Unmanned aerial vehicle airborne natural field source electromagnetic exploration data noise suppression method and system

Abstract

This invention relates to the field of geophysical electromagnetic detection technology, specifically disclosing a method and system for noise suppression of electromagnetic detection data from UAV-based natural field sources. The method includes: dividing the original time-domain three-component magnetic field signal into time windows and constructing a data matrix; constructing an overcomplete dictionary matrix containing typical noise morphology characteristics of various UAV platforms; selecting a subset of data as a training set and globally optimizing key hyperparameters of the denoising process to obtain the optimal denoising parameter set; performing sparse entropy adaptive denoising processing on the data of all time windows based on the optimal denoising parameter set, and concatenating and reconstructing the denoising results of each window to obtain the final denoised three-component magnetic field data. By introducing a "sparse entropy adaptive" mechanism, it achieves precise removal of structured interference and adaptive preservation of effective signals; combined with the "gray wolf optimization" algorithm, it overcomes the difficulty of balancing the sparse threshold and iteration depth in traditional methods.

Description

Technical Field

[0001] This invention relates to the field of geophysical electromagnetic detection technology, specifically to a method and system for suppressing noise in electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources. Background Technology

[0002] Unmanned aerial vehicle (UAV) electromagnetic field detection technology, as an emerging and efficient airborne geophysical exploration method, plays a crucial role in the analysis of deep geological structures and the prospecting of concealed mineral resources. This method relies on natural electromagnetic fields as the source, using a high-sensitivity receiving system mounted on a UAV platform to dynamically collect data on the response of the subsurface medium to electromagnetic waves, thereby inverting the subsurface electrical structure. Compared to traditional surface array-based exploration, this airborne detection mode completely overcomes the physical barriers of rugged terrain and dense vegetation, achieving non-contact data acquisition. It not only significantly improves the efficiency and coverage of field operations but also demonstrates excellent adaptability and application value in geological surveys in ecologically fragile areas and complex geographical environments.

[0003] However, in actual aerial exploration operations, data acquisition quality is often limited by both weak signal characteristics and complex environmental interference. Since the energy of natural field source signals is inherently limited, and the high-speed movement of the UAV platform introduces significant mechanical vibration noise, attitude change noise, and electromagnetic interference generated by the circuit system, these factors collectively lead to a substantial decrease in the signal-to-noise ratio of the raw data. Strong background noise can easily drown out weak secondary field signals from underground geological bodies, severely distorting the estimation results of the tilt vector and ultimately affecting the reliability of the inversion imaging. Therefore, considering the unique characteristics of the aerial mobile observation environment, researching and applying advanced signal denoising algorithms to eliminate environmental interference and reconstruct high-fidelity electromagnetic signals has become a core technical challenge for ensuring detection accuracy and improving the practical application effect of this technology. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for suppressing noise in electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for suppressing noise in electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources, the method comprising:

[0007] The original three-component time-domain magnetic field signal collected by the UAV platform is acquired, and the original three-component time-domain magnetic field signal is divided into time windows and a data matrix is ​​constructed.

[0008] An overcomplete dictionary matrix containing typical noise morphology features of various UAV platforms is constructed, and a sparse decomposition model is built based on the overcomplete dictionary matrix and the data matrix.

[0009] A portion of the data was selected as the training set. The Grey Wolf optimization algorithm was used to optimize the key hyperparameters of the noise reduction process globally, with the entropy characteristics, smoothness and energy distribution of the signal as the optimization objectives, so as to obtain the optimal set of noise reduction parameters.

[0010] Based on the optimal set of denoising parameters, sparse entropy adaptive denoising is performed on the data of all time windows, and the denoising results of each window are spliced ​​and reconstructed to obtain the final denoised three-component magnetic field data.

[0011] As a further embodiment of the present invention, the specific method for constructing the data matrix is ​​as follows:

[0012] The sampling points within each time window are stacked in columns to form the observation vector y. i ;

[0013] Arrange the observation vectors of the M windows column-wise to form a data matrix. .

[0014] As a further embodiment of the present invention, the atom types of the overcomplete dictionary matrix include: sine or cosine atoms for representing power frequency interference, damped oscillation atoms for representing mechanical vibration, square wave or triangular wave atoms for representing switching interference, and pulse atoms for representing impact interference.

[0015] As a further embodiment of the present invention, the sparse decomposition model is expressed as:

[0016] ;

[0017] Among them, a i Let r be a sparse coefficient vector, D be an overcomplete dictionary matrix, and r be a sparse coefficient vector. i It contains valid signals from natural field sources.

[0018] As a further embodiment of the present invention, the sparse entropy adaptive noise reduction processing is based on an iterative process of matching pursuit, including the following sub-steps:

[0019] Initialize the residuals and calculate the correlation between the current residuals and each atom in the dictionary in each iteration;

[0020] Atoms are screened based on sparse thresholds and sparse coefficients are updated to obtain new residuals;

[0021] Calculate the entropy-type complexity index of the current residual;

[0022] Determine the termination condition of the iteration, if And k <K max If the result is positive, then set k = k + 1 and re-execute the iteration; otherwise, stop the iteration and output the current residual as the denoised signal for the current time window.

[0023] in, E is an entropy-based complexity metric. th K is the residual entropy threshold. max is the maximum number of iterations, and k is the current number of iterations.

[0024] As a further embodiment of the present invention, the entropy complexity index is fuzzy entropy.

[0025] As a further aspect of the present invention, the use of the gray wolf optimization algorithm for global optimization specifically includes:

[0026] Define the noise reduction hyperparameter vector As for the position of an individual gray wolf, T s Here, L is the sparsity threshold factor, L is the number of atoms selected in each iteration, and E is the number of atoms selected in each iteration. th K is the residual entropy threshold. max This represents the maximum number of iterations.

[0027] Initialize the gray wolf population and configure the denoising process for each individual gray wolf based on the denoising hyperparameter vector;

[0028] The comprehensive evaluation index of the denoised signal is calculated as the fitness value. The evaluation index includes entropy value, frequency band energy balance and time domain smoothness.

[0029] The position of the gray wolf population is iteratively updated based on the fitness value until the convergence condition is met. The hyperparameter vector corresponding to the gray wolf individual with the best fitness is output as the optimal noise reduction parameter set.

[0030] The present invention also provides a noise suppression system for electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources, the system comprising:

[0031] The data matrix construction module is used to acquire the original three-component time-domain magnetic field signal collected by the UAV platform, divide the original three-component time-domain magnetic field signal into time windows and construct a data matrix.

[0032] An overcomplete dictionary construction module is used to construct an overcomplete dictionary matrix containing typical noise morphological characteristics of various UAV platforms, and to construct a sparse decomposition model based on the overcomplete dictionary matrix and the data matrix.

[0033] The optimization module is used to select a portion of the data as the training set and adopt the Grey Wolf optimization algorithm. With the entropy characteristics, smoothness and energy distribution of the signal as optimization objectives, it performs global optimization on the key hyperparameters of the noise reduction process to obtain the optimal set of noise reduction parameters.

[0034] The output module is used to perform sparse entropy adaptive denoising on the data of all time windows based on the optimal denoising parameter set, and to concatenate and reconstruct the denoising results of each window to obtain the final denoised three-component magnetic field data.

[0035] Compared with existing technologies, the beneficial effects of this invention are as follows: The method proposed in this application addresses the characteristics of large noise amplitude and strong structure in UAV platform data. By introducing a "sparse entropy adaptive" mechanism, it dynamically monitors signal complexity using residual entropy values, achieving precise removal of structured interference and adaptive preservation of effective signals. Simultaneously, combined with the "Grey Wolf Optimization" algorithm, it automatically searches for the optimal parameter combination, overcoming the difficulty in balancing sparse threshold and iteration depth in traditional methods. The processed data effectively suppresses strong noise such as mechanical vibration and electromagnetic interference while retaining weak low-frequency information reflecting underground electrical structures, significantly improving data smoothness consistency and the reliability of subsequent inversion interpretation. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0037] Figure 1 A flowchart illustrating a method for suppressing noise in electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources, provided in an embodiment of the present invention.

[0038] Figure 2 The flowchart illustrates the sparse entropy adaptive method provided in this embodiment of the invention. Detailed Implementation

[0039] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, 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 merely illustrative of the present invention and are not intended to limit the present invention.

[0040] Figure 1 This is a flowchart illustrating a method for suppressing noise in electromagnetic detection data from UAV-based natural field sources. In this embodiment of the invention, the method for suppressing noise in electromagnetic detection data from UAV-based natural field sources includes:

[0041] The original three-component time-domain magnetic field signal collected by the UAV platform is acquired, and the original three-component time-domain magnetic field signal is divided into time windows and a data matrix is ​​constructed.

[0042] An overcomplete dictionary matrix containing typical noise morphology features of various UAV platforms is constructed, and a sparse decomposition model is built based on the overcomplete dictionary matrix and the data matrix.

[0043] A portion of the data was selected as the training set. The Grey Wolf optimization algorithm was used to optimize the key hyperparameters of the noise reduction process globally, with the entropy characteristics, smoothness and energy distribution of the signal as the optimization objectives, so as to obtain the optimal set of noise reduction parameters.

[0044] Based on the optimal set of denoising parameters, sparse entropy adaptive denoising is performed on the data of all time windows, and the denoising results of each window are spliced ​​and reconstructed to obtain the final denoised three-component magnetic field data.

[0045] In this embodiment, the airborne natural field source data denoising method provided in this application is applied to the three-component magnetic field data of airborne natural field sources already acquired by the UAV platform, in order to at least solve the problem that the UAV platform has strong noise and it is difficult to retain effective information of airborne natural field sources while suppressing strong interference.

[0046] After completing the aerial survey, the UAV obtained the original three-component magnetic field signal B in the time domain at each measurement point. x (t), B y (t), B z (t). The original time-domain three-component magnetic field signal is preprocessed and segmented according to time windows to construct a data matrix for sparse decomposition. Considering the characteristics of large noise amplitude, strong structure, and obvious periodicity and pulse characteristics of the UAV platform, an overcomplete dictionary matrix is ​​pre-constructed, with column vectors consisting of different types of dictionary atoms. The pre-defined overcomplete dictionary matrix contains various noise morphological features. A portion of the data is selected as the training set, and the Grey Wolf optimization algorithm is used. The entropy characteristics, smoothness index, and energy distribution of the denoised signal are used as objective functions to drive the sparse entropy adaptive denoising process. Global optimization is performed on key hyperparameters, including the sparse threshold and entropy threshold, to obtain the optimal denoising parameter set. The sparse entropy adaptive denoising process represents the observed signal as a superposition of sparse noise components and residual signals. The sparse iteration depth is adaptively controlled by the entropy measure of the residual signal complexity. Based on the optimal denoising parameter set, the sparse entropy adaptive denoising process is executed for all time windows of the original time-domain three-component magnetic field signal, and the denoising results of each time window are spliced ​​and reconstructed to obtain the final denoised airborne natural field source three-component magnetic field data.

[0047] In a preferred embodiment of the present invention, the specific method for constructing the data matrix is ​​as follows:

[0048] The sampling points within each time window are stacked in columns to form the observation vector y. i ;

[0049] Arrange the observation vectors of the M windows column-wise to form a data matrix. .

[0050] In this embodiment, the raw three-component time-domain magnetic field signal collected by the UAV platform is acquired. First, the raw signal undergoes preprocessing, including removing the DC component, removing linear trends, bandpass filtering, time alignment, and amplitude normalization. Then, the data is divided into time windows, with fixed or non-overlapping windows. The sampling points within each time window are stacked column-wise to form an observation vector. , where N W Given the window length, the observation vectors of multiple windows are arranged column-wise to construct a data matrix for sparse decomposition. For three-component data, three data matrices can be constructed separately, or a multi-component joint data matrix can be formed by expanding the row dimension.

[0051] As a preferred embodiment of the present invention, the atom types of the overcomplete dictionary matrix include: sine or cosine atoms for representing power frequency interference, damped oscillation atoms for representing mechanical vibration, square wave or triangular wave atoms for representing switching interference, and pulse atoms for representing impact interference.

[0052] In this embodiment, an overcomplete dictionary matrix D containing various noise morphological features is pre-constructed, and a sparse decomposition model of the observed signal is established. The construction of the overcomplete dictionary D aims to cover typical interference sources during UAV flight. Its atom library specifically includes: sine or cosine atoms for representing power frequencies and their harmonics, damped oscillation atoms for representing mechanical vibrations and intermittent interference, square or triangular wave atoms for representing charging / discharging and switching interference, and pulse atoms for representing short-duration high-amplitude impacts. By combining the above-mentioned feature atoms, the dictionary can correspond to the complex electromagnetic environment noise of the UAV platform to the greatest extent.

[0053] In a preferred embodiment of the present invention, within each time window, the observation vector y i The representation in the dictionary satisfies a sparse decomposition model, which is expressed as:

[0054] ;

[0055] Among them, a i Let D be a sparse coefficient vector, with most elements being zero or close to zero. The dictionary atoms corresponding to the non-zero coefficients mainly represent structured strong interference components. Let D be an overcomplete dictionary matrix, and r be a sparse coefficient vector. i To include the effective signal from the natural field source, ideally it mainly includes the effective component of the natural field source and a small amount of random noise.

[0056] In this embodiment, the high correlation between dictionary atoms and structured strong interference is utilized to map the large-amplitude noise energy onto the sparse coefficient vector a. i Go up, that is, go up iCharacterize the structured strong disturbance components; simultaneously, define the residual r in the model. i For components that cannot be sparsely represented by a dictionary, the effective signal of the natural field source is physically classified into the residual r. i In this process, strong interference is separated from effective signals.

[0057] like Figure 2 As shown, in a preferred embodiment of the present invention, the sparse entropy adaptive noise reduction processing is based on an iterative process of matching pursuit, including the following sub-steps:

[0058] Initialize the residuals and calculate the correlation between the current residuals and each atom in the dictionary in each iteration;

[0059] Atoms are screened based on sparse thresholds and sparse coefficients are updated to obtain new residuals;

[0060] Calculate the entropy-type complexity index of the current residual;

[0061] Determine the termination condition of the iteration, if And k <K max If the result is positive, then set k = k + 1 and re-execute the iteration; otherwise, stop the iteration and output the current residual as the denoised signal for the current time window.

[0062] in, E is an entropy-based complexity metric. th K is the residual entropy threshold. max is the maximum number of iterations, and k is the current number of iterations.

[0063] In this embodiment, based on the optimal denoising parameter set, sparse entropy adaptive denoising processing is performed on all time windows of the original time-domain three-component magnetic field signal. This process adopts an iterative approach based on matching pursuit.

[0064] Iterative decomposition: Initializing residuals In each iteration, the correlation between the current residual and each atom in the dictionary is calculated. Based on a preset sparsity threshold factor, a candidate set of atoms with high correlation is selected. Then, the corresponding coefficients are estimated using the least squares method. The linear combination of these atoms is subtracted from the residual to obtain a new residual signal. By repeating the iterative steps of "correlation screening - coefficient update - residual update", structured strong interference components can be gradually extracted using dictionary atoms.

[0065] After each round of sparse iteration, the current residual signal Calculate entropy-based complexity metrics, such as fuzzy entropy. This entropy is used as a measure of the complexity of the residual. The entropy value reflects the regularity and randomness of the sequence: when the residual is still dominated by regular, periodic, and structured noise components, The value is relatively small; when the residual is mainly composed of natural field source signals and small random fluctuations. Too large. Set the residual entropy threshold E. th Compare after each iteration With E th ;when And the number of iterations k did not reach the preset upper limit K max When the residual still contains a significant amount of strong noise from the rules, the next round of sparse iteration continues; when Or the number of iterations reaches K max When the residual is determined to be dominated by the effective signal of the natural field source, sparse iteration is stopped, and the residual signal at this time is... As the initial denoising result for this time window, the residual signals retained from each time window are stitched together in chronological order to obtain continuous, high signal-to-noise ratio three-component magnetic field data of airborne natural field sources. This method utilizes a sparse representation model... With entropy measurement The combined "sparse entropy adaptive" mechanism can adaptively determine the iteration depth based on the actual data complexity, effectively suppressing strong structured interference while reducing excessive attenuation of the effective signal from the natural field source.

[0066] In a preferred embodiment of the present invention, the global optimization using the gray wolf optimization algorithm specifically includes:

[0067] Define the noise reduction hyperparameter vector As for the position of an individual gray wolf, T s Here, L is the sparsity threshold factor, L is the number of atoms selected in each iteration, and E is the number of atoms selected in each iteration. th K is the residual entropy threshold. max This represents the maximum number of iterations.

[0068] Initialize the gray wolf population and configure the denoising process for each individual gray wolf based on the denoising hyperparameter vector;

[0069] The comprehensive evaluation index of the denoised signal is calculated as the fitness value. The evaluation index includes entropy value, frequency band energy balance and time domain smoothness.

[0070] The position of the gray wolf population is iteratively updated based on the fitness value until the convergence condition is met. The hyperparameter vector corresponding to the gray wolf individual with the best fitness is output as the optimal noise reduction parameter set.

[0071] In this embodiment, based on the aforementioned "sparse entropy adaptive" denoising kernel, the Grey Wolf optimization algorithm is further employed to globally optimize key hyperparameters in the denoising process. This overcomes the problem that parameters such as the sparse threshold, the number of dictionary atoms selected in each round, the residual entropy threshold, and the upper limit of the iteration steps heavily rely on human experience. Specifically, the sparse threshold factor T is... s The number of dictionary atoms selected in each round of matching and tracking, L, and the residual entropy threshold, E. th Maximum number of iterations K max The noise reduction hyperparameter vector consists of optional entropy calculation parameters, etc. The vector is considered as the position of the individual gray wolf in the search space. First, a representative time period is selected from the collected airborne natural field source data to form a training dataset, which includes typical UAV platform noise and natural field source signals. A certain number of gray wolf individuals are randomly generated as the initial population, with each individual corresponding to a set of candidate hyperparameter vectors θ. For each individual in the population, a sparse entropy adaptive denoising kernel is configured using its corresponding θ, and denoising is performed on the training dataset to obtain the denoised three-component magnetic field signal. A set of comprehensive quality indicators is calculated, such as whether the entropy value of the denoised signal falls within a reasonable range corresponding to the natural field source signal, whether the energy distribution in different time-frequency bands is balanced, and the smoothness of the time-domain waveform. These indicators are combined into a fitness function with preset weights, so that the better the denoising quality, the better the fitness function. Subsequently, following the update rules of the gray wolf optimization algorithm, a leader individual is selected based on fitness to guide the update of the positions of other individuals, so that the hyperparameter vector θ approaches the optimal region generation by generation. When the number of iterations reaches a preset upper limit or the fitness change tends to stabilize, the gray wolf individual with the best fitness is selected, and its corresponding hyperparameter vector is... This serves as the optimal set of noise reduction parameters for this implementation method, which will be used for subsequent processing of all data.

[0072] Obtaining the optimal set of noise reduction parameters Subsequently, this parameter set was fixed and applied to the complete three-component magnetic field data of the airborne natural field source: the preprocessed data of the entire survey line or survey area was divided into time windows and the data matrix was constructed in the same way as in the training phase, and the determined hyperparameters were called for each time window. The sparse entropy adaptive denoising kernel is repeatedly executed to obtain the preliminary denoising results for each window. Finally, the denoised signals of all time windows are spliced ​​and reconstructed in chronological order to obtain a continuous, denoised time series of the three-component magnetic field of the aerospace natural field source.

[0073] Through the "Grey Wolf Optimized Sparse Entropy Adaptive" two-layer mechanism, the noise reduction intensity and iteration depth can be adaptively adjusted without relying on repeated manual adjustment of algorithm parameters. This effectively suppresses large amplitude and strongly structured interference noise generated by the UAV platform, attitude change noise, and electromagnetic interference generated by the circuit system. At the same time, it retains as much effective information about natural field sources related to underground electrical structures as possible, so that the processed aviation natural field source data has a higher signal-to-noise ratio and better smoothness consistency, providing a reliable data foundation for subsequent frequency domain analysis and electromagnetic interpretation.

[0074] This invention also provides a noise suppression system for electromagnetic detection data from UAV-based natural field sources, the system comprising:

[0075] The data matrix construction module is used to acquire the original three-component time-domain magnetic field signal collected by the UAV platform, divide the original three-component time-domain magnetic field signal into time windows and construct a data matrix.

[0076] An overcomplete dictionary construction module is used to construct an overcomplete dictionary matrix containing typical noise morphological characteristics of various UAV platforms, and to construct a sparse decomposition model based on the overcomplete dictionary matrix and the data matrix.

[0077] The optimization module is used to select a portion of the data as the training set and adopt the Grey Wolf optimization algorithm. With the entropy characteristics, smoothness and energy distribution of the signal as optimization objectives, it performs global optimization on the key hyperparameters of the noise reduction process to obtain the optimal set of noise reduction parameters.

[0078] The output module is used to perform sparse entropy adaptive denoising on the data of all time windows based on the optimal denoising parameter set, and to concatenate and reconstruct the denoising results of each window to obtain the final denoised three-component magnetic field data.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for suppressing noise in electromagnetic detection data from unmanned aerial vehicles (UAVs) of natural field sources, characterized in that, The method includes: The original three-component time-domain magnetic field signal collected by the UAV platform is acquired, and the original three-component time-domain magnetic field signal is divided into time windows and a data matrix is ​​constructed. An overcomplete dictionary matrix containing typical noise morphological characteristics of various UAV platforms is constructed, and a sparse decomposition model is built based on the overcomplete dictionary matrix and the data matrix. A portion of the data was selected as the training set. The Grey Wolf optimization algorithm was used to optimize the key hyperparameters of the noise reduction process globally, with the entropy characteristics, smoothness and energy distribution of the signal as the optimization objectives, so as to obtain the optimal set of noise reduction parameters. Based on the optimal set of denoising parameters, sparse entropy adaptive denoising is performed on the data of all time windows, and the denoising results of each window are spliced ​​and reconstructed to obtain the final denoised three-component magnetic field data.

2. The method for suppressing noise in electromagnetic detection data from UAV airborne natural field sources according to claim 1, characterized in that, The specific method for constructing the data matrix is ​​as follows: The sampling points within each time window are stacked in columns to form the observation vector y. i ; Arrange the observation vectors of the M windows column-wise to form a data matrix. .

3. The method for suppressing noise in electromagnetic detection data from UAV airborne natural field sources according to claim 1, characterized in that, The atom types of the overcomplete dictionary matrix include: sine or cosine atoms for representing power frequency interference, damped oscillation atoms for representing mechanical vibration, square or triangular wave atoms for representing switching interference, and pulse atoms for representing impact interference.

4. The method for suppressing noise in electromagnetic detection data from UAV airborne natural field sources according to claim 3, characterized in that, The sparse decomposition model is expressed as follows: ； Among them, a i Let r be a sparse coefficient vector, D be an overcomplete dictionary matrix, and r be a sparse coefficient vector. i It contains valid signals from natural field sources.

5. The method for suppressing noise in electromagnetic detection data from UAV-based natural field sources according to claim 1, characterized in that, The sparse entropy adaptive noise reduction process is based on an iterative process of matching pursuit, including the following sub-steps: Initialize the residuals and calculate the correlation between the current residuals and each atom in the dictionary in each iteration; Atoms are screened based on sparse thresholds and sparse coefficients are updated to obtain new residuals; Calculate the entropy-type complexity index of the current residual; Determine the termination condition of the iteration, if And k <K max If the result is positive, then set k = k + 1 and re-execute the iteration; otherwise, stop the iteration and output the current residual as the denoised signal for the current time window. in, E is an entropy-based complexity metric. th K is the residual entropy threshold. max is the maximum number of iterations, and k is the current number of iterations.

6. The method for suppressing noise in electromagnetic detection data from UAV airborne natural field sources according to claim 5, characterized in that, The entropy complexity index is fuzzy entropy.

7. The method for suppressing noise in electromagnetic detection data from UAV airborne natural field sources according to claim 1, characterized in that, The use of the Grey Wolf optimization algorithm for global optimization specifically includes: Define the noise reduction hyperparameter vector As for the position of an individual gray wolf, T s Here, E is the sparsity threshold factor, L is the number of atoms selected in each iteration, and E is the number of atoms selected in each iteration. th K is the residual entropy threshold. max This represents the maximum number of iterations. Initialize the gray wolf population and configure the denoising process for each individual gray wolf based on the denoising hyperparameter vector; The comprehensive evaluation index of the denoised signal is calculated as the fitness value. The evaluation index includes entropy value, frequency band energy balance and time domain smoothness. The position of the gray wolf population is iteratively updated based on the fitness value until the convergence condition is met. The hyperparameter vector corresponding to the gray wolf individual with the best fitness is output as the optimal noise reduction parameter set.

8. A noise suppression system for electromagnetic detection data from UAV-based natural field sources, used to implement the noise suppression method for electromagnetic detection data from UAV-based natural field sources as described in any one of claims 1-7, characterized in that, The system includes: The data matrix construction module is used to acquire the original three-component time-domain magnetic field signal collected by the UAV platform, divide the original three-component time-domain magnetic field signal into time windows and construct a data matrix. An overcomplete dictionary construction module is used to construct an overcomplete dictionary matrix containing typical noise morphological characteristics of various UAV platforms, and to construct a sparse decomposition model based on the overcomplete dictionary matrix and the data matrix. The optimization module is used to select a portion of the data as the training set and adopt the Grey Wolf optimization algorithm. With the entropy characteristics, smoothness and energy distribution of the signal as optimization objectives, it performs global optimization on the key hyperparameters of the noise reduction process to obtain the optimal set of noise reduction parameters. The output module is used to perform sparse entropy adaptive denoising on the data of all time windows based on the optimal denoising parameter set, and to concatenate and reconstruct the denoising results of each window to obtain the final denoised three-component magnetic field data.

Images