A three-dimensional metallogenic prediction method based on spatial correlation correction and related equipment

Abstract

The application provides a three-dimensional metallogenic prediction method based on spatial correlation correction and related equipment, relates to the technical field of metallogenic prediction, adopts a local neighborhood optimal estimation strategy to reduce the calculation complexity, uses residual errors instead of target values to compensate space, accurately captures spatial correlation information, and improves the accuracy of three-dimensional metallogenic prediction; the prediction result takes into account the non-linear fitting of attribute characteristics and the continuity of spatial distribution, avoids the problems of high-grade area fragmentation and oversmoothing, and improves the stability and spatial reasonableness of three-dimensional metallogenic prediction.

Description

Technical Field

[0001] This invention relates to the field of mineralization prediction technology, and in particular to a three-dimensional mineralization prediction method and related equipment based on spatial correlation correction. Background Technology

[0002] 3D geographic data prediction is a core task in the fields of geographic information science and resource environment. Its core requirement is to balance the nonlinear fitting ability of attribute features with the correlation constraints of spatial distribution. Existing technologies are mainly divided into two categories: The first category is machine learning methods, such as multilayer perceptrons, random forests, and deep learning models. These methods achieve nonlinear prediction by learning the mapping relationship between high-dimensional attribute features and target variables, but they completely ignore the spatial autocorrelation of geographic data, resulting in prediction results lacking spatial rationality. In scenarios such as mineral exploration, they are prone to the problem of "fragmentation of high-grade areas." The second category is spatial statistical interpolation methods, such as global kriging interpolation. These methods build interpolation models based on the principle of spatial autocorrelation and can reflect the spatial distribution pattern, but they cannot utilize the nonlinear information of attribute features. Moreover, global interpolation requires solving large-scale matrices, which is computationally complex and extremely inefficient when processing massive amounts of 3D geographic data. At the same time, they are easily affected by interference from distant samples, resulting in overly smooth prediction results that fail to reflect local spatial features.

[0003] In recent years, some studies have attempted to combine machine learning with spatial statistical methods, but most of them adopt a "serial combination" engineering approach, that is, first predict the target value through machine learning, and then correct the target value through spatial interpolation. The core flaw of this approach is that it does not build a unified model of "feature fitting-spatial compensation" at the theoretical level, but simply superimposes the two methods. At the same time, directly performing spatial interpolation on the target value ignores the fact that the machine learning prediction residual is the core carrier of spatial correlation, resulting in insufficient targeting of spatial compensation and limited improvement in prediction accuracy. Summary of the Invention

[0004] This invention provides a three-dimensional mineralization prediction method and related equipment based on spatial correlation correction, with the aim of improving the accuracy, stability and spatial rationality of three-dimensional mineralization prediction.

[0005] To achieve the above objectives, this invention provides a three-dimensional mineralization prediction method based on spatial correlation correction, comprising:

[0006] Step 1: Obtain exploration data for the target mining area and divide the exploration data into training and testing sets;

[0007] Step 2: Input the training set into the deep neural network model for training to obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model. Input the test set into the trained deep neural network model for prediction to obtain the predicted values ​​of the test set.

[0008] Step 3: Calculate the residual for each training sample in the training set using the target variable in the exploration data and the predicted values ​​of the feature-driven terms in the training set;

[0009] Step 4: Select several neighboring training samples for each test sample in the test set in the training set to construct a local neighborhood, and use the weight coefficients and residuals of each sample in the local neighborhood to construct a residual minimum variance unbiased estimation model.

[0010] Step 5: Spatial compensation is performed on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain the residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

[0011] Furthermore, before dividing the exploration data into training and test sets, the following steps are also included:

[0012] The target variable in the exploration data is decomposed as follows:

[0013] ;

[0014] in, Indicates the first The target variable for each sample Represents the attribute feature vector The nonlinear function driven The parameters to be learned represent the feature-driven function. Represents a residual random field. Indicates the first The three-dimensional spatial coordinates of each sample.

[0015] Furthermore, the objective function of a deep neural network model is:

[0016] ;

[0017] in, Represents the objective function value. Indicates the number of samples.

[0018] Furthermore, step 4 includes:

[0019] For each test sample in the test set, the KD tree algorithm is used to select the nearest neighbor sample in the training set whose three-dimensional spatial coordinates are closest to those of the test sample to construct a local neighborhood.

[0020] Based on the weight coefficients and residual values ​​of each neighboring sample in the local neighborhood, an unbiased estimation model with minimum variance of residuals is constructed.

[0021] Furthermore, the expression for the residual minimum variance unbiased estimation model is:

[0022] ;

[0023] in, Indicates the residual compensation term. This represents the number of neighboring samples within a local neighborhood. Represents the local neighborhood of the first The weight coefficients of each sample, For the local neighborhood of the first The residual values ​​of each sample.

[0024] Furthermore, by reconstructing the predicted values ​​of the test set with the residual compensation term, the expression for the three-dimensional mineralization prediction result of the target mining area is obtained as follows:

[0025] ;

[0026] in, This represents the three-dimensional mineralization prediction results for the target mining area. This represents the predicted value for the test set. This indicates the residual compensation term.

[0027] Furthermore, following step 5, the following also includes:

[0028] The mean field iterative optimization is performed on the three-dimensional mineralization prediction results of the target mining area. The iterative formula is as follows:

[0029] ;

[0030] in, Indicates the first The three-dimensional mineralization prediction results after the second iteration optimization Indicates the spatial constraint weights. Represents the local neighborhood of the first The true target value of the nearest neighboring samples, This indicates the number of neighboring samples within a local neighborhood.

[0031] This invention also provides a three-dimensional mineralization prediction device based on spatial correlation correction. The device applies a three-dimensional mineralization prediction method based on spatial correlation correction and includes:

[0032] The acquisition module is used to acquire exploration data of the target mining area and divide the exploration data into training and testing sets.

[0033] The prediction module is used to input the training set into the deep neural network model for training, obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model, and input the test set into the trained deep neural network model for prediction, obtain the predicted values ​​of the test set.

[0034] The calculation module is used to calculate the residual of each training sample in the training set by using the target variable in the exploration data and the predicted value of the feature-driven term in the training set;

[0035] The module is used to select several neighboring training samples for each test sample in the test set in the training set, construct a local neighborhood, and use the weight coefficients and residuals of each sample in the local neighborhood to construct a residual minimum variance unbiased estimation model.

[0036] The reconstruction module is used to perform spatial compensation on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain the residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

[0037] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a three-dimensional mineralization prediction method based on spatial correlation correction.

[0038] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a three-dimensional mineralization prediction method based on spatial correlation correction.

[0039] The above-described solution of the present invention has the following beneficial effects:

[0040] This invention divides the exploration data of the target mining area into a training set and a test set. The training set is input into a deep neural network model for training, yielding predicted values ​​of the feature-driven terms and the trained deep neural network model. The test set is then input into the trained deep neural network model for prediction, yielding predicted values ​​for the test set. The residual of each training sample in the training set is calculated using the target variable in the exploration data and the predicted values ​​of the feature-driven terms in the training set. Several neighboring training samples are selected for each test sample in the test set from the training set to construct a local neighborhood. The weight coefficients of each sample in the local neighborhood and the residuals are used to construct a minimum variance unbiased estimate of the residuals. The model is designed to spatially compensate the residuals of each training sample using a residual minimum variance unbiased estimation model, resulting in a residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results for the target mining area. Compared with existing technologies, this invention adopts a local neighborhood optimal estimation strategy to reduce computational complexity, uses residuals instead of target values ​​for spatial compensation, accurately captures spatial correlation information, and improves the accuracy of three-dimensional mineralization prediction. The prediction results take into account both the nonlinear fitting of attribute features and the continuity of spatial distribution, avoiding fragmentation and oversmoothing problems in high-grade areas, and improving the stability and spatial rationality of three-dimensional mineralization prediction.

[0041] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0043] Figure 2 (a) is a scatter plot comparing the predicted results and the true values ​​using the embodiments of the present invention; (b) is a scatter plot comparing the predicted results and the true values ​​using random forest; (c) is a scatter plot comparing the predicted results and the true values ​​using logistic regression; and (d) is a scatter plot comparing the predicted results and the true values ​​using support vector machine.

[0044] Figure 3 This is a schematic diagram of the structure of the three-dimensional mineralization prediction device in an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of the structure of the terminal device in an embodiment of the present invention. Detailed Implementation

[0046] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0047] In the description of this invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0048] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0049] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0050] This invention addresses existing problems by providing a three-dimensional mineralization prediction method and related equipment based on spatial correlation correction.

[0051] like Figure 1 As shown, embodiments of the present invention provide a three-dimensional mineralization prediction method based on spatial correlation correction, comprising:

[0052] Step 1: Obtain exploration data for the target mining area and divide the exploration data into training and testing sets;

[0053] Step 2: Input the training set into the deep neural network model for training to obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model. Input the test set into the trained deep neural network model for prediction to obtain the predicted values ​​of the test set.

[0054] Step 3: Calculate the residual for each training sample in the training set using the target variable in the exploration data and the predicted values ​​of the feature-driven terms in the training set;

[0055] Step 4: Select several neighboring training samples for each test sample in the test set in the training set to construct a local neighborhood, and use the weight coefficients and residuals of each sample in the local neighborhood to construct a residual minimum variance unbiased estimation model.

[0056] Step 5: Spatial compensation is performed on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain the residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

[0057] Specifically, this invention uses a gold mine as the target mining area to obtain exploration data for that gold mine. Exploration data Including the attribute feature vector of each sample Three-dimensional spatial coordinates Target variable In this embodiment of the invention, the attribute feature vector The 8-bit structural-fluid parameters include: orientation for expansion, tendency for expansion, fluid orientation towards precipitation-favorable sites, fluid tendency towards precipitation-favorable sites, upper interface distance of alteration zone, lower interface distance of alteration zone, fluid channel flux, and fluid flow distance. The objective variable is... Gold grade (g / t).

[0058] Ideally, before dividing the exploration data into training and test sets, the following steps are also included:

[0059] The attribute feature vectors are standardized to eliminate the influence of dimensions. The expression for standardization is:

[0060] ;

[0061] in, This represents the attribute vector after standardization. Represents the mean vector. This represents the standard deviation vector.

[0062] Specifically, before dividing the exploration data into training and test sets, the following steps are also included:

[0063] The target variable in the exploration data is decomposed as follows:

[0064] ;

[0065] in, Indicates the first The target variable for each sample Represents the attribute feature vector The nonlinear function driven by this is only related to the attribute features of the sample and is independent of its spatial location. The parameters to be learned represent the feature-driven function. This represents a residual random field, which is only related to spatial location and characterizes the spatially correlated portion of the target variable that is not explained by attribute features. Indicates the first The three-dimensional spatial coordinates of each sample.

[0066] Most preferably, the deep neural network model provided in this embodiment of the invention includes an input layer, a hidden layer, and an output layer. The input layer has a dimension of 'd', consistent with the dimension of the attribute feature vector. The hidden layer has at least two layers with dimensions of 128 and 64 respectively. Each layer uses the ReLU activation function to achieve non-linear mapping. The output layer has a dimension of 1 and outputs the predicted values ​​of the feature-driven terms from the training set. and the predicted values ​​of the test set .

[0067] Specifically, the prediction expression of a deep neural network model is:

[0068] ;

[0069] in, This represents the predicted value of the feature-driven terms in the training set.

[0070] Specifically, in order to learn model parameters, embodiments of the present invention... The objective function of the deep neural network model is defined as:

[0071] ;

[0072] in, Represents the objective function value. Indicates the number of samples.

[0073] In this embodiment of the invention, the Adam optimizer is used to minimize the above objective function to complete model training. After training, the attribute feature vectors of the test set are input into the trained deep neural network model for prediction to obtain the predicted values ​​of the test set.

[0074] Specifically, due to the feature-driven predicted values ​​of the training set Since only the influence of attribute feature vectors is fitted and spatial correlation is not considered, prediction residuals exist. In this embodiment of the invention, the expression for calculating the prediction residuals of the training set using the predicted values ​​of the target variable and the feature-driven terms of the training set is as follows:

[0075] ;

[0076] Among them, residual It represents the portion of the target variable not explained by attribute features; its essence is a function of three-dimensional spatial coordinates. Therefore, the residual set... Viewed as a random field defined in three-dimensional space ,Right now .

[0077] To construct the mathematical model of the random field, we assume that the residual random field satisfies the second-order stationarity assumption, and the expression is:

[0078] ;

[0079] This assumption has two implications: the mathematical expectation of the random field is 0, meaning that the residuals have no systematic bias in the global domain;

[0080] The covariance of a residual random field is only related to the spatial lag distance. It is relevant and independent of spatial location; this is the core assumption of spatial statistical modeling.

[0081] According to spatial statistics theory, the variogram With covariance function Satisfying Relationships The variogram is used to quantify the spatial correlation strength of the residuals: when the spatial lag distance... Less than range When, the value of the variogram changes with The increase of [a certain value] indicates that the residuals have spatial correlation; when [a certain value] increases, [the remaining values] increase. Greater than range When the value of the variogram stabilizes, it indicates that the residuals have no spatial correlation.

[0082] In this embodiment of the invention, the mutation function can be a spherical model, a Gaussian model, or an exponential model, wherein the expression for the spherical model is:

[0083] ;

[0084] in, Indicates the value of the nugget. Indicates the base value. Indicates a change in range.

[0085] Specifically, step 4 includes:

[0086] For each test sample in the test set, the KD tree algorithm is used to select the nearest neighbor sample in the training set whose three-dimensional spatial coordinates are closest to those of the test sample to construct a local neighborhood.

[0087] Based on the weight coefficients and residual values ​​of each neighboring sample in the local neighborhood, an unbiased estimation model with minimum variance of residuals is constructed.

[0088] Specifically, for each test sample in the test set, the KD tree algorithm is used to select the nearest neighbor sample in the training set whose three-dimensional spatial coordinates are closest to the test sample, constructing a local neighborhood, including:

[0089] For any test sample in the test set, construct a KD Tree index of the training sample coordinates, and calculate the Euclidean distance between the 3D coordinates of any test sample and all training samples. The calculation expression is as follows:

[0090] ;

[0091] Based on the above Euclidean distance selection, the one with the smallest distance is chosen. Using training samples, construct a local neighborhood. .

[0092] This step can reduce computational complexity from Reduce to ( (To increase the number of test samples), significantly improving computational efficiency.

[0093] In this embodiment of the invention, the number of samples in the local neighborhood ranges from 8 to 32, and the preferred value in this embodiment of the invention is 16.

[0094] Specifically, the expression for the residual minimum variance unbiased estimation model is:

[0095] ;

[0096] in, Indicates the residual compensation term. This represents the number of neighboring samples within a local neighborhood. Represents the local neighborhood of the first The weight coefficients of each sample, For the local neighborhood of the first The residual values ​​of each sample.

[0097] Specifically, the solution for the weight coefficients must satisfy a constrained optimization problem, expressed as:

[0098] ;

[0099] ;

[0100] Unbiasedness constraint This ensures that the expected value of the residual estimate is equal to the expected value of the actual residual.

[0101] This invention transforms the constrained optimization problem into a system of linear equations, as shown below:

[0102] ;

[0103] In the formula, The variogram matrix of the local neighborhood. These are Lagrange multipliers used to introduce unbiased constraints. Let be the variogram vector of the test point and its local neighborhood samples. , is the variogram function, satisfying By solving this system of linear equations, the optimal weight vector can be obtained. .

[0104] Specifically, by reconstructing the predicted values ​​of the test set with the residual compensation term, the expression for the three-dimensional mineralization prediction result of the target mining area is obtained as follows:

[0105] ;

[0106] in, This represents the three-dimensional mineralization prediction results for the target mining area. This represents the predicted value for the test set. This indicates the residual compensation term.

[0107] Specifically, after step 5, the following is also included:

[0108] The mean field iterative optimization is performed on the three-dimensional mineralization prediction results of the target mining area. The iterative formula is as follows:

[0109] ;

[0110] in, Indicates the first The three-dimensional mineralization prediction results after the second iteration optimization Indicates the spatial constraint weights. Represents the local neighborhood of the first The true target value of the nearest neighboring samples, This indicates the number of neighboring samples within a local neighborhood.

[0111] To verify the effectiveness of the proposed model, this invention designed multiple sets of comparative experiments. The model provided in this invention was compared quantitatively and qualitatively with three widely used classical machine learning models—random forest, logistic regression, and support vector machine—on the same 3D mineralization prediction task in the same mining area. The comparison results are as follows: Figure 2 As shown, the comparison results are presented as a scatter plot of predicted values ​​versus actual values. All subplots use the actual values ​​obtained from the actual exploration as the x-axis and the predicted values ​​output by each model as the y-axis. A uniform y=x ideal fitting reference line is also plotted to visually determine the closeness between the predicted results and the actual values. Figure 2(a) shows the prediction results of the model of this invention. It can be seen that the overall distribution of the data points output by the model is highly concentrated and closely follows the y=x reference line. Even in the high-quality range (true value > 5), it maintains extremely high prediction consistency without significant deviation or jumps. Quantitative evaluation indicators further show that the model provided in this embodiment of the invention has a determination coefficient R² of 0.918, while the mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) are only 0.257, 0.132, 0.363, and 0.151, respectively, which are the best among the four error indicators, demonstrating the strongest fitting ability and the smallest prediction bias. As a control, Figure 2 (b) Random Forest Figure 2 (c) Logistic Regression and Figure 2 (d) The prediction scatter points of the support vector machine are more discrete overall, and the data points deviate significantly from the y=x reference line, especially in the high-value region where the prediction bias increases significantly, and local outliers and discontinuous distribution phenomena are prone to occur. In terms of indicators, the errors of the three traditional models are higher than those of the model of this invention. Among them, the R2 of the support vector machine is only 0.862, and the fitting effect and prediction accuracy are significantly worse than the method provided by the embodiment of this invention. The above experimental results fully demonstrate that the embodiment of this invention, by introducing residual space compensation and local neighborhood optimal estimation strategies, and on the basis of using deep neural networks to achieve strong nonlinear fitting of geological attribute characteristics, can more accurately characterize the inherent spatial correlation and spatial structure characteristics of mining area data. It effectively improves the problems of fragmentation of high-grade areas, discontinuous spatial distribution, and excessive smoothing that are common in traditional machine learning models in three-dimensional mineralization prediction, thereby significantly improving the accuracy, stability and spatial rationality of three-dimensional mineralization prediction results, which are more in line with the actual mineralization geological laws and exploration application needs.

[0112] In summary, this embodiment of the invention divides the acquired exploration data of the target mining area into a training set and a test set; the training set is input into a deep neural network model for training, obtaining the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model; the test set is input into the trained deep neural network model for prediction, obtaining the predicted values ​​of the test set; the residual of each training sample in the training set is calculated using the target variable in the exploration data and the predicted values ​​of the feature-driven terms of the training set; several neighboring training samples are selected for each test sample in the test set from the training set to construct a local neighborhood, and the minimum variance residual is constructed using the weight coefficients of each sample in the local neighborhood and the residual. A partial estimation model is used; the residuals of each training sample are spatially compensated using a residual minimum variance unbiased estimation model to obtain a residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results for the target mining area. Compared with existing technologies, this embodiment of the invention adopts a local neighborhood optimal estimation strategy to reduce computational complexity, uses residuals instead of target values ​​for spatial compensation, accurately captures spatial correlation information, and improves the accuracy of three-dimensional mineralization prediction. The prediction results take into account both the nonlinear fitting of attribute features and the continuity of spatial distribution, avoiding fragmentation and oversmoothing problems in high-grade areas, and improving the stability and spatial rationality of three-dimensional mineralization prediction.

[0113] like Figure 3 As shown, this embodiment of the invention also provides a three-dimensional mineralization prediction device 100 based on spatial correlation correction, which applies the three-dimensional mineralization prediction method based on spatial correlation correction as described in the above embodiments. The three-dimensional mineralization prediction device 100 includes:

[0114] The acquisition module 101 is used to acquire exploration data of the target mining area and divide the exploration data into a training set and a test set.

[0115] The prediction module 102 is used to input the training set into the deep neural network model for training, to obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model, and to input the test set into the trained deep neural network model for prediction, to obtain the predicted values ​​of the test set.

[0116] The calculation module 103 is used to calculate the residual of each training sample in the training set by using the target variable in the exploration data and the predicted value of the feature-driven term in the training set;

[0117] The construction module 104 is used to select several neighboring training samples for each test sample in the test set in the training set, construct a local neighborhood, and construct a residual minimum variance unbiased estimation model using the weight coefficients and residuals of each sample in the local neighborhood.

[0118] The reconstruction module 105 is used to perform spatial compensation on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain the residual compensation term, and reconstruct the predicted values ​​of the test set with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

[0119] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0120] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0121] This invention also provides a terminal device, such as... Figure 4 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 4 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, it implements the above-described three-dimensional mineralization prediction method based on spatial correlation correction.

[0122] The terminal device D10 can be a desktop computer, laptop, handheld computer, server, server cluster, or cloud server, etc. This terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will understand that... Figure 4 This is merely an example of terminal device D10 and does not constitute a limitation on terminal device D10. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0123] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0124] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, boot loader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0125] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0126] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0127] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a three-dimensional mineralization prediction method based on spatial correlation correction.

[0128] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0129] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A three-dimensional mineralization prediction method based on spatial correlation correction, characterized in that, include: Step 1: Obtain exploration data for the target mining area and divide the exploration data into a training set and a test set; Step 2: Input the training set into the deep neural network model for training to obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model. Input the test set into the trained deep neural network model for prediction to obtain the predicted values ​​of the test set. Step 3: Calculate the residual of each training sample in the training set using the target variable in the exploration data and the predicted value of the feature-driven term in the training set; Step 4: In the training set, select several neighboring training samples for each test sample in the test set to construct a local neighborhood. Then, using the weight coefficients and residuals of each sample in the local neighborhood, construct the expression for the residual minimum variance unbiased estimation model: ； in, Indicates the residual compensation term. This represents the number of neighboring samples within a local neighborhood. Represents the local neighborhood of the first The weight coefficients of each sample, For the local neighborhood of the first The residual values ​​of each sample; Step 5: Spatial compensation is performed on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain the residual compensation term. The predicted values ​​of the test set are then reconstructed with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

2. The three-dimensional mineralization prediction method based on spatial correlation correction according to claim 1, characterized in that, Before dividing the exploration data into training and test sets, the following steps are also included: The target variable in the exploration data is decomposed as follows: ； in, Indicates the first The target variable for each sample Represents the attribute feature vector The nonlinear function driven The parameters to be learned represent the feature-driven function. Represents a residual random field. Indicates the first The three-dimensional spatial coordinates of each sample.

3. The three-dimensional mineralization prediction method based on spatial correlation correction according to claim 1, characterized in that, The objective function of the deep neural network model is: ； in, Represents the objective function value. Indicates the number of samples. Indicates the first The target variable for each sample Represents the attribute feature vector The nonlinear function driven The parameter to be learned represents the feature-driven function.

4. The three-dimensional mineralization prediction method based on spatial correlation correction according to claim 1, characterized in that, Step 4 includes: For each test sample in the test set, the nearest neighbor sample with the closest three-dimensional spatial coordinates to the test sample is selected in the training set using the KD tree algorithm to construct a local neighborhood. Based on the weight coefficients and residual values ​​of each neighboring sample in the local neighborhood, an unbiased estimation model with minimum variance of residuals is constructed.

5. The three-dimensional mineralization prediction method based on spatial correlation correction according to claim 1, characterized in that, The predicted values ​​of the test set are reconstructed with the residual compensation term to obtain the expression for the three-dimensional mineralization prediction result of the target mining area: ； in, This represents the three-dimensional mineralization prediction results for the target mining area. This represents the predicted value for the test set. This indicates the residual compensation term.

6. The three-dimensional mineralization prediction method based on spatial correlation correction according to claim 5, characterized in that, Following step 5, the following is also included: The mean field iterative optimization is performed on the three-dimensional mineralization prediction results of the target mining area. The iterative formula is as follows: ； in, Indicates the first The three-dimensional mineralization prediction results after the second iteration optimization Indicates the spatial constraint weights. Represents the local neighborhood of the first The true target value of the nearest neighboring samples, This indicates the number of neighboring samples within a local neighborhood.

7. A three-dimensional mineralization prediction device based on spatial correlation correction, characterized in that, The three-dimensional mineralization prediction device, which applies the spatial correlation correction-based three-dimensional mineralization prediction method as described in any one of claims 1-6, comprises: The acquisition module is used to acquire exploration data of the target mining area and divide the exploration data into a training set and a test set. The prediction module is used to input the training set into the deep neural network model for training, to obtain the predicted values ​​of the feature-driven terms of the training set and the trained deep neural network model, and to input the test set into the trained deep neural network model for prediction, to obtain the predicted values ​​of the test set. The calculation module is used to calculate the residual of each training sample in the training set using the target variable in the exploration data and the predicted value of the feature-driven term in the training set; The construction module is used to select several neighboring training samples for each test sample in the test set in the training set, construct a local neighborhood, and construct a residual minimum variance unbiased estimation model using the weight coefficients and residuals of each sample in the local neighborhood. The reconstruction module is used to perform spatial compensation on the residuals of each training sample using the residual minimum variance unbiased estimation model to obtain a residual compensation term, and to reconstruct the predicted values ​​of the test set with the residual compensation term to obtain the three-dimensional mineralization prediction results of the target mining area.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the three-dimensional mineralization prediction method based on spatial correlation correction as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the three-dimensional mineralization prediction method based on spatial correlation correction as described in any one of claims 1 to 6.

Images