An intelligent detection system for gold mine exploration based on deep learning

Abstract

A deep learning-based intelligent detection system for gold mine exploration relates to the technical field of geological exploration. The system constructs a historical mine model of a mine and divides it into several local ore bodies. Different first-class difference sets are obtained according to the extension law and lateral trend of each local ore body. Different second-class difference sets are obtained according to the element content of each local ore body. An ore body prediction model is constructed based on the first-class difference sets and the second-class difference sets. The ore body prediction model is used to obtain a predicted ore body. The predicted ore body is verified to obtain a distribution position. The predicted mine model is constructed in combination with the historical mine model. The distribution position is detected to obtain new resource reserves. The system can form an effective prediction mechanism for mineral exploration and improve the efficiency of mineral exploration.

Description

Technical Field

[0001] The present invention relates to the field of geological prospecting technology, and in particular to an intelligent detection system for gold mine exploration based on deep learning. Background Art

[0002] In the existing technology, the identified ore bodies are often analyzed as a whole to predict potential ore bodies. This method ignores the parameter differences between different areas within the ore body and fails to better utilize the existing data. In addition, most existing technologies use a single mineral exploration method and fail to integrate different technical means to accurately locate the ore body, resulting in excessively long mineral exploration time and low overall efficiency. In response to the shortcomings of the existing technology, the present invention provides a gold mine exploration intelligent detection system based on deep learning. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent detection system for gold mine exploration based on deep learning.

[0004] The purpose of the present invention can be achieved through the following technical solutions: A gold mine exploration intelligent detection system based on deep learning, including the following modules:

[0005] The data acquisition module is used to obtain the historical exploration data of the mine and build its historical mine model. The distribution characteristics of the ore body in the historical mine model are obtained and divided into several local ore bodies.

[0006] The first analysis module is used to obtain the extension law and lateral trend of each local ore body, and obtain different types of difference sets according to the extension law and lateral trend of adjacent local ore bodies;

[0007] The second analysis module is used to obtain the element content of each local ore body and obtain different second-class difference sets based on the element content of adjacent local ore bodies;

[0008] The ore body prediction module is used to construct an ore body prediction model based on the first-class difference set and the second-class difference set of adjacent local ore bodies, and obtain the corresponding predicted ore body using the ore body prediction model;

[0009] The ore body assessment module is used to verify the predicted ore body using the CSAMT method to obtain the corresponding distribution location, build a predicted mine model in combination with the historical mine model, and detect the distribution location to obtain the corresponding new resource reserves.

[0010] Furthermore, the process of obtaining historical exploration data of a mine and building its historical mine model includes:

[0011] The historical exploration data refers to the mineral data, drilling data, topographic data, geophysical data, and geochemical data of the ore bodies that have been identified in the current mine;

[0012] Use 3D modeling tools to construct a terrain model of the mine based on terrain data, a geological model of the mine based on drilling data, geophysical data, and geochemical data, and a distribution model of the ore bodies within the mine based on mineral data. Combine the terrain model, geological model, and distribution model to obtain a historical mine model.

[0013] Furthermore, the process of obtaining the distribution characteristics of the ore body in the historical mine model and dividing it into several local ore bodies includes:

[0014] Construct a corresponding three-dimensional coordinate system for the ore body in the historical mine model, and extract the distribution characteristics of the identified ore body, including coordinate information, morphological information, and scale information;

[0015] The coordinate information refers to the coordinate values ​​of each point on the distribution model surface of the ore body in the three-dimensional coordinate system, the morphological information refers to the shape and structure of the ore body, and the scale information refers to the size of the ore body;

[0016] In the historical mine model, the major axis, minor axis and center of the complete ore body are obtained by combining the distribution characteristics of the ore body. Several two-dimensional planes are set in sequence in the direction perpendicular to the major axis. The complete ore body is divided into different local ore bodies through the two-dimensional planes. The distances between adjacent two-dimensional planes are equal and parallel to each other.

[0017] Furthermore, the process of obtaining the extension law and lateral trend of each local ore body and obtaining different types of difference sets according to the extension law and lateral trend of adjacent local ore bodies includes:

[0018] The extension law refers to the strike, inclination and dip of the ore body in the mine, and the lateral trend refers to the lateral deviation characteristics of the ore body in the mine compared to the horizontal plane, including bending, turning, bifurcation and merging;

[0019] The local ore body where the center of the complete ore body is located is marked as K0, and each local ore body is numbered outward in sequence along the long axis direction of the complete ore body, and is respectively marked as K ai and K bj , where a and b represent different directions, which are completely opposite;

[0020] Adjacent local ore bodies are combined to obtain the corresponding combined ore body. Local ore bodies in different directions are not combined. Local ore bodies that are adjacent to the combined ore body in the same direction and far away from the center of the complete ore body are regarded as the extended ore body of the combined ore body.

[0021] The extension law and lateral trend of the combined ore body and its extended ore body are regarded as a type of difference set between the two, and the exhaustive method is used to obtain a type of difference set between different combined ore bodies and their extended ore bodies.

[0022] Furthermore, the element content of each local ore body is obtained, and the process of obtaining different second-class difference sets according to the element content of adjacent local ore bodies includes:

[0023] Sampling each local ore body separately to obtain local ore body samples, and using spectral quantitative analysis method to analyze each local ore body sample separately to obtain the element content of the corresponding local ore body;

[0024] The mean value of the element content of each local ore body contained in the combined ore body is taken as the element content of the combined ore body, and the element content of the combined ore body and its extended ore body is taken as the second-category difference set between the two. The exhaustive method is used to obtain the second-category difference sets between different combined ore bodies and their extended ore bodies.

[0025] Furthermore, an ore body prediction model is constructed based on the first-class difference set and the second-class difference set of the adjacent local ore bodies. The process of obtaining the corresponding predicted ore body using the ore body prediction model includes:

[0026] According to the first-class difference sets and second-class difference sets of different combined ore bodies and their extended ore bodies, a comprehensive difference set of the two is obtained, and the comprehensive difference set is divided into a training set and a test set;

[0027] Constructing a convolutional neural network, taking the extension laws, lateral trends, and element contents of different combined ore bodies in the training set as input data of the convolutional neural network, and taking the extension laws, lateral trends, and element contents of the corresponding extended ore bodies in the training set as output data of the convolutional neural network, and training the convolutional neural network to obtain an initial convolutional neural network;

[0028] The initial convolutional neural network is validated using the test set, and the initial convolutional neural network with a test error threshold less than or equal to the preset value is output as the ore body prediction model.

[0029] The extension law, lateral trend and element content of the complete ore body are input into the ore body prediction model to obtain the extension law, lateral trend and element content of the next predicted ore body, which refers to the next local ore body that has not yet been identified relative to the complete ore body.

[0030] Furthermore, the CSAMT method is used to verify the predicted ore body to obtain the corresponding distribution location. The process of building a predictive mine model in combination with the historical mine model includes:

[0031] In the historical mine model, the distribution area of ​​the predicted ore body is obtained according to the extension law and lateral trend of the predicted ore body, and corresponding element content thresholds are set for different element contents;

[0032] The element contents of the predicted ore body are compared with their corresponding element content thresholds. When at least k elements have contents greater than or equal to their element content thresholds, the predicted ore body is marked as a verification state, where k is a preset natural number.

[0033] A geophysical exploration profile is laid out on the surface corresponding to the distribution area of ​​the predicted ore body in the verification state. The CSAMT method is used to measure the resistivity of the subsurface from shallow to deep at each preset measuring point. The resistivity contour map is drawn using Surfer software, and the distribution position of the predicted ore body is obtained. The distribution position is uploaded to the historical mine model for synchronization to obtain the predicted mine model.

[0034] Furthermore, the process of detecting the distribution location to obtain the corresponding newly added resource reserves includes:

[0035] The distribution location of the predicted ore body is sampled to obtain the predicted ore body sample, and the corresponding element content is obtained. The available ore body is delineated according to the element content. The element content of the available ore body meets the preset element content requirements, and the new resource reserves of the delineated available ore body are obtained in the predicted mine model.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention constructs a historical mine model of a mine, divides the complete ore body that has been identified therein into different local ore bodies, and obtains different combined ore bodies and their corresponding extended ore bodies by arranging and combining the local ore bodies. According to the corresponding relationship between the extension law, lateral trend, and element content of the currently identified combined ore body and its extended ore body, a corresponding ore body prediction model can be constructed, which is conducive to obtaining the corresponding parameters of the potential predicted ore body based on the various parameters of the current complete ore body, and can form an effective prediction mechanism for mineral exploration, thereby improving the efficiency of mineral exploration;

[0038] By using the CSAMT method to verify the predicted ore body to obtain the corresponding distribution position, different means can be combined to obtain the specific location of the predicted ore body, which is conducive to improving the accuracy of mineral exploration. Compared with the traditional single mineral exploration method, the present invention can significantly improve the prospecting accuracy, save exploration time, and realize an efficient and green exploration mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0040] like Figure 1 As shown in the figure, a deep learning-based intelligent detection system for gold mine exploration includes the following modules:

[0041] The data acquisition module is used to obtain the historical exploration data of the mine and build its historical mine model. The distribution characteristics of the ore body in the historical mine model are obtained and divided into several local ore bodies.

[0042] The first analysis module is used to obtain the extension law and lateral trend of each local ore body, and obtain different types of difference sets according to the extension law and lateral trend of adjacent local ore bodies;

[0043] The second analysis module is used to obtain the element content of each local ore body and obtain different second-class difference sets based on the element content of adjacent local ore bodies;

[0044] The ore body prediction module is used to construct an ore body prediction model based on the first-class difference set and the second-class difference set of adjacent local ore bodies, and obtain the corresponding predicted ore body using the ore body prediction model;

[0045] The ore body assessment module is used to verify the predicted ore body using the CSAMT method to obtain the corresponding distribution location, build a predicted mine model in combination with the historical mine model, and detect the distribution location to obtain the corresponding new resource reserves.

[0046] It should be further explained that, in the specific implementation process, the process of obtaining the historical exploration data of the mine and building its historical mine model includes:

[0047] The historical exploration data refers to the mineral data, drilling data, topographic data, geophysical data, and geochemical data of the ore bodies that have been identified in the current mine;

[0048] The mineral data includes the location, reserves, and grade of the ore body; the drill hole data includes the location, depth, and lithology of the drill hole; the topographic data includes the mine's topographic map, elevation data, and remote sensing images; the geophysical data includes seismic data, magnetic data, and electrical data; and the geochemical data includes chemical analysis results of soil and rock;

[0049] Use 3D modeling tools to construct a terrain model of the mine based on terrain data, a geological model of the mine based on drilling data, geophysical data, and geochemical data, and a distribution model of the ore bodies within the mine based on mineral data. Combine the terrain model, geological model, and distribution model to obtain a historical mine model.

[0050] It should be further explained that, in the specific implementation process, the process of obtaining the distribution characteristics of the ore body in the historical mine model and dividing it into several local ore bodies includes:

[0051] Constructing a corresponding three-dimensional coordinate system for the ore body in the historical mine model, including x-axis, y-axis, and z-axis, and extracting the distribution characteristics of the identified ore body, including coordinate information, morphological information, and scale information;

[0052] The coordinate information refers to the coordinate values ​​of each point on the surface of the distribution model of the ore body in the three-dimensional coordinate system; the morphological information refers to the shape and structure of the ore body, including layered, vein-like, massive, and cystic; the scale information refers to the size of the ore body, including length, width, and thickness;

[0053] In the embodiments of the present invention, the ore body refers to a single, continuous, and complete ore body, which is regarded as an irregular three-dimensional object. The major axis, minor axis, and center of the complete ore body are obtained by combining the distribution characteristics of the ore body in the historical mine model;

[0054] The major axis refers to the longest one-dimensional dimension direction of the irregular three-dimensional object in space, and the minor axis refers to the shortest one-dimensional dimension direction of the irregular three-dimensional object in the direction perpendicular to the major axis. Several two-dimensional planes are arranged in sequence in the direction perpendicular to the major axis. The complete ore body can be divided into different local ore bodies through the two-dimensional planes, and the distances between adjacent two-dimensional planes are equal and parallel to each other.

[0055] It should be further explained that, in the specific implementation process, the process of obtaining the extension law and lateral trend of each local ore body and obtaining different types of difference sets according to the extension law and lateral trend of adjacent local ore bodies includes:

[0056] The extension law refers to the strike, inclination and dip of the ore body in the mine. The strike refers to the extension direction of the ore body, which is usually consistent with the long axis direction of the ore body;

[0057] The inclination refers to the state in which an ore body tilts in a certain direction, including stable (i.e., the inclination of an ore body is consistent at different depths or locations) and variable (i.e., the inclination of an ore body changes with depth or location). The dip angle refers to the angle between the ore body and the horizontal plane.

[0058] The local ore body where the center of the complete ore body is located is marked as K0, and each local ore body is numbered outward in sequence along the long axis direction of the complete ore body, and is respectively marked as K ai and K bj , where a and b represent different directions, which are completely opposite to each other, i = 1, 2, ..., n, where n is the number of local ore bodies in the a direction, and j = 1, 2, ..., m, where m is the number of local ore bodies in the b direction;

[0059] The lateral trend refers to the lateral deviation characteristics of the ore body in the mine compared to the horizontal plane, including bending, turning, bifurcation and merging, which are all common knowledge;

[0060] Combine adjacent local ore bodies to obtain the corresponding combined ore body. For example, combine local ore bodies K0, K a1 Combine to obtain combined ore body K 0,a1, or the local ore body K a1 , K a2 , K a3 Combine to obtain combined ore body K a1,2,3 , and even K b4,5,6,7 , local ore bodies in different directions are not combined;

[0061] The local ore body adjacent to the combined ore body in the same direction and far away from the center of the complete ore body is regarded as the extended ore body of the combined ore body. The extension law and lateral trend of the combined ore body and its extended ore body are obtained respectively, and regarded as a class of difference sets between the two. The exhaustive method is used to obtain a class of difference sets between different combined ore bodies and their extended ore bodies.

[0062] It should be further explained that, in a specific implementation process, the process of obtaining the element content of each local ore body and obtaining different second-class difference sets according to the element content of adjacent local ore bodies includes:

[0063] Each local ore body is sampled to obtain the corresponding local ore body sample, and the spectral quantitative analysis method is used to analyze each local ore body sample to obtain the element content of the corresponding local ore body, including the near-ore halo elements Au, Ag, Cu, Pb, Zn, As, the leading halo elements Sb, Hg, Ba, F, Mo, and the trailing halo elements Co, V, Ti, Mn, etc.

[0064] The mean value of the element content of each local ore body contained in the combined ore body is taken as the element content of the combined ore body, and the element content of the combined ore body and its extended ore body is taken as the second-category difference set between the two. The exhaustive method is used to obtain the second-category difference sets between different combined ore bodies and their extended ore bodies.

[0065] It should be further explained that, in the specific implementation process, the ore body prediction model is constructed based on the first-class difference set and the second-class difference set of adjacent local ore bodies. The process of obtaining the corresponding predicted ore body using the ore body prediction model includes:

[0066] Obtaining a comprehensive difference set of the two based on the first and second difference sets of different combined ore bodies and their extended ore bodies, wherein the comprehensive difference set includes the extension rules, lateral trends, and element contents of the different combined ore bodies and their extended ore bodies, and dividing the obtained comprehensive difference set into a training set and a test set;

[0067] Constructing a convolutional neural network, taking the extension laws, lateral trends, and element contents of different combined ore bodies in the training set as input data of the convolutional neural network, and taking the extension laws, lateral trends, and element contents of the corresponding extended ore bodies in the training set as output data of the convolutional neural network, and training the convolutional neural network to obtain an initial convolutional neural network;

[0068] The initial convolutional neural network is validated using the test set, and the initial convolutional neural network with a test error threshold less than or equal to the preset value is output as the ore body prediction model.

[0069] The extension law, lateral trend and element content of the complete ore body are input into the ore body prediction model, and the ore body prediction model is used to output the extension law, lateral trend and element content of the next predicted ore body. The predicted ore body refers to the next local ore body that has not yet been identified relative to the complete ore body.

[0070] It should be further explained that, in the specific implementation process, the CSAMT method is used to verify the predicted ore body to obtain the corresponding distribution location. The process of building a predictive mining model in combination with the historical mining model includes:

[0071] In the historical mine model, the distribution area of ​​the predicted ore body is obtained according to the extension law and lateral trend of the predicted ore body. The distribution area at this time is a rough area;

[0072] Set corresponding element content thresholds for different element contents, compare the element contents of the predicted ore body with the corresponding element content thresholds, and mark the predicted ore body as verified when at least k element contents are greater than or equal to the element content thresholds, where k is a preset natural number;

[0073] Geophysical exploration profiles are laid out on the surface corresponding to the distribution area of ​​the predicted ore body under verification, and the subsurface resistivity is measured from shallow to deep at each preset measuring point using the CSAMT (controlled source audio frequency magnetotelluric sounding) method;

[0074] Use Surfer software to draw the corresponding resistivity contour map and obtain the distribution position of the predicted ore body. The distribution position at this time is a specific location. The obtained distribution position is uploaded to the historical mine model for synchronization to obtain the corresponding predicted mine model.

[0075] It should be further explained that, in the specific implementation process, the process of detecting the distribution location to obtain the corresponding new resource reserves includes:

[0076] Sampling is performed on the distribution location of the predicted ore body to obtain the corresponding predicted ore body samples and the corresponding element content. The corresponding available ore body is delineated based on the obtained element content. The element content of the available ore body must meet the preset element content requirements. The reserves of the delineated available ore body are obtained in the predicted mine model and marked as new resource reserves.

[0077] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A gold mine exploration intelligent detection system based on deep learning, characterized in that: Includes the following modules: The data acquisition module is used to obtain the historical exploration data of the mine and build its historical mine model. The distribution characteristics of the ore body in the historical mine model are obtained and divided into several local ore bodies. The first analysis module is used to obtain the extension law and lateral trend of each local ore body, and obtain different types of difference sets according to the extension law and lateral trend of adjacent local ore bodies; The second analysis module is used to obtain the element content of each local ore body and obtain different second-class difference sets based on the element content of adjacent local ore bodies; The ore body prediction module is used to construct an ore body prediction model based on the first-class difference set and the second-class difference set of adjacent local ore bodies, and obtain the corresponding predicted ore body using the ore body prediction model; The ore body assessment module is used to verify the predicted ore body using the CSAMT method to obtain the corresponding distribution location, build a predicted mine model based on the historical mine model, and detect the distribution location to obtain the corresponding new resource reserves; The process of obtaining the extension law and lateral trend of the local ore body and obtaining different types of difference sets includes: The extension law refers to the strike, inclination and dip of the ore body in the mine, and the lateral trend refers to the lateral deviation characteristics of the ore body in the mine compared to the horizontal plane, including bending, turning, bifurcation and merging; The local ore body where the center of the complete ore body is located is marked as K0, and each local ore body is numbered outward in sequence along the long axis direction of the complete ore body, and is respectively marked as K ai and K bj , where a and b represent different directions, which are completely opposite; Adjacent local ore bodies are combined to obtain the corresponding combined ore body. Local ore bodies in different directions are not combined. Local ore bodies that are adjacent to the combined ore body in the same direction and far away from the center of the complete ore body are regarded as the extended ore body of the combined ore body. The extension law and lateral trend of the combined ore body and its extended ore body are regarded as a type of difference set between the two, and the exhaustive method is used to obtain a type of difference set between different combined ore bodies and their extended ore bodies.

2. The deep learning-based intelligent detection system for gold mine exploration according to claim 1, characterized in that: The process of obtaining historical exploration data and building a historical mine model includes: The historical exploration data refers to the mineral data, drilling data, topographic data, geophysical data, and geochemical data of the ore bodies that have been identified in the current mine; Use 3D modeling tools to construct a terrain model of the mine based on terrain data, a geological model of the mine based on drilling data, geophysical data, and geochemical data, and a distribution model of the ore bodies within the mine based on mineral data. Combine the terrain model, geological model, and distribution model to obtain a historical mine model.

3. The deep learning-based intelligent detection system for gold mine exploration according to claim 2 is characterized in that: The process of obtaining the distribution characteristics of the ore body and dividing it into several local ore bodies includes: Construct a corresponding three-dimensional coordinate system for the ore body in the historical mine model, and extract the distribution characteristics of the identified ore body, including coordinate information, morphological information, and scale information; The coordinate information refers to the coordinate values ​​of each point on the distribution model surface of the ore body in the three-dimensional coordinate system, the morphological information refers to the shape and structure of the ore body, and the scale information refers to the size of the ore body; In the historical mine model, the major axis, minor axis and center of the complete ore body are obtained by combining the distribution characteristics of the ore body. Several two-dimensional planes are set in sequence in the direction perpendicular to the major axis. The complete ore body is divided into different local ore bodies through the two-dimensional planes. The distances between adjacent two-dimensional planes are equal and parallel to each other.

4. The deep learning-based intelligent detection system for gold mine exploration according to claim 3 is characterized in that: The process of obtaining the element content of a local ore body and obtaining different second-class difference sets includes: Sampling each local ore body separately to obtain local ore body samples, and using spectral quantitative analysis method to analyze each local ore body sample separately to obtain the element content of the corresponding local ore body; The mean value of the element content of each local ore body contained in the combined ore body is taken as the element content of the combined ore body, and the element content of the combined ore body and its extended ore body is taken as the second-category difference set between the two. The exhaustive method is used to obtain the second-category difference sets between different combined ore bodies and their extended ore bodies.

5. The deep learning-based intelligent detection system for gold mine exploration according to claim 4 is characterized in that: The process of constructing an ore body prediction model based on the first-class difference set and the second-class difference set and obtaining the predicted ore body includes: According to the first-class difference sets and second-class difference sets of different combined ore bodies and their extended ore bodies, a comprehensive difference set of the two is obtained, and the comprehensive difference set is divided into a training set and a test set; Constructing a convolutional neural network, taking the extension laws, lateral trends, and element contents of different combined ore bodies in the training set as input data of the convolutional neural network, and taking the extension laws, lateral trends, and element contents of the corresponding extended ore bodies in the training set as output data of the convolutional neural network, and training the convolutional neural network to obtain an initial convolutional neural network; The initial convolutional neural network is validated using the test set, and the initial convolutional neural network with a test error threshold less than or equal to the preset value is output as the ore body prediction model. The extension law, lateral trend and element content of the complete ore body are input into the ore body prediction model to obtain the extension law, lateral trend and element content of the next predicted ore body, which refers to the next local ore body that has not yet been identified relative to the complete ore body.

6. The deep learning-based intelligent detection system for gold mine exploration according to claim 5, characterized in that: The process of obtaining the distribution location of the predicted ore body and building a predicted mine model includes: In the historical mine model, the distribution area of ​​the predicted ore body is obtained according to the extension law and lateral trend of the predicted ore body, and corresponding element content thresholds are set for different element contents; The element contents of the predicted ore body are compared with their corresponding element content thresholds. When at least k elements have contents greater than or equal to their element content thresholds, the predicted ore body is marked as a verification state, where k is a preset natural number. A geophysical exploration profile is laid out on the surface corresponding to the distribution area of ​​the predicted ore body in the verification state. The CSAMT method is used to measure the resistivity of the subsurface from shallow to deep at each preset measuring point. The resistivity contour map is drawn using Surfer software, and the distribution position of the predicted ore body is obtained. The distribution position is uploaded to the historical mine model for synchronization to obtain the predicted mine model.

7. The deep learning-based intelligent detection system for gold mine exploration according to claim 6, characterized in that: The process of acquiring new resource reserves includes: The distribution location of the predicted ore body is sampled to obtain the predicted ore body sample, and the corresponding element content is obtained. The available ore body is delineated according to the element content. The element content of the available ore body meets the preset element content requirements, and the new resource reserves of the delineated available ore body are obtained in the predicted mine model.

Images