A method and system for predicting prospecting target areas based on altered mineral analysis

Abstract

This invention discloses a method and system for predicting mineral exploration target areas based on alteration mineral analysis, belonging to the field of mineral exploration target area prediction technology. A mineral exploration target area prediction system based on alteration mineral analysis includes: a data acquisition module, a cluster analysis module, an alteration assemblage discrimination module, a spatial modeling module, a spatiotemporal coupling module, and a mineralization assessment module. This invention introduces mineral thermodynamic phase diagram constraints to perform hierarchical clustering of alteration minerals and their associated assemblages, effectively analyzing the alteration information superimposed in multiple phases in complex tectonic zones, and establishing a mineral associated assemblage structure model with the ability to express genetic differences. By constructing genetic phase labels and forming a temporal decoupling model, it achieves systematic differentiation of alteration forms formed under mineralization processes of different ages, providing a temporal basis for the identification of multi-stage mineralization processes.

Description

Technical Field

[0001] This invention relates to the field of mineral exploration target area prediction technology, and in particular to a method and system for mineral exploration target area prediction based on alteration mineral analysis. Background Technology

[0002] Alteration minerals are important alteration products of geological bodies under the influence of ore-forming fluids. Their types, assemblage relationships, and spatial distribution characteristics are of great significance in revealing mineralization processes and delineating prospecting targets. Current technologies for alteration mineral identification mainly rely on spectral feature extraction from remote sensing images, alteration zone delineation from geological mapping, or elemental enrichment indices from geochemical anomalies. However, the effectiveness of these methods is significantly limited in areas with frequent tectonic activity and overlapping multiple mineralization processes.

[0003] In alteration assemblages modeling, existing techniques mostly rely on thin rock sections, qualitative descriptions, or single spectral inversion methods, which struggle to systematically reflect the thermodynamic stability and symbiotic characteristics of altered minerals and lack quantitative means to express the genetic background of mineral assemblages. Regarding temporal sequence discrimination, current research has failed to establish phase identification rules associated with thermodynamic conditions, presenting significant difficulties in identifying and grading multi-stage alteration forms. In mineralization prediction methods, conventional models typically depend on elemental anomaly thresholds, artificial weighting, or simple statistical classification methods, failing to integrate uncertainties in geological evolution and lacking a unified genetic-driven discrimination logic. In terms of spatial representation, traditional methods often use two-dimensional anomaly maps or local profile interpolation to depict alteration form distribution, without incorporating occurrence information from tectonic measurement data as modeling constraints, resulting in limited three-dimensional spatial modeling capabilities.

[0004] Existing methods cannot distinguish information from multiple superimposed phases, lack spatiotemporal coupling modeling capabilities, rely on empirical judgment for mineralization identification, and have difficulty in quantifying the reliability of target area predictions. These problems make it difficult to meet the actual needs of refined identification and prediction of mineralization systems in deep mineral exploration and complex tectonic areas. Summary of the Invention

[0005] A method for predicting mineral exploration target areas based on alteration mineral analysis includes:

[0006] Collect multi-source geological data of the target area, including tectonic measurement data, remote sensing image data, geochemical data, alteration minerals and their associated assemblages;

[0007] Based on the constraints of mineral thermodynamic phase diagrams, hierarchical clustering analysis is performed on altered minerals and their associated assemblages, as well as mineral spectral features extracted from remote sensing image data, to generate mineral associated assemblages feature maps, assign genetic period labels, and construct a time-series decoupling model.

[0008] The temporal decoupling model is used as a prior knowledge base and input together with the elemental anomaly indicators in the geochemical data into the Bayesian network-based mineralization alteration combination probability discrimination model. The matching probability between the target alteration combination and the typical mineralization alteration mode is calculated, and the alteration identification results with matching probabilities higher than the preset threshold are output.

[0009] The spatial distribution trend features of alteration identification results are extracted, and the fault zone attitude information extracted from the construction measurement data is used as a spatial constraint to construct the input dataset; an implicit function field representing the boundary of alteration is generated through an implicit geological modeling algorithm, and a three-dimensional distribution boundary model of alteration is extracted.

[0010] A spatiotemporal weight matrix is ​​established, and the genetic period label, matching probability, and three-dimensional distribution boundary model of alteration are coupled to generate a spatiotemporal evolution probability field of alteration, which serves as the basic data input for mineralization confidence calculation and target area prediction.

[0011] Based on the random forest algorithm, the mineralization confidence of the target area is calculated by integrating geological, geophysical, thermodynamic, remote sensing and geochemical indicators, and the target area prediction results and graded scores are output.

[0012] As a preferred embodiment of the present invention, the hierarchical clustering analysis includes:

[0013] Based on the thermodynamic phase diagram stability domain characteristics of altered minerals and their symbiotic assemblage data, altered minerals are initially grouped. The mineral spectral features extracted from remote sensing image data are fused with the altered mineral grouping results using preset weighting coefficients. The fused data is then input into a clustering model for hierarchical clustering. Based on the grouping results output by the clustering model, typical mineral assemblages and their symbiotic relationships in each group are extracted to construct the corresponding mineral symbiotic assemblage feature map.

[0014] As a preferred embodiment of the present invention, the step of fusing the mineral spectral features extracted from remote sensing image data with the alteration mineral grouping results through a preset weighting coefficient includes:

[0015] Multiple characteristic reflection bands and spectral morphology parameters of altered minerals are extracted from remote sensing image data. The spectral features are then vectorized and encoded, and combined with altered minerals and their co-occurring data to form an input dataset. This input dataset is then used as input to a clustering model to perform hierarchical clustering.

[0016] As a preferred embodiment of the present invention, the construction of the causal period label includes:

[0017] Based on the characteristic map of mineral assemblages and combined with the stability domain of the thermodynamic phase diagram of each group of altered minerals, a period discrimination rule is established; and according to the rule, a unique genetic period label is assigned to each group of altered mineral assemblages to characterize the temporal attribute of the alteration corresponding to the assemblages.

[0018] As a preferred technical solution of the present invention, the mineralization alteration combination probability discrimination model is a Bayesian network model constructed through typical mineralization alteration combination samples. The model uses the genetic period labels generated in the time-series decoupling model as prior input nodes, and jointly constructs the input structure with the elemental anomaly indicators in the geochemical data. The matching probability between the target alteration combination and the typical mineralization alteration mode is used as the output node, and the discrimination results are filtered based on a preset probability threshold in the output stage, and the alteration identification results with matching probabilities higher than the threshold are output.

[0019] As a preferred embodiment of the present invention, the extraction of the three-dimensional distribution boundary model of the alteration form includes:

[0020] The spatial location and distribution trend characteristics of alteration variants are obtained, and the fault zone attitude information extracted from the structural measurement data is used as a spatial constraint to construct a modeling dataset containing a spatial point set, structural control lines, and distribution trend vectors. Based on the modeling dataset, an implicit geological modeling algorithm is used to generate an implicit function field, and the three-dimensional distribution boundary model of the alteration variants is obtained by extracting the function isosurface.

[0021] As a preferred embodiment of the present invention, the generation of the spatiotemporal evolution probability field of the alteration form includes:

[0022] Using genetic stage labels, the matching probability of mineralization alteration assemblages, and the three-dimensional distribution boundary model of alteration forms as input variables, a spatiotemporal weight matrix is ​​constructed. Among them, the temporal compatibility weight is allocated according to the chronological order of the genetic stage labels, with the weight decreasing in the early stage alteration assemblages; the spatial continuity weight is calculated based on the spatial density gradient in the three-dimensional distribution boundary model of alteration forms, with the weight increasing in high-density areas; and the mineralization matching probability weight is directly adopted from the matching probability value output by the Bayesian network.

[0023] A gridded interpolation algorithm is used to fit the spatiotemporal weight matrix in three dimensions. Within the spatial range defined by the three-dimensional distribution boundary model of the alteration morphology, a three-dimensional probability distribution model that simultaneously includes time attributes, spatial location, and mineralization probability values ​​is generated as the spatiotemporal evolution probability field of the alteration morphology.

[0024] As a preferred embodiment of the present invention, the acquisition of the mineralization confidence level includes:

[0025] The temporal distribution factor, spatial density index, and mineralization matching probability extracted from the spatiotemporal evolution probability field of alteration forms are used as core feature parameters. Combined with remote sensing image data, geochemical data, and geological structure information, a feature vector set is constructed. A random forest model is trained based on this feature vector set and applied to each grid cell in the target area. The mineralization confidence value of each cell is output as the basis for the probability score of target area prediction.

[0026] As a preferred embodiment of the present invention, the output target region prediction result includes:

[0027] The spatial distribution results of mineralization confidence are obtained, and multiple grid units within the target area are classified according to the confidence threshold setting rules to generate multiple prospecting target area candidate units. Adjacent units of the same confidence level are spatially aggregated to output a prospecting target area prediction layer with clear boundary range and level label.

[0028] A mineral exploration target area prediction system based on alteration mineral analysis includes:

[0029] The data acquisition module is used to collect multi-source geological data of the target area;

[0030] The clustering analysis module is used to perform hierarchical clustering analysis on altered minerals and their associated data, as well as mineral spectral features extracted from remote sensing image data, and to construct a time-series decoupling model.

[0031] The alteration combination discrimination module calculates the matching probability between the target alteration combination and the typical mineralized alteration mode based on the mineralized alteration combination probability discrimination model, and outputs the identification results of alteration types with matching probabilities higher than a preset threshold.

[0032] The spatial modeling module is used to generate implicit function fields representing alteration boundary through implicit geological modeling algorithms, and to extract the three-dimensional distribution boundary model of alteration.

[0033] The spatiotemporal coupling module is used to couple the causal period labels, matching probabilities, and the three-dimensional distribution boundary model of the alteration variants to generate the spatiotemporal evolution probability field of the alteration variants.

[0034] The mineralization assessment module is used to calculate the mineralization confidence level of the target area and output the target area prediction results and graded scores.

[0035] The present invention has the following advantages:

[0036] This invention employs a mineral thermodynamic phase diagram constraint to perform hierarchical clustering of altered minerals and their associated assemblages, effectively analyzing alteration information from multiple superimposed phases in complex tectonic zones and establishing a mineral assemblage structure model capable of expressing genetic differences. By constructing genetic phase labels and forming a temporal decoupling model, it enables the systematic differentiation of alteration forms created under mineralization processes of different eras, providing a temporal basis for identifying multi-stage mineralization processes.

[0037] This invention establishes a probability discrimination mechanism for mineralization alteration combinations by inputting genetic period labels and geochemical anomaly indicators into a Bayesian network model, which significantly improves the scientificity and accuracy of mineralization orientation identification of alteration types. By extracting the spatial distribution trend characteristics of alteration types and combining them with tectonic information such as fault zone occurrence as spatial constraints, an implicit geological modeling algorithm is used to reconstruct the three-dimensional boundaries of alteration types, thereby improving the continuity and structural rationality of alteration type spatial modeling.

[0038] This invention generates a spatiotemporal evolution probability field of alteration forms by constructing a spatiotemporal weight matrix that includes temporal, spatial, and mineralization probabilities, providing quantitative data support for subsequent target area prediction under a dynamic evolution background. By integrating features from multiple sources such as remote sensing, geochemistry, and geological structure, it calculates mineralization confidence ratings, enhancing the discriminative ability and credibility expression of mineral exploration results. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only schematic diagrams of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of a mineral exploration target area prediction system based on alteration mineral analysis used in an embodiment of the present invention.

[0041] Figure 2 This is a distribution map of alteration mineral assemblages used in the embodiments of the present invention;

[0042] Figure 3 This is a diagram of the mineralization-related alteration combination zoning pattern used in the embodiments of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0044] Example 1: A method for predicting mineral exploration target areas based on alteration mineral analysis, combined with... Figure 2 and Figure 3 This includes the following steps:

[0045] Step S1: Collect multi-source geological data of the target area, including tectonic measurement data, remote sensing image data, geochemical data, alteration minerals and their associated data, to build the data foundation for subsequent alteration identification and mineralization prediction.

[0046] Structural measurement data refers to the regional fault structure occurrence parameters extracted through field geological surveys or existing geological data, including fault strike, dip, dip angle, displacement characteristics, and fault control line distribution information, which are used as spatial constraint inputs in the subsequent implicit geological modeling process.

[0047] The remote sensing image data is medium-to-high resolution multispectral remote sensing data covering the target area, including visible light, shortwave infrared and thermal infrared bands, used to extract the spectral reflectance characteristics of altered minerals, such as characteristic absorption valleys, reflectance peak positions, spectral slopes, etc.

[0048] Geochemical data includes multi-element test data from rock, soil, water or sediment samples obtained through field sampling, with elements including at least Cu, Pb, Zn, Mo, As, Au, Ag, Fe, Si, Al, K, Mg, etc.; used for subsequent identification of anomalous enrichment areas of elements and participating in probabilistic discrimination modeling of mineralization alteration assemblages.

[0049] Alteration mineral data refers to the types of alteration minerals and their quantitative content determined based on field sampling or mineral spectral analysis, including typical alteration minerals such as chlorite, sericite, muscovite, kaolinite, hematite, and pyrite.

[0050] Co-occurrence assemblage data refers to a data structure based on the co-occurrence relationships of different alteration minerals in a single mineral sample. This data is obtained by statistically analyzing the combination patterns of alteration minerals at different sampling points and is used to subsequently construct a characteristic map of alteration mineral co-occurrence assemblages. Co-occurrence assemblages include not only the names of the minerals that appear simultaneously, but also their order of magnitude (such as the percentage relationship between the main and accessory minerals) and frequency of occurrence.

[0051] The above multi-source geological data were all obtained through the following methods:

[0052] The structural measurement data mainly comes from existing regional geological survey reports and field structural measurements;

[0053] Remote sensing image data is acquired through publicly available satellite platforms or purchased from commercial data service providers;

[0054] Geochemical data were obtained from field sampling and then tested using XRF (X-ray fluorescence spectroscopy), ICP-MS (inductively coupled plasma mass spectrometry), and other methods.

[0055] Alteration mineral data were acquired using XRD (X-ray diffraction), SWIR (shortwave infrared spectroscopy), and VNIR (visible-near infrared spectroscopy).

[0056] The symbiotic assemblage data are based on the statistics of the aforementioned mineral identification samples and processed into a database.

[0057] Step S2: Based on the constraints of the mineral thermodynamic phase diagram, perform hierarchical cluster analysis on the data of altered minerals and their symbiotic assemblages, as well as the mineral spectral features extracted from remote sensing image data, generate a mineral symbiotic assemblage feature map, assign genetic period labels, and construct a time-series decoupling model;

[0058] Among them, mineral thermodynamic phase diagrams are used to express the stability boundaries of minerals under different temperature and pressure conditions. In this embodiment, phase diagram data generated by a rock thermodynamic database (such as THERMOCALC) is used to analyze the stable occurrence range of altered minerals in the target area under different geological conditions, serving as genetic constraints in subsequent cluster analysis.

[0059] The hierarchical cluster analysis includes:

[0060] Alteration minerals are initially grouped based on the stability domain characteristics of thermodynamic phase diagrams of alteration minerals and their associated assemblages. This process involves screening minerals from all sampling points for thermodynamic stability and classifying them into several initial mineral groups based on their stable overlapping regions on the PT (pressure-temperature) diagram.

[0061] The mineral spectral features extracted from remote sensing image data are fused with the alteration mineral grouping results through a preset weighting coefficient. The preset weighting coefficient is determined based on the spectral-mineral consistency in the sample data. A typical setting method is to introduce a multimodal weighting factor into the distance function of the clustering model to fuse the contribution intensity of remote sensing features and ground mineral distribution.

[0062] The fused data is then input into a clustering model for hierarchical clustering. The clustering model can employ hierarchical clustering, density peak clustering, or spectral clustering algorithms. Its core principle is to classify alteration assemblages layer by layer based on the similarity and stability information between minerals. The purpose of this layering is to distinguish between major mineral assemblages, associated assemblages, and boundary assemblages, thereby improving the ability to identify weak mineralization and alteration signals.

[0063] Based on the grouping results output by the clustering model, typical mineral assemblages and their co-occurrence relationships in each group are extracted, and corresponding mineral co-occurrence feature maps are constructed. These maps are structured descriptions of mineral types, abundance proportions, and spatial co-occurrence frequencies within each assemblage, and are key inputs for the subsequent construction of genetic epoch models.

[0064] In the process of implementing hierarchical clustering in typical sample areas, such as Figure 2 As shown, different types of alteration minerals and their associated assemblages exhibit distinct spatial distribution characteristics and zonation patterns. This map serves as a crucial input for subsequent genetic determination and temporal decoupling modeling.

[0065] Multiple characteristic reflection bands and spectral morphology parameters of altered minerals were extracted from remote sensing image data. Typical bands include 2200nm (mica), 2330nm (chlorite), and 1900nm (hydroxyl water molecules). Spectral morphology parameters include slope, envelope angle, and symmetry factor.

[0066] The spectral features are vectorized and encoded (e.g., PCA dimensionality reduction + normalization), then combined with alteration minerals and their associated data to form an input dataset. This input dataset is then used as input to a clustering model for hierarchical clustering. The fused dataset expresses multi-source indicators in a unified feature space, enabling the model to simultaneously utilize the differences between mineral assemblage structure and spectral expression for clustering decisions.

[0067] The construction of the causal period labels includes:

[0068] Based on the characteristic map of mineral symbiotic assemblages, and combined with the stable domain intervals of the thermodynamic phase diagrams of each group of altered minerals, a period discrimination rule is established. The discrimination rule is generated by constraining the intersection of stable regions in PT space for each assemblage, the degree of overlap with known mineral assemblages of mineralization ages, and stable evolution paths (e.g., the evolutionary sequence from early sericite to late kaolinite).

[0069] According to the rules described above, a unique genetic period label is assigned to each alteration mineral assemblage to characterize the temporal attribute of the alteration process corresponding to that assemblage. The label uses discrete-time encoding (such as period numbers T1, T2, T3) and establishes a correspondence with known mineralization tectonic events in conjunction with the regional geochronological background, serving as the temporal prior input for subsequent probabilistic discrimination models.

[0070] The constructed temporal decoupling model is a ternary expression system of alteration combination, thermodynamic interval, and time attribute. It distinguishes alteration combinations with different causal backgrounds under the background of multiple superposition, providing a clear time dimension stratification basis for subsequent Bayesian discrimination.

[0071] Step S3: Use the time-series decoupling model as a prior knowledge base, and input it together with the elemental anomaly indicators in the geochemical data into the Bayesian network-based mineralization alteration combination probability discrimination model. Calculate the matching probability between the target alteration combination and the typical mineralization alteration mode, and output the alteration identification results where the matching probability is higher than the preset threshold.

[0072] The prior knowledge base refers to the temporal decoupling model constructed in step S2, whose core data includes: genetic stage labels for each mineral assemblage, structural characteristics of the co-existing assemblage, and thermodynamic stability background. The genetic stage labels possess both temporal and structural expressive capabilities, significantly enhancing the model's ability to classify multi-stage superimposed alteration as prior input.

[0073] The Bayesian network model is based on Figure 3 The typical alteration combination mineralization model shown is based on a priori structural foundation. This model was constructed using field mineralization samples and clarified the probabilistic relationships of various combinations under different time series and mineralization environments.

[0074] Elemental anomalies in geochemical data include typical mineralization-related elements such as Cu, Pb, Zn, Au, As, Sb, Hg, and Mo, which are identified in geochemical survey lines or sample points using background-anomaly discrimination methods (such as the 3σ method). Each sampling point can form a geochemical anomaly vector, representing the geochemical characteristics of its mineralization potential.

[0075] The probabilistic discrimination model for mineralization and alteration assemblages is a Bayesian network model constructed using typical mineralization and alteration assemblages. The model uses the genetic period labels generated in the time-series decoupling model as prior input nodes, and jointly constructs the input structure with elemental anomaly indicators in the geochemical data. The input structure is a network of nodes with conditional dependencies, where the prior nodes are discrete (periods T1 / T2), and the geochemical indicators are continuous or discretized inputs, representing the joint probability distribution structure among the data.

[0076] Bayesian networks employ a two-stage approach: structure learning and parameter learning.

[0077] Structural learning: Based on typical mineralization sample combinations (a priori tectonic zone mineralization points and their geochemical-mineral assemblage characteristics), a network topology is constructed, and greedy search (K2 algorithm) or scoring function method is used to determine the causal paths between nodes;

[0078] Parameter learning: Using maximum likelihood estimation or Bayesian estimation methods, determine the conditional probability distribution of each node under different states;

[0079] Finally, the matching probability between the target alteration combination and the typical mineralization alteration mode is used as the output node, and the output value is a continuous probability between [0,1], which is used to characterize the mineralization similarity of the current alteration combination.

[0080] In the output stage, the system filters the discrimination results based on a preset probability threshold (0.6), removing low-confidence samples that do not have obvious mineralization characteristics, and only outputs the alteration identification results with a matching probability higher than the threshold, which serve as the input basis for subsequent 3D modeling and spatiotemporal fusion. This result essentially constitutes a spatially discrete but probabilistically continuous alteration directional target distribution map. The preset probability threshold is obtained through machine learning fitting or expert experience.

[0081] Step S4: Extract the spatial distribution trend features of the alteration pattern identification results, use the fault zone attitude information extracted from the construction measurement data as spatial constraints, and construct the input dataset; generate an implicit function field representing the boundary of the alteration pattern through an implicit geological modeling algorithm, and extract the three-dimensional distribution boundary model of the alteration pattern;

[0082] The spatial distribution trend characteristics of the alteration identification results refer to the extraction of the distribution direction, distribution density gradient and change trend of alterations within the target area based on the spatial location data of high-matching probability alterations output in step S3, using kernel density estimation (KDE), directional gradient analysis (such as principal component axis fitting), and spatial trend surface regression methods.

[0083] The fault zone attitude information includes parameters such as strike, dip, and dip angle, which are usually derived from fault structure measurement data from field geological surveys, or obtained through inversion methods such as remote sensing interpretation, geophysical profiling, and DEM analysis. In this step, the geometric attitude information of the fault zone will be used to constrain the modeling boundary, making the final model closer to geological reality.

[0084] The process of constructing the input dataset includes: using the spatial location points in the erosion identification results as the initial spatial point set; extracting the distribution trend vector field to form a vector dataset reflecting the main control direction of the erosion; defining the fracture control line (extracted by constructing measurement data) as the structural constraint boundary; and constructing a three-dimensional data structure based on the above data as the input basis for three-dimensional modeling.

[0085] The implicit geological modeling algorithm refers to a method that continuously represents geological boundaries using mathematical functions. Typical methods include RBF (Radial Basis Function), SDF (Signed Distance Function), or Level-Set methods. Its core idea is to define an implicit function φ(x,y,z) in three-dimensional space, such that isosurfaces with φ=0 represent alteration boundaries.

[0086] After generating the implicit function field, the φ=0 isosurface is extracted by the numerical solution method (MarchingCubes voxel algorithm) to construct a three-dimensional alteration boundary model with geological continuity and structural constraints.

[0087] The extracted three-dimensional distribution boundary model of the alteration form includes:

[0088] The spatial location and distribution trend characteristics of alteration variants are obtained, and the fault zone attitude information extracted from the structural measurement data is used as a spatial constraint to construct a modeling dataset containing a spatial point set, structural control lines, and distribution trend vectors. Based on the modeling dataset, an implicit geological modeling algorithm is used to generate an implicit function field, and the three-dimensional distribution boundary model of the alteration variants is obtained by extracting the function isosurface.

[0089] The three-dimensional model of the alteration variants output in this step not only visualizes the spatial morphology of mineralization-related alteration variants, but also provides a spatial structural basis for the subsequent construction of a spatiotemporal evolution probability field.

[0090] Step S5: Establish a spatiotemporal weight matrix, couple the genetic period label, matching probability, and three-dimensional distribution boundary model of alteration to generate a spatiotemporal evolution probability field of alteration, which serves as the basic data input for mineralization confidence calculation and target area prediction.

[0091] The spatiotemporal evolution probability field of the generated alteration form includes:

[0092] Using genetic stage labels, the matching probability of mineralization alteration assemblages, and the three-dimensional distribution boundary model of alteration forms as input variables, a spatiotemporal weight matrix is ​​constructed, where:

[0093] Temporal compatibility weights are used to reflect the correlation strength of various alteration assemblages with mineralization at different genetic stages. A time weighting function is set according to the temporal order of the genetic stage labels, such as using an exponentially decreasing function. ,in Period number, To adjust the coefficient, so that early alteration has a smaller weight in the weighting;

[0094] Spatial continuity weights are used to quantify the degree of aggregation of alteration forms or structural integrity in a spatial structure. These weights are constructed based on the spatial density gradient extracted from a three-dimensional distribution boundary model. Specifically, the modeling region is divided into a three-dimensional voxel grid, and the weight is estimated by the number density of alteration form grids per unit volume, with denser regions assigned higher weights.

[0095] The mineralization matching probability weights are directly derived from the matching probability results output by the Bayesian network. They represent the statistical similarity and credibility between the current target alteration combination and the typical mineralization combination. Weight transformation functions, such as normalization functions or logarithmic mapping, can be set on the original probability values ​​to improve the gradient response.

[0096] The spatiotemporal weight matrix is ​​a four-dimensional data structure denoted as W(x,y,z,t), where x,y,z represent three-dimensional spatial coordinates, and t represents the causal time dimension. Each cell records the comprehensive weight value at a specific time and spatial location.

[0097] A gridded interpolation algorithm is used to fit the spatiotemporal weight matrix in three dimensions. Within the spatial range defined by the three-dimensional distribution boundary model of the alteration morphology, a three-dimensional probability distribution model that simultaneously includes time attributes, spatial location, and mineralization probability values ​​is generated as the spatiotemporal evolution probability field of the alteration morphology.

[0098] In the specific fitting process, commonly used spatial interpolation algorithms in geosciences, such as multivariate spline interpolation, inverse distance weighted interpolation (IDW), or kriging interpolation, can be selected to smoothly construct the mineralization probability distribution field and avoid local outliers from interfering with the expression of the overall evolution trend.

[0099] Step S6: Based on the random forest algorithm, calculate the mineralization confidence of the target area by integrating geological, geophysical, thermodynamic, remote sensing and geochemical indicators, and output the target area prediction results and graded scores.

[0100] The acquisition of the mineralization confidence level includes:

[0101] The temporal distribution factor, spatial density index and mineralization matching probability extracted from the spatiotemporal evolution probability field of alteration bodies are used as core feature parameters. Combined with remote sensing image data, geochemical data and geological structure information, a set of feature vectors is constructed.

[0102] Among them: the temporal distribution factor is derived from the superposition density of genetic stage labels within the target grid, reflecting the possibility that the location has undergone multiple alteration processes; the spatial density index is calculated from the voxel density generated by the three-dimensional distribution boundary model, used to characterize the enrichment degree of local alteration forms; the mineralization matching probability is extracted from the probability value output by the Bayesian model; the principal component analysis results of alteration-sensitive bands such as NDVI, Al-OH, and Fe-OH are extracted from remote sensing image features; geochemical indicators include the abundance and anomaly level of target elements (such as Cu, Mo, Pb, Zn, As, etc.) in surface samples; and geological structural features include structural complexity evaluation parameters derived from tectonic measurement data, such as fault intersection density and fold axis distribution intensity.

[0103] The completed feature vector set, with each spatial grid cell as the basic unit, forms the training sample set. Typical mining area data with supervised information is used as training labels to train a random forest model. This model performs ensemble learning across multiple decision trees to improve classification stability and prediction generalization ability.

[0104] After training, the model is applied to all grid cells in the area to be predicted, and their mineralization potential is classified and evaluated. The mineralization confidence value of each cell is output, and its value range is generally set to the interval of 0 to 1, which represents the magnitude of the mineralization probability.

[0105] The output target region prediction results include:

[0106] Obtain the spatial distribution results of mineralization confidence, classify and grade multiple grid units within the target area according to the confidence threshold setting rules, and generate multiple prospecting target area candidate units;

[0107] The confidence threshold can be preset according to the needs of different mineral exploration tasks. Commonly used classification standards can be set as follows: [0.8–1.0] for first-level target area, [0.6–0.8] for second-level target area, [0.4–0.6] for third-level target area, etc. The specific threshold can be dynamically adjusted in the system.

[0108] Spatially aggregate adjacent units of the same confidence level, and use a connectivity analysis method based on spatial adjacency graph to merge them into a single area, outputting a prospecting target area prediction layer with clear boundary range and level label.

[0109] This layer can be displayed graphically on the system terminal interface and can be exported to standardized spatial data file formats (such as GeoTIFF, Shapefile, GDB, etc.) for results submission, map making, and mineral exploration decision support.

[0110] Example 2: A mineral exploration target area prediction system based on alteration mineral analysis, see [link to example]. Figure 1 As shown, it includes the following modules:

[0111] The data acquisition module is used to collect multi-source geological data of the target area;

[0112] The clustering analysis module is used to perform hierarchical clustering analysis on altered minerals and their associated data, as well as mineral spectral features extracted from remote sensing image data, and to construct a time-series decoupling model.

[0113] The alteration combination discrimination module calculates the matching probability between the target alteration combination and the typical mineralized alteration mode based on the mineralized alteration combination probability discrimination model, and outputs the identification results of alteration types with matching probabilities higher than a preset threshold.

[0114] The spatial modeling module is used to generate implicit function fields representing alteration boundary through implicit geological modeling algorithms, and to extract the three-dimensional distribution boundary model of alteration.

[0115] The spatiotemporal coupling module is used to couple the causal period labels, matching probabilities, and the three-dimensional distribution boundary model of the alteration variants to generate the spatiotemporal evolution probability field of the alteration variants.

[0116] The mineralization assessment module is used to calculate the mineralization confidence level of the target area and output the target area prediction results and graded scores.

[0117] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of 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 predicting mineral exploration target areas based on alteration mineral analysis, characterized in that, include: Collect multi-source geological data of the target area, including tectonic measurement data, remote sensing image data, geochemical data, alteration minerals and their associated assemblages; Based on the constraints of mineral thermodynamic phase diagrams, hierarchical clustering analysis is performed on altered minerals and their associated assemblages, as well as mineral spectral features extracted from remote sensing image data, to generate mineral associated assemblages feature maps, assign genetic period labels, and construct a time-series decoupling model. The temporal decoupling model is used as a prior knowledge base and input together with the elemental anomaly indicators in the geochemical data into the Bayesian network-based mineralization alteration combination probability discrimination model. The matching probability between the target alteration combination and the typical mineralization alteration mode is calculated, and the alteration identification results with matching probabilities higher than the preset threshold are output. The spatial distribution trend features of alteration variant identification results are extracted, and the fault zone orientation information extracted from the measurement data is used as a spatial constraint to construct the input dataset. An implicit function field representing the boundary of alteration forms is generated using an implicit geological modeling algorithm, and a three-dimensional distribution boundary model of alteration forms is extracted. Using genetic stage labels, matching probabilities, and three-dimensional distribution boundary models of alteration forms as input variables, a spatiotemporal weight matrix is ​​constructed to generate a spatiotemporal evolution probability field of alteration forms, which serves as the basic data input for calculating mineralization confidence and predicting target areas. The spatiotemporal evolution probability field of the generated alteration form includes: The time series compatibility weight is assigned by a time weight function set according to the chronological order of the causal period labels. The time weight function adopts an exponentially decreasing function form. By dividing the modeling region of the three-dimensional distribution boundary model of the alteration form into a three-dimensional voxel mesh and extracting the spatial density gradient, the spatial continuity weight is calculated through the spatial density gradient. The mineralization matching probability weight is obtained by applying a weight transformation function to the value of the matching probability, wherein the weight transformation function includes a normalization function or a logarithmic mapping. A gridded interpolation algorithm is used to fit the spatiotemporal weight matrix in three dimensions. Within the spatial range defined by the three-dimensional distribution boundary model of the alteration morphology, a three-dimensional probability distribution model that simultaneously includes time attributes, spatial location and mineralization probability values ​​is generated as the spatiotemporal evolution probability field of the alteration morphology. Based on the random forest algorithm, the mineralization confidence of the target area is calculated by integrating geological, geophysical, thermodynamic, remote sensing and geochemical indicators, and the target area prediction results and graded scores are output.

2. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The hierarchical cluster analysis includes: Based on the thermodynamic phase diagram stability domain characteristics of altered minerals and their symbiotic assemblage data, altered minerals are initially grouped. The mineral spectral features extracted from remote sensing image data are fused with the altered mineral grouping results using preset weighting coefficients. The fused data is then input into a clustering model for hierarchical clustering. Based on the grouping results output by the clustering model, typical mineral assemblages and their symbiotic relationships in each group are extracted to construct the corresponding mineral symbiotic assemblage feature map.

3. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 2, characterized in that, The process of fusing the mineral spectral features extracted from remote sensing image data with the alteration mineral grouping results using preset weighting coefficients includes: Multiple characteristic reflection bands and spectral morphology parameters of altered minerals are extracted from remote sensing image data. The spectral features are then vectorized and encoded, and combined with altered minerals and their co-occurring data to form an input dataset. This input dataset is then used as input to a clustering model to perform hierarchical clustering.

4. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The construction of the causal period labels includes: Based on the characteristic map of mineral assemblages and combined with the stability domain of the thermodynamic phase diagram of each group of altered minerals, a period discrimination rule is established; and according to the rule, a unique genetic period label is assigned to each group of altered mineral assemblages to characterize the temporal attribute of the alteration corresponding to the assemblages.

5. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The mineralization alteration combination probability discrimination model is a Bayesian network model constructed using typical mineralization alteration combination samples. The model uses the genetic period labels generated in the time-series decoupling model as prior input nodes, and jointly constructs the input structure with the elemental anomaly indicators in the geochemical data. The matching probability between the target alteration combination and the typical mineralization alteration mode is used as the output node. In the output stage, the discrimination results are filtered based on a preset probability threshold, and the alteration identification results with matching probabilities higher than the threshold are output.

6. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The extracted three-dimensional distribution boundary model of the alteration form includes: The spatial location and distribution trend characteristics of alteration variants are obtained, and the fault zone attitude information extracted from the structural measurement data is used as a spatial constraint to construct a modeling dataset containing a spatial point set, structural control lines, and distribution trend vectors. Based on the modeling dataset, an implicit geological modeling algorithm is used to generate an implicit function field, and the three-dimensional distribution boundary model of the alteration variants is obtained by extracting the function isosurface.

7. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The acquisition of the mineralization confidence level includes: The temporal distribution factor, spatial density index, and mineralization matching probability extracted from the spatiotemporal evolution probability field of alteration forms are used as core feature parameters. Combined with remote sensing image data, geochemical data, and geological structure information, a feature vector set is constructed. A random forest model is trained based on this feature vector set and applied to each grid cell in the target area. The mineralization confidence value of each cell is output as the basis for the probability score of target area prediction.

8. The method for predicting mineral exploration target areas based on alteration mineral analysis according to claim 1, characterized in that, The output target region prediction results include: The spatial distribution results of mineralization confidence are obtained, and multiple grid units within the target area are classified according to the confidence threshold setting rules to generate multiple prospecting target area candidate units. Adjacent units of the same confidence level are spatially aggregated to output a prospecting target area prediction layer with clear boundary range and level label.

9. A mineral exploration target area prediction system based on alteration mineral analysis, characterized in that, The system applies a mineral exploration target area prediction method based on alteration mineral analysis as described in any one of claims 1 to 8, comprising: The data acquisition module is used to collect multi-source geological data of the target area; The clustering analysis module is used to perform hierarchical clustering analysis on altered minerals and their associated data, as well as mineral spectral features extracted from remote sensing image data, and to construct a time-series decoupling model. The alteration combination discrimination module calculates the matching probability between the target alteration combination and the typical mineralized alteration mode based on the mineralized alteration combination probability discrimination model, and outputs the identification results of alteration types with matching probabilities higher than a preset threshold. The spatial modeling module is used to generate implicit function fields representing alteration boundary through implicit geological modeling algorithms, and to extract the three-dimensional distribution boundary model of alteration. The spatiotemporal coupling module is used to couple the causal period labels, matching probabilities, and the three-dimensional distribution boundary model of the alteration variants to generate the spatiotemporal evolution probability field of the alteration variants. The mineralization assessment module is used to calculate the mineralization confidence level of the target area and output the target area prediction results and graded scores.

Images