Carbonate clay type lithium ore prediction method and system

Abstract

The invention provides a carbonate clay type lithium ore prediction method and system. The method comprises the following steps: firstly, acquiring multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of a target area; then, extracting and processing to obtain a metal estimation map, a lithium element concentration map, a carbonation alteration abnormal map and a geological factor density map; secondly, obtaining a sample data set and a prediction data set through normalization processing, meshing processing and data division, and performing positive and negative sample division and data enhancement on the sample data set; thirdly, training and optimizing the plurality of first lithium ore prediction models according to the enhanced sample data set, and determining an evaluation index value; and finally, based on the prediction data set, through the target lithium ore prediction model meeting the evaluation threshold, obtaining a carbonate lithium ore mineralization prediction result. According to the method, the characteristic expression capability of geological data is enhanced, and the carbonate clay type lithium ore can be effectively, efficiently and accurately predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium ore exploration, and in particular to a method and system for predicting carbonate clay-type lithium ore. Background Art

[0002] Carbonate clay-type lithium ore is an important lithium resource, widely used in industries such as batteries, ceramics, and glass. With the increasing demand for lithium resources, the exploration and prediction of carbonate clay-type lithium ore has become a research focus in the field of geological exploration.

[0003] Traditional methods for predicting the mineralization of carbonate clay-type lithium deposits typically employ statistical mineralization prediction methods, such as the weight-of-evidence method. Specifically, the weight-of-evidence method, based on Bayesian theory, quantitatively assesses the mineralization indices by calculating the conditional probability relationship between geological factors and known mineralization sites. This method is widely used in non-vegetated areas and can effectively utilize remote sensing, geological, and geochemical data to evaluate regional mineralization prospects.

[0004] However, when these prediction methods are applied to vegetated areas, the vegetation obscures geological features, making direct observation via conventional means difficult, increasing the difficulty of resource exploration. Furthermore, these prediction methods face challenges when dealing with complex surface environments: a single data source, insufficient model accuracy, and a scarcity of samples. This results in low efficiency and accuracy in predicting carbonate clay-type lithium deposits. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that it is impossible to efficiently and accurately predict carbonate clay-type lithium ore.

[0006] To solve the above technical problems, the present invention provides a method and system for predicting carbonate clay-type lithium ore, which specifically adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for predicting carbonate clay-type lithium deposits, comprising: first, obtaining multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data, and geological information data for a target area; the target area includes a sample area and a prediction area, and the sample area includes a carbonate clay-type lithium deposit area and a non-carbonate clay-type lithium deposit area. Then, a metal estimation map is determined based on the multispectral remote sensing image data, a lithium concentration map is determined based on the geochemical sampling data, a carbonation alteration anomaly map is determined based on the hyperspectral remote sensing image data, and a geological factor density map is determined based on the geological information data. Next, the metal estimation map, lithium concentration map, carbonation alteration anomaly map, and geological factor density map are normalized and gridded. Next, data partitioning is performed on the normalized and gridded metal estimation map, lithium concentration map, carbonation alteration anomaly map, and geological factor density map based on the sample area and the prediction area to obtain a sample dataset and a prediction dataset. Based on the carbonate clay type lithium ore area and the non-carbonate clay type lithium ore area, the sample data set is divided into positive and negative samples, and the sample data set after the positive and negative sample division is data enhanced to obtain an enhanced sample data set. Then, multiple first lithium ore prediction models are constructed, and the first lithium ore prediction model is a deep learning model for predicting the location of carbonate clay type lithium ore. According to the enhanced sample data set, the multiple first lithium ore prediction models are trained and optimized respectively, and the evaluation index values ​​corresponding to the multiple first lithium ore prediction models are determined. Finally, based on the prediction data set, the target carbonate lithium ore mineralization prediction result is obtained by the target lithium ore prediction model. Among them, the target lithium ore prediction model is the first lithium ore prediction model corresponding to the evaluation index value that meets the evaluation threshold.

[0008] This method first acquires multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data, and geological information data for the target area. It then extracts and processes the data to obtain metal estimation maps, lithium concentration maps, carbonate alteration anomaly maps, and geological factor density maps. Next, through normalization, gridding, and data partitioning, a sample dataset and a prediction dataset are obtained. The sample dataset is then partitioned into positive and negative samples and data augmented. Next, multiple first lithium ore prediction models are trained and optimized based on the augmented sample datasets, and evaluation index values ​​are determined. Finally, based on the prediction dataset, a target lithium ore prediction model that meets the evaluation threshold is used to obtain a target carbonate lithium ore mineralization prediction result. This method is capable of extracting multi-scale and spatially correlated features of geological data from complex geological environments. By extracting valuable information from the target area (e.g., vegetation cover) and stripping redundant information, and by integrating multiple lithium ore prediction models, the stability and generalization of the lithium ore prediction model are improved, the automated identification and classification of geological elements is achieved, and the feature expression capability of geological data is enhanced, thereby effectively and efficiently predicting carbonate clay-type lithium deposits.

[0009] In conjunction with the first aspect, in an optional implementation, when there are multiple target lithium ore prediction models, the method of obtaining a target lithium carbonate mineralization prediction result based on the prediction data set using the target lithium ore prediction model includes: first, according to the prediction data set, performing predictions using the multiple target lithium ore prediction models respectively to obtain multiple first lithium carbonate mineralization prediction results. Then, the multiple first lithium carbonate mineralization prediction results are integrated using an ensemble learning algorithm to obtain a target lithium carbonate mineralization prediction result.

[0010] In this implementation, when there are multiple target lithium ore prediction models, by integrating multiple first carbonate lithium ore mineralization prediction results, a target carbonate lithium ore mineralization prediction result with higher consistency and accuracy can be generated to further improve the accuracy of carbonate clay-type lithium ore prediction.

[0011] In conjunction with the first aspect, in an optional implementation, the determination of the metal estimation map based on the multispectral remote sensing image data includes: first, performing a first preprocessing on the multispectral remote sensing image data, the first preprocessing including: a first radiation calibration process, a first atmospheric correction process, and a band fusion preprocessing, to obtain the multispectral remote sensing image data after the first preprocessing. Then, based on the multispectral remote sensing image data after the first preprocessing, determining the metal stress vegetation index, the metal stress vegetation index being: a vegetation index considering green wave and shortwave infrared (Vegetation Index considering green wave and shortwave infrared)

[0012] Greenness and Shortwave infrared (VIGS), the expression of VIGS is:

[0013]

[0014] Where G represents the spectral data of the visible green band, S1 represents the spectral data of the first shortwave infrared band, S2 represents the spectral data of the second shortwave infrared band, N represents the spectral data of the near-infrared band, R represents the reflectance of the red band, w1 represents the first preset weight, w2 represents the second preset weight, w3 represents the third preset weight, and w4 represents the fourth preset weight. Finally, a metal assessment map is generated based on the Metal Stress Vegetation Index.

[0015] In this implementation, the metal stress vegetation index extracted based on multispectral remote sensing image data can accurately determine the metal assessment map to serve as the basis for lithium ore prediction.

[0016] In conjunction with the first aspect, in one optional implementation, determining a lithium concentration map based on geochemical sampling data includes: first, extracting lithium concentration values ​​and associated element concentration values ​​based on the geochemical sampling data; the associated elements include gallium, sodium, and calcium. Then, generating a lithium concentration map using a kriging interpolation method based on the lithium concentration values ​​and the associated element concentration values.

[0017] In this implementation, a lithium concentration map can be effectively determined through interpolation based on geochemical sampling data. This lithium concentration map can be used to characterize the spatial concentration distribution of lithium and associated elements in the target area. Therefore, the lithium concentration map provides an accurate geochemical basis for locating the core area of ​​​​the mineralized area of ​​​​the lithium deposit in the carbonate clay type.

[0018] In combination with the first aspect, in an optional implementation method, the above-mentioned determination of the carbonation alteration anomaly map based on the hyperspectral remote sensing image data includes: first, performing a second preprocessing on the hyperspectral remote sensing image data, the second preprocessing including: spectral band merging processing, bad band removal processing, second radiation calibration processing, second atmospheric correction processing and orthorectification processing, to obtain the hyperspectral remote sensing image data after the second preprocessing. Then, the hyperspectral remote sensing image data after the second preprocessing is subjected to minimum noise separation processing and spectral angle matching processing to obtain matched hyperspectral remote sensing image data. Next, the matched hyperspectral remote sensing image data is respectively subjected to principal component analysis and independent component analysis to determine the carbonation alteration anomaly information. Finally, a carbonation alteration anomaly map is generated based on the carbonation alteration anomaly information.

[0019] In this implementation, a carbonation alteration anomaly map can be effectively determined using principal component analysis and independent component analysis based on hyperspectral remote sensing image data. This map can indicate surface alteration zones associated with lithium mineralization, assisting in carbonate lithium mineralization prediction and thus improving the accuracy of carbonate lithium mineralization prediction.

[0020] In combination with the first aspect, in an optional implementation, determining the geological factor density map based on the geological information data includes: generating the geological factor density map through a buffer distance analysis method based on the geological information data.

[0021] In this implementation, a geological factor density map can be effectively determined using buffered distance analysis based on geological information data. This map reflects the control of geological structure on lithium ore occurrence and can therefore effectively assist in the prediction of carbonate lithium mineralization.

[0022] In combination with the first aspect, in an optional implementation method, the above-mentioned multiple first lithium ore prediction models are multiple of the following models: AlexNet network model, LeNet network model, MobileNet network model, ResNet network model, VGGNet network model and Vi T model.

[0023] In combination with the first aspect, in an optional implementation method, the above-mentioned evaluation index value includes one or more of the following indicators: confusion matrix, precision, accuracy, recall rate, F1 score value, ROC receiver operating characteristic curve, and area under the AUC curve value.

[0024] In this implementation, the prediction capability and prediction effect of the first lithium ore prediction model can be effectively and comprehensively evaluated through the above-mentioned evaluation index values, so as to accurately determine the target lithium ore prediction model.

[0025] In combination with the first aspect, in an optional implementation, the above-mentioned ensemble learning algorithm is: averaging method, or voting method.

[0026] In this implementation, the multiple first carbonate lithium mineralization prediction results can be effectively integrated through the averaging method or the voting method to obtain the target carbonate lithium mineralization prediction result, thereby further improving the accuracy of carbonate clay-type lithium ore prediction.

[0027] In a second aspect, the present invention provides a carbonate clay-type lithium deposit prediction method system, comprising: an acquisition module, an extraction module, a preprocessing module, a data partitioning module, a sample construction module, a model construction module, a training module, and a prediction module. The acquisition module can be used to acquire multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data, and geological information data for a target area; the target area includes a sample area and a prediction area, and the sample area includes a carbonate clay-type lithium deposit area and a non-carbonate clay-type lithium deposit area. The extraction module can be used to determine a metal estimation map based on the multispectral remote sensing image data, a lithium concentration map based on the geochemical sampling data, a carbonate alteration anomaly map based on the hyperspectral remote sensing image data, and a geological factor density map based on the geological information data. The preprocessing module can be used to normalize and grid the metal estimation map, lithium concentration map, carbonate alteration anomaly map, and geological factor density map. The data partitioning module can be used to partition the normalized and gridded metal assessment maps, lithium concentration maps, carbonate alteration anomaly maps, and geological factor density maps based on the sample area and the prediction area, thereby obtaining a sample dataset and a prediction dataset. The sample construction module can be used to partition the sample dataset into positive and negative samples based on the carbonate clay-type lithium ore area and the non-carbonate clay-type lithium ore area, and perform data enhancement on the sample dataset after the positive and negative sample partitioning to obtain an enhanced sample dataset. The model construction module can be used to construct multiple first lithium ore prediction models, each of which is a deep learning model for predicting the location of carbonate clay-type lithium ore. The training module can be used to train and optimize the multiple first lithium ore prediction models based on the enhanced sample dataset, and determine the evaluation index values ​​corresponding to each of the multiple first lithium ore prediction models. The prediction module can be used to obtain a target carbonate lithium ore mineralization prediction result using a target lithium ore prediction model based on the prediction dataset, wherein the target lithium ore prediction model is a first lithium ore prediction model corresponding to an evaluation index value that meets an evaluation threshold.

[0028] In a third aspect, the present invention provides an electronic device comprising: a memory, one or more processors; the memory is coupled to the processor; wherein the memory stores computer program code, the computer program code comprises computer instructions, and when the computer instructions are executed by the processor, the electronic device executes the method provided by the above-mentioned first aspect and any optional implementation thereof.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium comprising computer instructions, which, when executed on an electronic device, enables the electronic device to execute the method provided in the first aspect and any optional implementation thereof.

[0030] It can be understood that the beneficial effects that can be achieved by the carbonate clay-type lithium ore prediction method system provided by the second aspect, the electronic device of the third aspect, and the computer-readable storage medium of the fourth aspect can be referred to the beneficial effects of the first aspect and any possible design thereof, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the principle of the method for predicting carbonate clay-type lithium ore provided in the embodiment of the present application;

[0032] Figure 2 A schematic flow chart of a method for predicting carbonate clay-type lithium ore provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of the principle of determining a carbonation alteration anomaly map provided in an embodiment of the present application;

[0034] Figure 4 Schematic diagram of the structure of the carbonate clay-type lithium ore prediction system provided in the embodiment of the present application. DETAILED DESCRIPTION

[0035] The following embodiments are described in detail, with examples illustrated in the accompanying drawings. When the following description refers to the drawings, identical numbers in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following embodiments are not intended to represent all possible implementations consistent with the present application. They are merely examples of systems and methods consistent with certain aspects of the present application, as detailed in the claims.

[0036] Carbonate clay-type lithium ore is an important lithium resource, widely used in industries such as batteries, ceramics, and glass. With the increasing demand for lithium resources, the exploration and prediction of carbonate clay-type lithium ore has become a research focus in the field of geological exploration.

[0037] Traditional methods for predicting the mineralization of carbonate clay-type lithium deposits typically employ statistical mineralization prediction methods, such as the weight-of-evidence method. Specifically, the weight-of-evidence method, based on Bayesian theory, quantitatively assesses the mineralization indices by calculating the conditional probability relationship between geological factors and known mineralization sites. This method is widely used in non-vegetated areas and can effectively utilize remote sensing, geological, and geochemical data to evaluate regional mineralization prospects.

[0038] However, when these prediction methods are applied to vegetated areas, the vegetation obscures geological features, making direct observation via conventional means difficult, increasing the difficulty of resource exploration. Furthermore, these prediction methods face challenges when dealing with complex surface environments: a single data source, insufficient model accuracy, and a scarcity of samples. This results in low efficiency and accuracy in predicting carbonate clay-type lithium deposits.

[0039] In order to solve the above problems, the present application provides a method and system for predicting carbonate clay type lithium ore, which can be applied to the prediction of carbonate clay type lithium ore in vegetation covered areas. Specifically, Figure 1 The schematic diagram of the principle of the carbonate clay type lithium ore prediction method provided in the embodiment of the present application is as follows: Figure 1 As shown, the method first obtains multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of the target area. Then, metal estimation maps, lithium element concentration maps, carbonation alteration anomaly maps and geological factor density maps are extracted and processed. Secondly, the metal estimation maps, lithium element concentration maps, carbonation alteration anomaly maps and geological factor density maps are normalized, gridded and data divided in turn to obtain sample data sets and prediction data sets, and the sample data sets are divided into positive and negative samples and data enhanced. Next, multiple first lithium ore prediction models are trained and optimized based on the enhanced sample data sets, and evaluation index values ​​are determined. Finally, based on the prediction data set, the target lithium ore prediction model that meets the evaluation threshold is used to obtain the target carbonate lithium ore mineralization prediction result. This method can extract the multi-scale characteristics and spatial correlation characteristics of geological data from complex geological environments, realize the automatic identification and classification of geological elements, and enhance the feature expression ability of geological data. By extracting valuable information and stripping off redundant information from the target area (for example, vegetation coverage area), and predicting through a lithium ore prediction model based on deep learning, efficient and accurate prediction of carbonate clay-type lithium ore can be achieved.

[0040] The following describes the solution provided by the embodiments of the present application in conjunction with the accompanying drawings.

[0041] Specifically, Figure 2 A flow chart of the method for predicting carbonate clay-type lithium ore provided in the embodiment of the present application is shown in FIG. Figure 2 As shown, the carbonate clay type lithium ore prediction method provided in the embodiment of the present application includes the following steps S101-S108:

[0042] S101. Acquire multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data, and geological information data of a target area.

[0043] In an embodiment of the present application, the target area includes: a sample area and a prediction area. Among them, the sample area is an area where the location distribution of carbonate clay-type lithium mines is known, that is, the sample area includes: a carbonate clay-type lithium mine area and a non-carbonate clay-type lithium mine area. The multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data corresponding to the sample area can be used to train and optimize the lithium mine prediction model, so as to determine the target lithium mine prediction model that meets the evaluation threshold after training and optimization. The prediction area is an area where the location distribution of carbonate clay-type lithium mines is unknown, that is, the area to be predicted by the target lithium mine prediction model.

[0044] Specifically, multispectral remote sensing image data can be used to characterize the reflection and absorption characteristics of objects in the target area for multiple different bands (for example, 4-15 discontinuous spectral bands, such as visible light, near infrared, short-wave infrared, etc.). Geochemical sampling data can be used to characterize the soil, rocks, water bodies, sediments and other media on the surface sampled at regular intervals in the target area and perform chemical analysis to obtain the content of elements or compounds. Hyperspectral remote sensing image data is used to characterize the reflection and absorption characteristics of objects in the target area for multiple different bands (for example, hundreds of continuous narrow bands). Geological information data can be used to characterize geological information in the target area (for example, faults, lithologic boundaries, stratigraphic occurrence, tectonic units, etc.).

[0045] For example, the multispectral remote sensing image data may be Landsat8 OL I multispectral data, the hyperspectral remote sensing image may be GF-5AHS I hyperspectral data, the geochemical sampling data may be 1:200,000 soil geochemical data, and the geological information data may include: brittle-ductile shear zone information, structural fracture information, lithology information and other data.

[0046] S102. Determine a metal assessment map based on multispectral remote sensing image data, determine a lithium element concentration map based on geochemical sampling data, determine a carbonate alteration anomaly map based on hyperspectral remote sensing image data, and determine a geological factor density map based on geological information data.

[0047] Furthermore, feature information extraction is performed based on the multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data obtained in S101 to realize the identification and classification of geological elements and enhance the feature expression ability of geological data, so as to improve the accuracy of carbonate clay-type lithium deposit prediction.

[0048] In some embodiments, the metal assessment map can be determined based on the metal stress vegetation index extracted from multispectral remote sensing image data. Specifically, determining the metal assessment map based on the multispectral remote sensing image data includes:

[0049] First, a first preprocessing is performed on the multispectral remote sensing image data to obtain multispectral remote sensing image data after the first preprocessing, wherein the first preprocessing includes: a first radiometric calibration process, a first atmospheric correction process, and a band fusion preprocessing.

[0050] Specifically, the first radiometric calibration process converts the raw digital values ​​(DN values) in multispectral remote sensing image data into physically meaningful radiometric brightness (e.g., radiometric brightness values). This eliminates sensor errors (e.g., sensitivity drift, nonlinear response), ensuring the reliability of multispectral remote sensing image data. It also enables comparability of data across time, sensors, or platforms, supporting quantitative remote sensing applications (e.g., reflectivity inversion and temperature calculation).

[0051] The first atmospheric correction processing can eliminate the influence of atmospheric scattering (Rayleigh scattering, aerosol scattering) and absorption (such as water vapor, ozone) in multispectral remote sensing image data to obtain the true surface reflectance.

[0052] Band fusion preprocessing can fuse multi-source remote sensing images with different spatial resolutions or spectral characteristics to generate multispectral remote sensing image data with both high spatial resolution and high spectral information. This can enhance spatial detail and feature expression, combining the strengths of different sensors (such as thermal infrared and visible light) to assist in object classification or target recognition.

[0053] Then, the metal stress vegetation index is determined based on the multispectral remote sensing image data after the first preprocessing. In one implementation, the metal stress vegetation index is a vegetation index (VIGS) that takes into account green wave and short-wave infrared rays. The expression of VIGS is:

[0054]

[0055] Among them, G represents the spectral data of the green band of visible light, S1 represents the spectral data of the first short-wave infrared band, S2 represents the spectral data of the second short-wave infrared band, N represents the spectral data of the near-infrared band, R represents the reflectivity of the red band, w1 represents the first preset weight, w2 represents the second preset weight, w3 represents the third preset weight, and w4 represents the fourth preset weight.

[0056] Finally, a metal assessment map is generated based on the Metal Stress Vegetation Index. This map can be used to identify areas within the target region where vegetation is experiencing abnormal stress from underground metal elements (lithium and associated elements). This map can serve as a basis for lithium deposit prediction, reflecting the metal enrichment signals associated with carbonate clay-type lithium deposits underlying vegetation-covered areas.

[0057] In some embodiments, the lithium concentration map can be determined based on geochemical sampling data by interpolation. Specifically, determining the lithium concentration map based on geochemical sampling data includes:

[0058] First, based on geochemical sampling data, the concentration values ​​of lithium and associated elements are extracted, including gallium, sodium, and calcium.

[0059] Then, based on the lithium and associated element concentrations, a lithium concentration map is generated using Kriging interpolation. This map can be used to characterize the spatial concentration distribution of lithium and associated elements within the target area. This map provides a geochemical basis for locating the core mineralization area of ​​volcanic clay-type lithium deposits.

[0060] The principle of the above-mentioned Kriging interpolation method is an advanced geostatistical process that generates an estimated surface from a set of scattered points (i.e., sampling points) with measured values ​​(i.e., sampling data of sampling points in geochemical sampling data). It is assumed that the distance or direction between sampling points can reflect the spatial correlation that can be used to explain surface changes. A mathematical function can be fitted to a specified number of points or all points within a specified radius to determine the output value for each location. Kriging interpolation specifically includes: exploratory statistical analysis of data, variogram modeling, and surface creation, as well as the study of variance surfaces. The expression of Kriging interpolation is:

[0061]

[0062] in, represents the predicted interpolation value of the predicted position s0, Z(s i ) means s i The measured value at position, λ i Indicates s i The unknown weight of the measurement at the location, where N is the number of measurements.

[0063] In some embodiments, the carbonation alteration anomaly map can be determined based on hyperspectral remote sensing image data through principal component analysis and independent component analysis. Specifically, Figure 3 The schematic diagram of the principle of determining the carbonation alteration anomaly map provided in the embodiment of this application is as follows: Figure 3 As shown in Figure 2, the carbonation alteration anomaly map is determined based on hyperspectral remote sensing image data, including:

[0064] First, the hyperspectral remote sensing image data is subjected to a second preprocessing to obtain the hyperspectral remote sensing image data after the second preprocessing. The second preprocessing includes: spectral band merging processing, bad band removal processing, second radiometric calibration processing, second atmospheric correction processing, and orthorectification processing.

[0065] Specifically, spectral band merging can fuse multiple spectral bands (such as multispectral and panchromatic bands, or multitemporal data) acquired by different sensors or at different times into a unified dataset, preserving or enhancing specific information. Bad band removal can be used to remove invalid bands from hyperspectral remote sensing image data due to sensor noise, atmospheric absorption (such as water vapor and carbon dioxide absorption bands), or hardware failure. The methods and functions of the second radiometric calibration and second atmospheric correction processes are similar to those of the first radiometric calibration and first atmospheric correction processes described above and are not further described here. Orthorectification can eliminate geometric distortion caused by terrain undulations and sensor attitude (such as tilt and altitude variations) to generate images with unified geographic coordinates and vertical projection. This improves the geometric accuracy of hyperspectral remote sensing image data, making it more consistent with true surface coordinates to support accurate multi-source data registration (such as overlay with geological maps and DEMs). It can also eliminate terrain shadows in hyperspectral remote sensing images.

[0066] Then, the hyperspectral remote sensing image data after the second preprocessing is subjected to minimum noise separation processing and spectral angle matching processing to obtain matched hyperspectral remote sensing image data.

[0067] Minimum Noise Fractionation (MNF) processing can denoise, reduce dimensionality, and compress hyperspectral remote sensing image data after the second preprocessing step. Specifically, it can separate noise and signal in the data through orthogonal transformation, prioritizing components with high signal-to-noise ratios. Furthermore, it can reduce data redundancy, providing more concise and effective input for subsequent classification or target detection.

[0068] Spectral Angle Matching (SAM) processing can achieve spectral similarity measurement and classification decisions. That is, by calculating the "angle" difference between the unknown spectrum and the reference spectrum, it is determined whether the two belong to the same ground feature category, which can be used for mineral identification, vegetation type distinction, etc.

[0069] Next, the matched hyperspectral remote sensing image data were used to extract carbonation alteration anomaly information by principal component analysis (PCA) and independent component analysis (ICA).

[0070] Finally, a carbonation alteration anomaly map is generated based on the carbonation alteration anomaly information. This carbonation alteration anomaly map can be used to characterize the spatial distribution and intensity of surface carbonate mineral alteration (such as calcite and dolomite) in the target area. This carbonation alteration anomaly map can indicate surface alteration zones associated with lithium mineralization, assisting in the prediction of carbonate lithium mineralization.

[0071] In some embodiments, a geological factor density map can be obtained based on analysis of geological information data. Specifically, determining a geological factor density map based on geological information data includes: generating a geological factor density map based on the geological information data using a buffer distance analysis method. The geological factor density map can be used to characterize the quantitative contribution of geological structures (such as faults and lithologic contact zones) in the target area to the control of mineralization, reflecting the control of geological structures on the occurrence of lithium ore, and can assist in the prediction of carbonate lithium ore mineralization.

[0072] S103. Normalize and grid the metal assessment map, lithium element concentration map, carbonation alteration anomaly map, and geological factor density map.

[0073] Next, the metal assessment map, lithium element concentration map, carbonation alteration anomaly map and geological factor density map determined in S102 are normalized and gridded. This converts the above data into standardized, spatially grid-consistent image data, providing a reliable data basis for the input of the lithium ore prediction model and facilitating the training, optimization and prediction processing of the lithium ore prediction model.

[0074] S104. Based on the sample area and the prediction area, the metal estimation map, lithium element concentration map, carbonation alteration anomaly map, and geological factor density map after normalization and gridding are divided to obtain a sample data set and a prediction data set.

[0075] Furthermore, based on the location division of the sample area and the prediction area in the target area, the processing results of S103 are divided into data. The sample data set includes: a metal estimation map, a lithium element concentration map, a carbonation alteration anomaly map, and a geological factor density map corresponding to the sample area after normalization and gridding. The prediction data set includes: a metal estimation map, a lithium element concentration map, a carbonation alteration anomaly map, and a geological factor density map corresponding to the prediction area after normalization and gridding. The sample data set can be used for subsequent training and optimization of the lithium ore prediction model, and the prediction data set is used to predict carbonate clay-type lithium ore based on the target lithium ore prediction model.

[0076] S105. Based on the carbonate clay type lithium ore region and the non-carbonate clay type lithium ore region, the sample data set is divided into positive and negative samples, and the sample data set after the positive and negative sample division is enhanced to obtain an enhanced sample data set.

[0077] Specifically, to facilitate training and optimization of the first lithium ore prediction model and evaluate its predictive capabilities, the sample dataset must first be divided into positive and negative samples. Positive sample data corresponds to carbonate clay-type lithium ore regions, while negative sample data corresponds to non-carbonate clay-type lithium ore regions.

[0078] In one implementation, a carbonate clay-type lithium ore distribution map of the sample area can be obtained. The carbonate clay-type lithium ore distribution map is used to describe the location distribution of carbonate clay-type lithium ore, that is, to characterize carbonate clay-type lithium ore areas and non-carbonate clay-type lithium ore areas. Based on the carbonate clay-type lithium ore distribution map, the sample data set can be divided into positive and negative samples. In addition, the carbonate clay-type lithium ore distribution map can also be used as positive sample data and negative sample data, respectively, to serve as the true value for training and optimizing the first lithium ore prediction model, so as to accurately evaluate the prediction ability based on the prediction results of the first lithium ore prediction model.

[0079] Furthermore, data enhancement is performed on the sample data set after the positive and negative samples are divided to obtain an enhanced sample data set, thereby increasing the number of samples. Exemplarily, data enhancement methods may include rotation, cropping, mirroring, and adding noise.

[0080] S106. Construct multiple first lithium ore prediction models.

[0081] In an embodiment of the present application, a deep learning model based on a convolutional neural network and a visual self-attention mechanism is used to construct multiple first lithium ore prediction models. Specifically, the first lithium ore prediction model is a deep learning model for predicting the location of carbonate clay-type lithium ore. Thus, by training and optimizing the multiple first lithium ore prediction models, a target lithium ore prediction model with strong predictive ability and good results is selected based on the evaluation index values, and ultimately used to predict carbonate clay-type lithium ore in the prediction area.

[0082] In some embodiments, the above-mentioned multiple first lithium ore prediction models are multiple of the following models: AlexNet network model, LeNet network model, lightweight mobile network (MobileNet) model, residual network (ResNet) model, VGG network (VGGNet) model and Vi T (Visi on Transformer) model.

[0083] S107. Training and optimizing the plurality of first lithium ore prediction models respectively according to the enhanced sample data set, and determining evaluation index values ​​corresponding to the plurality of first lithium ore prediction models respectively.

[0084] Specifically, the enhanced sample data set in S105 is used to train and optimize the multiple first lithium ore prediction models constructed in S106, and the prediction results of the first lithium ore prediction models are evaluated according to preset evaluation indicators, and the evaluation indicator values ​​corresponding to the multiple first lithium ore prediction models are determined to be used for screening the multiple first lithium ore prediction models and determining the target lithium ore prediction model that is ultimately used to predict carbonate clay-type lithium ore.

[0085] In some embodiments, the evaluation index value includes one or more of the following indicators: confusion matrix, precision, accuracy, recall rate, F1 score value, ROC receiver operating characteristic curve, and area under the AUC curve value.

[0086] In some embodiments, the evaluation index value includes one or more of the following indicators: average precision, average accuracy, average recall, and average F1 score.

[0087] In this way, the prediction ability and prediction effect of the first lithium ore prediction model can be effectively and comprehensively evaluated through the above-mentioned evaluation index values, so as to accurately determine the target lithium ore prediction model.

[0088] S108. Obtain target carbonate lithium ore mineralization prediction results through a target lithium ore prediction model based on the prediction data set.

[0089] Finally, based on the evaluation index values ​​corresponding to the multiple first lithium ore prediction models determined in S107, the model is screened using the evaluation threshold to determine a target lithium ore prediction model. The target lithium ore prediction model is the first lithium ore prediction model corresponding to the evaluation index value that meets the evaluation threshold. The evaluation threshold is preset based on actual application requirements and is not specifically limited in this application.

[0090] Next, the prediction data set divided in S104 can be input into the target lithium ore prediction model for prediction, and the target carbonate lithium ore mineralization prediction result can be output. The target carbonate lithium ore mineralization prediction result can be used to characterize the location distribution of carbonate clay-type lithium ore in the prediction area.

[0091] In some embodiments, there may be multiple evaluation index values ​​of the first lithium ore prediction model that meet the evaluation threshold. That is, when there are multiple target lithium ore prediction models, S108, based on the prediction data set, obtains the target carbonate lithium ore mineralization prediction result through the target lithium ore prediction model, specifically including:

[0092] First, based on the prediction data set, predictions were made using multiple target lithium ore prediction models to obtain multiple first carbonate lithium ore mineralization prediction results.

[0093] Then, multiple first carbonate lithium ore mineralization prediction results are integrated through an ensemble learning algorithm to obtain a target carbonate lithium ore mineralization prediction result.

[0094] In this way, by integrating multiple first carbonate lithium mineralization prediction results, a target carbonate lithium mineralization prediction result with higher consistency and accuracy can be generated to further improve the accuracy of carbonate clay-type lithium ore prediction.

[0095] In some embodiments, the ensemble learning algorithm is an averaging method or a voting method. Thus, the averaging method or the voting method can effectively integrate the multiple first lithium carbonate mineralization prediction results to obtain the target lithium carbonate mineralization prediction result.

[0096] Using the carbonate clay-type lithium ore prediction method provided in the embodiment of the present application, first, multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of the target area are obtained. Then, metal assessment maps, lithium element concentration maps, carbonation alteration anomaly maps and geological factor density maps are extracted and processed. Secondly, a sample data set and a prediction data set are obtained through normalization processing, gridding processing and data partitioning, and the sample data set is divided into positive and negative samples and data enhancement is performed. Next, multiple first lithium ore prediction models are trained and optimized respectively according to the enhanced sample data set, and the evaluation index value is determined. Finally, based on the prediction data set, a target lithium ore prediction model that meets the evaluation threshold is used to obtain the target carbonate lithium ore mineralization prediction result. This method can extract multi-scale characteristics and spatial correlation characteristics of geological data from a complex geological environment. By extracting valuable information and stripping off redundant information from target areas (e.g., vegetation-covered areas), and integrating multiple lithium ore prediction models, the stability and generalization capabilities of the lithium ore prediction model are improved, the automatic identification and classification of geological elements are realized, and the feature expression capabilities of geological data are enhanced, thereby effectively and accurately predicting carbonate clay-type lithium deposits.

[0097] The present application also provides a carbonate clay type lithium ore prediction system. Specifically, Figure 4 This is a schematic diagram of the structure of the carbonate clay type lithium ore prediction system provided in the embodiment of the present application, as shown in FIG. Figure 4 As shown, the carbonate clay-type lithium ore prediction system 400 includes: an acquisition module 401, an extraction module 402, a preprocessing module 403, a data partitioning module 404, a sample construction module 405, a model construction module 406, a training module 407 and a prediction module 408.

[0098] The acquisition module 401 can be used to acquire multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data, and geological information data of the target area. The target area includes: a sample area and a prediction area. The sample area includes: a carbonate clay-type lithium ore area and a non-carbonate clay-type lithium ore area.

[0099] The extraction module 402 can be used to determine a metal assessment map based on multispectral remote sensing image data, determine a lithium element concentration map based on geochemical sampling data; determine a carbonation alteration anomaly map based on hyperspectral remote sensing image data; and determine a geological factor density map based on geological information data.

[0100] The pre-processing module 403 can be used to perform normalization and gridding processing on the metal assessment map, lithium element concentration map, carbonation alteration anomaly map and geological factor density map.

[0101] The data partitioning module 404 can be used to partition the metal assessment map, lithium concentration map, carbonation alteration anomaly map and geological factor density map after normalization and gridding based on the sample area and the prediction area to obtain a sample data set and a prediction data set.

[0102] The sample construction module 405 can be used to divide the sample data set into positive and negative samples based on the carbonate clay type lithium mineral area and the non-carbonate clay type lithium mineral area, and perform data enhancement on the sample data set after the positive and negative sample division to obtain an enhanced sample data set.

[0103] The model building module 406 can be used to build multiple first lithium ore prediction models, where the first lithium ore prediction model is a deep learning model for predicting the location of carbonate clay-type lithium ore.

[0104] The training module 407 can be used to train and optimize the multiple first lithium ore prediction models according to the enhanced sample data set, and determine the evaluation index values ​​corresponding to the multiple first lithium ore prediction models.

[0105] The prediction module 408 can be used to obtain a target carbonate lithium ore mineralization prediction result through a target lithium ore prediction model based on the prediction data set, where the target lithium ore prediction model is a first lithium ore prediction model corresponding to an evaluation index value that meets an evaluation threshold.

[0106] Using the carbonate clay type lithium ore prediction system provided by the embodiment of the present application, first, the multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of the target area are acquired through the acquisition module. Then, the metal assessment map, lithium element concentration map, carbonate alteration anomaly map and geological factor density map are extracted and processed by the extraction module. Secondly, normalization processing, gridding processing and data partitioning are performed through the preprocessing module and the data partitioning module to obtain a sample data set and a prediction data set, and the sample data set is divided into positive and negative samples and data enhancement is performed through the sample construction module. Next, the training module is used to train and optimize the multiple first lithium ore prediction models according to the enhanced sample data set, and the evaluation index value is determined. Finally, the prediction module obtains the target carbonate lithium ore mineralization prediction result based on the prediction data set through the target lithium ore prediction model that meets the evaluation threshold. The system can effectively realize the efficient and accurate prediction of carbonate clay type lithium ore.

[0107] An embodiment of the present invention further provides an electronic device, which may include: a display screen, a memory, and one or more processors. The display screen, memory, and processor are coupled. The memory is used to store computer program code, which includes computer instructions. When the processor executes the computer instructions, the electronic device can perform the various methods or steps described in the aforementioned carbonate clay-type lithium ore prediction method embodiment. Of course, the electronic device includes, but is not limited to, the aforementioned display screen, memory, and one or more processors.

[0108] An embodiment of the present invention also provides a computer-readable storage medium for storing computer instructions for running the above-mentioned carbonate clay-type lithium ore prediction method.

[0109] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0110] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0111] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0112] Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the overall concept of this application and do not constitute a limitation on the scope of protection of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without expending creative work shall fall within the scope of protection of this application.

Claims

1. A method for predicting carbonate clay-type lithium ore, characterized in that: include: Acquire multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of the target area; The target area includes: a sample area and a prediction area, and the sample area includes: a carbonate clay type lithium ore area and a non-carbonate clay type lithium ore area; Determine a metal assessment map based on the multispectral remote sensing image data, determine a lithium element concentration map based on the geochemical sampling data, determine a carbonation alteration anomaly map based on the hyperspectral remote sensing image data, and determine a geological factor density map based on the geological information data; performing normalization and gridding processing on the metal assessment map, the lithium element concentration map, the carbonation alteration anomaly map, and the geological factor density map; Based on the sample area and the prediction area, data segmentation is performed on the metal estimation map, lithium element concentration map, carbonation alteration anomaly map, and geological factor density map after the normalization and gridding processing to obtain a sample data set and a prediction data set; Based on the carbonate clay type lithium ore region and the non-carbonate clay type lithium ore region, the sample data set is divided into positive and negative samples, and data enhancement is performed on the sample data set after the positive and negative sample division to obtain an enhanced sample data set; Constructing a plurality of first lithium ore prediction models, wherein the first lithium ore prediction models are deep learning models for predicting the location of carbonate clay-type lithium ore; Training and optimizing the plurality of first lithium ore prediction models respectively according to the enhanced sample data set, and determining evaluation index values ​​corresponding to the plurality of first lithium ore prediction models respectively; A target lithium carbonate mineralization prediction result is obtained based on the prediction data set through a target lithium ore prediction model, and the target lithium ore prediction model is a first lithium ore prediction model corresponding to the evaluation index value that meets the evaluation threshold.

2. The method according to claim 1, characterized in that In the case where there are multiple target lithium ore prediction models, the target carbonate lithium ore mineralization prediction result is obtained by the target lithium ore prediction model based on the prediction data set, including: According to the prediction data set, prediction is performed using the multiple target lithium ore prediction models to obtain multiple first carbonate lithium ore mineralization prediction results; The multiple first carbonate lithium ore mineralization prediction results are integrated through an integrated learning algorithm to obtain the target carbonate lithium ore mineralization prediction result.

3. The method according to claim 1, characterized in that Determining a metal assessment map based on the multispectral remote sensing image data includes: Performing a first preprocessing on the multispectral remote sensing image data to obtain multispectral remote sensing image data after the first preprocessing; the first preprocessing includes: a first radiation calibration process, a first atmospheric correction process, and a band fusion preprocessing; According to the multispectral remote sensing image data after the first preprocessing, a metal stress vegetation index is determined. The metal stress vegetation index is: a vegetation index VIGS considering green wave and short-wave infrared rays. The expression of the VIGS is: Wherein, G represents the spectral data of the green band of visible light, S1 represents the spectral data of the first short-wave infrared band, S2 represents the spectral data of the second short-wave infrared band, N represents the spectral data of the near-infrared band, R represents the reflectivity of the red band, w1 represents the first preset weight, w2 represents the second preset weight, w3 represents the third preset weight, and w4 represents the fourth preset weight; The metal assessment map is generated based on the metal stress vegetation index.

4. The method according to claim 1, wherein Determining a lithium element concentration map based on the geochemical sampling data includes: Extracting the concentration value of lithium and the concentration value of associated elements based on the geochemical sampling data; wherein the associated elements include gallium, sodium and calcium; The lithium element concentration map is generated by Kriging interpolation method according to the concentration value of the lithium element and the concentration value of the associated element.

5. The method according to claim 1, characterized in that Determining a carbonation alteration anomaly map based on the hyperspectral remote sensing image data includes: Performing a second preprocessing on the hyperspectral remote sensing image data to obtain hyperspectral remote sensing image data after the second preprocessing; the second preprocessing includes: spectral band merging processing, bad band removal processing, second radiometric calibration processing, second atmospheric correction processing and orthorectification processing; performing minimum noise separation processing and spectral angle matching processing on the hyperspectral remote sensing image data after the second preprocessing to obtain matched hyperspectral remote sensing image data; The matched hyperspectral remote sensing image data are respectively subjected to principal component analysis and independent component analysis to determine abnormal carbonation alteration information; The carbonation alteration anomaly map is generated based on the carbonation alteration anomaly information.

6. The method according to claim 1, characterized in that Determining a geological factor density map based on the geological information data includes: The geological factor density map is generated by a buffer distance analysis method based on the geological information data.

7. The method according to claim 1, characterized in that The multiple first lithium ore prediction models are multiple of the following models: AlexNet network model, LeNet network model, MobileNet network model, ResNet network model, VGGNet network model and ViT model.

8. The method according to claim 1, characterized in that The evaluation index value includes one or more of the following indicators: confusion matrix, precision, accuracy, recall rate, F1 score, ROC receiver operating characteristic curve, and area under the AUC curve.

9. The method according to claim 2, characterized in that The ensemble learning algorithm is: averaging method or voting method.

10. A carbonate clay type lithium ore prediction method system, characterized in that: include: Acquisition module, extraction module, preprocessing module, data partitioning module, sample construction module, model construction module, training module and prediction module; among them, The acquisition module is used to acquire multispectral remote sensing image data, geochemical sampling data, hyperspectral remote sensing image data and geological information data of the target area; the target area includes: a sample area and a prediction area, and the sample area includes: a carbonate clay type lithium ore area and a non-carbonate clay type lithium ore area; The extraction module is used to determine a metal assessment map based on the multispectral remote sensing image data, a lithium element concentration map based on the geochemical sampling data, a carbonation alteration anomaly map based on the hyperspectral remote sensing image data, and a geological factor density map based on the geological information data; The pre-processing module is used to perform normalization and gridding processing on the metal assessment map, the lithium element concentration map, the carbonation alteration anomaly map and the geological factor density map; The data partitioning module is used to partition the metal estimation map, lithium element concentration map, carbonation alteration anomaly map and geological factor density map after normalization and gridding based on the sample area and the prediction area to obtain a sample data set and a prediction data set; The sample construction module is used to divide the sample data set into positive and negative samples based on the carbonate clay-type lithium ore region and the non-carbonate clay-type lithium ore region, and perform data enhancement on the sample data set after the positive and negative sample division to obtain an enhanced sample data set; The model building module is used to build a plurality of first lithium ore prediction models, wherein the first lithium ore prediction model is a deep learning model for predicting the location of carbonate clay-type lithium ore; The training module is used to train and optimize the multiple first lithium ore prediction models respectively according to the enhanced sample data set, and determine the evaluation index values ​​corresponding to the multiple first lithium ore prediction models respectively; The prediction module is used to obtain a target carbonate lithium ore mineralization prediction result through a target lithium ore prediction model based on the prediction data set, and the target lithium ore prediction model is a first lithium ore prediction model corresponding to the evaluation index value that meets the evaluation threshold.

Images