Aeromagnetic local anomaly extraction method based on BP neural network

Abstract

The invention discloses an aeromagnetic local anomaly extraction method based on a BP neural network. The method comprises the following steps: S1, obtaining actually measured aeromagnetic data; s2, performing gridding and normalization preprocessing on the actually measured aeromagnetic data to obtain preprocessed actually measured aeromagnetic data; and S3, inputting the preprocessed actually measured aeromagnetic data into the trained BP neural network model, and outputting local abnormal data extracted from the actually measured aeromagnetic data. According to the scheme, the recognition precision of shallow concealed ore bodies or structures in aeromagnetic data interpretation is improved.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration data processing technology, and in particular to a method for extracting local aeromagnetic anomalies based on a BP neural network. Background Technology

[0002] Airborne magnetic surveying is an important tool for geological surveys and mineral exploration. In airborne magnetic survey data, the observed magnetic field is typically a superposition of regional fields (background fields) generated by deep regional structures, local fields generated by shallow local geological bodies, and various types of noise. To identify shallow mineral resources or engineering geological targets, it is essential to effectively extract local anomalies from the superimposed field.

[0003] Traditional anomaly separation methods mainly include polynomial fitting, matched filtering, Wiener filtering, and upward continuation. These methods are mostly based on linear transformations in the frequency or spatial domains. However, when the spectra of the deep background field and the shallow local field overlap, or when the background field changes drastically and is highly nonlinear, traditional linear filters often struggle to preserve the complete shape of the local anomaly while removing the background. This can easily lead to anomaly amplitude attenuation or the generation of false anomalies, affecting the accuracy of subsequent geological interpretation. Summary of the Invention

[0004] The purpose of this invention is to provide a method for extracting local aeromagnetic anomalies based on a BP neural network.

[0005] To achieve the above objectives, the embodiments of the present invention are implemented in the following ways:

[0006] In a first aspect, the present invention provides a method for extracting local aeromagnetic anomalies based on a BP neural network, comprising the following steps:

[0007] S1: Obtain measured aeromagnetic data;

[0008] S2: Perform gridding and normalization preprocessing on the measured aeromagnetic data to obtain preprocessed measured aeromagnetic data;

[0009] S3: Input the preprocessed measured aeromagnetic data into the trained BP neural network model, and output the local anomaly data extracted from the measured aeromagnetic data.

[0010] In one embodiment of the present invention, the training dataset of the BP neural network model is generated by forward modeling based on the aeromagnetic anomaly forward modeling model, which includes a background field model for simulating magnetic anomalies of deep, large-scale geological bodies and a local anomaly model for simulating magnetic anomalies of shallow, small-scale geological bodies.

[0011] In one embodiment of the present invention, the background field model adopts a horizontal plate-like body as the background field, and it is assumed that the observation surface of the horizontal plate-like body is... The center is ;

[0012] The magnetic anomaly generated by the background field model The forward modeling formula for cuboid magnetic anomalies is used:

[0013]

[0014] in, The permeability of free space, H / m, The deep background field magnetization intensity; Here are the coordinates of the observation point. The coordinate difference corresponding to the integral variable. , , , The coordinates of the field source point; ; These are respectively related to the geomagnetic tilt angle Magnetic declination And the direction cosine coefficient related to the magnetization direction; the geometric boundary of the plate-like body corresponding to the integration limit is: , It is half the length of the plate-like body. It is half the width of the plate-like body. The thickness of the plate-like body. Determine the burial depth of the interface for the plate-like body. , , , , , .

[0015] In one embodiment of the present invention, the local anomaly body in the local anomaly model is a sphere;

[0016] The anomaly generated by the local anomaly model for:

[0017]

[0018] in, The permeability of free space, The depth of the sphere; The radius of the sphere; The magnetization intensity of the local anomalous body is set. Coordinates of the projection point of the observation plane at the center of the spherical anomaly .

[0019] In one embodiment of the present invention, the forward model magnetic anomaly The following calculations were performed using the aforementioned forward modeling of aeromagnetic anomalies:

[0020] (3)

[0021] The set of all the magnetic anomalies in the forward model is used as the training dataset.

[0022] In one embodiment of the present invention, the BP neural network model includes: an input layer, a hidden layer, and an output layer;

[0023] The input layer uses a sliding window mode to collect spatial data features, sets the window size to N×N grid points, the number of nodes in the input layer to N×N, and the input of the input layer is the normalized value of the training dataset.

[0024] The hidden layer nodes are based on empirical formulas. It is confirmed that, among them, The number of nodes in the input layer. This represents the number of nodes in the output layer. A constant between 1 and 10;

[0025] The output layer is configured with one node; the target output is the theoretical local outlier value corresponding to the center point of the window.

[0026] In one embodiment of the present invention, training the BP neural network model includes:

[0027] The full-field data generated by the forward model is gridded and normalized, and samples are extracted using a sliding window.

[0028] The sample is input into the established BP neural network model, forward propagation is performed, and the output value of the BP neural network model is calculated.

[0029] Calculate the mean square error between the output value and the theoretical local outlier;

[0030] Based on the mean square error, the gradients of each weight and bias parameter in the BP neural network model are calculated using the backpropagation algorithm;

[0031] The weights and bias parameters are updated using the gradient descent algorithm based on the gradient.

[0032] In one embodiment of the present invention, the method for extracting aeromagnetic local anomalies based on a BP neural network further includes:

[0033] All the local anomaly data are combined into a grid, and local anomaly contour maps are drawn.

[0034] In a second aspect, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method for extracting aeromagnetic local anomalies based on a BP neural network as described in the first aspect.

[0035] Thirdly, this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for extracting aeromagnetic local anomalies based on a BP neural network as described in the first aspect.

[0036] Compared with existing technologies, the present invention provides a method for extracting local aeromagnetic anomalies based on a BP neural network. By utilizing the powerful nonlinear approximation capability of a pre-trained BP neural network, it effectively overcomes the shortcomings of traditional filtering methods in separating spectral overlap regions, and significantly improves the accuracy of identifying shallow concealed ore bodies or structures in the interpretation of aeromagnetic data. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a method for extracting local aeromagnetic anomalies based on a BP neural network according to an embodiment of the present invention.

[0038] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0039] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0040] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0041] like Figure 1 As shown, a method for extracting local aeromagnetic anomalies based on a BP neural network according to a preferred embodiment of the present invention includes the following steps:

[0042] S1: Obtain measured aeromagnetic data.

[0043] Specifically, measured aeromagnetic data refers to the total magnetic field strength or magnetic anomaly data actually obtained through measurements using an airborne magnetometer. Measured aeromagnetic data can be obtained by reading the aeromagnetic measurement data file to be processed from a data storage device or a real-time acquisition device.

[0044] S2: Perform gridding and normalization preprocessing on the measured aeromagnetic data to obtain preprocessed measured aeromagnetic data.

[0045] Specifically, Kriging interpolation or the minimum curvature method can be used to grid the measured data to generate regular grid data.

[0046] The Min-Max normalization method is used to transform each value in the grid data to the interval [0,1] or [-1,1], which is consistent with the interval when training the BP neural network model, thus obtaining the preprocessed grid data.

[0047] S3: Input the preprocessed measured aeromagnetic data into the trained BP neural network model, and output the local anomaly data extracted from the measured aeromagnetic data.

[0048] Specifically, the preprocessed grid data is traversed sequentially using a sliding window (e.g., 5×5) similar to the training time. The data within each window is flattened into a row vector and input into the trained BP neural network model. The BP neural network model outputs a local anomaly prediction value for the center point of that window. After traversal, the prediction values ​​for all center points are rearranged into a grid of the same size as the input grid, thus obtaining the extracted local anomaly data grid.

[0049] The aeromagnetic local anomaly extraction method based on BP neural network provided in this embodiment utilizes the powerful nonlinear approximation capability of the pre-trained BP neural network to effectively overcome the shortcomings of traditional filtering methods in separating overlapping spectral regions, and significantly improves the accuracy of identifying shallow concealed ore bodies or structures in the interpretation of aeromagnetic data.

[0050] In one embodiment, the training dataset for training the BP neural network model is generated by forward modeling based on an aeromagnetic anomaly forward modeling model, which includes a background field model for simulating magnetic anomalies in deep, large-scale geological bodies and a local anomaly model for simulating magnetic anomalies in shallow, small-scale geological bodies.

[0051] Specifically, the forward modeling model of aeromagnetic anomalies is a theoretical model that calculates the distribution of the magnetic field generated on the surface or in the air based on the known geometric parameters, physical properties, and observation system of the source (geological body). The background field model is used to simulate the mathematical model of regionally varying magnetic anomaly components caused by deep, large-scale geological bodies (such as magnetic basements and deep igneous beds). The local anomaly model is used to simulate the mathematical model of localized high-frequency magnetic anomaly components caused by shallow, small-scale geological bodies (such as ore bodies, dikes, and structures).

[0052] Forward modeling allows for precise control of the relationship between geological bodies and magnetic fields. Constructing a combined model of deep background field and shallow local anomalies can realistically simulate the magnetic field superposition encountered in actual exploration, providing high-quality and diverse learning samples for neural networks.

[0053] In one embodiment, the background field model uses a horizontal plate-like body as an example. A large, deeply buried horizontal plate-like body (such as a deep igneous bedrock or magnetic basement) with a certain thickness and horizontal extension is designed to generate a regional background field. The resulting magnetic anomaly... The forward modeling formula for cuboid magnetic anomalies (Bhattacharyya, 1964) is adopted. The observation surface is assumed to be... The center is The formula for calculating the total magnetic anomaly of a cuboid is:

[0054]

[0055] in, The permeability of free space, H / m, The deep background field magnetization intensity; Here are the coordinates of the observation point. The coordinate difference corresponding to the integral variable. , , , The coordinates of the field source point; ; These are respectively related to the geomagnetic tilt angle Magnetic declination And the direction cosine coefficient related to the magnetization direction; the geometric boundary of the plate-like body corresponding to the integration limit is: , It is half the length of the plate-like body. It is half the width of the plate-like body. The thickness of the plate-like body. Determine the burial depth of the interface for the plate-like body. , , , , , .

[0056] The local anomaly is defined as multiple shallowly buried, strongly magnetic bodies randomly superimposed on the background field. The local anomaly bodies in the local anomaly model are assumed to be spheres. The anomalies they generate... for:

[0057]

[0058] in, The permeability of free space, The depth of the sphere; The radius of the sphere; The magnetization intensity of the local anomalous body is set. Coordinates of the projection point of the observation plane at the center of the spherical anomaly .

[0059] The aforementioned background field model and local anomaly model, combining horizontal plate-like bodies and spheres, can simulate the main regional background morphology and represent common local anomaly sources (such as skarn-type deposits, magmatic nodules, etc.), making the training data typical and representative.

[0060] In one embodiment, the forward model magnetic anomaly The following calculations were performed using the aforementioned forward modeling of aeromagnetic anomalies:

[0061] (3)

[0062] The set of all the magnetic anomalies in the forward model is used as the training dataset.

[0063] Specifically, the magnetic anomalies generated by the aforementioned background field model... All anomalies generated by the local anomaly model The summation yields a noise-free theoretical superposition field. Gaussian white noise is then added, with the noise level set to 1%-5% of the maximum amplitude of the theoretical superposition field, to generate the final magnetic anomaly data.

[0064] By randomly varying various parameters in the background field model and local anomaly model, the magnetic anomaly of the forward model is repeatedly calculated thousands of times to generate a training dataset containing thousands of forward model simulation data.

[0065] In one embodiment, the BP neural network model includes: an input layer, a hidden layer, and an output layer;

[0066] The input layer uses a sliding window mode to collect spatial data features, sets the window size to N×N grid points, the number of input layer nodes to N×N, and the input of the input layer is the normalized value of the training dataset.

[0067] Specifically, a sliding window is a method for extracting local neighborhood features from two-dimensional grid data. The window slides point by point on the grid, and each time it captures the values ​​of all grid points within the window's coverage area as an input sample. Assuming the window size is set to 5×5 or 7×7 grid points, the corresponding number of input layer nodes is 25 or 49.

[0068] The Min-Max normalization method can be used to normalize the data in the training dataset.

[0069] The hidden layer nodes are based on empirical formulas. It is confirmed that, among them, The number of nodes in the input layer. This represents the number of nodes in the output layer. It is a constant between 1 and 10.

[0070] The output layer is configured with one node; the target output is the theoretical local outlier value corresponding to the center point of the window.

[0071] Specifically, if the number of nodes in the input layer is 49, the number of nodes in the output layer... =1. Let α=5. According to the formula, the number of hidden layer nodes m≈12, so the number of hidden layer nodes is set to 12.

[0072] The data grid in the generated training dataset is normalized. A 7×7 window is used for sliding cropping, and the data in each window is flattened into a 49-dimensional vector as input. The theoretical local outlier corresponding to the center point of the window is used as the target output.

[0073] In one embodiment, training the BP neural network model includes:

[0074] The full-field data generated by the forward model is gridded and normalized, and samples are extracted using a sliding window.

[0075] The sample is input into the established BP neural network model, forward propagation is performed, and the output value of the BP neural network model is calculated.

[0076] Calculate the mean square error between the output value and the theoretical local outlier;

[0077] Based on the mean square error, the gradients of each weight and bias parameter in the BP neural network model are calculated using the backpropagation algorithm;

[0078] The weights and bias parameters are updated using the gradient descent algorithm based on the gradient.

[0079] Specifically, a) initialize all network weights and biases to small random numbers.

[0080] b) Input a batch of samples from the training dataset into the network, perform forward propagation, and obtain the predicted output.

[0081] c) Calculate the mean square error (MSE) of this batch of samples.

[0082] d) Perform backpropagation to compute the gradient of MSE with respect to all weights and biases.

[0083] e) Use the Adam optimizer to update the network parameters based on the gradient.

[0084] f) Repeat step be, iterating multiple times (epochs) on the entire training dataset until the MSE on the validation set no longer decreases significantly or reaches the preset number of iterations (e.g., 1000 epochs).

[0085] In this embodiment, through detailed training steps, it is ensured that the BP neural network model can stably and effectively learn the mapping rules for extracting local anomalies from massive synthetic data, and finally obtain a high-performance nonlinear filter that can be directly applied to measured data.

[0086] In one embodiment, the method for extracting aeromagnetic local anomalies based on a BP neural network further includes: combining all the local anomaly data into a grid and drawing local anomaly contour maps.

[0087] Specifically, a local anomaly contour map is a graphic representation of the intensity distribution of local anomalies on a two-dimensional plane, using curves (contours) formed by connecting points with equal numerical values.

[0088] All local anomaly data are combined into a grid, where the value of each point represents the intensity of the local anomaly extracted at that location. Using specialized plotting software or programming libraries, the grid data is read, appropriate contour intervals are set, and contour lines are automatically drawn. Scale bars, legends, coordinate axes, and other information are added to the map to create a thematic map suitable for geological interpretation. Further calculations of anomaly amplitude, gradient, and other characteristics can be performed, or the data can be overlaid with geological, geochemical, and other data for analysis.

[0089] In this embodiment, the digital results extracted by the BP neural network are transformed into intuitive graphical results, which greatly facilitates geologists in identifying anomalies, delineating areas, and inferring geological origins. This directly serves actual production work such as mineral exploration and geological mapping, completing a closed loop from data processing to geological application.

[0090] This application provides a method for extracting local aeromagnetic anomalies based on a BP neural network. Utilizing the powerful nonlinear mapping capability of the BP neural network, it learns the characteristics of forward model data to achieve high-precision extraction of local anomalies in complex backgrounds. A combined model of "deep, large background + shallow, strong anomalies" is specifically designed to simulate real, complex geological conditions. "Input-target" pairs are generated using forward model data, allowing the network to learn how to "see through" the background. The trained BP network is then used as a nonlinear filter and directly applied to measured data.

[0091] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 2 The diagram shows a structural schematic of an electronic device 300 suitable for implementing embodiments of this application.

[0092] like Figure 2As shown, the electronic device 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the device 300. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0093] The following components are connected to I / O interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 306 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0094] In particular, according to embodiments of this disclosure, the above references Figure 1 The described process can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program containing program code for performing the above-described method for extracting aeromagnetic local anomalies based on a BP neural network. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311.

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0096] The units or modules described in the embodiments of this application can be implemented in software or hardware. The described units or modules can also be located in a processor. The names of these units or modules do not necessarily constitute a limitation on the unit or module itself.

[0097] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, a laptop computer, a mobile phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0098] In another aspect, this application also provides a storage medium, which may be the storage medium included in the aforementioned apparatus in the above embodiments; or it may be a standalone storage medium not assembled into the device. The storage medium stores one or more programs, which are used by one or more processors to execute the aeromagnetic local anomaly extraction method based on a BP neural network described in this application.

[0099] Storage media, including both permanent and non-permanent, removable and non-removable media, can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0100] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0101] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0102] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for extracting local anomalies of aeromagnetic data based on BP neural network, characterized in that, The method comprises the following steps: S1: obtaining measured aeromagnetic data; S2: performing gridding and normalization preprocessing on the measured aeromagnetic data to obtain preprocessed measured aeromagnetic data; S3: inputting the preprocessed measured aeromagnetic data into a trained BP neural network model to output local anomaly data extracted from the measured aeromagnetic data.

2. The method of claim 1, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, The training data set of the BP neural network model is generated by forward calculation according to an aeromagnetic anomaly forward model, and the aeromagnetic anomaly forward model comprises a background field model for simulating magnetic anomalies of deep large-scale geological bodies and a local anomaly model for simulating magnetic anomalies of shallow small-scale geological bodies.

3. The method of claim 2, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, The background field model adopts a horizontal plate-shaped body, and assumes that the observation surface of the horizontal plate-shaped body is , and the center is ; The background field model produces magnetic anomalies The cuboid magnetic anomaly forward formula is adopted: wherein, is the vacuum permeability, H / m, is the deep background field magnetization; is the observation point coordinate, is the coordinate difference corresponding to the integral variable, , , is the field source point coordinate; ; are the direction cosine coefficients related to the geomagnetic inclination , the geomagnetic declination and the magnetization direction, respectively; the integral limits correspond to the geometric boundaries of the plate-shaped body: , is half the length of the plate-shaped body, is half the width of the plate-shaped body, is the thickness of the plate-shaped body, is the depth of the plate-shaped body interface, , , , , , .​ 4. The method of claim 3, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, The local anomaly body in the local anomaly model adopts a sphere; The local anomaly model generates an anomaly is: wherein, is the vacuum permeability, is the sphere buried depth; is the sphere radius; is the magnetization of the local anomaly, and is set , the sphere anomaly center observation plane projection point coordinates .

5. The method of claim 4, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, Forward model magnetic anomaly The forward model magnetic anomaly is calculated using the aeromagnetic anomaly forward model. (3) All sets of magnetic anomalies of the forward models are used as the training data set.

6. The method of claim 2, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, The BP neural network model comprises an input layer, a hidden layer and an output layer. The input layer adopts a sliding window mode to collect spatial data features, a window size of N*N grid points is set, the number of input layer nodes is N*N, and the input of the input layer is the normalized value of the training data set data. The hidden layer nodes are determined according to an empirical formula wherein, is the number of input layer nodes, is the number of output layer nodes, is a constant between 1 and 10; The output layer is provided with one node, and the target output is a theoretical local anomaly value corresponding to the center point of the window.

7. The method of claim 6, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, Training the BP neural network model comprises: performing gridding and normalization processing on the full-field data generated by the forward model, and using a sliding window to intercept samples; inputting the samples into the established BP neural network model, performing forward propagation, and calculating the output value of the BP neural network model; calculating the mean square error between the output value and the theoretical local anomaly value; based on the mean square error, calculating the gradient of each weight and bias parameter in the BP neural network model by using a back propagation algorithm; according to the gradient, updating the weight and bias parameters by using a gradient descent algorithm.

8. The method of claim 1, wherein the BP neural network-based extraction of local aeromagnetic anomalies is characterized by, Further comprising: combining all the local anomaly data into a grid to draw a local anomaly contour map.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the method for extracting aeromagnetic local anomalies based on a BP neural network according to any one of claims 1-8.

10. A readable storage medium, having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the method for extracting aeromagnetic local anomalies based on a BP neural network according to any one of claims 1-8.

Images