A method, device, equipment and medium for exploring and evaluating a carbonatite type rare earth ore deposit

Abstract

The application discloses a kind of carbonatite type rare earth ore deposit exploration evaluation method, device, equipment and medium, belong to mineral exploration technical field, method includes: obtaining the multi-source exploration data of target area;Multi-source exploration data is extracted to feature, and obtain multidimensional exploration feature set;Multi-dimensional exploration feature set is input into preset ore-forming prediction model, and preset ore-forming prediction model is known carbonatite type rare earth ore deposit area with the multi-source exploration identification feature vector of sample, with the random feature vector of non-mineralization area as negative sample, and is obtained by random forest or XGBoost algorithm training;Unit classification of each exploration unit is output by preset ore-forming prediction model;Based on unit classification, each exploration unit is carried out three-dimensional geological modeling and spatial positioning, and three-dimensional ore body model is obtained;Based on three-dimensional ore body model, the carbonatite type rare earth ore deposit of target area is carried out resource evaluation, and evaluation result is obtained.The present application can improve the exploration efficiency and exploration accuracy of carbonatite type rare earth ore deposit.

Description

Technical Field

[0001] This invention relates to the field of mineral exploration technology, and more specifically, to a method, apparatus, equipment, and medium for the exploration and evaluation of carbonate rock-type rare earth deposits. Background Technology

[0002] Rare earth elements are key mineral resources supporting high-end industries such as new energy, new materials, and advanced manufacturing. Carbonate-type rare earth deposits, with their large reserves, ease of mining, and high economic value, have become the primary source of global rare earth resources, contributing over 50% of global rare earth reserves and production. my country's super-large deposits such as Bayan Obo, Mianning, Maoniuping, and Weishan belong to this category.

[0003] Currently, with the decreasing availability of easily identifiable surface deposits, concealed and semi-concealed deposits have become the primary focus of exploration. Remote sensing or geophysical exploration is commonly used to explore and identify these types of deposits. However, traditional geophysical methods are not only limited in scope but also susceptible to interference from overburden and complex mineralization markers, leading to complex anomaly responses, difficult interpretation, and insufficient identification accuracy. Furthermore, they struggle to achieve global-scale mineralization comparison and rapid regional scanning, impacting exploration efficiency and reducing accuracy. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a method, apparatus, equipment, and medium for the exploration and evaluation of carbonate rock-type rare earth deposits.

[0005] A method for exploring and evaluating carbonate rock-type rare earth deposits, comprising:

[0006] S101: Acquire multi-source exploration data of the target area; the target area includes multiple exploration units; the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data;

[0007] S102: Extract features from the multi-source exploration data to obtain a multi-dimensional exploration feature set;

[0008] S103: Input the multidimensional exploration feature set into the preset mineralization prediction model and output the unit classification of each exploration unit;

[0009] S104: Based on the unit classification, perform three-dimensional geological modeling and spatial positioning for each exploration unit to obtain a three-dimensional ore body model;

[0010] S105: Based on the three-dimensional ore body model, a resource evaluation of the carbonate-type rare earth deposit in the target area is conducted to obtain the evaluation results.

[0011] Preferably, S101 includes:

[0012] The spectral remote sensing image data is sequentially subjected to radiometric calibration, atmospheric correction and mineral mapping to obtain preprocessed spectral remote sensing impact data.

[0013] The geophysical spectral data is filtered, polarized, extended, and anomaly separated to generate preprocessed geophysical spectral data; wherein the preprocessed geophysical spectral data includes magnetic anomaly maps, gravity anomaly maps, and radioactive anomaly maps.

[0014] Calculate the spatial correlation coefficient between the fluorine content and the rare earth element content in the geochemical element content data. If the spatial correlation coefficient is higher than a preset threshold, the fluorine content and its spatial coupling with rare earth and iron elements are determined as additional features.

[0015] Based on the preprocessed spectral remote sensing impact data, the preprocessed geophysical spectral data, and / or the additional features, preprocessed multi-source exploration data is generated.

[0016] Preferably, in step S103, the training process of the preset mineralization prediction model includes:

[0017] A training sample set is obtained, which includes positive samples of the training multidimensional exploration identifier feature vectors of mineralized units and negative samples of the training multidimensional exploration identifier feature vectors of units without mineralization records. The training multidimensional exploration identifier feature vectors include the gradient value of the annular magnetic anomaly, the difference of the negative gravity anomaly, the thorium anomaly count rate in the gamma spectrum, the depth of the characteristic absorption peak of rare earth elements in the remote sensing spectrum, the spatial correlation coefficient of the content of fluorine and rare earth elements in the geochemical assemblage anomaly, and the characteristic ratio of geochemical elements.

[0018] The training multidimensional exploration identifier feature vectors corresponding to the positive and negative samples are standardized to obtain standard training samples.

[0019] The standard training samples are input into the initial mineralization prediction model. In the preset initial mineralization prediction model, the Bayesian optimization method is used with the area under the ROC curve or the F1 score as the optimization objective. The combination of hyperparameters is iteratively adjusted until the area under the ROC curve converges or the preset number of iterations is reached. The optimal hyperparameters are then determined, and the preset mineralization prediction model is obtained.

[0020] Using a pre-defined mineralization prediction model, the contribution of each trained multidimensional exploration identifier feature vector to the model prediction results is calculated.

[0021] Based on the aforementioned contribution level, core exploration identification features are determined;

[0022] Based on the predicted probability distribution of the preset mineralization prediction model on the validation set, a precision-recall curve is plotted, and the probability value corresponding to the balance point of precision and recall is selected as a preset threshold. The preset threshold is used to classify the exploration unit.

[0023] Preferably, in S103, the following is further included:

[0024] Isolated anomalies are filtered downgraded or upgraded based on the spatial neighborhood classification consistency of each exploration unit to obtain the filtering correction result;

[0025] The predicted probability is adjusted by confidence weighting based on the spatial correlation coefficients of fluorine, rare earth, and iron elements to obtain the weighted adjustment result.

[0026] Identify low-probability magnetic anomaly regions, reclassify the original low-probability units within them, and attach candidate labels for deep concealed mineral deposits to obtain anomaly discrimination results;

[0027] The unit classification is corrected based on the filtering correction result, the weighted correction result, and the anomaly discrimination result to obtain the final unit classification; wherein, the unit classification includes a first probability mineralization unit, a second probability mineralization unit, and a third probability mineralization unit, and the first probability, the second probability, and the third probability decrease in sequence.

[0028] Preferably, in S104, the following is included:

[0029] Using the spatial range of the first probabilistic mineralization unit as the constraint boundary, the three-dimensional seismic inversion wave impedance data volume and magnetotelluric sounding resistivity data volume of the target area are input, and an implicit modeling algorithm is used to construct a three-dimensional geological structure model including the top and bottom interfaces of the carbonate rock mass, the fracture structure surface, and the envelope surface of the alteration zone.

[0030] The preset mineralization prediction model outputs the mineralization probability of each exploration unit based on the multidimensional exploration feature set.

[0031] Using the mineralization probability value as a covariate, and combining borehole core hyperspectral logging, gamma logging and resistivity logging data, the three-dimensional spatial morphology and grade distribution of rare earth mineralization bodies are predicted within the three-dimensional geological structure model using Kriging or co-co-Kriging interpolation methods.

[0032] Based on the aforementioned three-dimensional geological structure model and three-dimensional spatial morphology and grade distribution, the three-dimensional ore body model is constructed.

[0033] Preferably, in S105, the following is included:

[0034] Extract the volume, average grade, thickness, and burial depth parameters of the three-dimensional ore body model, and call the pre-established grade-tonnage database of carbonate rock-type rare earth deposits for similarity matching to obtain the first resource quantity;

[0035] A linear regression equation is established based on borehole data of the target area or typical mineral deposits in the area. The spatial coupling degree between the fluorine and iron anomalies in the geochemical element content data is used to estimate the second resource quantity.

[0036] The first resource quantity and the second resource quantity are weighted and fused according to their respective confidence levels, wherein the confidence level is determined by the spatial correlation coefficient of fluorine-iron-rare earth and the matching sample size in the grade-tonnage database.

[0037] Based on the weighted and fused resource quantity values ​​and their corresponding probability intervals, an evaluation result is generated that includes the resource quantity range, quality classification, and confidence level.

[0038] Preferably, after S105, the method further includes:

[0039] Using the centroid projection point of the three-dimensional ore body model as the initial first priority drilling position, calculate the planar coordinates of the centroid, and mark the vertical projection point of the planar coordinates on the top surface of the three-dimensional ore body model as P0.

[0040] Extract the spatial gradient field of the predicted grade of each exploration unit in the three-dimensional ore body model, calculate the grade change rate of each exploration unit in the east-west, north-south and vertical directions, and generate a grade gradient vector map.

[0041] Sort the units by grade gradient modulus from high to low, select the N units with the largest modulus, and use the spatial cluster center as the projection position of the high grade gradient region, marked as P1, P2...Pn; where N is positively correlated with the area of ​​the target region.

[0042] Extract the distribution boundary of the first probability mineralization unit in the unit classification, and combine it with the fault structure surface in the three-dimensional geological model to calculate the structural ore-controlling gradient zone. The structural ore-controlling gradient zone is defined as the buffer zone of the intersection line between the boundary of the high probability mineralization unit and the fault surface.

[0043] The equally spaced sampling points along the center line of the buffer zone are projected onto the ground surface and marked as Q1, Q2...Qm;

[0044] The initial borehole location set {P0, P1…Pn, Q1…Qm} is deduplicated and optimized by sorting: the Euclidean distance between any two points is calculated, points whose distance is less than the preset minimum borehole spacing are deleted, and points with large grade gradient modulus or falling into the structural ore-controlling gradient zone are retained first.

[0045] Starting from P0, the minimum spanning tree algorithm is used to connect the remaining points to generate a shortest path covering all recommended borehole locations, thus obtaining the drilling path.

[0046] The design depth is dynamically calculated based on the grade at each recommended borehole location;

[0047] A drilling layout suggestion map is generated based on the design depth, borehole number, construction sequence, and expected mineralization depth range.

[0048] This invention also provides an exploration and evaluation device for carbonate rock-type rare earth deposits, comprising:

[0049] The acquisition module is used to acquire multi-source exploration data of a target area; the target area includes multiple exploration units; the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data;

[0050] The extraction module is used to extract features from the multi-source exploration data to obtain a multi-dimensional exploration feature set;

[0051] The input module is used to input the multidimensional exploration feature set into the preset mineralization prediction model and output the unit classification of each exploration unit;

[0052] The modeling and positioning module is used to perform three-dimensional geological modeling and spatial positioning of each exploration unit based on the unit classification, so as to obtain a three-dimensional ore body model.

[0053] The evaluation module is used to evaluate the resources of the carbonate-type rare earth deposits in the target area based on the three-dimensional ore body model, and obtain the evaluation results.

[0054] The present invention also provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, characterized in that the computer program, when executed by the processor, implements the steps in the above-described method for exploring and evaluating carbonate rock-type rare earth deposits.

[0055] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps in the above-described method for exploring and evaluating carbonate rock-type rare earth deposits.

[0056] The beneficial effects of the carbonate rock-type rare earth deposit exploration and evaluation method provided by this invention are as follows: Compared with the prior art, this invention can acquire multi-source exploration data of the target area and integrate spectral remote sensing, geophysical and geochemical data. Then, it extracts features from the multi-source exploration data to obtain a multi-dimensional exploration feature set. This can reduce the interpretation difficulties caused by interference from overburden and complex mineralization representation from multiple dimensions. Then, it classifies the exploration units through a preset mineralization prediction model to achieve a refined classification of the target area. Then, through unit classification, it performs three-dimensional geological modeling and spatial positioning to generate a three-dimensional ore body model. Through the overall evaluation of the three-dimensional ore body model, it can conduct refined exploration of rare earth deposits from multiple perspectives, further improving exploration efficiency and accuracy.

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0058] 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 described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a method for exploring and evaluating carbonate rock-type rare earth deposits according to an embodiment of the present invention.

[0060] Figure 2 This is a schematic diagram of the structure of a carbonate rock-type rare earth deposit exploration and evaluation device provided in another embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the structure of an electronic device provided in another embodiment of the present invention. Detailed Implementation

[0062] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0063] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0064] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0065] Please see Figure 1 A method for exploring and evaluating carbonate rock-type rare earth deposits, comprising:

[0066] To address the problems of existing technologies, this application provides a method, apparatus, equipment, and storage medium for the exploration and evaluation of carbonate rock-type rare earth deposits. In this application, multi-source exploration data of the target area is acquired and integrated with spectral remote sensing, geophysical, and geochemical data. Then, feature extraction is performed on the multi-source exploration data to obtain a multi-dimensional exploration feature set. This reduces the difficulty in interpretation caused by interference from overburden layers and complex mineralization representations from multiple dimensions. Subsequently, exploration units are classified using a preset mineralization prediction model, achieving refined classification of the target area. Based on unit classification, three-dimensional geological modeling and spatial positioning are performed to generate a three-dimensional ore body model. Through a holistic evaluation of the three-dimensional ore body model, refined exploration of rare earth deposits can be conducted from multiple perspectives, further improving exploration efficiency and accuracy.

[0067] The following section first introduces the exploration and evaluation method for carbonate rock-type rare earth deposits provided in the embodiments of this application.

[0068] Figure 1 A schematic flowchart of an embodiment of the method for exploring and evaluating carbonate rock-type rare earth deposits provided in this application is shown. Figure 1 As shown, the exploration and evaluation method for carbonate rock-type rare earth deposits may include steps S101 to S105:

[0069] S101, acquire multi-source exploration data of the target area.

[0070] In this embodiment, the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data. The target area includes multiple exploration units. The spectral remote sensing image data, geophysical spectral data, and geochemical element content data can be collected manually or directly imported from existing databases. The spectral remote sensing image data can be acquired by optical sensors, which record electromagnetic spectrum information reflected or emitted by the Earth's surface. The geophysical spectral data can be directly measured by geophysical instruments such as magnetometers, gravimeters, or gamma-ray spectrometers. The geochemical element content data can be obtained through laboratory analysis of surface soil, rock, or stream sediment samples.

[0071] As an example, in order to conduct detailed exploration of rare earth deposits in the future, the target area can be divided into multiple exploration units. These units can be regular grid units or irregular polygonal units divided based on geological features.

[0072] S102, feature extraction is performed on multi-source exploration data to obtain a multi-dimensional exploration feature set.

[0073] In this embodiment, the multidimensional exploration features include geometric morphological feature values, spectral absorption depth values, magnetic anomaly gradient values, gravity anomaly differences, gamma-ray energy spectrum count rates, geochemical element content values, and element ratios. These multidimensional exploration features can be directly extracted from multi-source exploration data. For example, geometric morphological feature values ​​can be obtained from remote sensing images through visual interpretation or simple image processing algorithms to acquire the boundary and shape information of geological bodies; spectral absorption depth values ​​can be directly read from the absorption valley depth of the band of interest in the original spectral curve; magnetic anomaly gradient values ​​and gravity anomaly differences can be obtained from the original magnetic field and gravity field data through simple difference operations or filtering; gamma-ray energy spectrum count rates can be directly read from gamma-ray energy spectrum data; and geochemical element content values ​​and element ratios can be directly obtained from geochemical analysis reports.

[0074] S103, input the multi-dimensional exploration feature set into the preset mineralization prediction model. The preset mineralization prediction model is trained by using the multi-source exploration identifier feature vector of the known carbonate rock type rare earth deposit area as positive samples and the random feature vector of the non-mineralized area as negative samples. The preset mineralization prediction model outputs the unit classification of each exploration unit.

[0075] In this embodiment, the preset mineralization prediction model can be a classifier built based on a machine learning algorithm. During the model training phase, exploration data from known carbonate-type rare earth deposit areas can be collected as positive samples, and exploration data from unmineralized areas can be collected as negative samples. These sample data are used to train the model; for example, a random forest algorithm or an XGBoost algorithm can be used, and parameters can be selected using default parameters or through a simple cross-validation method. After the model is trained, it can output the unit classification of each exploration unit based on the multidimensional features of the input exploration unit, for example, directly outputting a binary classification result (mineralized or unmineralized) or a preliminary mineralization probability value.

[0076] S104. Based on unit classification, three-dimensional geological modeling and spatial positioning are performed on each exploration unit to obtain a three-dimensional ore body model.

[0077] In this embodiment of the application, the unit classification results output by the model can be used as the preliminary boundary of the mineralized body, for example, the two-dimensional plane range of all exploration units classified as "mineralized". Then, combined with the existing geological profile map of the area or the simple geological body extension law, these two-dimensional boundaries can be extended upward or downward to form a preliminary three-dimensional ore body model.

[0078] S105, Based on the three-dimensional ore body model, the resource evaluation of the carbonate-type rare earth deposit in the target area was carried out, and the evaluation results were obtained.

[0079] In this embodiment, the preliminary resource quantity of the ore body can be directly calculated based on the volume of the initially constructed three-dimensional ore body model and an empirically estimated average grade. The evaluation result can be presented simply as a numerical value of resource quantity, or accompanied by a general grade range.

[0080] When exploring and evaluating carbonate-type rare earth deposits, multi-source exploration data of the target area can be acquired and integrated with spectral remote sensing, geophysical, and geochemical data. Feature extraction from this multi-source data yields a multi-dimensional exploration feature set, which can reduce the interpretation difficulties caused by interference from overburden and complex mineralization representations from multiple dimensions. Then, exploration units are classified using a pre-set mineralization prediction model, achieving refined classification of the target area. Based on unit classification, three-dimensional geological modeling and spatial positioning are performed to generate a three-dimensional ore body model. A comprehensive evaluation of the three-dimensional ore body model allows for refined exploration of rare earth deposits from multiple perspectives, further improving exploration efficiency and accuracy.

[0081] In some other embodiments, prior to S102, the method may further include:

[0082] Radiometric calibration, atmospheric correction and mineral mapping were performed on the spectral remote sensing image data to obtain preprocessed spectral remote sensing impact data.

[0083] The geophysical spectral data is filtered, polarized, extended, and anomaly separated to generate preprocessed geophysical spectral data, which includes magnetic anomaly maps, gravity anomaly maps, and radioactive anomaly maps.

[0084] The spatial correlation coefficient between fluorine and rare earth element content in geochemical elemental data is calculated. If the spatial correlation coefficient is higher than a preset threshold, the fluorine content and its spatial coupling with rare earth and iron elements are identified as additional features.

[0085] Based on the preprocessed spectral remote sensing impact data, the preprocessed geophysical spectral data and / or additional features, preprocessed multi-source exploration data is generated.

[0086] Feature extraction is performed on multi-source exploration data to obtain a multi-dimensional exploration feature set, including:

[0087] Feature extraction is performed on the preprocessed multi-source exploration data to obtain a multi-dimensional exploration feature set.

[0088] In this embodiment, radiometric calibration, atmospheric correction, and mineral mapping are performed on the spectral remote sensing image data to obtain preprocessed spectral remote sensing impact data.

[0089] The geophysical spectral data is filtered, polarized, extended, and anomaly separated to generate preprocessed geophysical spectral data, which includes magnetic anomaly maps, gravity anomaly maps, and radioactive anomaly maps.

[0090] The spatial correlation coefficient between fluorine and rare earth element content in geochemical elemental data is calculated. If the spatial correlation coefficient is higher than a preset threshold, the fluorine content and its spatial coupling with rare earth and iron elements are identified as additional features.

[0091] Based on the preprocessed spectral remote sensing impact data, the preprocessed geophysical spectral data and / or additional features, preprocessed multi-source exploration data is generated.

[0092] Feature extraction is performed on multi-source exploration data to obtain a multi-dimensional exploration feature set, including:

[0093] Feature extraction is performed on the preprocessed multi-source exploration data to obtain a multi-dimensional exploration feature set.

[0094] Specifically, when performing radiometric calibration, atmospheric correction, and mineral mapping on spectral remote sensing image data, radiometric calibration can correct for systematic errors and environmental influences generated by remote sensing sensors during data acquisition, and convert the digital quantization values ​​received by the sensor into physically meaningful radiance or reflectance values.

[0095] Atmospheric correction can eliminate the absorption, scattering, and reflection effects of the atmosphere on remote sensing signals to obtain accurate surface reflectance information. For example, methods based on radiative transfer models (such as MODTRAN and 6S models) can be used to perform physical correction by inputting atmospheric and geometric parameters; or empirical or semi-empirical methods based on dark target methods or histogram matching methods can be used for correction.

[0096] Algorithms such as Spectral Angle Mapping (SAM) and Spectral Matched Filtering (SMF) can be used to match the spectral curve of each pixel in the image with the standard spectra in a known mineral spectral library; or transformation methods such as Principal Component Analysis (PCA) and Independent Component Analysis (ICA) can be used to enhance mineral feature information and perform classification. Through the above processing, preprocessed spectral remote sensing impact data can be obtained, and its accuracy and usability are significantly improved.

[0097] Then, in order to eliminate random noise and high-frequency interference in geophysical data, filtering, polarization, extension and anomaly separation processing can be performed on the geophysical spectral data to generate preprocessed geophysical spectral data, including magnetic anomaly maps, gravity anomaly maps and radioactive anomaly maps. These data have higher signal-to-noise ratio and interpretability.

[0098] It is worth noting that spectral remote sensing impact data can include satellite spectral remote sensing data; geophysical spectral data can include airborne geophysical data, airborne hyperspectral rapid surface scanning data, and ground-based high-precision geophysical data. These can be obtained through relevant technologies, which will not be elaborated upon here.

[0099] For geochemical element content data, the spatial correlation coefficient between fluorine content and rare earth element content can be calculated. Data preprocessing can be performed using the spatial correlation coefficient. Specifically, if the spatial correlation coefficient is higher than a preset threshold, the fluorine content and its spatial coupling with rare earth and iron elements are identified as additional features. The calculation of the spatial correlation coefficient aims to quantify the degree of correlation between the spatial distribution of fluorine content and rare earth element content in geochemical element content data.

[0100] As an example, statistical methods such as Pearson correlation coefficient and Spearman rank correlation coefficient can be used to calculate the element content data after spatial interpolation; or geographically weighted regression (GWR) or spatial autocorrelation analysis (such as Moran's I) can be used to assess local spatial correlation to reveal the strength of the association between elements in different regions; for example, when the absolute value of the correlation coefficient is higher than 0.7 or the p value is less than 0.05, the correlation is considered significant.

[0101] This coupling relationship can be quantified by calculating the covariance matrix of the multi-element combination or by performing principal component analysis; alternatively, multivariate statistical methods (such as factor analysis and cluster analysis) can be used to identify the symbiotic patterns of the element combination.

[0102] By integrating the preprocessed spectral remote sensing image data, geophysical spectral data, and / or additional features, a unified, high-quality multi-source exploration dataset is formed. This ensures that all optimized data and potential additional features can work synergistically, providing a reliable foundation for subsequent feature extraction.

[0103] In some other embodiments, prior to S103, the method may further include:

[0104] A training sample set was obtained, which included positive samples of the training multidimensional exploration identifier feature vectors of mineralized units and negative samples of the training multidimensional exploration identifier feature vectors of non-mineralized units. The training multidimensional exploration identifier feature vectors included the gradient value of the annular magnetic anomaly, the difference of the negative gravity anomaly, the thorium anomaly count rate in the gamma spectrum, the depth of the characteristic absorption peak of rare earth elements in the remote sensing spectrum, the spatial correlation coefficient of the content of fluorine and rare earth elements in the geochemical assemblage anomaly, and the characteristic ratio of geochemical elements.

[0105] The training multidimensional exploration identifier feature vectors corresponding to positive and negative samples are standardized to obtain standard training samples.

[0106] Standard training samples are input into the initial mineralization prediction model. In the preset initial mineralization prediction model, the Bayesian optimization method is used to iteratively adjust the hyperparameter combination with the area under the ROC curve or the F1 score as the optimization objective until the area under the ROC curve converges or the preset number of iterations is reached. The optimal hyperparameters are then determined, and the preset mineralization prediction model is obtained.

[0107] Using a pre-defined mineralization prediction model, the contribution of each trained multidimensional exploration identifier feature vector to the model prediction results is calculated.

[0108] Based on contribution, core exploration identifier features are determined;

[0109] Based on the predicted probability distribution of the preset mineralization prediction model on the validation set, a precision-recall curve is plotted. The probability value corresponding to the balance point between precision and recall is selected as the preset threshold, which is used to classify exploration units.

[0110] In this embodiment, the training sample set may include positive samples of the training multidimensional exploration identifier feature vector of mineralized units and negative samples of the training multidimensional exploration identifier feature vector of non-mineralized records. Positive samples represent areas where carbonate rock-type rare earth mineralization is known to exist, while negative samples represent areas where mineralization is known to be non-mineralized or with extremely low mineralization. In order to better train the preset mineralization prediction model, it can be trained by training multidimensional exploration identifier features. Specifically, the training multidimensional exploration identifier feature vector may include the gradient value of the annular magnetic anomaly, the difference of the negative gravity anomaly, the thorium anomaly count rate in the gamma spectrum, the depth of the characteristic absorption peak of rare earth elements in the remote sensing spectrum, the spatial correlation coefficient of the content of fluorine and rare earth elements in the geochemical assemblage anomaly, and the characteristic ratio of geochemical elements.

[0111] These feature vectors can be obtained through comprehensive analysis and extraction of historical exploration data, borehole data, geological survey reports, etc., which will not be elaborated further here.

[0112] After obtaining the training sample set, in order to ensure the fairness and rationality of the training, the training multidimensional exploration identifier feature vectors corresponding to the positive and negative samples can be standardized to obtain standard training samples, thereby improving the training efficiency and prediction accuracy of the model.

[0113] When training the initial mineralization prediction model, the area under the ROC curve or the F1 score can be used as the optimization objective. The combination of hyperparameters can be continuously adjusted until the training stopping condition is met, thus obtaining the preset mineralization prediction model.

[0114] To quantify the importance of each feature in the model decision-making process, methods such as SHAP (SHapley Additive exPlanations) values ​​or permutation importance can be used to calculate the feature contribution. SHAP values ​​can explain the contribution of each feature in a single prediction, while permutation importance assesses the impact on model performance by randomly shuffling the order of individual features.

[0115] Subsequently, by analyzing the contribution of each feature, the key features that have the greatest impact on the model's prediction results can be identified, thereby reducing feature redundancy and improving the model's interpretability and operational efficiency.

[0116] Finally, based on the predicted probability distribution of the pre-defined mineralization prediction model on the validation set, a precision-recall curve is plotted. The probability value corresponding to the balance point between precision and recall is selected as the pre-defined threshold, which is used to classify exploration units. The precision-recall curve (PR curve) is an effective tool for evaluating the performance of classification models on imbalanced datasets. By analyzing the PR curve, a balance point can be found that achieves a good balance between the model's precision (the proportion of actual positive samples among those predicted as positive) and recall (the proportion of actual positive samples that are predicted as positive). The probability value corresponding to this balance point is selected as the classification threshold, used to convert the continuous probability values ​​output by the model into discrete unit classification results (such as mineralization or non-mineralization).

[0117] In some other embodiments, after the preset mineralization prediction model outputs the unit classification for each exploration unit, the method further includes:

[0118] Isolated anomalies are filtered downgraded or upgraded based on the spatial neighborhood classification consistency of each exploration unit to obtain the filtering correction result;

[0119] The predicted probability is adjusted by confidence weighting based on the spatial correlation coefficients of fluorine, rare earth, and iron elements to obtain the weighted adjustment result.

[0120] Identify low-probability magnetic anomaly regions, reclassify the original low-probability units within them, and attach candidate labels for deep concealed mineral deposits to obtain anomaly discrimination results;

[0121] The cell classification is corrected based on the filtering correction results, weighted correction results, and anomaly detection results to obtain the final cell classification.

[0122] In this embodiment, a majority voting method can be used. That is, for a target exploration unit, the classification results of other exploration units in its preset neighborhood (e.g., a 3x3 or 5x5 grid centered on the unit) are counted. If the classification of the target unit is inconsistent with the classification of the majority of units in the neighborhood, its classification is adjusted, for example, downgraded from mineralized to non-mineralized, or upgraded from non-mineralized to mineralized, so as to enhance the spatial coherence of the classification results.

[0123] Furthermore, based on spatial autocorrelation analysis, the local spatial autocorrelation index of each unit and its neighboring units can be calculated to identify statistically significant outliers and adjust the classification according to the overall trend of its neighborhood.

[0124] When performing weighted correction, the spatial correlation coefficients between fluorine, rare earth, and iron elements within each exploration unit can be calculated first and normalized. Then, the normalized correlation coefficients are used as weights and weighted averaged or multiplied with the original mineralization probability output by the preset mineralization prediction model to obtain the corrected probability value.

[0125] In addition, a correction method based on fuzzy logic can be adopted to define fuzzy membership functions of the spatial correlation coefficients and mineralization probabilities of these elements, and to correct the original mineralization probabilities through fuzzy inference rules so that they are more in line with the geological genesis.

[0126] When identifying anomalies, magnetic anomaly data can be filtered or analyzed to identify areas where the magnetic anomaly intensity is below a preset threshold and the spatial variation is gradual. For exploration units within these areas, if their original mineralization probability is low, their probability values ​​are increased and they are explicitly marked as "deep concealed mineralization candidates".

[0127] Image segmentation or clustering algorithms can also be used to identify regions with low-slow magnetic anomaly characteristics, and dynamically calculate bonus values ​​based on the characteristics of these regions (such as area, shape, distance from known mineral deposits, etc.), reclassify low-probability units, and attach deep concealed mineral candidate labels.

[0128] After obtaining the filtering correction results, weighted correction results, and anomaly detection results, the unit classification can be further modified to form a more comprehensive and accurate final classification result. For example, a decision fusion approach can be adopted to treat the filtering correction results, weighted correction results, and anomaly detection results as independent decision layers and set priorities.

[0129] In one example, if a cell is marked as a "deep concealed ore candidate" by the anomaly detection result, this result is adopted first; otherwise, the weighted correction result and the filtered correction result are considered. Another method is probability fusion, which converts all correction results into probability values ​​and fuses these probability values ​​through methods such as weighted averaging, Bayesian fusion, or Dempster-Shafer evidence theory, and finally classifies them according to the fused probability values.

[0130] In this embodiment, isolated anomalies can be filtered downgraded or upgraded based on spatial neighborhood classification consistency. Then, the predicted probabilities are weighted and corrected based on the spatial correlation coefficients of fluorine, rare earth, and iron elements. Low-gradient magnetic anomaly regions are identified, and the original low-probability units within them are reclassified with added scores and additional deep-seated concealed mineral candidate labels. Finally, the unit classification is comprehensively corrected based on the filtering correction results, weighted correction results, and anomaly discrimination results, integrating multi-source correction information to make the final unit classification results more accurate, complete, and reliable, thereby significantly improving the overall accuracy and efficiency of the exploration and evaluation of carbonate rock-type rare earth deposits.

[0131] In other embodiments, the unit classification includes a first probability mineralization unit, a second probability mineralization unit, and a third probability mineralization unit, wherein the probabilities decrease sequentially from first to third. The first probability mineralization unit represents the area with the highest probability of mineralization, followed by the second probability mineralization unit, and the third probability mineralization unit represents the area with a lower probability of mineralization but still has some potential. This classification can be based on the mineralization probability values ​​output by a preset mineralization prediction model, and can be divided by setting different probability thresholds. For example, units with probability values ​​higher than 0.8 can be classified as first probability mineralization units, those between 0.5 and 0.8 as second probability mineralization units, and those between 0.2 and 0.5 as third probability mineralization units. Alternatively, these classifications can be determined based on expert experience or historical mineral deposit data, combined with statistical methods (such as cluster analysis). This classification helps to prioritize high-potential areas and optimize resource allocation during subsequent modeling.

[0132] When generating a three-dimensional ore body model, S104 may include:

[0133] Using the spatial range of the first probabilistic ore-forming unit as the constraint boundary, the three-dimensional seismic inversion wave impedance data volume and magnetotelluric sounding resistivity data volume of the target area are input, and an implicit modeling algorithm is used to construct a three-dimensional geological structure model including the top and bottom interfaces of the carbonate rock mass, the fracture structure surface, and the envelope surface of the alteration zone.

[0134] The preset mineralization prediction model is based on a multi-dimensional exploration feature set and outputs the mineralization probability of each exploration unit;

[0135] Using the mineralization probability value as a covariate, and combining borehole core hyperspectral logging, gamma logging and resistivity logging data, the three-dimensional spatial morphology and grade distribution of rare earth mineralization bodies are predicted within the three-dimensional geological structure model using kriging or co-co-kriging interpolation methods.

[0136] A three-dimensional ore body model is constructed based on a three-dimensional geological structure model and three-dimensional spatial morphology and grade distribution.

[0137] In this embodiment, a set of all exploration units classified as first probabilistic ore-forming units can be extracted, and their outer contours or envelopes can be used as geometric constraints for modeling. This allows the subsequent 3D geological model to focus on high-probability ore-forming units. To comprehensively reflect the geological environment, elastic parameter information of underground rocks can be obtained through 3D seismic inversion impedance data to reflect structural features such as lithological interfaces and fracture structures. This is crucial for identifying the contact relationship and internal structure between carbonate rocks and surrounding rocks. Magnetotelluric resistivity data can reflect the electrical characteristics of the underground medium, offering advantages in identifying alteration zones, fluid channels, and electrical differences between different lithologies. In addition to seismic and magnetotelluric data, 3D gravity inversion density data or 3D magnetic anomaly inversion magnetic susceptibility data can also be input to provide more comprehensive geophysical constraints. The 3D geological structure model constructed through implicit modeling algorithms can comprehensively reflect the geological conditions of the target area, providing a more refined geological basis for subsequent exploration and evaluation.

[0138] When predicting the probability of mineralization, the pre-set mineralization prediction model can comprehensively utilize a multi-dimensional exploration feature set (including geometric morphological features, spectral absorption depth values, magnetic anomaly gradient values, gravity anomaly differences, gamma-ray energy spectrum count rates, geochemical element content values, and element ratios, etc.) to assess the mineralization potential of each exploration unit and output a mineralization probability value between 0 and 1. This probability value intuitively reflects the likelihood of rare earth mineralization in the unit, providing a quantitative basis for subsequent mineralization prediction.

[0139] In this embodiment, the mineralization probability value is used as a covariate, meaning that when interpolating the grade of rare earth mineralization, in addition to considering the grade value and spatial autocorrelation of the borehole measured data itself, the mineralization probability value is also introduced as auxiliary information into the interpolation model. For example, in co-kriging interpolation, the mineralization probability value can be used as a secondary variable, participating in the interpolation calculation together with the main variable (borehole grade), thereby utilizing its spatial distribution trend to optimize grade prediction. Hyperspectral logging data from borehole cores can provide mineral composition information, which helps in identifying rare earth minerals; gamma-ray logging data can reflect the content of radioactive elements (such as Th and U), which are often associated with rare earth mineralization; resistivity logging data reflects the conductivity of rocks and can be used to identify alteration zones or mineralization zones.

[0140] The three-dimensional geological structure model, constructed subsequently using a three-dimensional geological structure model and its three-dimensional spatial morphology and grade distribution, not only includes the geological background of the ore body (such as carbonate rocks, faults, and alteration zones), but also accurately depicts the geometry, spatial location, and internal grade distribution of the mineralization body. The construction process typically involves overlaying the predicted mineralization body boundaries and grade framework onto the geological structure model, forming a complete digital model that can be used for resource quantity calculation and mining planning.

[0141] The unit classification is refined into first-probability mineralization units, second-probability mineralization units, and third-probability mineralization units, with the probabilities decreasing sequentially. This hierarchical design allows subsequent modeling processes to prioritize areas with the highest mineralization potential, avoiding unnecessary computational resource investment in low-probability units, thus optimizing the focus and efficiency of modeling. Based on this, using the spatial range of the first-probability mineralization unit as the constraint boundary, and combining the 3D seismic inversion impedance data and magnetotelluric sounding resistivity data of the target area, an implicit modeling algorithm is used to construct a 3D geological structure model including the top and bottom interfaces of the carbonate rock mass, fault structural surfaces, and alteration zone envelopes. This method ensures that the modeling boundary is based on the highest-probability area, reduces the interference of irrelevant data on model accuracy, and utilizes seismic and electromagnetic data to provide deep geological information, enhancing the geological realism of the model. The adoption of implicit modeling algorithms efficiently captures complex geological interfaces. Furthermore, the mineralization probability value of each exploration unit output from the multi-dimensional exploration feature set is cleverly used as a covariate in the pre-defined mineralization prediction model. Combined with borehole core hyperspectral logging, gamma logging, and resistivity logging data, kriging or co-co-kriging interpolation methods are used to predict the three-dimensional spatial morphology and grade distribution of rare earth mineralization within the three-dimensional geological structure model. This optimizes the interpolation weights, fully utilizes the advantages of multi-source data, and significantly improves the accuracy of mineralization morphology and grade prediction. Finally, based on the accurately constructed three-dimensional geological structure model and the highly accurate predicted three-dimensional spatial morphology and grade distribution, a complete and reliable three-dimensional orebody model is constructed, laying a solid foundation for subsequent resource evaluation.

[0142] In some other embodiments, S105 may include:

[0143] Extract the volume, average grade, thickness and burial depth parameters of the three-dimensional ore body model, call the pre-established grade-tonnage database of carbonate rock type rare earth deposits for similarity matching, and obtain the first resource quantity;

[0144] A linear regression equation is established based on borehole data of the target area or typical mineral deposits in the area. The spatial coupling degree between fluorine and iron anomalies in geochemical element content data is used to estimate the second resource quantity.

[0145] The first and second resource quantities are weighted and fused according to their respective confidence levels. The confidence level is determined by the spatial correlation coefficient of fluorine-iron-rare earth and the matching sample size in the grade-tonnage database.

[0146] Based on the weighted and fused resource quantity values ​​and their corresponding probability intervals, an evaluation result is generated that includes the resource quantity range, quality classification, and confidence level.

[0147] In this embodiment, when extracting the volume, average grade, thickness, and burial depth parameters of the three-dimensional ore body model, the volume parameter can be obtained by meshing or voxelizing the three-dimensional ore body model and then summing the volumes of all voxels; the average grade parameter can be calculated based on the predicted grade data of each exploration unit in the three-dimensional ore body model, using a weighted average or geostatistical interpolation method (such as Kriging interpolation); the thickness parameter can be obtained by cutting the ore body model along the vertical direction or a specific exploration direction, measuring the vertical or true thickness of the ore body at different locations, and calculating its average value; the burial depth parameter can be determined by determining the depth of the top surface of the ore body relative to the surface or a certain reference surface, and calculating its average burial depth.

[0148] Then, using known statistical patterns and empirical data of mineral deposits, a preliminary resource estimate is provided for the target mineral deposit. Then, using a multi-dimensional feature matching algorithm (e.g., Euclidean distance, cosine similarity), the similarity of the target ore body's volume, average grade, thickness, and burial depth parameters with the deposits in the pre-established grade-tonnage database of carbonate rock-type rare earth deposits is compared. The most similar deposit is then selected as a reference for resource estimation. The grade-tonnage database can contain statistical data on the grade, tonnage, ore body morphology, and burial depth of explored carbonate rock-type rare earth deposits globally or regionally.

[0149] To compensate for the shortcomings of single geological model estimation, geochemical data, especially indicator elements closely related to rare earth mineralization, can be introduced. Resource quantity can be estimated through statistical models, and then a multiple linear regression model can be established by statistically analyzing the rare earth grade, fluorine and iron content and their spatial coupling degree in borehole data. That is, the calculated fluorine-iron spatial coupling degree can be substituted into the established linear regression equation to estimate the second resource quantity.

[0150] After obtaining the first and second resource quantities, the first and second resource quantities are weighted and summed. By combining the advantages of different estimation methods and considering their respective confidence levels, the accuracy and robustness of the final resource quantity estimation are improved.

[0151] It is worth noting that the confidence level is determined by the spatial correlation coefficient of fluorine-iron-rare earth elements and the matching sample size in the grade-tonnage database to ensure the rationality of the weighted fusion. The spatial correlation coefficient of fluorine-iron-rare earth elements can be used to assess the reliability of the geochemical estimate by calculating the Pearson correlation coefficient, Spearman correlation coefficient, or geostatistical correlation among fluorine, iron, and rare earth elements in the target area. The higher the correlation coefficient, the higher the confidence level of the second resource quantity. The larger the matching sample size in the grade-tonnage database, the more sufficient the similar deposit data on which the first resource quantity estimate is based, and the higher its confidence level.

[0152] When generating evaluation results that include resource range, grade classification, and confidence level based on the weighted and fused resource quantity values ​​and their corresponding probability intervals, the probability intervals can be obtained by performing uncertainty analysis on the weighted and fused resource quantity using methods such as Monte Carlo simulation and Bootstrap resampling to obtain its 90% or 95% confidence interval. The grade classification can be based on industry standards or preset thresholds (e.g., high grade, medium grade, low grade) to classify the average grade of the deposit. The confidence level can be classified based on the overall confidence level of the final resource quantity estimate (e.g., high, medium, low), combined with factors such as geological reliability and data integrity.

[0153] This application integrates the geometric and grade parameters of a three-dimensional orebody model, statistical experience from a grade-tonnage database, and the spatial coupling degree of geochemical elements (fluorine, iron, and rare earth elements). Then, it performs a confidence-based weighted fusion of the first and second resource quantities, quantifying their probability intervals, grade grades, and confidence levels. This application can provide more reliable, comprehensive resource evaluation results with uncertain quantitative information, significantly improving the accuracy of resource estimation and decision support capabilities for concealed mineral deposits.

[0154] In other embodiments, after evaluating the resources of carbonate-type rare earth deposits in the target area based on a three-dimensional ore body model and obtaining the evaluation results, the method further includes:

[0155] Using the centroid projection point of the three-dimensional ore body model as the initial first priority drilling position, calculate the planar coordinates of the centroid, and mark the vertical projection point of the planar coordinates on the top surface of the three-dimensional ore body model as P0;

[0156] Extract the spatial gradient field of the predicted grade of each exploration unit in the three-dimensional ore body model, calculate the grade variation rate of each exploration unit in the east-west, north-south and vertical directions, and generate a grade gradient vector map.

[0157] Sort the units by grade gradient modulus from high to low, select the N units with the largest modulus, and use the spatial cluster center as the projection position of the high grade gradient region, marked as P1, P2...Pn; where N is positively correlated with the area of ​​the target region.

[0158] Extract the distribution boundary of the first probability mineralization unit in the unit classification, and combine it with the fault structure surface in the three-dimensional geological model to calculate the tectonic ore-controlling gradient zone. The tectonic ore-controlling gradient zone is defined as the buffer zone of the intersection line between the boundary of the high probability mineralization unit and the fault surface.

[0159] Project equally spaced sampling points along the center line of the buffer zone onto the ground surface and label them as Q1, Q2…Qm;

[0160] The initial borehole location set {P0, P1…Pn, Q1…Qm} is deduplicated and optimized by sorting: the Euclidean distance between any two points is calculated, points whose distance is less than the preset minimum borehole spacing are deleted, and points with large grade gradient modulus or falling into the structural ore-controlling gradient zone are retained first.

[0161] Starting from P0, the minimum spanning tree algorithm is used to connect the remaining points to generate a shortest path covering all recommended borehole locations, thus obtaining the drilling path.

[0162] The design depth is dynamically calculated based on the grade at each recommended borehole location;

[0163] A drilling layout suggestion map is generated based on the design depth, borehole number, construction sequence, and expected mineralization depth range.

[0164] In this embodiment, when generating the drilling layout proposal map, the three-dimensional ore body model can first be discretized into volumetric elements. The weighted average of the coordinates of all volumetric elements can be calculated as the centroid, and then it can be vertically projected onto the surface. Subsequently, three-dimensional difference operations can be performed on the grade data in the three-dimensional ore body model to calculate the grade difference between each exploration unit and its adjacent units in the X, Y, and Z directions, thereby obtaining the gradient components. Alternatively, the finite element method or finite difference method can be used to fit a local grade function at each grid point, and then the partial derivative of the function can be obtained to obtain the gradient.

[0165] Next, the grade gradient modulus (i.e., the square root of the sum of the squares of the gradient components in each direction) of each exploration unit is calculated, and then the units are sorted in descending order, selecting the top N units. K-means clustering is then performed on these N units, with the cluster centers designated as P1...Pn. N can be determined based on the target area and a preset borehole density per unit area to ensure reasonable coverage.

[0166] Then, using GIS or 3D geological modeling software, the boundary line of the first probabilistic mineralization unit is extracted and spatially superimposed with the fracture structure surface in the 3D geological model to find the intersection line. Then, with the intersection line as the center, a preset distance is extended to both sides to form a buffer zone.

[0167] Furthermore, equidistant sampling can be performed along the center line of the buffer zone at preset sampling intervals to obtain the three-dimensional coordinates of the sampling points, and then these points can be vertically projected onto the ground surface.

[0168] Subsequently, all point pairs can be traversed to perform deduplication and optimization sorting on the initial borehole location set {P0, P1…Pn, Q1…Qm}: calculate the Euclidean distance between any two points, delete points whose distance is less than the preset minimum borehole spacing, and prioritize keeping points with large grade gradient modulus or those falling within the structural ore-controlling gradient zone; if the distance is less than the minimum borehole spacing, compare the priorities of the two points (grade gradient modulus and whether they are within the structural ore-controlling zone), keep the point with higher priority, and delete the point with lower priority.

[0169] Based on this, starting from P0, the minimum spanning tree algorithm is used to connect the remaining points to generate a shortest path covering all recommended borehole locations, thus obtaining the drilling path. The remaining borehole locations are regarded as nodes of the graph, and the Euclidean distance between nodes is used as the edge weight. Then, Prim's algorithm or Kruskal's algorithm is applied to construct the minimum spanning tree.

[0170] Next, based on the grade distribution predicted in the three-dimensional ore body model at the borehole location, combined with the preset minimum industrial grade boundary and ore body thickness, the drilling depth is determined, such as drilling through the entire ore body or reaching the preset grade cutoff depth.

[0171] Finally, GIS or professional geological software can be used to overlay all optimized borehole locations, numbers, design depths, construction sequences, and other information onto a surface topographic map or 3D model, and mark the expected mineral depth range to generate a drilling layout suggestion map.

[0172] This application uses the ore body's center of gravity projection point P0 as the initial reference point, then extracts the grade spatial gradient field and identifies high gradient regions P1-Pn. It combines the first probability ore-forming unit boundary and fault structure surface to calculate the structural ore-controlling gradient zone Q1-Qm, integrating the control effect of geological structural factors on ore body distribution. It deduplicates and optimizes the ranking of all recommended borehole locations, then uses the minimum spanning tree algorithm to plan the shortest drilling path. Based on the grade of each recommended borehole location, it dynamically calculates the design depth, ensuring precise matching between drilling depth and ore body distribution. The resulting drilling layout suggestion map integrates all optimized parameters, providing intuitive, systematic, and executable guidance for drilling operations. This significantly improves the efficiency and accuracy of drilling verification, reduces exploration costs, and enhances the accuracy of resource discovery.

[0173] Based on the method for exploring and evaluating carbonate rock-type rare earth deposits provided in the above embodiments, this application also provides specific implementation methods for the apparatus for exploring and evaluating carbonate rock-type rare earth deposits. Please refer to the following embodiments.

[0174] like Figure 2 As shown, the present invention also provides an exploration and evaluation device for carbonate rock-type rare earth deposits, comprising:

[0175] The acquisition module 201 is used to acquire multi-source exploration data of a target area; the target area includes multiple exploration units; the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data;

[0176] Extraction module 202 is used to extract features from the multi-source exploration data to obtain a multi-dimensional exploration feature set;

[0177] Input module 203 is used to input the multidimensional exploration feature set into the preset mineralization prediction model and output the unit classification of each exploration unit;

[0178] The modeling and positioning module 204 is used to perform three-dimensional geological modeling and spatial positioning of each exploration unit based on the unit classification, so as to obtain a three-dimensional ore body model.

[0179] Evaluation module 205 is used to evaluate the resources of carbonate-type rare earth deposits in the target area based on the three-dimensional ore body model and obtain evaluation results.

[0180] Figure 3 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0181] An electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0182] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0183] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 302 may include removable or non-removable (or fixed) media, or memory 302 may be non-volatile solid-state memory. Memory 302 may be internal or external to the integrated gateway disaster recovery device.

[0184] In one instance, memory 302 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0185] Memory 302 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method for exploring and evaluating carbonate-type rare earth deposits according to the first aspect of this disclosure.

[0186] The processor 301 reads and executes computer program instructions stored in the memory 302 to achieve... Figure 1 The embodiment shown illustrates a method for exploring and evaluating carbonate rock-type rare earth deposits.

[0187] In one example, the electronic device may also include a communication interface 303 and a bus 304. For example, Figure 3 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 304 and complete communication with each other.

[0188] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0189] Bus 304 includes hardware, software, or both, that couples components of an electronic device together. For example, and not as a limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0190] This electronic device can execute the carbonate rock-type rare earth deposit exploration and evaluation method described in the embodiments of this application, thereby achieving a combination of Figures 1-2 The method and apparatus for the exploration and evaluation of carbonate rock-type rare earth deposits are described.

[0191] Furthermore, in conjunction with the carbonate rock-type rare earth deposit exploration and evaluation methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the carbonate rock-type rare earth deposit exploration and evaluation methods in the above embodiments.

[0192] In an optional embodiment, in conjunction with the carbonate rock-type rare earth deposit exploration and evaluation method in the above embodiments, this application embodiment can provide a computer program product to implement it. The instructions in the computer program product are executed by the processor of an electronic device, enabling the electronic device to implement any of the carbonate rock-type rare earth deposit exploration and evaluation methods in the above embodiments.

[0193] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0194] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0195] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0196] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0197] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for exploring and evaluating carbonate rock-type rare earth deposits, characterized in that, include: S101: Acquire multi-source exploration data of the target area; the target area includes multiple exploration units; the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data; S102: Extract features from the multi-source exploration data to obtain a multi-dimensional exploration feature set; S103: Input the multidimensional exploration feature set into the preset mineralization prediction model and output the unit classification of each exploration unit; S104: Based on the unit classification, perform three-dimensional geological modeling and spatial positioning for each exploration unit to obtain a three-dimensional ore body model; S105: Based on the three-dimensional ore body model, a resource evaluation of the carbonate-type rare earth deposit in the target area is conducted to obtain the evaluation results.

2. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 1, characterized in that, S101 includes: The spectral remote sensing image data is sequentially subjected to radiometric calibration, atmospheric correction and mineral mapping to obtain preprocessed spectral remote sensing impact data. The geophysical spectral data is filtered, polarized, extended, and anomaly separated to generate preprocessed geophysical spectral data; wherein the preprocessed geophysical spectral data includes magnetic anomaly maps, gravity anomaly maps, and radioactive anomaly maps. Calculate the spatial correlation coefficient between the fluorine content and the rare earth element content in the geochemical element content data. If the spatial correlation coefficient is higher than a preset threshold, the fluorine content and its spatial coupling with rare earth and iron elements are determined as additional features. Based on the preprocessed spectral remote sensing impact data, the preprocessed geophysical spectral data, and / or the additional features, preprocessed multi-source exploration data is generated.

3. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 2, characterized in that, In step S103, the training process of the preset mineralization prediction model includes: A training sample set is obtained, which includes positive samples of the training multidimensional exploration identifier feature vectors of mineralized units and negative samples of the training multidimensional exploration identifier feature vectors of units without mineralization records. The training multidimensional exploration identifier feature vectors include the gradient value of the annular magnetic anomaly, the difference of the negative gravity anomaly, the thorium anomaly count rate in the gamma spectrum, the depth of the characteristic absorption peak of rare earth elements in the remote sensing spectrum, the spatial correlation coefficient of the content of fluorine and rare earth elements in the geochemical assemblage anomaly, and the characteristic ratio of geochemical elements. The training multidimensional exploration identifier feature vectors corresponding to the positive and negative samples are standardized to obtain standard training samples. The standard training samples are input into the initial mineralization prediction model. In the preset initial mineralization prediction model, the Bayesian optimization method is used with the area under the ROC curve or the F1 score as the optimization objective. The combination of hyperparameters is iteratively adjusted until the area under the ROC curve converges or the preset number of iterations is reached. The optimal hyperparameters are then determined, and the preset mineralization prediction model is obtained. Using a pre-defined mineralization prediction model, the contribution of each trained multidimensional exploration identifier feature vector to the model prediction results is calculated. Based on the aforementioned contribution level, core exploration identification features are determined; Based on the predicted probability distribution of the preset mineralization prediction model on the validation set, a precision-recall curve is plotted, and the probability value corresponding to the balance point of precision and recall is selected as a preset threshold. The preset threshold is used to classify the exploration unit.

4. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 3, characterized in that, S103 also includes: Isolated anomalies are filtered downgraded or upgraded based on the spatial neighborhood classification consistency of each exploration unit to obtain the filtering correction result; The predicted probability is adjusted by confidence weighting based on the spatial correlation coefficients of fluorine, rare earth, and iron elements to obtain the weighted adjustment result. Identify low-probability magnetic anomaly regions, reclassify the original low-probability units within them, and attach candidate labels for deep concealed mineral deposits to obtain anomaly discrimination results; The unit classification is corrected based on the filtering correction result, the weighted correction result, and the anomaly discrimination result to obtain the final unit classification; wherein, the unit classification includes a first probability mineralization unit, a second probability mineralization unit, and a third probability mineralization unit, and the first probability, the second probability, and the third probability decrease in sequence.

5. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 4, characterized in that, S104 includes: Using the spatial range of the first probabilistic mineralization unit as the constraint boundary, the three-dimensional seismic inversion wave impedance data volume and magnetotelluric sounding resistivity data volume of the target area are input, and an implicit modeling algorithm is used to construct a three-dimensional geological structure model including the top and bottom interfaces of the carbonate rock mass, the fracture structure surface, and the envelope surface of the alteration zone. The preset mineralization prediction model outputs the mineralization probability of each exploration unit based on the multidimensional exploration feature set. Using the mineralization probability value as a covariate, and combining borehole core hyperspectral logging, gamma logging and resistivity logging data, the three-dimensional spatial morphology and grade distribution of rare earth mineralization bodies are predicted within the three-dimensional geological structure model using Kriging or co-co-Kriging interpolation methods. Based on the aforementioned three-dimensional geological structure model and three-dimensional spatial morphology and grade distribution, the three-dimensional ore body model is constructed.

6. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 5, characterized in that, S105 includes: Extract the volume, average grade, thickness, and burial depth parameters of the three-dimensional ore body model, and call the pre-established grade-tonnage database of carbonate rock-type rare earth deposits for similarity matching to obtain the first resource quantity; A linear regression equation is established based on borehole data of the target area or typical mineral deposits in the area. The spatial coupling degree between the fluorine and iron anomalies in the geochemical element content data is used to estimate the second resource quantity. The first resource quantity and the second resource quantity are weighted and fused according to their respective confidence levels, wherein the confidence level is determined by the spatial correlation coefficient of fluorine-iron-rare earth and the matching sample size in the grade-tonnage database. Based on the weighted and fused resource quantity values ​​and their corresponding probability intervals, an evaluation result is generated that includes the resource quantity range, quality classification, and confidence level.

7. The method for exploring and evaluating carbonate rock-type rare earth deposits according to claim 6, characterized in that, Following S105, the following is also included: Using the centroid projection point of the three-dimensional ore body model as the initial first priority drilling position, calculate the planar coordinates of the centroid, and mark the vertical projection point of the planar coordinates on the top surface of the three-dimensional ore body model as P0. Extract the spatial gradient field of the predicted grade of each exploration unit in the three-dimensional ore body model, calculate the grade change rate of each exploration unit in the east-west, north-south and vertical directions, and generate a grade gradient vector map. Sort the units by grade gradient modulus from high to low, select the N units with the largest modulus, and use the spatial cluster center as the projection position of the high grade gradient region, marked as P1, P2...Pn; where N is positively correlated with the area of ​​the target region. Extract the distribution boundary of the first probability mineralization unit in the unit classification, and combine it with the fault structure surface in the three-dimensional geological model to calculate the structural ore-controlling gradient zone. The structural ore-controlling gradient zone is defined as the buffer zone of the intersection line between the boundary of the high probability mineralization unit and the fault surface. The equally spaced sampling points along the center line of the buffer zone are projected onto the ground surface and marked as Q1, Q2...Qm; The initial borehole location set {P0, P1…Pn, Q1…Qm} is deduplicated and optimized by sorting: the Euclidean distance between any two points is calculated, points whose distance is less than the preset minimum borehole spacing are deleted, and points with large grade gradient modulus or falling into the structural ore-controlling gradient zone are retained first. Starting from P0, the minimum spanning tree algorithm is used to connect the remaining points to generate a shortest path covering all recommended borehole locations, thus obtaining the drilling path. The design depth is dynamically calculated based on the grade at each recommended borehole location; A drilling layout suggestion map is generated based on the design depth, borehole number, construction sequence, and expected mineralization depth range.

8. A device for exploring and evaluating carbonate rock-type rare earth deposits, characterized in that, include: The acquisition module is used to acquire multi-source exploration data of a target area; the target area includes multiple exploration units; the multi-source exploration data includes spectral remote sensing image data, geophysical spectral data, and geochemical element content data; The extraction module is used to extract features from the multi-source exploration data to obtain a multi-dimensional exploration feature set; The input module is used to input the multidimensional exploration feature set into the preset mineralization prediction model and output the unit classification of each exploration unit; The modeling and positioning module is used to perform three-dimensional geological modeling and spatial positioning of each exploration unit based on the unit classification, so as to obtain a three-dimensional ore body model. The evaluation module is used to evaluate the resources of the carbonate-type rare earth deposits in the target area based on the three-dimensional ore body model, and obtain the evaluation results.

9. An electronic device comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, characterized in that, When the computer program is executed by the processor, it implements the steps in the method for exploring and evaluating carbonate rock-type rare earth deposits as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the method for exploring and evaluating carbonate rock-type rare earth deposits as described in any one of claims 1-7.

Images