data is the j-th in this group of data, e is the base of the natural logarithm, and n is the total number of data groups.

[0028] Preferably, the data loss item is:

[0029] ;

[0030] Among them, L Date is the data loss item; is the magnetic interference value of the i-th UAV, is the predicted value of the i-th magnetic interference.

[0031] Preferably, the physical information loss item is:

[0032] ;

[0033]

[0034]

[0035] ;

[0036] Among them, L PDE is the physical information loss item; The calculated background field value is calculated by removing the predicted interference site from the aeromagnetic data to eliminate the influence of geomagnetic gradient; H* is the ideal geomagnetic background field without interference; is the predicted value of magnetic interference; H x 、H y 、H z Calculate the magnetic field strength in the X, Y, and Z directions of the background field respectively; The three-component magnetic field intensity of the background field calculated after compensation of the measured aeromagnetic data is used to calculate its divergence to characterize the strength of the magnetic field divergence at each point in space. When the divergence approaches 0, it means there is no external magnetic field interference here.

[0037] Another object of an embodiment of the present invention is to provide an aeromagnetic data compensation system for a multi-rotor UAV, for implementing the above-mentioned aeromagnetic data compensation method, which comprises:

[0038] A data acquisition module is used to acquire aeromagnetic data of a multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data;

[0039] a total geomagnetic field determination module, configured to determine total geomagnetic field data based on the three-component fluxgate data;

[0040] A direction cosine matrix calculation module is used to calculate the direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data;

[0041] a magnetic interference prediction module, configured to output a magnetic interference prediction value based on a preset PINN neural network, taking the direction cosine matrix and the total geomagnetic field data as input values;

[0042] A data compensation module is used to compensate the aeromagnetic data according to the magnetic interference prediction value.

[0043] Compared with existing traditional compensation methods, the present invention provides a method for compensating aeromagnetic data of a multi-rotor drone. The PINN neural network used in combination with the physical information contained in the data can have a strong fitting compensation ability, while not requiring reliance on a large amount of data support, and the compensation accuracy of aeromagnetic data is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic flow chart of a method for compensating aeromagnetic data of a multi-rotor UAV provided in an embodiment of the present invention.

[0045] Figure 2 A schematic diagram of a PINN neural network provided by an embodiment of the present invention.

[0046] Figure 3 A schematic diagram of the structure of an aeromagnetic data compensation system for a multi-rotor UAV provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] It should be noted that, if no explanation is given for the various parameters in the formulas involved in the embodiments of the present invention, they are assumed to have existing publicly known meanings and will not be elaborated on here.

[0049] There are generally two methods for aeromagnetic compensation of multi-rotor UAV aeromagnetic survey platforms: one is traditional linear compensation, which has a poor fitting effect and low accuracy; the other is data-driven nonlinear compensation. This method regards the relationship between data as a nonlinear problem. However, since this neural network is completely data-driven and relies heavily on large amounts of high-quality data, it cannot generate accurate predicted interference values.

[0050] like Figure 1 As shown, in one embodiment of the present invention, in order to solve the above technical problems, a method for compensating aeromagnetic data of a multi-rotor UAV is provided, which includes the following steps:

[0051] S100, obtaining aeromagnetic data of a multi-rotor UAV; the aeromagnetic data including three-component fluxgate data;

[0052] S200, determining total geomagnetic field data according to the three-component fluxgate data;

[0053] S300, calculating a direction cosine matrix according to the three-component fluxgate data and the total geomagnetic field data;

[0054] S400, based on a preset PINN neural network, taking the direction cosine matrix and the total geomagnetic field data as input values, and outputting a magnetic interference prediction value;

[0055] S500: Compensate the aeromagnetic data according to the magnetic interference prediction value.

[0056] In an embodiment of the present invention, to address the issue of aeromagnetic interference (AMI) for multi-rotor UAV platforms, a physical information neural network (PINN) is constructed by introducing physical information into a neural network to predict magnetic interference. The PINN typically consists of a feedforward, fully connected neural network with a direction cosine matrix as input and magnetic interference as output. The PINN performs a nonlinear fit to the magnetic interference, and a physical information loss term is incorporated into the loss function, allowing the laws of physics to guide model training. The PINN neural network employed in this embodiment of the present invention, combined with the physical information contained in the data, offers strong fitting and compensation capabilities while not requiring extensive data support, resulting in high compensation accuracy for aeromagnetic data.

[0057] In a preferred embodiment of the present invention, the three-component fluxgate data is data measured by a fluxgate magnetometer, which includes the components of the geomagnetic field in three directions, respectively. Indicates; total geomagnetic field data is expressed as Total geomagnetic field data The calculation formula is as follows:

[0058] .

[0059] In a preferred embodiment of the present invention, the direction cosine matrix includes multiple direction cosines in the TL equation; the direction cosines are determined by the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field; the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field are respectively expressed as 、 、 express.

[0060] In practical applications, according to the theory of Tolles and Lawson, magnetic interference can be divided into three parts according to the causes of magnetic interference: constant field interference, induction field interference and eddy current field interference.

[0061] Based on this, the above TL equation is:

[0062]

[0063]

[0064] ;

[0065] Where, It is the preset 18 compensation coefficients; H p is the constant field interference value; H i is the induction field interference value; H ec is the eddy current field interference value; H t is the total magnetic interference value.

[0066] According to the above formula, the TL equation can be organized as:

[0067] ;

[0068] Wherein, X is the direction cosine matrix, and C is the compensation coefficient matrix containing 18 compensation coefficients.

[0069] like Figure 2 As shown, in a preferred embodiment of the present invention, the construction method of the PINN neural network is as follows: construct a feedforward fully connected neural network (DNN), where each neuron is connected to all neurons in the previous layer and the next layer. After receiving the input, each neuron activates and generates an output signal based on its own judgment, and then summarizes it to achieve training of the information source; the product of the direction cosine matrix and the total geomagnetic field data is selected as the input, the magnetic interference value is used as the output, and the softmax function is selected as the nonlinear activation function of the feedforward fully connected neural network, and the output value is calculated using the feedforward fully connected neural network; the total loss function (L) of the feedforward fully connected neural network includes a physical information loss term (L PDE ) and data loss term (L Date ), the output value is constrained by the total loss function, and the threshold of the total loss function is set Compare and determine whether re-iteration is needed to achieve efficient convergence.

[0070] In a preferred embodiment of the present invention, for a multi-layer feedforward fully connected neural network, the model parameters of the feedforward fully connected neural network are expressed as:

[0071] ;

[0072] in, represents the output of the lth layer, f represents the nonlinear activation function of the layer, is the weight matrix of this layer, is the output of the previous layer, is the bias of this layer.

[0073] In a preferred embodiment of the present invention, the expression of the nonlinear activation function is:

[0074] ;

[0075] Among them, Xi It represents the i-th group of data of the input direction cosine matrix, j represents that the current data is the j-th in this group of data, e is the base of the natural logarithm, and n is the total number of data groups.

[0076] In a preferred embodiment of the present invention, the data loss item is:

[0077] ;

[0078] Among them, L Date is the data loss item; is the magnetic interference value of the i-th UAV, is the predicted value of the i-th magnetic interference.

[0079] In an embodiment of the present invention, by introducing the data loss term, the difference between the predicted output value of the feedforward fully connected neural network and the actual observed data can be measured, with the aim of enabling the feedforward fully connected neural network to fit the data as much as possible.

[0080] In practical applications, the measurement principle of aerial survey data is to treat the geomagnetic background field of the work area as a constant value under ideal conditions during the preprocessing phase. However, under actual flight conditions, in addition to the essential geomagnetic gradient, other interference factors may also exist. Therefore, the magnetic field strength of the embodiments of the present invention can be constrained by adding Maxwell equations constraints to the airborne magnetic field noise, making the compensated results closer to the actual geomagnetic field value. Specifically, by designing a physical information loss term, the physical quantity predicted by the feedforward fully connected neural network is substituted into the Maxwell equations to calculate the residual error that constitutes this part of the loss function.

[0081] In a preferred embodiment of the present invention, the physical information loss term is:

[0082] ;

[0083]

[0084]

[0085] ;

[0086] Among them, L PDE is the physical information loss item; The calculated background field value is calculated by removing the predicted interference site from the aeromagnetic data to eliminate the influence of geomagnetic gradient; H* is the ideal geomagnetic background field without interference; is the predicted value of magnetic interference; H x 、H y 、H z Calculate the magnetic field strength in the X, Y, and Z directions of the background field respectively; The three-component magnetic field intensity of the background field calculated after compensation of the measured aeromagnetic data is used to calculate its divergence to characterize the strength of the magnetic field divergence at each point in space. When the divergence approaches 0, it means that there is no interference from external magnetic field sources, and it also shows that the predicted value is accurate.

[0087] In summary, the total loss function of the feedforward fully connected neural network provided by the embodiment of the present invention is:

[0088] ;

[0089] Where L is the total loss function; and are hyperparameters that weigh the importance of physical information loss term and data loss term respectively.

[0090] like Figure 3 As shown, in another embodiment of the present invention, an aeromagnetic data compensation system for a multi-rotor UAV is provided, which is used to implement the above-mentioned aeromagnetic data compensation method, and includes:

[0091] The data acquisition module 10 is used to acquire aeromagnetic data of the multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data;

[0092] a total geomagnetic field determination module 20, configured to determine total geomagnetic field data based on the three-component fluxgate data;

[0093] A direction cosine matrix calculation module 30 is used to calculate the direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data;

[0094] a magnetic interference prediction module 40 for outputting a magnetic interference prediction value based on a preset PINN neural network and taking the direction cosine matrix and the total geomagnetic field data as input values;

[0095] The data compensation module 50 is configured to compensate the aeromagnetic data according to the magnetic interference prediction value.

[0096] It should be noted that the above modules can be implemented in the form of a computer program, which can be run on a computer device. The computer program constituting each module can be stored in the memory of the computer device so that the processor executes each step of the above method.

[0097] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0098] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0099] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for compensating aeromagnetic data of a multi-rotor UAV, characterized in that: The following steps are involved: Acquire aeromagnetic data of a multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data; Determining total geomagnetic field data based on the three-component fluxgate data; Calculating a direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data; Based on a preset PINN neural network, the direction cosine matrix and the total geomagnetic field data are used as input values ​​to output a magnetic interference prediction value; The aeromagnetic data is compensated according to the magnetic interference prediction value.

2. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 1, characterized in that: The three-component fluxgate data includes the components of the geomagnetic field in three directions, respectively. Indicates; total geomagnetic field data is expressed as Total geomagnetic field data The calculation formula is as follows: 。 3. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 2, characterized in that: The direction cosine matrix includes multiple direction cosines in the TL equation; the direction cosines are determined by the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field; the cosine values ​​of the angles between the components of the geomagnetic field in three directions and the geomagnetic field are respectively expressed as 、 、 express; The TL equation is: ； ； ； Where, It is the preset 18 compensation coefficients; H p is the constant field interference value; H i is the induction field interference value; H ec is the eddy current field interference value; H t is the total magnetic interference value.

4. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 3, characterized in that: The PINN neural network construction method is as follows: constructing a feedforward fully connected neural network, in which each neuron is connected to all neurons in the previous and next layers. After receiving input, each neuron activates and generates output signals based on its own judgment, and then aggregates them to achieve training of the information source; selecting the product of the direction cosine matrix and the total geomagnetic field data as input, the magnetic interference value as output, and selecting the softmax function as the nonlinear activation function of the feedforward fully connected neural network, and using the feedforward fully connected neural network to calculate the output value; the total loss function of the feedforward fully connected neural network includes a physical information loss term and a data loss term, and the output value is constrained by the total loss function, and compared with a set threshold of the total loss function to determine whether re-iteration is needed, thereby achieving efficient convergence.

5. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 4, characterized in that: The model parameters of the feedforward fully connected neural network are expressed as: ； in, represents the output of the lth layer, f represents the nonlinear activation function of the layer, is the weight matrix of this layer, is the output of the previous layer, is the bias of this layer.

6. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 4 or 5, characterized in that: The expression of the nonlinear activation function is: ； Among them, X i It represents the i-th group of data of the input direction cosine matrix, j represents that the current data is the j-th in this group of data, e is the base of the natural logarithm, and n is the total number of data groups.

7. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 4, characterized in that: The data loss items are: ； Among them, L Date is the data loss item; is the magnetic interference value of the i-th UAV, is the predicted value of the i-th magnetic interference.

8. The aeromagnetic data compensation method for a multi-rotor UAV according to claim 4, characterized in that: The physical information loss term is: ； ； ； ； ； ； Among them, L PDE is the physical information loss item; The calculated background field value is calculated by removing the predicted interference site from the aeromagnetic data to eliminate the influence of geomagnetic gradient; H* is the ideal geomagnetic background field without interference; is the predicted value of magnetic interference; H x 、H y 、H z Calculate the magnetic field strength in the X, Y, and Z directions of the background field respectively; The three-component magnetic field intensity of the background field calculated after compensation of the measured aeromagnetic data is used to calculate its divergence to characterize the strength of the magnetic field divergence at each point in space. When the divergence approaches 0, it means there is no external magnetic field interference here.

9. An aeromagnetic data compensation system for a multi-rotor UAV, used to implement the aeromagnetic data compensation method for a multi-rotor UAV according to any one of claims 1 to 8, characterized in that: include: A data acquisition module is used to acquire aeromagnetic data of a multi-rotor UAV; the aeromagnetic data includes three-component fluxgate data; a total geomagnetic field determination module, configured to determine total geomagnetic field data based on the three-component fluxgate data; A direction cosine matrix calculation module is used to calculate the direction cosine matrix based on the three-component fluxgate data and the total geomagnetic field data; a magnetic interference prediction module, configured to output a magnetic interference prediction value based on a preset PINN neural network, taking the direction cosine matrix and the total geomagnetic field data as input values; A data compensation module is used to compensate the aeromagnetic data according to the magnetic interference prediction value.

Images