Geological mineral exploration data extraction method and system

Abstract

The invention belongs to the technical field of geological information processing, and discloses a geological mineral exploration data extraction system which comprises the following steps: constructing an air-ground-hole three-in-one acquisition network, and respectively acquiring earth surface mineral spectrum data, seismic longitudinal wave and transverse wave reflection data and rock core main quantity and trace element composition data; processing the collected data to form a five-dimensional data cube containing space coordinates, spectrum parameters, physical property parameters, element parameters and time indexes, performing physical layer attribute extraction on the five-dimensional data cube, performing analysis to obtain a wave impedance value and a lithology category probability array, performing semantic layer attribute extraction on the five-dimensional data cube, and performing semantic layer attribute extraction on the five-dimensional data cube; an alteration type probability array is obtained through analysis, an alteration zoning graph layer is generated, the mineralization potential index and a five-dimensional data cube are fused, a four-dimensional dynamic database is constructed, a mineralization potential prediction graph layer and a drilling priority graph are generated based on the four-dimensional dynamic database, and therefore the decision-making level of ore body drilling deployment is improved.

Description

Technical Field

[0001] The present invention relates to the field of geological information processing technology, and more particularly to a method and system for extracting data from geological and mineral exploration. Background Art

[0002] In the exploration process of geological mineral resources, the precise identification of lithologic boundaries is one of the key factors restricting the accuracy of prospecting predictions and drilling efficiency. Currently, a variety of information collection methods have been widely used in the mineral exploration process. Remote sensing technology is used to obtain the spectral reflection characteristics of surface minerals, seismic reflection and electromagnetic detection methods are used to obtain underground geological structures and medium physical parameters, and drill core analysis is used to extract elemental composition and mineral combination data. The geological lithology is studied and analyzed by combining and analyzing these multi-source data.

[0003] However, at present, comprehensive exploration data are often processed through independent analysis processes, and the data collection links from different sources lack systematic planning. The spatial resolution and sampling density of different equipment vary greatly, resulting in difficulty in effectively fusing multi-source data and achieving accurate registration and unified modeling. As a result, the results of lithologic boundary identification are fragmented and inconsistent. In addition, the traditional data analysis process mainly relies on expert experience and two-dimensional plane judgment, which makes it difficult to fully depict the complex three-dimensional structural pattern and alteration evolution process of the underground, and cannot dynamically express the evolution path and spatiotemporal characteristics of the mineralization system, thereby limiting the comprehensive identification ability of potential ore bodies and the scientific decision-making level of drilling deployment.

[0004] In view of this, the present invention proposes a method and system for extracting data from geological and mineral exploration to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: A method for extracting data from geological and mineral exploration comprises:

[0006] Step S1: Construct an air-ground-hole trinity acquisition network to collect surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data;

[0007] Step S2: Processing the collected data to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index;

[0008] Step S3: Extract physical layer attributes from the five-dimensional data cube, analyze and obtain wave impedance values ​​and lithology category probability arrays. Simultaneously, extract semantic layer attributes from the five-dimensional data cube, analyze and obtain alteration type probability arrays and generate alteration zoning layers.

[0009] Step S4: performing simulation calculations on the wave impedance value, the lithology category probability value array, and the alteration type probability value array to obtain a mineralization potential index;

[0010] Step S5: Fusing the mineralization potential index with the five-dimensional data cube to construct a four-dimensional dynamic database, and generating a mineralization potential prediction layer and a drilling priority map based on the four-dimensional dynamic database.

[0011] Furthermore, the method of constructing an air-ground-hole trinity data collection network includes:

[0012] Preset the target exploration area and deploy various acquisition equipment in the target exploration area, including deploying an array of drones above the target exploration area. The flight path of the drone array is pre-designed using the principle of "vertical structural trend + grid coverage". The wireless seismic detector array is laid out on the ground in the target exploration area. The wireless seismic detector array is pre-arranged using the principle of "cross-shaped main axis + concentric circle auxiliary". Drilling points and drilling measurement equipment are deployed in the target exploration area.

[0013] Standardize sampling parameters for drones, geophones, and borehole survey equipment, including spatial resolution, sampling density, and sampling intervals;

[0014] Unified configuration of WGS84 spatial coordinate system for multiple types of acquisition equipment, and real-time positioning of spatial coordinates through satellite navigation and positioning services;

[0015] Connect various acquisition devices to the network time synchronization mechanism, use the network time protocol to obtain UTC timestamps, and embed timestamp calibration programs in various acquisition devices to automatically correct time offsets.

[0016] Furthermore, the method of separately collecting surface mineral spectrum data, seismic P-wave and S-wave reflection data, and core major and trace element composition data includes:

[0017] Use drone arrays to collect hyperspectral images of the area covered by the pre-designed flight path, and analyze the hyperspectral images to obtain surface mineral spectral data;

[0018] Acquire seismic longitudinal and shear wave reflection data through wireless seismic geophone arrays;

[0019] Drilling is carried out at the borehole point using drilling measurement equipment to collect continuous core samples, which are then analyzed to obtain the major and trace element composition data of the core;

[0020] The acquired surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are embedded with spatial coordinates and UTC timestamps respectively.

[0021] Furthermore, the method of processing the collected data to form a five-dimensional data cube including spatial coordinates, spectral parameters, physical property parameters, element parameters and time index includes:

[0022] A regular three-dimensional voxel grid is established in the target exploration area based on the WGS84 spatial coordinate system, and the center point coordinates of each voxel unit are used as its unique spatial index in the three-dimensional space to form a spatial dimension;

[0023] Based on spatial coordinate difference mapping, surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are assigned to corresponding voxel units, constructing an attribute dimension that includes spectral parameters, physical parameters, and elemental parameters.

[0024] Assign a time index to each voxel unit based on the UTC timestamp to form a time dimension;

[0025] The spatial dimension, attribute dimension and time dimension of each voxel unit are integrated to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index.

[0026] Furthermore, the method for extracting physical layer attributes from the five-dimensional data cube and analyzing and obtaining wave impedance values ​​and lithology category probability arrays includes:

[0027] Build a 3D-CNN model based on voxel feature learning, which includes an input layer, multi-layer 3D convolutional layers, batch normalization layers, activation layers, pooling layers, and fully connected layers;

[0028] The five-dimensional data cube is fed into the 3D-CNN model as input, and the voxel unit is used as the processing unit to extract its deep fusion properties.

[0029] Perform nonlinear transformation and spatial dimensionality reduction on the deep fusion attributes of each voxel unit and output its corresponding wave impedance value;

[0030] The deep fusion attributes of each voxel unit are input into the fully connected layer and passed through the Softmax function to output the lithology category probability array corresponding to each voxel unit.

[0031] Furthermore, the method of extracting semantic layer attributes from the five-dimensional data cube, analyzing to obtain an alteration type probability array and generating an alteration zoning layer includes:

[0032] Taking the five-dimensional data cube as the input data source, the spectral parameters, physical parameters, element parameters and time index of each voxel unit are extracted to form a fusion attribute vector;

[0033] Slice the five-dimensional data cube based on any two dimensions in the three-dimensional space coordinate axis, extract a representative two-dimensional slice sequence, and retain the fusion attribute vector corresponding to each voxel unit in the slice;

[0034] The retained fusion attribute vector is divided into channels according to parameter categories and re-encoded into a multi-channel image, where each channel corresponds to spectral features, physical properties, and elemental features to match the ResNet50 model input format;

[0035] Multi-channel images are input to the ResNet50 model, which performs deep semantic segmentation and classification prediction on the two-dimensional slices and outputs an array of alteration type probabilities for each voxel unit;

[0036] The alteration type probability array of each voxel unit is mapped to a three-dimensional voxel grid according to spatial coordinates to form an alteration zoning layer.

[0037] Furthermore, a method for simulating and calculating the wave impedance value, the lithology category probability value array, and the alteration type probability value array to obtain the mineralization potential index includes:

[0038] The wave impedance value, lithology category probability value array, and alteration type probability value array are used as input variables for mineralization potential analysis. Their corresponding values ​​are extracted in voxel units and normal distribution functions are fitted respectively.

[0039] Construct a mineralization potential estimation function with wave impedance, lithology probability and alteration probability as input;

[0040] For each voxel unit, N random samples are taken from the normal distribution of the three types of input variables, and the sampling results are substituted into the mineralization potential function to obtain N mineralization potential values;

[0041] The N-times mineralization potential values ​​of each voxel unit are statistically analyzed, and the expected value is calculated as the mineralization potential index of the voxel unit.

[0042] Furthermore, the method for constructing a four-dimensional dynamic database includes:

[0043] Extract the spatial coordinates and corresponding time index of each voxel unit in the five-dimensional data cube to construct a joint index key of space + time;

[0044] The mineralization potential index of each voxel unit is associated and matched with the corresponding space + time joint index key to generate a four-dimensional dynamic data record;

[0045] All four-dimensional dynamic data records are uniformly organized and stored to form a four-dimensional dynamic database.

[0046] Furthermore, the method for generating a mineralization potential prediction layer and a drilling priority map based on a four-dimensional dynamic database includes:

[0047] Index the four-dimensional dynamic database, extract the mineralization potential index of each voxel unit at different time indexes, and generate a mineralization potential prediction layer;

[0048] Perform regional cluster analysis on the mineralization potential prediction layer, extract potential value clustering areas and calculate their clustering strength indicators;

[0049] Based on the spatial coupling analysis of the mineralization potential prediction layer and the alteration zoning layer, the spatial synergy score of each area is calculated;

[0050] Comprehensively analyze the mineralization potential index, clustering intensity index, and synergy score to build a drilling priority assessment model;

[0051] Based on the drilling priority evaluation model, the target exploration area is spatially optimized and classified into drilling priority areas, suboptimal areas and low priority areas, and a drilling priority map is output.

[0052] A geological and mineral exploration data extraction system, comprising:

[0053] Acquisition module: used to build an air-ground-hole trinity acquisition network to collect surface mineral spectrum data, seismic longitudinal and shear wave reflection data, and core major and trace element composition data;

[0054] Data fusion module: used to process the collected data to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index;

[0055] Feature extraction module: used to extract attributes at the physical and semantic layers of the five-dimensional data cube, and obtain wave impedance values, lithology category probability arrays, alteration type probability arrays, and alteration zoning layers respectively;

[0056] Decision analysis module: simulates and calculates wave impedance values, lithology probability arrays, and alteration type probability arrays to obtain the mineralization potential index;

[0057] Result output module: used to fuse the mineralization potential index with the five-dimensional data cube, build a four-dimensional dynamic database, and generate mineralization potential prediction layers and drilling priority maps based on the four-dimensional dynamic database.

[0058] The technical effects and advantages of the method and system for extracting data from geological and mineral exploration of the present invention are as follows:

[0059] 1. This invention collaboratively deploys surface drones, ground seismometers, and underground borehole measurement equipment to construct a multi-scale observation system covering "air, ground, and borehole." This system can simultaneously acquire surface mineral spectral data, seismic P- and S-wave reflection data, and core major and trace element composition data. By adopting a unified WGS84 spatial coordinate system and embedding a GNSS positioning system, it can achieve high-precision spatial alignment of multiple types of equipment across air, ground, and boreholes, avoiding spatial drift during the integration of heterogeneous data and providing a consistent spatial indexing system for the subsequent construction of a five-dimensional data cube. All types of equipment are connected to a network time synchronization mechanism, and by embedding UTC timestamps, multi-source data have a unified time reference coordinate, ensuring timeliness for subsequent data processing. By standardizing the sampling parameters of various acquisition devices, this system effectively overcomes the fusion barriers caused by differences in resolution, coordinate system, and acquisition method in traditional exploration, ensuring comparability between different data types at the same scale and laying the foundation for multi-source fusion modeling.

[0060] 2. By constructing a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, elemental parameters, and time index, the present invention can fully express the spatiotemporal evolution of various geological information in the target area, solving the one-sidedness problem of traditional single-source data analysis and improving the systematicity and spatial consistency of data fusion. A 3D-CNN model is used at the physical layer to extract fusion attributes such as wave impedance and lithology probability value arrays, and a ResNet50 model is used at the semantic layer to identify alteration types. This can effectively improve the ability to express underground medium structure and alteration spatial distribution. By integrating the influencing factors related to wave impedance, lithology probability, and alteration type probability to construct a potential function, and performing large-sample deduction on the fused wave impedance, lithology probability value array, and alteration type probability value array, the mineralization potential index is made more credible and discriminative.

[0061] 3. By introducing the time dimension on the basis of the traditional three-dimensional voxel structure and mapping the potential index to the voxel grid, the present invention constructs a four-dimensional dynamic database with time series properties. It can not only support single-period spatial analysis, but also perform dynamic modeling of the mineralization process. It introduces spatial statistical indicators such as clustering intensity and spatial synergy scoring, and constructs a comprehensive scoring function with adjustable weights in combination with mineralization potential. It realizes the refined classification of drilling priority areas, suboptimal areas and low priority areas, provides a data-driven priority area division scheme for engineering deployment, and improves the scientific nature of drilling layout and the efficiency of resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of the structure of a method for extracting data from geological and mineral exploration according to the present invention;

[0063] Figure 2 This is a flow chart of a geological and mineral exploration data extraction system of the present invention;

[0064] Figure 3 Flowchart for generating a five-dimensional data cube for the present invention. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] Example 1

[0067] See also Figure 1 and Figure 3 As shown, the main design contents of the method for extracting data from geological and mineral exploration described in this embodiment are as follows:

[0068] The geological and mineral exploration process usually involves the collection of multiple data on the surface, underground, and in-hole. However, the original spatial resolution, coordinate system, and time system of the data collected by different collection equipment may be inconsistent. The lack of a unified coordinate and time reference causes spatial dislocation and time drift between the data, causing confusion in the time logic during geological analysis, and seriously restricting the accuracy of subsequent data fusion and modeling.

[0069] Based on this, a method for extracting data from geological and mineral exploration is designed, including:

[0070] Step S1: Construct an air-ground-hole trinity acquisition network to collect surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data;

[0071] The methods for constructing an air-ground-hole trinity data collection network include:

[0072] Preset the target exploration area and deploy various acquisition equipment in the target exploration area, including deploying an array of drones above the target exploration area. The flight path of the drone array is pre-designed using the principle of "vertical structural trend + grid coverage". The wireless seismic detector array is laid out on the ground in the target exploration area. The wireless seismic detector array is pre-arranged using the principle of "cross-shaped main axis + concentric circle auxiliary". Drilling points and drilling measurement equipment are deployed in the target exploration area.

[0073] Standardize sampling parameters for drones, geophones, and borehole survey equipment, including spatial resolution, sampling density, and sampling intervals;

[0074] Unified configuration of WGS84 spatial coordinate system for multiple types of acquisition equipment, and real-time positioning of spatial coordinates through satellite navigation and positioning services;

[0075] Connect various acquisition devices to the network time synchronization mechanism, use the network time protocol to obtain UTC timestamps, and embed timestamp calibration programs in various acquisition devices to automatically correct time offsets;

[0076] Methods for separately collecting surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data include:

[0077] Use drone arrays to collect hyperspectral images of the area covered by the pre-designed flight path, and analyze the hyperspectral images to obtain surface mineral spectral data;

[0078] Acquire seismic longitudinal and shear wave reflection data through wireless seismic geophone arrays;

[0079] Drilling is carried out at the drilling points using drilling measurement equipment to collect continuous core samples, and the core samples are analyzed to obtain the major and trace element composition data of the cores;

[0080] The acquired surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are embedded with spatial coordinates and UTC timestamps respectively.

[0081] It should be explained that the flight path of the drone array is pre-designed using the principle of "perpendicular to the structural direction + grid coverage". That is, the drone's route is as perpendicular to the structural direction as possible. The data profile obtained in this way can "cut" the geological body to the maximum extent, thus clearly showing the boundaries, contact relationships and structural deformation characteristics between different geological units. Grid coverage is based on the perpendicular structural direction and fully covers the entire target exploration area in the form of a regular grid to avoid data omissions.

[0082] It should be explained that the principle of "cross-shaped main axis + concentric circle auxiliary" is to arrange the seismic detectors in a cross shape along the main structural axis, and at the same time arrange several concentric circles in a ring to improve the accuracy of structural identification and reflection acquisition;

[0083] For example, the target exploration area is delineated according to the research objectives. The area of ​​this target exploration area is 10 square kilometers. The UAV track line is arranged vertically in the northwest-southeast direction to form a regular grid. At the same time, the grid spacing is set to 200 meters to ensure that the entire area is covered and no data is missed. With the center of the area as the origin, a cross-shaped main axis is arranged along the east-west and north-south directions. The spacing between the detectors on the main axis is set to 100 meters. Then, with the origin as the center, concentric circle auxiliary survey lines with radii of 300 meters and 600 meters respectively are set. The spacing between the detectors on the circle is also 100 meters, so as to effectively receive seismic longitudinal and shear wave reflection data from different directions. At the same time, three boreholes are deployed in the target exploration area with depths of 200 meters, 300 meters and 400 meters respectively to sample and analyze lithologic samples and obtain major and trace element composition data of the core.

[0084] It should be explained that hyperspectral images must first undergo preprocessing such as radiation correction, geometric correction, and atmospheric correction to eliminate the effects of solar altitude angle, atmospheric scattering, etc., to obtain the true surface reflectance, and then extract the spectral curve of each pixel from the preprocessed image. For each pixel, the reflectance value is extracted in band order to form a continuous spectral curve. When the pixel contains multiple minerals, the spectrum is a linear combination of the spectra of each mineral. The end-member spectra are separated by linear spectral unmixing, and the extracted spectra are compared with the known mineral spectral library to determine the surface mineral type and spectral parameters to form surface mineral spectral data. After collecting continuous core samples, the content of elements in the core samples is determined by X-ray fluorescence spectroscopy. The surface mineral spectral data includes band reflectance, principal component decomposition value, characteristic band, seismic longitudinal and shear wave reflection data include longitudinal wave velocity, shear wave velocity, Poisson's ratio, reflection energy, etc. The core major and trace element composition data include SiO2, Fe2O3, Cu, Pb, Zn, As, etc.

[0085] It needs to be explained that the sampling parameters of drones, seismic detectors and drilling measurement equipment are standardized to ensure the consistency and comparability of the data. For example, the spatial resolution of drone hyperspectral images is set to 1 meter, the sampling density is every 200 meters × 200 meters, and hyperspectral images are collected every 10 seconds; the spatial sampling interval of seismic detectors is 100 meters, and the sampling frequency of seismic longitudinal and shear wave reflection data is set to 1000Hz; the measurement accuracy of drilling measurement equipment for core samples is at the centimeter level, and the drilling measurement equipment collects a core sample at a certain speed and distance.

[0086] It should be explained that the coordinates of each pixel in the surface mineral spectral data correspond to the drone track acquisition location, and the timestamp is the moment the image was captured. The seismic data, including the longitudinal and shear wave reflection data recorded by each geophone, are associated with their spatial coordinates and acquisition time. Each analysis result in the core data corresponds to the drilling location and sampling time, thus enabling the corresponding spatial coordinates and UTC timestamps to be embedded.

[0087] It should be explained that by embedding a timestamp calibration program in the device, the time offset is automatically detected at regular intervals, and if the offset exceeds a certain range, it is automatically corrected.

[0088] By building an air-ground-hole trinity acquisition network and unifying WGS84 coordinates and UTC timestamps, the three types of data from aerial hyperspectral, ground seismic waves, and in-hole rock core can be seamlessly integrated in three-dimensional space and time dimensions. Preset layout rules and standardized sampling parameter settings, combined with a network time synchronization mechanism, greatly improve acquisition efficiency and reduce human intervention. Data embedded with coordinates and timestamps can be directly used for voxel mapping, reducing the difficulty of subsequent data processing and facilitating subsequent source tracking, cross-validation, and data reuse, thereby improving the overall reliability and reproducibility of the system.

[0089] Step S2: Processing the collected data to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index;

[0090] The method for processing the collected data to form a five-dimensional data cube including spatial coordinates, spectral parameters, physical property parameters, element parameters and time index includes:

[0091] A regular three-dimensional voxel grid is established in the target exploration area based on the WGS84 spatial coordinate system, and the center point coordinates of each voxel unit are used as its unique spatial index in the three-dimensional space to form a spatial dimension;

[0092] Based on spatial coordinate difference mapping, surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are assigned to corresponding voxel units, constructing an attribute dimension that includes spectral parameters, physical parameters, and elemental parameters.

[0093] Assign a time index to each voxel unit based on the UTC timestamp to form a time dimension;

[0094] The spatial dimension, attribute dimension and time dimension of each voxel unit are integrated to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index.

[0095] It should be explained that the coordinates of the center point of each voxel unit are generally selected as the longitude, latitude and elevation of the voxel center point;

[0096] It should be explained that the data acquisition positions and voxel center points of the original surface mineral spectral data, seismic longitudinal and shear wave reflection data, and core major and trace element composition data usually do not completely coincide with each other. Therefore, the difference between the center point coordinates and the spatial coordinates of the original data is calculated through spatial coordinate difference mapping, so that the original data can be embedded in the nearest voxel unit.

[0097] The representation of a 3D voxel grid is:

[0098] V={v x,y,z |x=1,...,N x ; y=1,...,N y ; z=1,...,N z ;}

[0099] The five-dimensional data cube can be represented as Each voxel unit data is represented by D (x,y,z,t) ={S v ,P v ,E v};

[0100] Among them: x, y, z are the index variables of voxel units in three-dimensional space, v represents the voxel unit, N x ,N y ,N z , respectively represents the number of voxels or resolution of the grid in three spatial dimensions, S v 、P v and E v They are respectively represented as the spectral parameters, physical property parameters and element parameters of the voxel unit v, d is the attribute dimension containing the spectral parameters, physical property parameters and element parameters, T is the time layer of the time dimension, and t means that each layer T corresponds to a time index, that is, t∈{1,2,3...,T}.

[0101] Step S3: Extract physical layer attributes from the five-dimensional data cube, analyze and obtain wave impedance values ​​and lithology category probability arrays. Simultaneously, extract semantic layer attributes from the five-dimensional data cube, analyze and obtain alteration type probability arrays and generate alteration zoning layers.

[0102] Methods for extracting physical layer attributes from a five-dimensional data cube and analyzing and obtaining wave impedance values ​​and lithology category probability arrays include:

[0103] Build a 3D-CNN model based on voxel feature learning, which includes an input layer, multi-layer 3D convolutional layers, batch normalization layers, activation layers, pooling layers, and fully connected layers;

[0104] The five-dimensional data cube is fed into the 3D-CNN model as input, and the voxel unit is used as the processing unit to extract its deep fusion properties.

[0105] Perform nonlinear transformation and spatial dimensionality reduction on the deep fusion attributes of each voxel unit and output its corresponding wave impedance value;

[0106] The deep fusion attributes of each voxel unit are input into the fully connected layer and passed through the Softmax function to output the lithology category probability array corresponding to each voxel unit.

[0107] The method for extracting semantic layer attributes from a five-dimensional data cube, analyzing to obtain an alteration type probability array, and generating an alteration zoning layer includes:

[0108] Taking the five-dimensional data cube as the input data source, the spectral parameters, physical parameters, element parameters and time index of each voxel unit are extracted to form a fusion attribute vector;

[0109] Slice the five-dimensional data cube based on any two dimensions in the three-dimensional space coordinate axis, extract a representative two-dimensional slice sequence, and retain the fusion attribute vector corresponding to each voxel unit in the slice;

[0110] The retained fusion attribute vector is divided into channels according to parameter categories and re-encoded into a multi-channel image, where each channel corresponds to spectral features, physical properties, and elemental features to match the ResNet50 model input format;

[0111] Multi-channel images are input to the ResNet50 model, which performs deep semantic segmentation and classification prediction on the two-dimensional slices and outputs an array of alteration type probabilities for each voxel unit;

[0112] The alteration type probability array of each voxel unit is mapped to a three-dimensional voxel grid according to spatial coordinates to form an alteration zoning layer.

[0113] The input layer in the 3D-CNN model receives It is converted into a tensor data format recognizable by the 3D-CNN model. Subsequently, multiple layers of 3D convolutional layers process the input data layer by layer to extract the deep fusion attributes between each voxel and its spatial neighborhood. After each convolutional layer, the batch normalization layer standardizes the feature output to accelerate network convergence and improve training stability. The activation layer performs nonlinear transformation on the normalized features to enhance the model's expressive power, effectively highlighting geological anomaly responses and suppressing invalid features. The pooling layer then performs spatial downsampling on the local feature area to further compress the data dimension, extract the main control information, and reduce computational complexity. Finally, the connection layer integrates the features output by all convolutional layers to output the wave impedance value and lithology category probability array for each voxel.

[0114] Set the Deep Fusion property to X v , and Xv ={S v ,P v ,E v}, set the wave impedance value of each voxel unit to Z v , Z v The expression for Z is: v =f pool (σ(F v ));

[0115] Where: σ(·) is the activation function, f pool (·) is the pooling function;

[0116] Set the lithology category probability array to R v , R v The expression to get is:

[0117] R v =Softmax(f fc (F v );

[0118] Where: f fc is the fully connected layer mapping function, and the lithology category probability array R v The sum of all class probabilities in is 1.

[0119] It should be explained that when slicing a five-dimensional data cube, any two dimensions of the three-dimensional coordinate axis can be selected. For example, when the three-dimensional coordinates are set to (X, Y, Z), (XY), (XZ), and (YZ) can be selected for slicing;

[0120] It should be explained that representative 2D slice sequences include those that can cover key geological structures or alteration-concentrated areas (fault zones, mineralized zones, alteration shells), span areas with large spatial parameter gradients (abrupt mineral reflectivity changes, element anomalous accumulation zones), and areas with high data density or strong information integrity (dense measured points, drill hole intersection areas);

[0121] It should be explained that the fusion attribute vector is divided into multiple independent image channels according to the parameter category, for example: channel 1 corresponds to spectral parameters, channel 2 corresponds to physical parameters, and channel 3 corresponds to element parameters;

[0122] Set the fusion attribute vector to X1 v , X1 v ={S v ,P v ,E v ,T v}, the alteration type probability array of each voxel unit is set to A v , A v The expression for obtaining is:

[0123]

[0124] Where: f ResTet is the ResNet50 network function, is the semantic feature vector of voxel v, Semantic fully connected mapping function.

[0125] Then, the two-dimensional slice results are mapped to the three-dimensional voxel position according to the original spatial coordinates to generate the alteration zoning layer, which is expressed as:

[0126]

[0127] Step S4: performing simulation calculations on the wave impedance value, the lithology category probability value array, and the alteration type probability value array to obtain a mineralization potential index;

[0128] The wave impedance value, lithology category probability value array, and alteration type probability value array are used as input variables for mineralization potential analysis. Their corresponding values ​​are extracted in voxel units and normal distribution functions are fitted respectively.

[0129] Construct a mineralization potential estimation function with wave impedance, lithology probability and alteration probability as input;

[0130] For each voxel unit, N random samples are taken from the normal distribution of the three types of input variables, and the sampling results are substituted into the mineralization potential function to obtain N mineralization potential values;

[0131] The N-times mineralization potential values ​​of each voxel unit are statistically analyzed, and the expected value is calculated as the mineralization potential index of the voxel unit.

[0132] Gaussian distribution models are constructed for the wave impedance value, lithology category probability value array, and alteration type probability value array respectively:

[0133]

[0134] Where: σ is the standard deviation, μ is the mean, both parameters can be obtained through fitting;

[0135] The comprehensive potential function is defined as:

[0136] Φ(Z v ,R v ,A v )=ω1·Z v +ω2·R v +ω3·A v ;

[0137] Where: ω1, ω2, ω3 are weighted coefficients. The weighted coefficients can be assigned based on actual geological experience, and ω1+ω2+ω3=1;

[0138] Set the mineralization potential index as MPI v , mineralization potential index is MPI v The formula for obtaining is:

[0139]

[0140] Where: N represents the number of simulations, are the kth group of samples randomly sampled, k=1,2,...,N,MPI k It represents the mineralization potential value obtained for the kth group of samples;

[0141] For example, the distribution parameters of a certain voxel, such as the wave impedance value, obey: Π(2.5,0.3 2 ), the lithology probability obeys Π(2.5,0.3 2 ), the alteration probability obeys Π(2.5,0.3 2 ), then for each k simulation sampling, the triple Each is sampled once from the above distribution and substituted into the comprehensive potential function Φ.

[0142] Step S5: Fusing the mineralization potential index with the five-dimensional data cube to construct a four-dimensional dynamic database, and generating a mineralization potential prediction layer and a drilling priority map based on the four-dimensional dynamic database.

[0143] The methods of the four-dimensional dynamic database include:

[0144] Extract the spatial coordinates and corresponding time index of each voxel unit in the five-dimensional data cube to construct a joint index key of space + time;

[0145] The mineralization potential index of each voxel unit is associated and matched with the corresponding space + time joint index key to generate a four-dimensional dynamic data record;

[0146] All four-dimensional dynamic data records are uniformly organized and stored to form a four-dimensional dynamic database.

[0147] The method for generating mineralization potential prediction layers and drilling priority maps based on a four-dimensional dynamic database includes:

[0148] Index the four-dimensional dynamic database, extract the mineralization potential index of each voxel unit at different time indexes, and generate a mineralization potential prediction layer;

[0149] Perform regional cluster analysis on the mineralization potential prediction layer, extract potential value clustering areas and calculate their clustering strength indicators;

[0150] Based on the spatial coupling analysis of the mineralization potential prediction layer and the alteration zoning layer, the spatial synergy score of each area is calculated;

[0151] Comprehensively analyze the mineralization potential index, clustering intensity index, and synergy score to build a drilling priority assessment model;

[0152] Based on the drilling priority evaluation model, the target exploration area is spatially optimized and classified into drilling priority areas, suboptimal areas and low priority areas, and a drilling priority map is output.

[0153] The spatial + temporal joint index key is represented as (x, y, z, t), and each dynamic data record is represented as The four-dimensional dynamic database representation is:

[0154]

[0155] At a fixed time t, the mineralization potential index of all spatial voxels at that time point is extracted to form a mineralization potential prediction layer. The mineralization potential prediction layer is expressed as:

[0156] The DBSCAN spatial density clustering algorithm is used to cluster the potential areas in the mineralization potential layer, and the cluster area is defined as C h , define the cluster area as C h Clustering strength index CI h , clustering intensity index CI h The way to obtain is:

[0157]

[0158] Where CI h represents the clustering strength index of the h-th clustering area, C h It is represented as a cluster area numbered h, which contains several voxel units, b v Indicates the vth voxel unit to the cluster area C h The spatial distance between the centers, ε is a small constant to prevent division by zero;

[0159] It needs to be explained that CI h The higher the value of , the more concentrated and densely distributed the high-potential voxels are in the cluster area;

[0160] It needs to be explained that b v It can be expressed using Euclidean distance;

[0161] Alteration zoning layer A (x,y,z) Perform spatial overlap analysis with the mineralization potential layer MPI (x, y, z) and define the synergistic scoring function as:

[0162]

[0163] In the formula: Indicates whether the voxel unit is located in a specific alteration type layer, where 1 indicates existence and 0 indicates non-existence; MPI v (x, y, z) represents the mineralization potential value at the corresponding position, and S(x, y, z) represents the spatial synergy score of a voxel unit, S ∈ [0, 1];

[0164] It should be noted that the higher the s value, the stronger the spatial synergy;

[0165] Construct a voxel comprehensive priority scoring function by integrating the three indicators:

[0166] DO(x, y, z) = α1·MPI(x, y, z) + α2·CI h + α3·S(x, y, z);

[0167] In the formula: α1 + α2 + α3 = 1, α1, α2, α3 are weighting coefficients, and the weighting coefficients can be assigned based on actual geological experience;

[0168] Set the upper threshold L1 and the lower threshold L2, compare the comprehensive score value DO(x, y, z) with the set upper and lower thresholds (L1, L2), and perform priority division according to the following rules:

[0169] When the comprehensive score value DO(x, y, z) > the upper threshold L1, the corresponding voxel unit is divided into the drilling priority area;

[0170] When the upper threshold L1 < DO(x, y, z) < the lower threshold L2, it is divided into the sub-optimal area;

[0171] When the comprehensive score value DO(x, y, z) < the lower threshold L2, it is divided into the low priority area;

[0172] After completing the priority division of all voxel units, sort all the comprehensive score values DO(x, y, z) in descending order, and map their corresponding spatial positions to the three-dimensional voxel grid to generate a drilling priority map reflecting the drilling priorities of different regions.

[0173] In this embodiment, a multi-scale observation system covering "air-ground-hole" is constructed by collaboratively deploying surface drones, ground seismometers, and underground borehole measurement equipment. Surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data can be simultaneously acquired. The unified WGS84 spatial coordinate system and embedded GNSS positioning system enable high-precision spatial alignment of multiple types of equipment across air, ground, and hole, avoiding spatial drift during the integration of heterogeneous data and providing a consistent spatial index system for the subsequent construction of a five-dimensional data cube. All types of equipment are connected to a network time synchronization mechanism. By embedding UTC timestamps, multi-source data have a unified time reference coordinate, providing timeliness for subsequent data processing. Standardized sampling parameters for various acquisition devices effectively address the fusion barriers caused by differences in resolution, coordinate system, and acquisition method in traditional exploration, ensuring comparability of data at the same scale, and laying the foundation for multi-source fusion modeling. By constructing a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, elemental parameters, and time index, it is possible to fully express various geological conditions in the target area. The evolution of information in time and space solves the one-sidedness problem of traditional single-source data analysis and improves the systematicness and spatial consistency of data fusion. The 3D-CNN model is used at the physical layer to extract fusion attributes such as wave impedance and lithology probability value array. The ResNet50 model is used at the semantic layer to identify the alteration type, which can effectively improve the expression ability of underground medium structure and alteration spatial distribution. By fusing the influencing factors related to wave impedance, lithology probability and alteration type probability to construct a potential function, a large sample deduction is performed on the fusion wave impedance, lithology probability value array and alteration type probability value array, so that The obtained mineralization potential index is more credible and discriminative; by introducing the time dimension on the basis of the traditional three-dimensional voxel structure, the potential index is mapped to the voxel grid, and a four-dimensional dynamic database with time series attributes is constructed, which can not only support single-period spatial analysis, but also perform dynamic modeling of the mineralization process. Spatial statistical indicators such as clustering intensity and spatial collaborative scoring are introduced, and combined with the mineralization potential, a comprehensive scoring function with adjustable weights is constructed to achieve refined classification of drilling priority areas, suboptimal areas and low priority areas, providing a data-driven priority area division plan for engineering deployment, and improving the scientific nature of drilling layout and resource allocation efficiency.

[0174] Example 2

[0175] See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of Example 1. A geological and mineral exploration data extraction system is provided, including:

[0176] Acquisition module: used to collect surface mineral spectrum data, seismic P-wave and S-wave reflection data, and core major and trace element composition data through the air-ground-hole trinity acquisition network, and perform spatial registration of multi-source data;

[0177] Data fusion module: used to construct the multi-source data after spatial registration into a five-dimensional data cube containing spatial, spectral, physical property, element and time indexes;

[0178] Feature extraction module: used to extract attributes at the physical and semantic layers of the five-dimensional data cube, and obtain wave impedance values, lithology category probability values, alteration type probability values, and alteration zoning layers respectively;

[0179] Decision analysis module: Based on the wave impedance value, lithology category probability value and alteration type probability value as input, the mineralization potential index is obtained through simulation calculation;

[0180] Result output module: Build a four-dimensional dynamic database and generate mineralization potential prediction layers and drilling priority maps based on the database.

[0181] In this embodiment, the data acquisition module builds a three-in-one acquisition network and pre-sets the time and space of the acquisition equipment, so that multi-source data can be effectively integrated in terms of resolution, coordinate system and acquisition method, ensuring the comparability of the same scale between various types of data and reducing the secondary processing of various types of collected data. The feature fusion module solves the one-sidedness problem of traditional single-source data analysis by constructing a five-dimensional data cube containing spatial coordinates, spectral parameters, physical parameters, element parameters and time index, and improves the systematicity and spatial consistency of data fusion, effectively improving the expression ability of underground medium structure and alteration spatial distribution. The feature extraction module integrates the semantic layer and physical layer of the five-dimensional data cube. The feature extraction of the layer, the acquisition of wave impedance value, lithology category probability value and alteration type probability value provide a data basis for the subsequent analysis of mineralization potential. The decision analysis module conducts large-sample deduction of the fused wave impedance, lithology probability and alteration type probability, making the mineralization potential index more credible and discriminative. The result output module introduces the time dimension on the basis of the traditional three-dimensional voxel structure to construct a four-dimensional dynamic database with time series attributes, which can not only support single-period spatial analysis, but also perform dynamic modeling of the mineralization process. It introduces spatial statistical indicators such as clustering intensity and spatial collaborative scoring, and combines the mineralization potential to construct a comprehensive scoring function with adjustable weights to achieve refined classification of drilling priority areas, suboptimal areas and low priority areas.

[0182] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0183] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only one type. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0184] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

[0185] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for extracting data from geological and mineral exploration, characterized in that: The method for extracting data from geological and mineral exploration comprises: Step S1: Construct an air-ground-hole trinity acquisition network to collect surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data; Step S2: Processing the collected data to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index; Step S3: Extract physical layer attributes from the five-dimensional data cube, analyze and obtain wave impedance values ​​and lithology category probability arrays. Simultaneously, extract semantic layer attributes from the five-dimensional data cube, analyze and obtain alteration type probability arrays and generate alteration zoning layers. Step S4: performing simulation calculations on the wave impedance value, the lithology category probability value array, and the alteration type probability value array to obtain a mineralization potential index; Step S5: Fusing the mineralization potential index with the five-dimensional data cube to construct a four-dimensional dynamic database, and generating a mineralization potential prediction layer and a drilling priority map based on the four-dimensional dynamic database.

2. A method for extracting data from geological and mineral exploration according to claim 1, characterized in that: The method for constructing an air-ground-hole trinity acquisition network comprises: Predetermine the target exploration area and deploy various data acquisition equipment there. This includes deploying an array of drones over the target exploration area. The flight paths of the drone arrays are pre-designed using the principle of "perpendicular to the structural trend + grid coverage." Wireless geophone arrays are laid out on the ground in the target exploration area using the principle of "cross-shaped main axis + concentric circle assistance." Drilling points and drilling measurement equipment are deployed in the target exploration area. Standardize sampling parameters for drones, geophones, and borehole survey equipment, including spatial resolution, sampling density, and sampling intervals; Unified configuration of WGS84 spatial coordinate system for multiple types of acquisition equipment, and real-time positioning of spatial coordinates through satellite navigation and positioning services; Connect various acquisition devices to the network time synchronization mechanism, use the network time protocol to obtain UTC timestamps, and embed timestamp calibration programs in various acquisition devices to automatically correct time offsets.

3. The method for extracting data from geological and mineral exploration according to claim 1, characterized in that: The method of separately collecting surface mineral spectrum data, seismic longitudinal and shear wave reflection data, and core major and trace element composition data includes: Use drone arrays to collect hyperspectral images of the area covered by the pre-designed flight path, and analyze the hyperspectral images to obtain surface mineral spectral data; Acquire seismic longitudinal and shear wave reflection data through wireless seismic geophone arrays; Drilling is carried out at the borehole point using drilling measurement equipment to collect continuous core samples, which are then analyzed to obtain the major and trace element composition data of the core; The acquired surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are embedded with spatial coordinates and UTC timestamps respectively.

4. A method for extracting data from geological and mineral exploration according to claim 3, characterized in that: The method of processing the collected data to form a five-dimensional data cube including spatial coordinates, spectral parameters, physical property parameters, element parameters and time index includes: A regular three-dimensional voxel grid is established in the target exploration area based on the WGS84 spatial coordinate system, and the center point coordinates of each voxel unit are used as its unique spatial index in the three-dimensional space to form a spatial dimension; Based on spatial coordinate difference mapping, surface mineral spectral data, seismic P-wave and S-wave reflection data, and core major and trace element composition data are assigned to corresponding voxel units, constructing an attribute dimension that includes spectral parameters, physical parameters, and elemental parameters. Assign a time index to each voxel unit based on the UTC timestamp to form a time dimension; The spatial dimension, attribute dimension and time dimension of each voxel unit are integrated to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index.

5. The method for extracting data from geological and mineral exploration according to claim 1, characterized in that: The method of extracting physical layer attributes from a five-dimensional data cube and analyzing and obtaining wave impedance values ​​and lithology category probability arrays includes: Build a 3D-CNN model based on voxel feature learning, which includes an input layer, multi-layer 3D convolutional layers, batch normalization layers, activation layers, pooling layers, and fully connected layers; The five-dimensional data cube is fed into the 3D-CNN model as input, and the voxel unit is used as the processing unit to extract its deep fusion properties. Perform nonlinear transformation and spatial dimensionality reduction on the deep fusion attributes of each voxel unit and output its corresponding wave impedance value; The deep fusion attributes of each voxel unit are input into the fully connected layer and passed through the Softmax function to output the lithology category probability array corresponding to each voxel unit.

6. A method for extracting data from geological and mineral exploration according to claim 1, characterized in that: The method of extracting semantic layer attributes from a five-dimensional data cube, analyzing to obtain an alteration type probability array, and generating an alteration zoning layer comprises: Taking the five-dimensional data cube as the input data source, the spectral parameters, physical parameters, element parameters and time index of each voxel unit are extracted to form a fusion attribute vector; Slice the five-dimensional data cube based on any two dimensions in the three-dimensional space coordinate axis, extract a representative two-dimensional slice sequence, and retain the fusion attribute vector corresponding to each voxel unit in the slice; The retained fusion attribute vector is divided into channels according to parameter categories and re-encoded into a multi-channel image, where each channel corresponds to spectral features, physical properties, and elemental features to match the ResNet50 model input format; Multi-channel images are input to the ResNet50 model, which performs deep semantic segmentation and classification prediction on the two-dimensional slices and outputs an array of alteration type probabilities for each voxel unit; The alteration type probability array of each voxel unit is mapped to a three-dimensional voxel grid according to spatial coordinates to form an alteration zoning layer.

7. A method for extracting data from geological and mineral exploration according to claim 6, characterized in that: The method of simulating and calculating the wave impedance value, the lithology category probability value array, and the alteration type probability value array to obtain the mineralization potential index includes: The wave impedance value, lithology category probability value array, and alteration type probability value array are used as input variables for mineralization potential analysis. Their corresponding values ​​are extracted in voxel units and normal distribution functions are fitted respectively. Construct a mineralization potential estimation function with wave impedance, lithology probability and alteration probability as input; For each voxel unit, N random samples are taken from the normal distribution of the three types of input variables, and the sampling results are substituted into the mineralization potential function to obtain N mineralization potential values; The N-times mineralization potential values ​​of each voxel unit are statistically analyzed, and the expected value is calculated as the mineralization potential index of the voxel unit.

8. A method for extracting data from geological and mineral exploration according to claim 1, characterized in that: The method for constructing a four-dimensional dynamic database comprises: Extract the spatial coordinates and corresponding time index of each voxel unit in the five-dimensional data cube to construct a joint index key of space + time; The mineralization potential index of each voxel unit is associated and matched with the corresponding space + time joint index key to generate a four-dimensional dynamic data record; All four-dimensional dynamic data records are uniformly organized and stored to form a four-dimensional dynamic database.

9. A method for extracting data from geological and mineral exploration according to claim 8, characterized in that: The method for generating a mineralization potential prediction layer and a drilling priority map based on a four-dimensional dynamic database includes: Index the four-dimensional dynamic database, extract the mineralization potential index of each voxel unit at different time indexes, and generate a mineralization potential prediction layer; Perform regional cluster analysis on the mineralization potential prediction layer, extract potential value clustering areas and calculate their clustering strength indicators; Based on the spatial coupling analysis of the mineralization potential prediction layer and the alteration zoning layer, the spatial synergy score of each area is calculated; Comprehensively analyze the mineralization potential index, clustering intensity index, and synergy score to build a drilling priority assessment model; Based on the drilling priority evaluation model, the target exploration area is spatially optimized and classified into drilling priority areas, suboptimal areas and low priority areas, and a drilling priority map is output.

10. A geological and mineral exploration data extraction system, which is used to implement a geological and mineral exploration data extraction method according to any one of claims 1 to 9, characterized in that: include: Acquisition module: used to build an air-ground-hole trinity acquisition network to collect surface mineral spectrum data, seismic longitudinal and shear wave reflection data, and core major and trace element composition data; Data fusion module: used to process the collected data to form a five-dimensional data cube containing spatial coordinates, spectral parameters, physical property parameters, element parameters and time index; Feature extraction module: used to extract attributes at the physical and semantic layers of the five-dimensional data cube, and obtain wave impedance values, lithology category probability arrays, alteration type probability arrays, and alteration zoning layers respectively; Decision analysis module: simulates and calculates wave impedance values, lithology probability arrays, and alteration type probability arrays to obtain the mineralization potential index; Result output module: used to fuse the mineralization potential index with the five-dimensional data cube, build a four-dimensional dynamic database, and generate mineralization potential prediction layers and drilling priority maps based on the four-dimensional dynamic database.

Images