Geological mineral intelligent detection system and method based on multi-source data fusion

Abstract

The invention discloses an intelligent geological mineral detection system and method based on multi-source data fusion, and belongs to the technical field of geological mineral detection. The system comprises a data acquisition module, a data preprocessing module, a data fusion module and an intelligent analysis module; wherein the data acquisition module is used for acquiring geological mineral data through various sensors; performing cleaning, standardization and dimension reduction processing on the acquired data by using a data preprocessing module; fusing the preprocessed multi-source data by adopting a Bayesian network through a data fusion module; the intelligent analysis module identifies a potential mineral product area through a YOLO target detection algorithm, then classifies geological units by using a K-means clustering algorithm, and finally optimizes a drilling path and position selection through a genetic algorithm; according to the system and the method provided by the invention, information of various data sources such as ground, air and satellites is integrated through multi-source data fusion, and the limitation of a single data source is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of geological and mineral exploration, and in particular relates to an intelligent geological and mineral exploration system and method based on multi-source data fusion. Background Art

[0002] In the current field of geological and mineral exploration, with the development of technology and the continuous growth of the demand for mineral resources, it is of crucial significance to develop more efficient and accurate exploration methods.

[0003] Traditional geological and mineral exploration often relies on single-type data and methods. Specifically, the following deficiencies exist: First, there is the deficiency of a single data source. Specifically, the data collected only by ground sensors has great limitations. For example, although a gravimeter can measure the gravity changes on the Earth's surface to infer the density differences of underground geological bodies, its resolution for geological bodies with small density differences is low. A magnetometer mainly detects underground magnetic geological bodies. However, in a complex geological environment, affected by geomagnetic interference and non-magnetic geological bodies, its detection results are often inaccurate. Resistivity logging equipment can only obtain resistivity information within a limited range around the borehole and is difficult to comprehensively reflect the underground structure of the entire exploration area. In terms of air sensors, although the LiDAR scanner carried by a drone can quickly obtain topographic data on the Earth's surface, its direct detection ability for underground mineral resources is limited. A hyperspectral camera can obtain spectral information of ground objects, but it is difficult to accurately judge the distribution of underground minerals solely based on this information. Satellite remote sensing data such as synthetic aperture radar (SAR) and optical satellite images, although they can provide large-area surface information, due to the limitations of their resolution and penetration depth, they cannot accurately determine the location and reserves of underground deep mineral resources. Second, there is a lack of data integration. In traditional geological and mineral exploration, data from different sources are often used independently. For example, ground sensor data, air sensor data, and satellite remote sensing data are usually analyzed by different teams, lacking an effective integration mechanism. This leads to a waste of data resources and the inability to fully utilize the complementarity between multi-source data. Therefore, there is an urgent need for an intelligent geological and mineral exploration system and method based on multi-source data fusion to detect and analyze geological and minerals from multiple perspectives, overcome the limitations of a single data source, and thus improve the accuracy and efficiency of detection. Summary of the Invention

[0004] The object of the present invention is to provide an intelligent geological and mineral exploration system and method based on multi-source data fusion, which solves the problem in the prior art that traditional geological and mineral exploration analyzes from a single data source, resulting in low accuracy and efficiency of detection.

[0005] To achieve the above-mentioned purpose, the present invention provides a geological mineral intelligent detection system based on multi-source data fusion, including a data acquisition module, a data preprocessing module, a data fusion module and an intelligent analysis module;

[0006] Data acquisition module: collects geological and mineral data through a variety of sensors;

[0007] Data preprocessing module: cleans, standardizes and reduces the dimension of the collected data to improve the data quality and make it suitable for subsequent fusion and analysis;

[0008] Data fusion module: Use Bayesian network to fuse pre-processed multi-source data and mine the intrinsic connection and complementary information between data;

[0009] Intelligent analysis module: Identify potential mineral areas through the YOLO target detection algorithm, then use the K-means clustering algorithm to classify geological units, and finally optimize the drilling path and location selection through the genetic algorithm.

[0010] Preferably, the sensors include ground sensors, aerial sensors and satellite remote sensing data sensors; wherein the ground sensors are gravimeters, magnetometers or resistivity logging equipment; the aerial sensors are LiDAR scanners or hyperspectral cameras carried by drones; and the satellite remote sensing data sensors are synthetic aperture radars or optical satellite images.

[0011] Preferably, the collected data is cleaned by removing outliers in the data by setting a threshold; missing values ​​are filled by using mean filling or interpolation methods.

[0012] Preferably, the collected data is standardized by standardizing the gravity value, magnetic value, resistivity value, and various values ​​in the air sensor and satellite remote sensing data sensor to make them in the same order of magnitude. The calculation expression is as follows:

[0013]

[0014] In the formula, x' represents the transformed data, x is the original data, μ is the mean, and σ is the standard deviation. The calculation expressions of the mean μ and the standard deviation σ are as follows:

[0015]

[0016] In the formula, the data set is {x 1 ,x 2 ,…,x n}.

[0017] Preferably, the process of performing dimensionality reduction processing on the collected data is as follows:

[0018] S1. Data centralization: Mean centralization is performed on each feature, and the formula is as follows:

[0019]

[0020] In the formula, X is the original data matrix, is the mean vector of each column feature, and X centered represents the data centralization matrix;

[0021] S2. Calculate the covariance matrix Cov(X) of the data matrix, and the formula is as follows:

[0022]

[0023] In the formula, m is the number of samples;

[0024] S3. Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix, and the formula is as follows:

[0025] Cov(X) = QΛQ T

[0026] In the formula, Q is the eigenvector matrix, and Λ is the diagonal eigenvalue matrix;

[0027] S4. Select the principal components: Select the first k eigenvectors with the largest eigenvalues as the principal components;

[0028] S5. Construct the projection matrix Q selected , and the formula is

[0029] Q selected = [v 1 , v 2 ,..., v k

[0030] where v k represents the k-th eigenvector;

[0031] S6. Project the data onto the selected principal components and visualize the data. The formula is as follows:

[0032] X reduced = X centered ×Q selected

[0033] In the formula, X reduced is the projected matrix.

[0034] The present invention also provides a method for an intelligent geological and mineral exploration system based on multi-source data fusion, including the following steps:

[0035] Step 1. Collect data: Collect geological and mineral data through ground sensors, aerial sensors, and satellite remote sensing data sensors;​

[0036] Step 2: Clean, standardize, and reduce the dimensionality of the data collected in Step 1 through the data preprocessing module;

[0037] Step 3: Fuse the data preprocessed in Step 2; extract the features of ground objects from the hyperspectral remote sensing image through a convolutional neural network, then use a long short-term memory network to capture the dynamic change rules, and finally use a Bayesian network to perform weighted fusion on geological and mineral data from different sources;

[0038] Step 4: Use the intelligent analysis module to analyze potential mineral areas and optimize the drilling path and location selection through a genetic algorithm.

[0039] Preferably, the specific process of Step 3 is as follows:

[0040] S31: Extract the features of ground objects from the hyperspectral remote sensing image through a convolutional neural network, and the calculation expression is as follows:

[0041]

[0042] In the formula, c ij represents the result of the convolution operation, a is the input image element, and k is the convolution kernel element. Among them, the size of the convolution kernel is p×q;

[0043] S32: Use a long short-term memory network to capture the dynamic change rules. The long short-term memory network is controlled by three gates, including the forget gate, the input gate, and the output gate;

[0044] Determine which information is forgotten or retained from the memory unit at each time step through the forget gate; the calculation expression is as follows:

[0045] f t =σ(W f ·[h t-1 ,x t +b f );

[0046] Among them, σ is the logical activation function, W f and b f are the weight matrix and bias term of the forget gate respectively, [h t-1 ,x t is the concatenation of the hidden state of the previous time step and the input of the current time step, and f t is the output of the forget gate. f t determines how much information of the memory unit state at the previous moment should be "forgotten" or "discarded";

[0047] The input gate consists of two parts: a sigmoid layer that determines which values will be updated, and a tanh layer that creates a new candidate value vector which will be added to the state. The calculation expression of the input gate is as follows:

[0048]

[0049] where, i t is the output of the input gate, is the candidate memory cell state, W i , W C and b i , b C are the relevant weights and biases respectively, and σ represents the sigmoid function;

[0050] The output gate is responsible for determining which part of the memory cell state will be output to the hidden state, and the calculation expression is as follows:

[0051]

[0052] o t = σ(W o ·[h t-1 , x t +b o );

[0053] h t = o t *tanh(C t );

[0054] In the formula, C t combines the information of the forget gate and the input gate, o t is the output of the sigmoid function of the output gate, C t is the updated memory cell state, h t is the final hidden state output, W o is the weight matrix of the output gate, which is used to adjust the influence of the input data and the hidden state at the previous moment on the current output, and b o is the bias term of the output gate, which is used to adjust the offset of the calculation;

[0055] S33. Use a Bayesian network to perform weighted fusion on geological and mineral data from different sources.

[0056] Preferably, the process of using a Bayesian network to perform weighted fusion on geological and mineral data from different sources in S33 is as follows: Let D 1 , D 2 , …, D n be data from different sources. According to Bayes' theorem Among them, A is the fused event, and B is the event related to different data sources; first, determine the data D of each data source i The prior probability P(D i ) in the fusion, and the conditional probability P(A|D i ) of the data of each data source given the fusion result A, and calculate the posterior probability through the Bayes formula Determine the weight w of each data in the fusion result according to the posterior probability i = P(D i |A), and finally obtain the result of weighted fusion

[0057] Preferably, the process of using the intelligent analysis module to analyze potential mineral areas is as follows:

[0058] S41. Identify potential mineral areas through the YOLO object detection algorithm; divide the input image into S×S grids, each grid is responsible for predicting B bounding boxes, and each bounding box contains 5 prediction values (x, y, w, h, c), where (x, y) is the offset of the center coordinates of the bounding box relative to the grid, w and h are the width and height of the bounding box, and c is the confidence that the bounding box contains the target; for each grid, also predict C class probabilities p c , and the final output of each grid is a vector with a length of 5B + C; through the forward propagation of the network, obtain the prediction results on each network, and then screen out potential mineral areas according to a preset threshold;

[0059] S42. Use the K-means clustering algorithm to classify geological units; first, randomly select K initial clustering centers μ 1 , μ 2 , …, μ K , and then calculate the distance from each data point x j to these K centers based on the Euclidean distance formula, and the calculation expression is where x jl and μ il are the l-th coordinate components of the data point x j and the clustering center μ i respectively, and assign the data point to the class where the nearest clustering center belongs, and then recalculate the center of each class, and the calculation expression is where, C i is the set of data points included in the i-th class, and repeat the above process until the clustering center no longer changes.

[0060] Preferably, optimizing the drilling path and position selection through the genetic algorithm is specifically as follows: Let the drilling path consist of m decision variables x 1 , x2 , …, x m which is composed of and encoded into a binary chromosome C = c 1 , c 2 , …, c L of length L; through selection, crossover, and mutation operations, the population is continuously evolved to obtain the optimal drilling path and location; among them, the selection operation is specifically to calculate the fitness value F(C) of each chromosome, and the fitness function is constructed based on the drilling cost and the expected mineral reserve index, and the expression is: F(C) = a×V - b×C d , where V is the expected mineral reserve and C d is the drilling cost, a and b are weight coefficients, and excellent chromosomes are selected into the next-generation population using the roulette wheel selection method according to the fitness value; the crossover operation is specifically to randomly select two chromosomes C 1 and C 2 , randomly select a crossover point k on the chromosome, and exchange the parts of C 1 and C 2 after the crossover point to obtain new chromosomes C′ 1 and C′ 2 ; the mutation operation is specifically for each chromosome, and with a certain mutation probability p m , randomly change a certain gene position in the chromosome, for example, change c i to 1 - c i ; after multiple generations of evolution, the optimal drilling path and location are obtained.

[0061] Therefore, the present invention adopts the above-mentioned intelligent geological and mineral exploration system and method based on multi-source data fusion, and has the following beneficial effects:

[0062] (1) The accuracy of detection is improved; through multi-source data fusion, the information of various data sources such as ground, air, and satellite is integrated, the advantages of different data are fully utilized, and the limitations of a single data source are reduced; by combining the macroscopic information of satellite remote sensing data with the microscopic data of ground sensors, the location and type of mineral resources can be determined more accurately;

[0063] (2) The data processing ability is enhanced; the data cleaning, standardization, and dimensionality reduction operations in the data preprocessing module improve the quality and processability of the data. Intelligent algorithms such as deep learning models and time series models can better mine the hidden information in the data in terms of feature extraction and spatio-temporal correlation modeling, thereby improving the performance of the entire detection system;

[0064] (3) Optimize exploration decisions; the object detection, classification and clustering, as well as prediction and optimization algorithms in the intelligent analysis module can provide a scientific basis for the selection of drilling paths and locations; optimizing the drilling path through genetic algorithms can reduce drilling costs, improve exploration efficiency, and reduce unnecessary drilling work.

[0065] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic structural diagram of a geological and mineral intelligent detection system based on multi-source data fusion according to the present invention;

[0067] Figure 2 It is an overall flow block diagram of the method of a geological and mineral intelligent detection system based on multi-source data fusion according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0069] Please refer to Figure 1 , a geological and mineral intelligent detection system based on multi-source data fusion, including a data acquisition module, a data preprocessing module, a data fusion module, and an intelligent analysis module;

[0070] Data acquisition module: Collect geological and mineral data through a variety of sensors; the sensors include ground sensors, aerial sensors, and satellite remote sensing data sensors; among them, the ground sensors are gravimeters, magnetometers, or resistivity logging equipment; the aerial sensors are LiDAR scanners or hyperspectral cameras carried by unmanned aerial vehicles; the satellite remote sensing data sensors are synthetic aperture radars or optical satellite images.

[0071] Data preprocessing module: Clean, standardize, and reduce the dimension of the collected data to improve data quality and make it suitable for subsequent fusion and analysis; specifically, cleaning the collected data is to remove outliers in the data by setting thresholds; for missing values, the method of mean filling or interpolation is used to fill them.

[0072] Specifically, standardizing the collected data is to standardize the gravity values, magnetic values, resistivity values, and various values in the aerial sensors and satellite remote sensing data sensors so that they are in the same order of magnitude. The calculation expression is as follows:

[0073]

[0074] Where x' represents the data after transformation, x is the original data, μ is the mean, and σ is the standard deviation. The calculation expressions for the mean μ and the standard deviation σ are as follows:

[0075]

[0076] Where the data set is {x 1 , x 2 , …, x n}.

[0077] The process of dimensionality reduction for the collected data is as follows:

[0078] S1. Data centering. Mean centering is performed on each feature. The formula is:

[0079]

[0080] Where X is the original data matrix, is the mean vector of each column feature, and X centered represents the data centering matrix;

[0081] S2. Calculate the covariance matrix Cov(X) of the data matrix. The formula is:

[0082]

[0083] Where m is the number of samples;

[0084] S3. Eigenvalue decomposition. Perform eigenvalue decomposition on the covariance matrix. The formula is:

[0085] Cov(X) = QΛQ T

[0086] Where Q is the eigenvector matrix and Λ is the diagonal eigenvalue matrix;

[0087] S4. Select the principal components. Select the first k eigenvectors with the largest eigenvalues as the principal components;

[0088] S5. Construct the projection matrix Q selected , and the formula is

[0089] Q selected = [v 1 , v 2 ,..., v k

[0090] Where v k represents the k-th eigenvector;

[0091] ​S6. Project the data onto the selected principal component and visualize the data. The formula is:

[0092] X reduced =X centered ×Q selected

[0093] Where, X reduced is the projection matrix.

[0094] Data fusion module: Use Bayesian network to fuse pre-processed multi-source data and mine the intrinsic connection and complementary information between data;

[0095] Intelligent analysis module: Identify potential mineral areas through the YOLO target detection algorithm, then use the K-means clustering algorithm to classify geological units, and finally optimize the drilling path and location selection through the genetic algorithm.

[0096] See also Figure 2 , a method for a geological mineral intelligent detection system based on multi-source data fusion, comprising the following steps:

[0097] Step 1: Collect data, collect geological and mineral data through ground sensors, aerial sensors and satellite remote sensing data sensors;

[0098] Step 2: The data collected in step 1 is cleaned, standardized and dimensionally reduced through a data preprocessing module;

[0099] Step 3: Fuse the preprocessed data in step 2; extract the features of the ground objects from the hyperspectral remote sensing images through the convolutional neural network, then use the long short-term memory network to capture the dynamic change rules, and finally use the Bayesian network to perform weighted fusion of geological and mineral data from different sources; the specific process is as follows:

[0100] S31. Extract ground features from hyperspectral remote sensing images through convolutional neural networks. The calculation expression is as follows:

[0101]

[0102] In the formula, c ij represents the result of convolution operation, a is the input image element, k is the convolution kernel element, and the size of the convolution kernel is p×q;

[0103] S32. Use the long short-term memory network to capture dynamic changes. The long short-term memory network is controlled by three gates: forget gate, input gate and output gate.

[0104] The forget gate determines which information is forgotten or retained from the memory unit at each time step; the calculation expression is as follows:

[0105] f t = σ(W f · [h t-1 , x t + b f );

[0106] Where σ is the logical activation function, W f and b f are the weight matrix and bias term of the forget gate respectively, [h t-1 , x t is the concatenation of the hidden state of the previous time step and the input of the current time step, and f t is the output of the forget gate, and f t determines how much information of the memory cell state at the previous moment should be "forgotten" or "discarded";

[0107] The input gate consists of two parts: a sigmoid layer determines which values will be updated, and a tanh layer creates a new candidate value vector that will be added to the state; the calculation expression of the input gate is as follows:

[0108]

[0109] Where i t is the output of the input gate, is the candidate memory cell state, W i , W C and b i , b C are the relevant weights and biases respectively, and σ represents the sigmoid function;

[0110] The output gate is responsible for determining which part of the memory cell state will be output to the hidden state, and the calculation expression is as follows:

[0111]

[0112] o t = σ(W o · [h t-1 , x t + b o );

[0113] h t = o t * tanh(C t );

[0114] In the formula, C t combines the information of the forget gate and the input gate, o t is the output of the sigmoid function of the output gate, and C tis the updated memory cell state, h t is the final hidden state output, W o is the weight matrix of the output gate, used to adjust the influence of the input data and the hidden state at the previous moment on the current output, b o is the bias term of the output gate, used to adjust the offset of the calculation;

[0115] S33. Use a Bayesian network to perform weighted fusion on geological and mineral data from different sources; the specific process is as follows: Let D 1 , D 2 , …, D n be data from different sources. According to Bayes' theorem where A is the fused event and B is the event related to different data sources; first determine the prior probability P(D i ) of each data source data D in the fusion, and the conditional probability P(A|D i ) of each data source data given the fusion result A. Calculate the posterior probability through Bayes' formula i ) Determine the weight w of each data in the fusion result according to the posterior probability i = P(D i |A), and finally obtain the result of weighted fusion

[0116] Step 4. Use the intelligent analysis module to analyze potential mineral areas, and optimize the drilling path and location selection through the genetic algorithm; among them, the process of using the intelligent analysis module to analyze potential mineral areas is as follows:

[0117] S41. Identify potential mineral areas through the YOLO object detection algorithm; divide the input image into S×S grids, each grid is responsible for predicting B bounding boxes, and each bounding box contains 5 prediction values (x, y, w, h, c), where (x, y) is the offset of the center coordinates of the bounding box relative to the grid, w and h are the width and height of the bounding box, and c is the confidence that the bounding box contains the target; for each grid, also predict C class probabilities p c , and the final output of each grid is a vector with a length of 5B + C; through the forward propagation of the network, obtain the prediction results on each network, and then screen out potential mineral areas according to a preset threshold;

[0118] S42. Use the K-means clustering algorithm to classify geological units; first randomly select K initial clustering centers μ 1 , μ 2 , …, μ K , and then calculate the Euclidean distance formula for each data point x jThe distance to these K centers, and the calculation expression is where x jl and μ il are respectively the l-th coordinate components of the data point x j and the clustering center μ i . The data points are assigned to the class to which the nearest clustering center belongs, and then the center of each class is recalculated. The calculation expression is where C i is the set of data points included in the i-th class. Repeat the above process until the clustering centers no longer change.

[0119] Optimizing the drilling path and location selection through the genetic algorithm is specifically as follows: Suppose the drilling path consists of m decision variables x 1 , x 2 , …, x m . It is encoded into a binary chromosome C = c 1 , c 2 , …, c L with a length of L. Through selection, crossover, and mutation operations, the population is continuously evolved to obtain the optimal drilling path and location. Among them, the selection operation is specifically to calculate the fitness value F(C) of each chromosome. The fitness function is constructed based on the drilling cost and the expected mineral reserve index, and the expression is: F(C) = a×V - b×C d , where V is the expected mineral reserve, C d is the drilling cost, and a and b are weight coefficients. Excellent chromosomes are selected into the next-generation population using the roulette wheel selection method according to the fitness value. The crossover operation is specifically to randomly select two chromosomes C 1 and C 2 , randomly select a crossover point k on the chromosome, and exchange the parts of C 1 and C 2 after the crossover point to obtain the new chromosomes C′ 1 and C′ 2 . The mutation operation is specifically to randomly change a certain gene position in each chromosome with a certain mutation probability p m , for example, changing c i to 1 - c i . After multiple generations of evolution, the optimal drilling path and location are obtained.

[0120] Therefore, the present invention adopts the above-mentioned intelligent geological and mineral exploration system and method based on multi-source data fusion. Through the mutual coordination and cooperation among the data acquisition module, data preprocessing module, data fusion module, and intelligent analysis module, the multi-source data is fused, integrating information from various data sources such as ground, air, and satellite, making full use of the advantages of different data, reducing the limitations of a single data source, and at the same time enhancing the data processing ability. By optimizing the drilling path through the genetic algorithm, the drilling cost can be reduced, the exploration efficiency can be improved, and unnecessary drilling work can be reduced.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements do not enable the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A geological and mineral intelligent detection system based on multi-source data fusion, characterized in that: It includes data acquisition module, data preprocessing module, data fusion module and intelligent analysis module; Data acquisition module: collects geological and mineral data through a variety of sensors; Data preprocessing module: cleans, standardizes and reduces the dimension of the collected data; Data fusion module: uses Bayesian network to fuse pre-processed multi-source data; Intelligent analysis module: Identify potential mineral areas through the YOLO target detection algorithm, then use the K-means clustering algorithm to classify geological units, and finally optimize the drilling path and location selection through the genetic algorithm.

2. The intelligent geological and mineral exploration system based on multi-source data fusion according to claim 1 is characterized by: Sensors include ground sensors, aerial sensors and satellite remote sensing data sensors; among them, ground sensors are gravimeters, magnetometers or resistivity logging equipment; aerial sensors are LiDAR scanners or hyperspectral cameras carried by drones; satellite remote sensing data sensors are synthetic aperture radars or optical satellite images.

3. The intelligent geological and mineral exploration system based on multi-source data fusion according to claim 2 is characterized in that: The collected data is cleaned by removing outliers in the data by setting thresholds; missing values ​​are filled by mean filling or interpolation.

4. The intelligent geological and mineral exploration system based on multi-source data fusion according to claim 3 is characterized in that: The specific standardization of the collected data is: standardize the gravity value, magnetic value, resistivity value, and various values ​​in the air sensor and satellite remote sensing data sensor to make them in the same order of magnitude. The calculation expression is as follows: In the formula, x ′ represents the transformed data, x is the original data, μ is the mean, σ is the standard deviation, and the calculation expressions of the mean μ and the standard deviation σ are as follows: In the formula, the data set is {x1,x2,…,x n }.

5. The intelligent geological and mineral exploration system based on multi-source data fusion according to claim 4 is characterized in that: The process of dimensionality reduction of the collected data is as follows: S1. Data centering, mean centering for each feature, the formula is: Where X is the original data matrix, is the mean vector of each column feature, X centered represents the data center matrix; S2. Calculate the covariance matrix Cov(X) of the data matrix. The formula is: Where m is the number of samples; S3, eigenvalue decomposition, perform eigenvalue decomposition on the covariance matrix, the formula is: Cov(X)=QΛQ T Where Q is the eigenvector matrix, Λ is the diagonal eigenvalue matrix; S4, select the principal component, select the eigenvectors with the largest eigenvalues ​​as the principal component; S5. Construct the projection matrix Q selected , the formula is Q selected =[v1,v2,...,v k ] Among them, v k represents the kth eigenvector; S6. Project the data onto the selected principal component and visualize the data. The formula is: X reduced =X centered ×Q selected Where, X reduced is the projection matrix.

6. A method of a geological mineral intelligent detection system based on multi-source data fusion as described in any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Collect data, collect geological and mineral data through ground sensors, aerial sensors and satellite remote sensing data sensors; Step 2: The data collected in step 1 is cleaned, standardized and dimensionally reduced through a data preprocessing module; Step 3: Fuse the preprocessed data in step 2; extract the features of the ground objects from the hyperspectral remote sensing images through convolutional neural networks, then use long short-term memory networks to capture the dynamic change rules, and finally use Bayesian networks to perform weighted fusion of geological and mineral data from different sources; Step 4: Use the intelligent analysis module to analyze potential mineral areas and optimize drilling paths and location selection through genetic algorithms.

7. The method of a geological mineral intelligent detection system based on multi-source data fusion according to claim 6 is characterized in that: The specific process of step 3 is as follows: S31. Extract ground features from hyperspectral remote sensing images through convolutional neural networks. The calculation expression is as follows: In the formula, c ij represents the result of convolution operation, a is the input image element, k is the convolution kernel element, and the size of the convolution kernel is p×q; S32. Use the long short-term memory network to capture dynamic changes. The long short-term memory network is controlled by three gates: forget gate, input gate and output gate. The forget gate determines which information is forgotten or retained from the memory unit at each time step; the calculation expression is as follows: f t =σ(W f ·[h t-1 ,x t ]+b f ); Where σ is the logistic activation function, W f and b f are the weight matrix and bias term of the forget gate respectively, [h t-1 ,x t ] is the concatenation of the hidden state of the previous time step and the input of the current time step, f t is the output of the forget gate, f t Determines how much information about the memory cell state at the previous moment should be "forgotten" or "discarded"; The input gate consists of two parts: a sigmoid layer that determines which values ​​will be updated, and a tanh layer that creates a new vector of candidate values ​​that will be added to the state; the calculation expression of the input gate is as follows: Among them, i t is the output of the input gate, is the candidate memory cell state, W i , W C and b i , b C are the relevant weights and biases respectively, σ represents the sigmoid function; The output gate is responsible for determining which part of the memory cell state will be output to the hidden state. The calculation expression is as follows: the t =σ(W o ·[h t-1 ,x t ]+b o ); h t =o t *tanh(C t ); In the formula, C t Combining the information of the forget gate and the input gate, o t is the output of the sigmoid function of the output gate, C t is the updated memory cell state, h t is the final hidden state output, W o is the weight matrix of the output gate, which is used to adjust the influence of the input data and the hidden state of the previous moment on the current output, b o is the bias term of the output gate, used to adjust the calculated offset; S33. Use Bayesian network to perform weighted fusion of geological and mineral data from different sources.

8. The method of a geological mineral intelligent detection system based on multi-source data fusion according to claim 7 is characterized in that: The process of weighted fusion of geological and mineral data from different sources using Bayesian network in S33 is as follows: Let D1, D2, …, D n For data from different sources, according to Bayes' theorem A is the fused event, and B is the event related to different data sources. First, determine the data D of each data source. i The prior probability P(D i ), and the conditional probability P(A|D) of each data source under a given fusion result A i ), calculate the posterior probability by Bayes formula Determine the weight w of each data in the fusion result according to the posterior probability i =P(D i |A), and finally obtain the result of weighted fusion 9. The method of a geological mineral intelligent detection system based on multi-source data fusion according to claim 8, characterized in that: The process of analyzing potential mineral areas using the intelligent analysis module is as follows: S41. Identify potential mineral areas through the YOLO target detection algorithm; divide the input image into S×S grids, each grid is responsible for predicting B bounding boxes, each bounding box contains 5 prediction values ​​(x, y, w, h, c), where (x, y) is the offset of the center coordinates of the bounding box relative to the grid, w and h are the width and height of the bounding box, and c is the confidence that the bounding box contains the target; for each grid, C category probabilities p are also predicted c , the final output of each grid is a vector of length 5B+C; through the forward propagation of the network, the prediction results are obtained on each network, and then the potential mineral areas are screened out according to the preset threshold; S42, use K-means clustering algorithm to classify geological units; first randomly select K initial cluster centers μ1,μ2,…,μ K , and then calculate each data point x based on the Euclidean distance formula j The distance to these K centers is calculated as where x jl and μ il The data points x j and cluster center μ i The lth coordinate component of , assigns the data point to the class to which the nearest cluster center belongs, and then recalculates the center of each class. The calculation expression is Among them, C i is the set of data points contained in the i-th class, and the above process is repeated until the cluster center no longer changes.

10. The method of a geological mineral intelligent detection system based on multi-source data fusion according to claim 9, characterized in that: The optimization of drilling path and location selection by genetic algorithm is as follows: Assume that the drilling path consists of m decision variables x1, x2, …, x m It is composed of a binary chromosome C = [c1, c2, ..., c L ]; through selection, crossover and mutation operations, the population is continuously evolved to obtain the optimal drilling path and position; the selection operation is specifically to calculate the fitness value F(C) of each chromosome, and the fitness function is constructed according to the drilling cost and expected mineral reserve indicators, and the expression is: F(C) = a×Vb×C d , where V is the expected mineral reserve, C d is the drilling cost, a and b are weight coefficients, and the roulette wheel selection method is used to select excellent chromosomes to enter the next generation population according to the fitness value; the crossover operation is to randomly select two chromosomes C1 and C2, randomly select a crossover point k on the chromosome, and exchange the parts of C1 and C2 after the crossover point to obtain new chromosomes C′1 and C′2; the mutation operation is to perform a mutation on each chromosome with a certain mutation probability p m By randomly changing a gene position in the chromosome, the optimal drilling path and position are obtained after multiple generations of evolution.

Images