Local geomagnetic field database building method based on improved super learning device

Abstract

The invention provides a local geomagnetic field database building method based on an improved super learning device. The method comprises the following steps: acquiring magnetic field distribution of a task area; fitting the Gaussian semi-variation function parameter and the index semi-variation function parameter, and interpolating the regional magnetic field by using a Kriging interpolation method to obtain a Gaussian interpolation result and an index interpolation result; interpolating the regional magnetic field by using the five machine learning sub-models to obtain five interpolation results, adjusting the weight of the interpolation result of each sub-model by using a GBM learning device, and combining to obtain a GBM weighted interpolation result of the machine learning sub-model; the GBM learner is used for adjusting the Gaussian interpolation result, the index interpolation result and the weight of the weighted interpolation result of the machine learning sub-model finally obtained in the fifth step, and a final GBM learner interpolation result is obtained; and establishing a local geomagnetic field database. According to the technical scheme, the technical problems that in the prior art, Kriging interpolation is poor in precision in a complex environment with strong magnetic field nonlinearity, and a single machine learning model is poor in universality are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of aeromagnetic exploration, and in particular to a method for building a local geological magnetic field database based on an improved super learner. Background Art

[0002] A geological magnetic field database is a magnetic field database that encompasses the magnetic fields on a plane at a specific altitude with specific longitude and latitude coordinates in a specific area due to geological magnetic field anomaly sources. The magnetic field near the Earth's surface comes from the main magnetic field (accounting for 95%, described by models such as IGRF and WFF), the geological magnetic field from the lithosphere (accounting for 4%), and the diurnal variation field from the universe and the sun (accounting for 1%). The geological magnetic field originates from metal deposits such as iron, cobalt, and nickel inside the lithosphere. Specifically, due to different rock types, compositions, and distributions within the Earth's crust lithosphere, the ferromagnetic substances in the lithosphere are magnetized by the geomagnetic field, and more significant magnetic anomalies will appear within a small scale range. The potential magnetic field sources are generally metal deposits such as iron, cobalt, and nickel. The depth of iron ore deposits ranges from dozens of meters to hundreds of meters, generally located in the upper part of the crust; cobalt and nickel ores are mostly associated with iron ores. The nickel ore is vertically distributed at about several hundred meters to one kilometer, and the cobalt ore is concentrated in copper-nickel sulfide deposits and certain types of sedimentary deposits. The vertical extension of these deposits is about several hundred meters. Although such minerals also appear in the deep part, the temperature and pressure conditions for mineralization are more suitable in the upper layer of the crust, and hydrothermal mineralization and sedimentary mineralization processes are more likely to occur in the upper part. Therefore, such interfering source deposits are generally distributed in the upper part of the crust and vertically distributed between several hundred meters and one kilometer.

[0003] In addition, if the area to be measured is located in a geological fault zone, the continuity of the rock types in this area is interrupted due to formation fractures in space, and the distribution of geological magnetic sources is discontinuous, resulting in sudden geological magnetic field anomalies. In addition, magma activity and rock fractures are more likely to occur within the crust around the fault zone, leading to a significant increase in the probability of magma hydrothermal action, and the ferromagnetic ore deposits are further magnetized. Around volcanoes, the geological magnetic field sources are also affected by the lava activities in this area. At the boundaries of the Earth's plates, there are plate newborn zones or plate extinction zones. The plate extinction zones cause the crust to be squeezed, uplifted, or one plate to be squeezed under another plate and subducted into the mantle, while the plate newborn zones are located at mid-ocean ridges, resulting in intense volcanic activities, making the distribution of the geological magnetic field in this area more complex and unpredictable.

[0004] In summary, the geological magnetic field is a type of magnetic field that accounts for a large proportion of the total magnetic field except for the main magnetic field, is relatively unstable in time, and is distributed more arbitrarily in space. Due to the above characteristics, it is difficult to establish a model for the geological magnetic field. Therefore, it is necessary to find a suitable method to establish a database for the geological magnetic field of the area to be measured. This requires interpolation based on the sample points of the geological magnetic field measured during the survey as known points to obtain the geological magnetic field of this area. Common interpolation methods include linear interpolation, inverse distance weighted method, etc.

[0005] The linear interpolation method uses a linear model to interpolate by extrapolating the measured values ​​of known measuring point coordinates. This method is simple in logic, but due to the characteristics of severe mutation and complex structure of the distribution of geological magnetic field sources, the error generated by interpolation using a simple linear model is large, especially when the survey lines are sparsely arranged, and the geological magnetic field anomaly sources between the survey lines may even be missed.

[0006] The inverse distance weighting method is based on the following basic assumptions: the value of an unknown point is more affected by nearby points than by distant points, and is inversely proportional to several powers of the distance between the points, which are related to the properties of the field to be interpolated. The advantages of this idea are clear ideas and simple algorithm structure, but the disadvantages are also very obvious: since the geological magnetic field source often comes from the geological structure in the lithosphere, the spatial correlation of the magnetic field brought by this structure is not applied in the inverse distance weighting method, so this method has a poor overall grasp effect, and the results are significantly affected by the uneven distribution of measurement points.

[0007] The goal of the minimum curvature method is to generate the smoothest surface to fit the known measured points. This method is suitable for surface interpolation and does not rely on specific statistical models and assumptions. However, this method is easy to ignore these details at extreme values ​​or sharp changes, and the interpolation result is smoother than the actual situation. Since the mineral deposits corresponding to the geological magnetic field source are more likely to appear near special geological structures such as fault zones and plate boundaries, the geological magnetic field is more likely to have jump-type anomalies, so it is not the most suitable choice in the application scenario of geological magnetic field.

[0008] Machine learning interpolation is a method of interpolating the interpolation area after learning the survey sampling point training model based on the machine learning model. Compared with the various interpolation methods mentioned above, its biggest advantage is that it can be well applied to complex nonlinear scenarios, such as situations with drastic changes in a small area. However, although machine learning can well capture the complex nonlinear relationship between input features and target values, it cannot directly capture the correlation of magnetic fields in space like geostatistical methods, and the versatility of a single machine model is not strong enough. For example, a model has strong noise resistance but its accuracy decreases when the measurement points are scarce, making it difficult to achieve accurate interpolation in various complex environments.

[0009] The Kriging method is a method that uses known measurement points as samples and fits a semivariogram that describes the spatial correlation of data. This function depicts the relationship between the field value and distance between two points in the region based on spatial correlation, and contains the spatial structure information of the field to be determined in space. It is the core of the Kriging method. After obtaining this function through fitting of known measurement points, the weights of each measurement value can be obtained through this function, thereby completing the interpolation of the points to be interpolated.

[0010] Compared with other interpolation methods, the Kriging method has the following advantages in this scenario: First, based on the principle of spatial autocorrelation, this interpolation method generates more accurate predictions by quantitatively describing spatial correlation through semivariogram, which is especially applicable in the case of geological magnetic field where the spatial variation of magnetic field is very significant. In contrast, the known measuring points of traditional interpolation methods are only affected by distance and ignore spatial correlation; second, the Kriging method can provide uncertainty estimates for each measuring point, and secondary collection can be performed for high uncertainty areas to ensure accuracy, while other interpolation methods cannot provide uncertainty assessment; third, when the measuring points are unevenly distributed, the Kriging method can automatically adjust the interpolation weight according to the distribution of measuring points, reducing the possibility of large interpolation errors in sparse measuring point areas. However, the Kriging method has certain limitations when the interpolation field is complex, mainly because its premise is that the interpolation quantity is stable in space, and steep interference sources such as underground iron ore veins will destroy this premise in a small area.

[0011] In summary, the Kriging method can capture the spatial correlation of the field to be interpolated, but it has limitations when the field to be interpolated is relatively complex and changes dramatically within a small scale. A single machine learning method can be well applied in complex nonlinear scenarios, but it cannot capture the spatial correlation of the field to be interpolated, and its versatility in various complex environments is low. Therefore, it is necessary to find a high-precision method for interpolation of regional geological magnetic fields. Summary of the invention

[0012] The present invention provides a method for building a local geological magnetic field database based on an improved super learner, which can solve the technical problems in the prior art that Kriging interpolation has poor accuracy in complex environments with strong magnetic field nonlinearity and a single machine learning model has poor versatility.

[0013] According to one aspect of the present invention, a method for building a local geological magnetic field database based on an improved super learner is provided. The method for building a local geological magnetic field database based on the improved super learner includes: step one, using a drone as an aerial magnetic exploration platform to collect magnetic field distribution in a mission area; step two, selecting the semivariogram function as an exponential and a Gaussian function, using the sampled data to fit the Gaussian semivariogram function parameters and the exponential semivariogram function parameters, and using the Kriging interpolation method based on the Gaussian semivariogram function and the exponential semivariogram function to interpolate the regional magnetic field to obtain Gaussian interpolation results and exponential interpolation results; step three, using the magnetic field sampling data obtained in step one Train each machine learning sub-model, use the five machine learning sub-models to interpolate the regional magnetic field, obtain five interpolation results, use the GBM learner to adjust the weight of the interpolation results of each sub-model, and combine them to obtain the machine learning sub-model GBM weighted interpolation result; step four, use the GBM learner to adjust the weights of the Gaussian interpolation result, the exponential interpolation result and the weighted interpolation result of the machine learning sub-model finally obtained in step five, and then use the GBM learner to weight the three interpolation results to obtain the final GBM learner interpolation result; step five, draw a geological magnetic interference distribution map according to the final GBM learner interpolation result in step six, and establish a local geological magnetic field database.

[0014] Furthermore, in step 1, since the aerial magnetic exploration uses a drone as a detection platform, a survey line is arranged to fly over the area to be measured along the survey line at the height to be measured to collect the magnetic field distribution of the area, ensuring that the survey line is distributed north-south, and the survey line interval is determined according to the required scale; the geological magnetic field B Terrain According to B Tot =B plat&earth +B Terrain +B Sun +B grad Calculate and obtain, where B Tot is the total interference, B Sun is the diurnal interference, B grad is the geomagnetic gradient interference, B plat&earth It is platform interference and geomagnetic gradient interference.

[0015] Furthermore, in step 2, the regional magnetic field is interpolated using the Kriging interpolation method based on the Gaussian semivariogram, specifically including: obtaining the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated based on the Gaussian semivariogram; calculating and obtaining the weight vector according to the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; calculating and obtaining the field value of the point to be interpolated according to the weight vector and the vector composed of the known point measurements, and completing the interpolation of each point outside the survey line according to the field value of the point to be interpolated to obtain the Gaussian interpolation result.

[0016] Further, in step two, the specific steps of using the Kriging interpolation method based on the exponential semivariogram to interpolate the regional magnetic field are as follows: based on the exponential semivariogram, obtain the covariance matrix of each point on the survey line and the covariance vector between each measurement point on the survey line and the point to be interpolated; calculate and obtain the weight vector according to the covariance matrix of each point on the survey line and the covariance vector between each measurement point on the survey line and the point to be interpolated; calculate and obtain the field value of the point to be interpolated according to the weight vector and the vector composed of the measured values of the known points, and complete the interpolation of each point outside the survey line according to the field value of the point to be interpolated to obtain the exponential interpolation result.

[0017] Further, the field value of the point u to be interpolated can be calculated according to zλ = Z(u), where z is the vector composed of the measured values of the known points, λ is the weight vector, and Z(u) is the field value of the point u to be interpolated.

[0018] Further, the weight vector λ is calculated according to Cλ = c, where C is the covariance matrix based on the covariance matrix of each point on the survey line, and c is the covariance vector between each measurement point on the survey line and the point to be interpolated.

[0019] Further, the five machine sub-models are Random Forest Model, Support Vector Machine (SVM), Gaussian Process Regression (GPR), K-Nearest Neighbor Model (KNN), and Generalized Additive Model (GAM).

[0020] According to another aspect of the present invention, there is provided a local geological magnetic field database building system based on an improved super learner. The local geological magnetic field database building system based on the improved super learner uses the above-mentioned local geological magnetic field database building method based on the improved super learner to build a local geological magnetic field database.

[0021] According to still another aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the steps of the above-mentioned local geological magnetic field database building method based on the improved super learner.

[0022] According to yet another aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the steps of the above-mentioned local geological magnetic field database building method based on the improved super learner.

[0023] Applying the technical solution of the present invention, a method for building a local geological magnetic field database based on an improved super learner is provided. This method adopts five machine learning models and the Kriging interpolation method, and innovatively introduces a GBM meta-learner for dynamically adjusting the weights of each interpolation result to integrate each machine learning sub-model to establish a super learner. This method can not only intelligently identify the weights of each sub-model in the super learner through the GBM meta-learner, but also adaptively adjust the weights of Kriging interpolation or super learner interpolation under different geological magnetic field conditions, systematically solving the problems of poor accuracy of Kriging interpolation in complex environments with strong magnetic field nonlinearity and poor versatility of a single machine learning model. The interpolation accuracy has been greatly improved compared with traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings included are used to provide a further understanding of the embodiments of the present invention, which form a part of the specification, are used to illustrate the embodiments of the present invention, and are used to explain the principles of the present invention together with the text description. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0025] Figure 1 Shows the overall steps of the specific implementation of the present invention;

[0026] Figure 2 Shows the process of Kriging interpolation;

[0027] Figure 3 Shows the establishment process of the super learner of the present invention;

[0028] Figure 4 Shows the data processing process of the GBM meta-learner adopted by the present invention;

[0029] Figure 5 Shows the magnetic field distribution map and sampling point distribution of a certain area in Yukon, Canada;

[0030] Figure 6 and Figure 7 Shows the interpolation results of the magnetic field distribution in this area using the Gaussian and exponential semi-variogram functions adopted by traditional Kriging Figure 5 for the measuring points in;

[0031] Figure 8 Shows the interpolation results of the magnetic field distribution in this area using the super learner with the traditional meta-learner integrating sub-models Figure 5 for the measuring points in;

[0032] Figure 9 Shows the use of the present invention Figure 5The interpolation result of the magnetic field distribution in this area by the mid-measurement point. Specific implementation manners

[0033] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way limits the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0034] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or their combinations.

[0035] Unless otherwise specifically stated, the relative arrangements of the components and steps, numerical expressions, and values set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that, for the sake of convenience of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationships. Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the authorized specification. In all the examples shown and discussed here, any specific value should be interpreted as merely exemplary and not as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that: like reference numerals and letters denote like items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0036] Such as Figures 1 to 4As shown in the figure, according to a specific embodiment of the present invention, a method for building a local geological magnetic field database based on an improved super learner is provided. The method for building a local geological magnetic field database based on an improved super learner includes: Step 1, using a drone as an airborne magnetic exploration platform to collect the magnetic field distribution in the task area; Step 2, selecting the semi-variogram as the exponential and Gaussian functions, using the sampling data to fit the Gaussian semi-variogram parameters and the exponential semi-variogram parameters, and using the Kriging interpolation method based on the Gaussian semi-variogram and the exponential semi-variogram to interpolate the regional magnetic field to obtain the Gaussian interpolation result and the exponential interpolation result; Step 3, using the magnetic field sampling data obtained in Step 1 to train each sub-model of machine learning, using five sub-models of machine learning to interpolate the regional magnetic field to obtain five interpolation results, and using the GBM learner to adjust the weights of the interpolation results of each sub-model, and combining to obtain the GBM weighted interpolation result of the machine learning sub-model; Step 4, using the GBM learner to further adjust the weights of the Gaussian interpolation result, the exponential interpolation result, and the weighted interpolation result of the machine learning sub-model finally obtained in Step 5. The three interpolation results are weighted by the GBM learner to obtain the final GBM learner interpolation result; Step 5, draw a geological magnetic interference distribution map according to the final GBM learner interpolation result in Step 6, and establish a local geological magnetic field database.

[0037] With this configuration method, a method for building a local geological magnetic field database based on an improved super learner is provided. This method adopts five machine learning models and the Kriging interpolation method, and innovatively introduces the GBM meta-learner for dynamically adjusting the weights of each interpolation result to integrate each machine learning sub-model to establish a super learner. This method can not only intelligently identify the weights of each sub-model in the super learner through the GBM meta-learner, but also adaptively adjust the weights of the Kriging interpolation or the super learner interpolation under different geological magnetic field conditions, systematically solving the problems of poor accuracy of the Kriging interpolation in a complex environment with strong magnetic field non-linearity and poor universality of a single machine learning model. The interpolation accuracy has been greatly improved compared with the traditional method.

[0038] The method proposed by the present invention includes the following steps:

[0039] (1) Using a drone as an airborne magnetic exploration platform to collect the magnetic field distribution in the task area.

[0040] (2) Determining the semi-variogram by adaptive Kriging interpolation and pre-interpolating the regional magnetic field.

[0041] (3) Using a machine learning model to train the regional magnetic field and perform fine interpolation on the regional magnetic field.

[0042] (4) Adopting GBM as a meta-learner to establish a super learner and fusing each interpolation result to obtain the super learner interpolation result.

[0043] (5) Draw a geological magnetic interference distribution map based on the interpolation results of the super learner and establish a database.

[0044] Specifically, in step 1, since the aerial magnetic exploration uses drones as detection platforms, the survey line is arranged to fly over the area to be measured along the survey line at the height to be measured to collect the magnetic field distribution of the area, ensuring that the survey line is distributed north-south, and the survey line interval is determined according to the required scale; geological magnetic field B Terrain According to B Tot =B plat&earth +B Terrain +B Sun +B grad Calculate and obtain, where B Tot is the total interference, B Sun is the diurnal interference, B grad is the geomagnetic gradient interference, B plat&earth It is platform interference and geomagnetic gradient interference.

[0045] Furthermore, in step 2, the regional magnetic field is interpolated using the Kriging interpolation method based on the Gaussian semivariogram, specifically including: obtaining the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated based on the Gaussian semivariogram; calculating and obtaining the weight vector according to the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; calculating and obtaining the field value of the point to be interpolated according to the weight vector and the vector composed of the known point measurements, and completing the interpolation of each point outside the survey line according to the field value of the point to be interpolated to obtain the Gaussian interpolation result.

[0046] In step 2, the regional magnetic field is interpolated using the Kriging interpolation method based on the exponential semivariogram function, specifically including: based on the exponential semivariogram function, obtaining the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; calculating and obtaining the weight vector according to the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; calculating and obtaining the field value of the point to be interpolated according to the weight vector and the vector composed of the known point measurements, and completing the interpolation of each point outside the survey line according to the field value of the point to be interpolated to obtain the exponential interpolation result.

[0047] The field value of the interpolation point u can be calculated according to zλ=Z(u), where z is a vector of known point measurements, λ is a weight vector, and Z(u) is the field value of the interpolation point u. The weight vector λ is calculated according to Cλ=c, where C is the covariance matrix based on the covariance matrix of each point on the survey line, and c is the covariance vector between each point on the survey line and the interpolation point.

[0048] In the present invention, the five machine sub-models are a random forest model, a support vector machine (SVM), a Gaussian process regression (GPR), a K-nearest neighbor model (KNN), and a generalized additive model (GAM).

[0049] According to another aspect of the present invention, there is provided a local geological magnetic field database building system based on an improved super learner. This local geological magnetic field database building system based on an improved super learner uses the local geological magnetic field database building method based on an improved super learner as described above to build a local geological magnetic field database.

[0050] Applying this configuration method, there is provided a local geological magnetic field database building system based on an improved super learner. This system adopts five machine learning models and the Kriging interpolation method, and innovatively introduces a GBM meta-learner for dynamically adjusting the weights of each interpolation result to integrate each machine learning sub-model to establish a super learner. This method can not only intelligently identify the weights of each sub-model in the super learner through the GBM meta-learner, but also adaptively adjust the weights of Kriging interpolation or super learner interpolation under different geological magnetic field conditions, systematically solving the problems of poor accuracy of Kriging interpolation in complex environments with strong magnetic field non-linearity and poor versatility of a single machine learning model. The interpolation accuracy has been greatly improved compared with traditional methods.

[0051] According to yet another aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the steps of the local geological magnetic field database building method based on an improved super learner as described above.

[0052] According to still another aspect of the present invention, there is provided a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the local geological magnetic field database building method based on an improved super learner as described above.

[0053] For a further understanding of the present invention, the following Figures 1 to 9 provides a detailed description of the local geological magnetic field database building method based on an improved super learner provided by the present invention.

[0054] As Figures 1 to 9 shown, the present invention provides a method for establishing a database of the geological magnetic field in a task area. First, magnetic field distribution samples are collected, and then interpolation is performed, and finally a local geological magnetic field database is established.

[0055] (1) Using an unmanned aerial vehicle as an airborne magnetic exploration platform, collect the magnetic field distribution in the task area.

[0056] Since an unmanned aerial vehicle (UAV) is used as the detection platform for airborne magnetic exploration, survey lines are arranged to fly circuitously along the survey lines at the required height over the area to be measured to collect the magnetic field distribution of the area. Ensure that the survey lines are distributed north-south, and determine the survey line interval according to the required scale. Generally, the survey line spacing is one percent of the scale. For example, for a scale of 1:10,000, the survey line spacing is 100 m.

[0057] (2) Determine the semivariogram by adaptive Kriging interpolation and pre-interpolate the magnetic field of the area.

[0058] Kriging interpolation is a geological interpolation method that interpolates based on the coordinates and measured values of several known points in an unknown area, grasping its spatial correlation. It is necessary to determine the parameters of the semivariogram, so as to achieve higher accuracy. The semivariogram parameters can be determined using the maximum likelihood parameter method. After determination, Kriging interpolation can be completed.

[0059] (3) Adopt a machine learning model to train the magnetic field of the area and perform fine interpolation on the regional magnetic field.

[0060] In order to make this method adapt to various adverse and complex environments as much as possible, such as uneven distribution of measurement points, large noise, and strong magnetic field non-linearity, different machine learning models are adopted to interpolate this area to obtain the interpolation results of each machine learning. The reason for adopting several machine learning models is that different machine learning models have their own advantages and are respectively suitable for different adverse environments.

[0061] (4) Adopt GBM as the meta-learner to establish a super-learner and fuse the interpolation results to obtain the super-learner interpolation result.

[0062] The interpolation results of each machine learning and the Kriging interpolation result need to be fused. The GBM meta-learner can adaptively learn through the training set and learn how to adaptively assign weights to each interpolation method. After the GBM meta-learner is trained, this learner is used to integrate the interpolation results to obtain the super-learner interpolation result.

[0063] (5) Draw a geological magnetic interference distribution map based on the improved super-learner interpolation result and establish a database.

[0064] In the previous step, the improved super-learner is based on the fusion of the final interpolation results of each machine learning sub-model and the Kriging pre-interpolation result, and a geological magnetic interference distribution map is drawn according to this result. For the convenience of data access and search, the Database Toolbox in MATLAB is adopted, and the interpolation results are stored in the database table in the form of SQL statements to complete the effective management of the data.

[0065] The following is a detailed description of the method for constructing a local geological magnetic field database based on an improved super learner provided by the present invention in combination with specific embodiments.

[0066] (1) The UAV platform flies along the survey line over the area to be measured, and after collection and processing, the geological magnetic field data of the area is obtained.

[0067] During the airborne magnetic exploration process, the distribution interval of the survey lines needs to be determined according to the required scale. Generally speaking, the line spacing is one percent of the scale. For example, for a scale of 1:10,000, the line spacing is 100 m. In this step, the interference from the fuselage platform has been compensated. The geomagnetic field is obtained during the operation of the platform due to the platform position and altitude, and can be compensated by a perfect geomagnetic field model. The diurnal variation field is compensated by the accompanying platform or the diurnal variation station. The remaining is the geological magnetic field. Specifically, the geological magnetic field B Terrain , the total interference B Tot , the diurnal variation interference B Sun , the geomagnetic gradient interference B grad and the platform interference, geomagnetic gradient interference B plat&earth satisfy the following relationship:

[0068] B Tot = B plat&earth + B Terrain + B Sun + B grad

[0069] And the platform interference, geomagnetic main field interference B plat&earth can be solved in the first step of calibration flight and the platform longitude, latitude coordinates, and altitude. The geomagnetic gradient interference B grad can be solved from the longitude, latitude, and altitude data. The diurnal variation field B Sun is compensated by a formation composed of several platforms, and the remaining geological magnetic field B Terrain can be obtained immediately.

[0070] The geological magnetic field data obtained in this process will be used as samples for geological magnetic field interpolation to establish a geological magnetic field database.

[0071] (2) Adaptive Kriging interpolation is used to determine the semi-variogram and pre-interpolate the magnetic field of the area.

[0072] The principle of the Kriging interpolation method is as follows:

[0073]

[0074] Where Z * (s 0 ) is the value to be interpolated at the point s 0 to be interpolated, and Z(s i ) is the known measured point s iObservation value, λ i is the weight of the measurement point s i and n is the number of known measurement points on the measurement line for interpolation. The key of the Kriging method lies in how to determine the weights, which depend on the spatial correlation between the known points on the measurement line and the point to be interpolated.

[0075] The semivariogram describes the relationship of the spatial random process changing according to the spatial coordinates and step length. If the spatial process is stationary, then its value only depends on the distance h from the measurement point. At this time, the semivariogram is a unary function γ(h). In terms of form, it can be expressed as:

[0076]

[0077] Z(x) and Z(x + h) are the values of the random process at two points x and x + h, that is, the value of the semivariogram is half of the mathematical expectation of the square of the difference of the random process at two points x and x + h. However, since in fact the overall spatial distribution of the random process within the region cannot be known a priori, the semivariogram cannot actually be obtained as above. In practice, it can only be obtained by fitting the semivariogram through the measured values at several known measurement points.

[0078] The semivariogram is a general expression form of a random process. Similar to any random process, its semivariogram can be obtained according to this formula. In fact, because the overall situation of the entire random process cannot be known in advance, it can only be fitted with a suitable model according to the properties of this random process. Due to the cubic attenuation property of geological magnetic interference, the Gaussian model and the exponential model can fit this process of geological magnetic interference better than other models. There are eight commonly used semivariograms, but the Gaussian and exponential models are the most suitable for this scenario.

[0079] Commonly used semivariograms include the exponential model, spherical model, Gaussian model, linear model, etc. Among them, the models that show exponential decay with distance are the exponential model and the Gaussian model. First is the exponential model:

[0080]

[0081] Then is the Gaussian model:

[0082]

[0083] where θ i is the correlation parameter of the point to be interpolated and the sampling point in the i-th direction, and d i is the distance between the point to be interpolated and the sampling point in the i-th direction.

[0084] The relevant parameters of the semivariogram are solved through the DACEToolbox, a toolbox for the Kriging method. This toolbox integrates seven semivariograms, namely spherical, Gaussian, exponential, linear, generalized exponential, Matern, cubic, and linear models, as well as stationary and first-order polynomial background models. The toolbox internally integrates the maximum likelihood parameter (MLE) algorithm to solve the parameters of the semivariogram.

[0085] Specifically, the exponential and Gaussian functions are selected as the semivariograms. The sampling data is used to fit the parameters of the Gaussian semivariogram and the exponential semivariogram. The Kriging interpolation method based on the Gaussian semivariogram is used to pre-interpolate the regional magnetic field to obtain the Gaussian interpolation result. The Kriging interpolation method based on the exponential semivariogram is used to pre-interpolate the regional magnetic field to obtain the exponential interpolation result.

[0086] After the semivariogram parameters are confirmed, the covariance between the measured values at each measurement point can be obtained immediately. The covariance between the measured values of any two points satisfies:

[0087] Cov(Z(s i ),Z(s j ))=σ 2 -γ(h i,j )

[0088] That is, the covariance Cov(Z(s i ),Z(s j )) between the measured values at point s i and point s j and the distance h i,j between the two points satisfy the above relationship. Z(s i ) and Z(s j ) are the measured values at point i and point j. σ 2 is the total variance of the known data. The covariance between the values at point i and point j can be obtained by this method. The meaning of γ is the semivariogram. Here, it means that: the covariance between the measured values Z(s i ) and Z(s j ) at points s i and s j is the difference between the variance of the known data and the value of the semivariogram when h ij is taken.

[0089] When interpolating an unknown point, point i is a known point and point j is an unknown point. At this time, σ 2 is still the total variance of the known data, and the distance h i,j between the two points is known. Therefore, the covariance between the known point and the point to be interpolated can still be obtained.

[0090] The covariance matrix is determined based on the covariance matrix C of each point on the survey line and the covariance vector c between each measurement point on the survey line and the point to be interpolated. The relationship between them satisfies the following formula:

[0091] Cλ = c

[0092] Where C is an n×n covariance matrix, and λ represents the vector composed of the weights of each known measurement point when interpolating the unknown point. The elements satisfy:

[0093] C i,j = Cov(Z(s i ), Z(s j )) = σ 2 -γ(h i,j )

[0094] Where h i,j is the distance between the measurement points s i and s j . The covariance of the measured value Z(s i ) of the i-th measurement point s i and the measured value Z(s j ) of the j-th measurement point s j can be solved through the total data variance σ 2 and the semivariogram γ.

[0095] And c is an n×1 covariance vector. The j-th element c j of this vector represents the covariance between the measured value Z(s j ) at the j-th known point s i and the field value Z(u) of the point u to be interpolated:

[0096]

[0097] Each element of the covariance vector c can be indirectly obtained through the semivariogram and the sample variance based on the distances between the known measurement points and the point to be interpolated as shown above without directly knowing the field value Z(u) of the point to be interpolated. Thus, the linear equation system Cλ = c can be solved, and finally the weight vector λ is obtained. Then, interpolation can be performed on the target point, thereby completing the establishment of the spatial geological magnetic field at this point, as follows.

[0098] zλ = Z(u)

[0099] Where z is the vector composed of the measured values of the known points, λ is the weight vector, and Z(u) is the field value of the point u to be interpolated.

[0100] Interpolate each point outside the survey line in this way repeatedly, and finally obtain the two-dimensional geological magnetic field data of the spatial geological magnetic field at this height.

[0101] The exponential semivariogram is suitable for continuous regions where the spatial variation of the magnetic field is relatively fast, and the Gaussian semivariogram is suitable for continuous regions where the magnetic field changes gently. Both can capture the spatial correlation of the magnetic field.

[0102] (3) Adopt a machine learning model to train the magnetic field in this area. For the complex situation of complex and steeply changing magnetic fields, fine interpolation of the regional magnetic field is required. Generally, there are interference sources such as iron veins underground at this time, and the spatial stationarity and continuity of the magnetic field are damaged, and the data distribution shows a complex non-linear relationship. Therefore, a machine learning model needs to be used for processing. The role of the super learner is to be based on ensemble learning and comprehensively use various methods to make the prediction effect more robust and accurate than each method alone. To achieve this goal, first, appropriate sub-models need to be selected to obtain a set of interpolation results respectively, and then an appropriate meta-learner is selected to assign different weights to the results. There are several ways to implement the meta-learner: linear regression, weighted linear regression, GBM, etc. After experiments here, GBM is more suitable as the meta-learner.

[0103] The super learner is a special machine learning model that integrates multiple machine learning models and flexibly combines each sub-machine learning model using a meta-learner according to specific situations. Since the super learner is used to flexibly use different machine learning models to process the magnetic field interpolation of complex non-linear relationships, it is very important to select appropriate machine learning models to overcome various adverse situations. Common adverse situations include sparse measurement points, different variation laws of the magnetic field in each dimension, large noise, strong non-linearity, and violent magnetic field changes, etc.

[0104] The present invention selects the following five sub-machine learning models:

[0105] The random forest model is based on the integration of multiple decision trees and can capture the high-dimensional non-linear relationship between input and output data.

[0106] The support vector machine (SVM) can process high-dimensional non-linear data by selecting an appropriate kernel function.

[0107] Gaussian process regression (GPR) is a probability-based model that can obtain the predicted value and the uncertainty interval of the predicted value.

[0108] The K-nearest neighbor model (KNN) is a model that predicts the point to be interpolated based on the distances of adjacent measurement points around the point to be interpolated.

[0109] The generalized additive model (GAM) models and superimposes the additive effects of different variables.

[0110] The present invention designs a meta-learner to combine the interpolation results of each sub-model according to the applicability of each sub-model under various adverse conditions, realizes the adaptive assignment of weights to each sub-model at each point, and improves the interpolation accuracy.

[0111] Specifically, the applicability of each machine learning model is as follows:

[0112] The random forest model has high robustness, but has large deviations when the measurement points are sparse or uneven. It performs well in complex nonlinear magnetic field environments and has strong noise tolerance.

[0113] The support vector machine model (SVM) performs well when the measurement points are sparse or unevenly distributed, the field changes differently in multiple dimensions, and the magnetic field changes are complex and nonlinear, but the calculation time is long.

[0114] Gaussian process regression (GPM) performs well in areas where the measurement points are sparse and the magnetic field variation pattern is more variable, but the calculation time is longer.

[0115] The K-nearest neighbor model (KNN) is suitable for processing smooth changes and can grasp the changes of the magnetic field in a single dimension, but it performs poorly in areas where the magnetic field changes nonlinearly and drastically and where there are few sampling points.

[0116] The generalized additive model (GAM) is suitable for capturing the relationship between a single input variable and output, but it has greater limitations when dealing with the relationship between multidimensional inputs. It is suitable for dealing with changes in a single dimension in spatial data, such as simply considering changes in the magnetic field in the east-west direction.

[0117] These five machine learning models cover a variety of unfavorable situations where traditional interpolation models perform poorly, including situations where the magnetic field changes differently in multiple dimensions, situations where measurement points are sparse, situations where background noise is strong, situations where the magnetic field changes gently and dramatically, and situations where the magnetic field has strong nonlinearity. For these unfavorable conditions, at least one machine learning model performs well, has a wider range of applicability, and the sub-models have good complementary capabilities.

[0118] All five models use the grid search algorithm to optimize hyperparameters, and use five-fold cross-validation to train the models to reduce the possibility of overfitting and improve the generalization ability of the models.

[0119] (4) Use the GBM learner as a meta-learner to fuse the interpolation results to obtain the final interpolation result.

[0120] In step (3), five machine learning models with different strengths are selected. In order to determine how to assign appropriate weights to each interpolation result in different locations to increase the interpolation accuracy, a meta-learner is required to integrate the interpolation results to generate the final interpolation result.

[0121] The present invention selects the Gradient Boosting Machine (GBM) as the meta-learner. This learner uses decision trees as basic units and can integrate various sub-machine learning models with their own advantages in applicability to establish a super-learner, thus effectively handling the complex non-linear changes of the spatial magnetic field. The super-learner adopts the meta-learner to integrate the interpolation results of various machine learning models, uses each measurement point as a training set for training, learns the characteristics of this area, and thus learns how to assign different weights to multiple machine sub-learning models according to the specific interpolation points to be interpolated, and adaptively combines the results of each sub-model to intelligently generate strong prediction results. In short, it adaptively determines how to assign the weights of each sub-model at the current point, and assigns weights according to local conditions at each point to obtain more accurate interpolation results. During the GBM training process, the LSBoost algorithm is used to optimize the hyperparameters, and then five-fold cross-validation is adopted to reduce the possibility of overfitting. The trained GBM flexibly assigns weights to the prediction results of each sub-model according to the input features at any interpolation point, and the weights of each sub-machine learning model at each point are dynamically adjusted according to the position. Since the geological magnetic field involved in the present invention is stable in a large range, but when it comes to geological interference sources such as iron ore veins, the magnetic field is no longer stable and is replaced by a complex non-linear relationship, so it is adaptively and dynamically adjusted near each interpolation point, giving a higher weight to the sub-model suitable for the local area, and significantly improving the interpolation accuracy.

[0122] The innovation of this method lies in that by using the dynamic weighting mechanism of the GBM meta-learner, it can flexibly adjust the weights of each sub-model at each spatial point, and thus has excellent performance in both the stable area with gentle magnetic field changes and the non-stable area with drastic magnetic field changes. Compared with the traditional single machine learning interpolation method and the Kriging interpolation method, the present invention has higher interpolation stability and accuracy in complex geological environments.

[0123] (5) Draw a geological magnetic interference distribution map based on the interpolation results of the super-learner and establish a database.

[0124] The present invention uses MATLAB to draw a heat map. The interpolation results use the Database Toolbox of MATLAB to automatically store the interpolation coordinates, interpolation results, and interpolation methods. Specifically, after interpolation, the encapsulated function in the Database Toolbox is used to insert into the pre-established database table in the form of an SQL statement. This process is completely automated and also facilitates the comparison and expansion of future interpolation results with the current data.

[0125] Effect verification

[0126] To actually verify that the present invention has performance improvements compared to traditional methods, experimental verification was carried out. The experimental results show that the interpolation method of the present invention has an error reduction of 10% to 30% in terms of the average relative error compared to the traditional method, and an error reduction of 5% to 10% in the root mean square error, fully demonstrating the superiority and versatility of the present invention under complex magnetic field conditions.

[0127] The experimental dataset selected the magnetic map distribution of the Canadian national land publicly released by the Canadian Geomatics department. The location is a mountain range near Yukon, the vegetation is virgin forest, and there are underground mineral deposits in this area, with a relatively complex magnetic environment. A rectangular area with a length of 8 km in the east-west direction, a width of 7 km in the north-south direction, and an area of 56 km 2 was selected as the experimental site.

[0128] The magnetic field distribution in this rectangular area has the following characteristics: it includes a "plain" area with a gentle magnetic field distribution, within which there are several local magnetic anomalies, and also includes a "rugged" area where the magnetic field changes violently with space, the spatial stationarity is destroyed, and it shows a complex non-linear change with space.

[0129] When taking measurement points in this area, the distribution of measurement points was simulated according to the requirements in the aviation magnetic exploration specification. The north-south distribution of the survey lines was adopted, the east-west survey line interval was 600 m, the measurement point interval on each survey line was 150 m, and a total of 12 survey lines were distributed in the entire area, and 44 measurement points were arranged on each survey line, evenly and comprehensively covering this rectangular area, as specifically shown in Figure 5 the figure.

[0130] Each model will use these measurement points to interpolate this area and compare with the data officially released by the Canadian government to evaluate the performance of each model.

[0131] The following models were used to interpolate this area respectively:

[0132] 1. Ordinary Kriging Gaussian semi-variogram interpolation (abbreviated as the KG model);

[0133] 2. Ordinary Kriging exponential semi-variogram interpolation (abbreviated as the KE model);

[0134] 3. Using a linear meta-learner to combine random forest, SVM, GPR, GAM, KNN (abbreviated as the SL-L5 model);

[0135] 4. Using a GBM meta-learner to combine random forest, SVM, GPR, GAM, KNN (abbreviated as the SL-GBM5 model);

[0136] 5. Using a linear meta-learner to combine the SL-GBM5 model and the KE and KG models (abbreviated as the KEG-SL-GBM5-L model);

[0137] 6. Adopt the GBM meta-learner, combine the SL-GBM5 model with the KE and KG models, and use GBM as the meta-learner for the super-learner model (abbreviated as the KEG-SL-GBM5-GBM model).

[0138] For the quality of the interpolation results, the following three indicators are used for evaluation:

[0139] 1. Maximum absolute error (ME): Compare the interpolation results with the geological magnetic map officially released by Canada, and the maximum absolute value of the error is the maximum absolute error.

[0140] 2. Mean absolute error (MAE): Compare the interpolation results with the geological magnetic map officially released by Canada, and take the average of the absolute values of the errors at each point to obtain the mean absolute error MAE.

[0141] 3. Root mean square error (RMSE): Compare the interpolation results with the geological magnetic map officially released by Canada, take the square root of the average of the squares of the errors at each point.

[0142] Since the interpolation results are prone to distortion when there are few measurement points at the edges, all interpolation results are trimmed by 10% of the size at the edges, and the middle area with a length of 7.2 km in the east-west direction and a width of 5.6 km in the north-south direction is taken. The official geological magnetic map is also trimmed in the same way, and the two are aligned for comparison to avoid the edge error caused by the distortion of the interpolation at the edges affecting the evaluation of the interpolation quality.

[0143] The interpolation errors of the six methods are shown in Table 1.

[0144] For the KEG-SL-GBM5-GBM model obtained by combining the hybrid Kriging Gaussian and exponential semi-variogram functions of the present invention with the super-learner using the GBM meta-learner weighted, compared with the traditional Kriging interpolation and the interpolation using the super-learner alone:

[0145] The ME is reduced by 16.9% compared with the traditional Kriging Gaussian interpolation, and is equivalent to the interpolation of the Kriging exponential interpolation, the super-learner model using the linear meta-learner or the GBM meta-learner.

[0146] The MAE is reduced by 30.3% compared with the traditional Kriging Gaussian or exponential interpolation, 21.5% lower than the interpolation using the super-learner adopting the linear meta-learner alone, and 10.8% lower than the interpolation using the super-learner adopting the GBM meta-learner alone.

[0147] The RMSE is 19.2% lower than that of the traditional Kriging Gaussian interpolation, 12.5% lower than that of the traditional Kriging exponential interpolation, 14.8% lower than that of the super learner interpolation using only the linear meta-learner, and 8.9% lower than that of the super learner interpolation using only the GBM meta-learner.

[0148] Table 1

[0149] Method category KE KG SL-L5 ME 169.42 214.39 175.70 MAE 22.83 22.14 20.37 RMSE 31.90 34.87 32.76 Method category SL-GBM5 KEG-SL-GBM5-L KEG-SL-GBM5-GBM ME 203.77 174.90 180.71 MAE 16.92 19.46 15.94 RMSE 30.62 29.78 27.90

[0150] In summary, the present invention provides a method for building a local geological magnetic field database based on an improved super learner. This method adopts five machine learning models and the Kriging interpolation method, and innovatively introduces the GBM meta-learner for dynamically adjusting the weights of each interpolation result to integrate each machine learning sub-model to establish a super learner. This method can not only intelligently identify the weights of each sub-model in the super learner through the GBM meta-learner, but also adaptively adjust the weights of the Kriging interpolation or the super learner interpolation under different geological magnetic field conditions, systematically solving the problems of poor accuracy of the Kriging interpolation in complex environments with strong magnetic field non-linearity and poor generality of a single machine learning model. The interpolation accuracy has been greatly improved compared with the traditional method.

[0151] For the sake of description, spatial relative terms such as "above", "over", "on the upper surface", "upper" etc. can be used here to describe the spatial position relationship between a device or feature shown in the figure and other devices or features. It should be understood that the spatial relative terms are intended to include different orientations in use or operation in addition to the orientation described in the figure for the device. For example, if the device in the figure is inverted, the device described as "above" or "over" other devices or structures will be positioned "below" or "under" other devices or structures after inversion. Thus, the exemplary term "above" can include both the orientations of "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the corresponding explanations for the spatial relative descriptions used here will be made.

[0152] In addition, it should be noted that using words such as "first", "second" etc. to limit components is only for the convenience of differentiating the corresponding components. Without additional statements, the above words have no special meanings, so they cannot be understood as limiting the protection scope of the present invention.

[0153] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for building a local geological magnetic field database based on an improved super learner, characterized in that: The local geological magnetic field database construction method based on the improved super learner includes: Step 1: Use the drone as an aerial magnetic exploration platform to collect magnetic field distribution in the mission area; Step 2: Select the semivariogram as exponential and Gaussian function, use the sampling data to fit the parameters of Gaussian semivariogram and exponential semivariogram, use the Kriging interpolation method based on Gaussian semivariogram and exponential semivariogram to interpolate the regional magnetic field, and obtain Gaussian interpolation results and exponential interpolation results; Step 3, using the magnetic field sampling data obtained in the step 1 to train each sub-model of machine learning, using the five sub-models of machine learning to interpolate the regional magnetic field to obtain five interpolation results, using the GBM learner to adjust the weight of the interpolation results of each sub-model, and combining to obtain the machine learning sub-model GBM weighted interpolation result; Step 4: Use the GBM learner to adjust the weights of the Gaussian interpolation result, the exponential interpolation result, and the weighted interpolation result of the machine learning sub-model finally obtained in step 5, and then weight the three interpolation results using the GBM learner to obtain the final GBM learner interpolation result; Step 5: Draw a geological magnetic interference distribution map based on the final GBM learner interpolation result in step 6 and establish a local geological magnetic field database.

2. The method for building a local geological magnetic field database based on an improved super learner according to claim 1, characterized in that: In the step 1, since the aerial magnetic exploration uses a drone as a detection platform, a survey line is arranged to fly over the area to be measured along the survey line at the height to be measured to collect the magnetic field distribution of the area, ensure that the survey line is distributed north-south, and determine the survey line interval according to the required scale; the geological magnetic field B Terrain According to B Tot =B plat&earth +B Terrain +B Sun +B grad Calculate and obtain, where B Tot is the total interference, B Sun is the diurnal interference, B grad is the geomagnetic gradient interference, B plat&earth It is platform interference and geomagnetic gradient interference.

3. The method for building a local geological magnetic field database based on an improved super learner according to claim 2, characterized in that: In the step 2, using the Kriging interpolation method based on the Gaussian semivariogram function to interpolate the regional magnetic field specifically includes: Based on the Gaussian semivariogram, the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated are obtained; The weight vector is obtained by calculating the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; According to the weight vector and the vector composed of the known point measurement values, the field value of the point to be interpolated is calculated, and the interpolation of each point outside the measurement line is completed according to the field value of the point to be interpolated to obtain the Gaussian interpolation result.

4. The method for building a local geological magnetic field database based on an improved super learner according to claim 2, characterized in that: In the step 2, interpolating the regional magnetic field using the Kriging interpolation method based on the exponential semivariogram function specifically includes: Based on the exponential semivariogram, the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated are obtained; The weight vector is obtained by calculating the covariance matrix of each point on the survey line and the covariance vector between each survey point on the survey line and the point to be interpolated; According to the weight vector and the vector composed of the known point measurement values, the field value of the point to be interpolated is calculated, and the interpolation of each point outside the measurement line is completed according to the field value of the point to be interpolated to obtain the exponential interpolation result.

5. The method for building a local geological magnetic field database based on an improved super learner according to claim 3 or 4, characterized in that: The field value of the interpolation point u can be calculated and obtained according to zλ=Z(u), wherein z is a vector composed of known point measurement values, λ is a weight vector, and Z(u) is the field value of the interpolation point u.

6. The method for building a local geological magnetic field database based on an improved super learner according to claim 5, characterized in that: The weight vector λ is calculated according to Cλ=c, wherein C is a covariance matrix based on each point on the survey line, and c is a covariance vector between each survey point on the survey line and the point to be interpolated.

7. The method for building a local geological magnetic field database based on an improved super learner according to claim 6, characterized in that: The five machine sub-models are random forest model support vector machine (SVM), Gaussian process regression (GPR), K nearest neighbor model (KNN) and generalized additive model (GAM).

8. A local geological magnetic field database construction system based on an improved super learner, characterized in that: The local geological magnetic field database building system based on the improved super learner uses the local geological magnetic field database building method based on the improved super learner as described in any one of claims 1 to 7 to build the local geological magnetic field database.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the computer program to implement the steps of the method for building a local geological magnetic field database based on an improved super learner according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for building a local geological magnetic field database based on an improved super learner are implemented as described in any one of claims 1 to 7.

Images