An intelligent metallogenic prediction method fusing multi-element data source-sink process analysis

Abstract

The application belongs to the field of uranium exploration, and particularly relates to an intelligent metallogenic prediction method fusing multi-element data source-converging process analysis, comprising data acquisition; adopting a bidirectional feature fusion network to delineate a potential uranium source area in a target area; obtaining mineral composition information of the potential uranium source area through mineral inversion; obtaining a uranium anomaly area in the potential uranium source area through an aerial gamma energy spectrum and a ground verification collaborative framework; extracting surface water flow direction information in the uranium anomaly area, inverting a basement relief form and a tectonic stress field, and constructing a coupling analysis process of the surface water flow, the basement relief form and the tectonic stress field to analyze uranium migration channels and migration driving mechanisms; and constructing a sandstone type uranium metallogenic prediction model based on remote sensing data features. The application realizes the system integration of multi-source remote sensing data in the whole process of uranium exploration, and provides key technical support and theoretical innovation for uranium resource security.

Description

Technical Field

[0001] This invention belongs to the field of uranium exploration, specifically involving an intelligent mineralization prediction method that integrates multiple data sources and sink process analysis. Background Technology

[0002] Against the backdrop of the "dual carbon" goals driving the large-scale development of nuclear energy and the severe global energy crisis, uranium resource supply has become a significant and long-term demand. Sandstone-type uranium deposits, as an important type of uranium deposit globally and the most economically recoverable type, are directly related to national energy security and the low-carbon transition process through efficient exploration and development.

[0003] To address the limitations of traditional exploration methods, such as surface cover, complex geological structures, and high costs, which hinder the efficient and accurate location of sandstone-type uranium mineralization target areas, this study aims to fully leverage the high-precision advantages of remote sensing technology in surface exploration. It integrates high-resolution optical imagery and SAR data to systematically analyze topographic features, catchment basin patterns, and valley network distribution. Based on this, tectonic stress field analysis is conducted using InSAR deformation fields to invert the direction of the tectonic stress field and quantitatively assess the intensity of tectonic activity, thereby inferring the migration trends of surface water and shallow groundwater. Simultaneously, by constraining the basement undulations using gravity and magnetic data, and through basin-scale tectonic evolution analysis and regional-scale tectonic difference studies, the study comprehensively characterizes the characteristics of the water flow system and the control mechanism of geological structure on uranium migration paths under the background of neotectonic movement.

[0004] Based on this, conducting research on basin-wide multi-source data fusion analysis and artificial intelligence-based mineral exploration prediction is of great strategic significance for breaking through the bottlenecks of traditional exploration technologies, promoting the advancement of mineral exploration methods and technologies, improving exploration efficiency, and achieving breakthroughs in mineral exploration in new areas. Specifically, sandstone-type uranium deposits in the Erlian Basin are generally characterized by shallow burial, near-source mineralization in ancient river valleys, and concentrated distribution in belts. Establishing comprehensive detection technologies and methods for sandstone-type uranium deposits in such basins, tailored to local conditions, has important demonstration and reference value for guiding mineral exploration practices. Summary of the Invention

[0005] Based on this, this invention proposes an intelligent mineralization prediction method that integrates multi-source data and sinking process analysis. Addressing the intelligent detection needs of the entire "source-transport-accumulation" process in sandstone-type uranium deposits, a three-stage progressive mineralization prediction method is constructed: First, through a bidirectional feature fusion framework and diffusion model-guided hyperspectral inversion technology, the bottleneck of global information capture in convolutional neural networks is overcome. Second, a collaborative framework combining airborne gamma spectroscopy and ground verification is used to accurately delineate uranium source areas. Third, multi-source remote sensing (spectral / InSAR / topographic), gravity, magnetic, and tectonic deformation data is integrated to construct a coupled process of paleogeomorphological restoration, basement undulation inversion, and stress field modeling, quantitatively analyzing the dynamic mechanism of uranium migration channels. Finally, the results of the first two stages are integrated to establish a high-precision database of ore-controlling elements. Finally, by combining deep learning and multi-source geoscientific data, a "source-transport-accumulation" mineralization prediction model is constructed, enabling intelligent delineation of prospecting target areas and forming a complete remote sensing prospecting technology system.

[0006] Firstly, this application provides an intelligent mineralization prediction method that integrates multiple data sources and sink process analysis, the method comprising: Step S1: Acquire multi-source remote sensing data, airborne gamma spectral data, gravity and magnetic data, and geological data covering the target area; Step S2: Delineate potential uranium source regions in the target area using a bidirectional feature fusion network; obtain mineral composition information of potential uranium source regions by performing mineral inversion using hyperspectral images; and obtain uranium anomaly regions in potential uranium source regions through a collaborative framework of airborne gamma spectroscopy and ground verification. Step S3: Within the uranium anomaly region, surface water flow direction information is extracted based on topographic data from multi-source remote sensing data. The undulating morphology of the basement is inverted using gravity and magnetic data, and the tectonic stress field is inverted based on InSAR deformation data. A coupled analysis process of surface water flow-basement morphology-tectonic stress field is constructed to analyze the uranium migration channel and transport mechanism. Step S4 focuses on the complex correlation and synergistic response mechanism between multi-source remote sensing data and key ore-controlling elements of sandstone-type uranium deposits, and constructs a mineralization prediction model for sandstone-type uranium deposits based on remote sensing data characteristics.

[0007] In one possible implementation, step S2 includes: Step S21: Preprocess the multi-source optical and SAR images of the target area, select the main information bands and stack them to form a multi-source remote sensing fusion feature map; input the multi-source remote sensing fusion feature map into a bidirectional feature fusion network, which includes a multi-scale feature extraction backbone network, a context and detail co-enhancement module and a cross-scale feature fusion network. The multi-scale feature extraction backbone network is composed of multiple stacked C3k2 modules. PPA modules are embedded in the key layers of the multi-scale feature extraction backbone network to form a context and detail co-enhancement module. The cross-scale feature fusion network is a BiFPN. The multi-level feature map enhanced by the PPA module is fused bidirectionally from top to bottom and from bottom to top to output the potential uranium source area delineation result. Step S22: In the potential uranium source region, mineral inversion is performed using hyperspectral images, and the mineral inversion results are output to obtain mineral composition information of the potential uranium source region; Step S23: Based on airborne gamma-ray spectroscopy data and combined with spatial clustering algorithms, delineate uranium anomaly regions in potential uranium source areas; within the uranium anomaly regions, combine mineral inversion results with ground verification to achieve high-precision prediction of the source rock distribution in the uranium anomaly regions.

[0008] In one possible implementation, step S22 specifically involves: constructing a diffusion model-guided hyperspectral mineral inversion network PMNet, which includes a multi-scale depth-separable convolutional branch, a VisionTransformer branch, a frequency domain feature interaction module, and a Vision Mamba sequence modeling module; inputting hyperspectral images of potential uranium source regions into the hyperspectral mineral inversion network PMNet; extracting local spectral detail features through the multi-scale depth-separable convolutional branch; extracting global spatial context features through the Vision Transformer branch; performing cross-domain fusion of the output features of the multi-scale depth-separable convolutional branch and the Vision Transformer branch through the frequency domain feature interaction module; and performing selective state updates and sequence modeling by the Vision Mamba sequence modeling module to output the mineral inversion results.

[0009] In one possible implementation, step S23 specifically includes: Step S231: Denoise and correct the airborne gamma spectrum data. Use the NASVD method to perform singular value decomposition on the airborne gamma spectrum data, select the principal component components with a cumulative variance contribution rate of not less than 95% for signal reconstruction, and obtain the reconstructed airborne gamma spectrum data. Use the MNF method to sort the reconstructed airborne gamma spectrum data according to the signal-to-noise ratio, and obtain the airborne gamma reconstruction data. Step S232: By extracting background samples from potential uranium source areas, the mean and standard deviation of uranium content and Th / U ratio are calculated respectively to construct a dual threshold model for uranium content and Th / U ratio; based on airborne gamma reconstruction data, the grid cells that satisfy the dual threshold model for uranium content and Th / U ratio in potential uranium source areas are spatially clustered using the K-means clustering algorithm to extract uranium anomaly regions. Step S233: Select exposed surface areas within the uranium anomaly region, and verify the anomaly by combining high-resolution surveys, geological mapping, and detailed evaluation to achieve high-precision prediction of the source rock distribution in the uranium anomaly region.

[0010] In one possible implementation, in the dual-threshold model of uranium content and Th / U ratio, the uranium content threshold is defined as: ,in, Indicates the uranium content threshold. This represents the average uranium content. The standard deviation of uranium content is represented; the threshold for the Th / U ratio is defined as: ,in, Indicates the threshold value of the Th / U ratio. This represents the mean Th / U ratio. This represents the standard deviation of the Th / U ratio.

[0011] In one possible implementation, the high-resolution survey specifically involves using a high-resolution handheld gamma spectrometer for fine surveying, achieving quantitative monitoring of U, Th, and K anomalies through dense sampling and high-frequency sampling; the geological mapping specifically involves conducting geological mapping at a scale of 1:10,000; and the fine evaluation specifically involves quantitatively analyzing the in-situ content, occurrence form, and mineralogical characteristics of uranium.

[0012] In one possible implementation, step S3 specifically includes: Step S31: Based on high-resolution optical data and SAR remote sensing data, the surface water flow direction in the uranium anomaly area is analyzed by remote sensing identification of key topographic parameters and division of topographic and geomorphic units, extraction of valley network and intelligent division of catchment basins, and estimation and verification of surface water flow direction. Step S32: Use gravity and magnetic data to invert the basement undulation morphology of the uranium anomaly region and characterize the spatial distribution of the local uplift in the uranium anomaly region; based on the tectonic stress field inversion and activity intensity assessment of InSAR deformation data, interpret the activity of the faults around the local uplift in the uranium anomaly region and the direction of the tectonic stress field. Step S33: Integrate the surface water flow direction and tectonic stress field direction in the uranium anomaly region, and conduct comparative analysis with Cenozoic tectonic events in the uranium anomaly region to determine the flow direction characteristics of ore-forming fluids in the Cenozoic ore-bearing strata; comprehensively analyze the spatial distribution of local uplifts in the surface water flow direction and the activity of faults around the local uplifts to analyze the constraints of local uplifts on the runoff of ore-forming fluids and reconstruct the migration path of ore-forming fluids in the Cenozoic ore-bearing strata.

[0013] In one possible implementation, the remote sensing identification of key terrain parameters and the division of terrain and geomorphological units specifically involve: generating 12.5m resolution DEM data using sentinel SAR data and preprocessing the DEM data; remotely identifying key terrain parameters in the DEM data; based on the key terrain parameters, extracting feature lines such as valley lines and ridge lines from the DEM data using geometric morphology methods, and determining the slope threshold in the DEM data using natural discontinuity grading methods; based on the extracted ridge lines and valley lines, dividing the uranium anomaly area into linear feature areas and non-linear feature areas; delineating ridges, valleys, and saddles within the linear feature areas based on the feature line morphology and local slope transitions; and classifying steep slopes and gentle slopes within the non-linear feature areas based on the slope thresholds, ultimately forming five types of terrain and geomorphological units: ridges, valleys, saddles, steep slopes, and gentle slopes.

[0014] In one possible implementation, preprocessing the DEM data includes: performing Gaussian filtering on the DEM data, and performing depression filling on the Gaussian-filtered DEM data.

[0015] In one possible implementation, step S4 specifically involves defining the oxidation front by the iron staining anomaly boundary, identifying the reduction barrier using shortwave infrared / radar, and accurately locating the oxidation-reduction transition zone; verifying the spatial configuration relationship between the uranium source area, migration path, and oxidation-reduction transition zone using GIS spatial overlay and buffer analysis; and using machine learning methods to fuse and quantify variables derived from multi-source remote sensing to generate a mineralization favorability probability map.

[0016] This application focuses on the entire process of the "source-migration-accumulation" mineralization system of sandstone-type uranium deposits, aiming to construct an intelligent exploration technology system that deeply integrates multi-source remote sensing and artificial intelligence. It develops a bidirectional feature fusion framework and a diffusion model-driven hyperspectral inversion method, combined with an airborne gamma-ray spectroscopy and ground-based verification framework, to achieve precise delineation of potential uranium source areas. It integrates multi-source remote sensing (spectral / InSAR / topographic), gravity, magnetic, and tectonic deformation data to construct a coupled analysis process of paleogeomorphic reconstruction, basement undulation inversion, and stress field modeling, quantitatively analyzing uranium migration channels and dynamic mechanisms. Based on ore-controlling elements (uranium source / structure / ... A high-precision database of sand bodies / oxidation zones is used to establish a synergistic response model between remote sensing features and ore-forming elements. This model integrates geological, geophysical, and geochemical data with deep learning technology to form a comprehensive "source-transport-accumulation" correlation prediction system. This system overcomes the limitations of traditional single-technology methods, achieving system integration of multi-source remote sensing data throughout the entire uranium exploration process. Intelligent algorithms improve the automation level of structural identification and anomaly delineation. Multi-scale data fusion reveals the deep-surface coupling mechanism, ultimately establishing a generalizable remote sensing prospecting technology standard, providing key technical support and theoretical innovation for uranium resource security. Compared with existing technologies, the beneficial effects of this invention are: (1) This application constructs a mineral inversion network driven by bidirectional feature fusion and diffusion model, and combines airborne gamma spectrum dual threshold clustering and ground verification to achieve accurate delineation of uranium source areas and high-precision identification of source rocks.

[0017] (2) This application quantitatively analyzed the uranium migration channels and dynamic mechanisms by coupling multiple factors such as surface water flow, basement undulation morphology and tectonic stress field, and revealed the migration path of ore-forming fluids.

[0018] (3) This application achieves precise positioning of the oxidation-reduction transition zone and intelligent prediction of mineral exploration target areas by extracting mineral control elements from multiple sources and integrating machine learning modeling. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating an intelligent mineralization prediction method integrating multiple data sources and sink process analysis, provided in an embodiment of the present invention; Figure 2 This invention provides a uranium mineralization prediction heatmap for a uranium mineralization potential area in the Erlian Basin, based on an intelligent mineralization prediction method that integrates multiple data sources and sink process analysis. Detailed Implementation

[0021] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0022] It should be understood that the described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0023] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0025] See Figure 1 This is a flowchart illustrating an intelligent mineralization prediction method that integrates multiple data sources and the analysis of the sinking process, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes: Step S1: Acquire multi-source remote sensing data, airborne gamma spectral data, gravity and magnetic data, and geological data of the target area.

[0026] Step S2 involves delineating potential uranium source regions in the target area using a bidirectional feature fusion network; obtaining mineral composition information of potential uranium source regions through mineral inversion using hyperspectral images; and obtaining uranium anomaly regions in potential uranium source regions through a collaborative framework of airborne gamma spectroscopy and ground verification, thereby achieving high-precision prediction of source rock distribution in uranium anomaly regions.

[0027] Due to the complex environment of potential uranium source regions, significant semantic gaps exist between low-level features such as texture and edges and high-level features such as contextual information during cross-level fusion. Existing algorithms struggle to balance the overall semantic information of large-scale geophysics with the fine details of small-scale targets when extracting associated features of targets at different scales. This makes uranium source information evaluation using multi-source remote sensing information a key challenge. This embodiment constructs a cross-scale uranium source information evaluation method based on prior knowledge fusion of multi-source remote sensing information through step S2. Specifically: Step S21: Preprocess the multi-source optical and SAR images of the target area, select the main information bands through transformations such as principal component analysis, and stack them to form a multi-source remote sensing fusion feature map; input the multi-source remote sensing fusion feature map into a bidirectional feature fusion network (BFMF-Net), which includes a multi-scale feature extraction backbone network, a context and detail co-enhancement module, and a cross-scale feature fusion network. The multi-scale feature extraction backbone network is composed of multiple stacked C3k2 modules and is used to extract multi-level features from the multi-source remote sensing fusion feature map; in the key layers of the multi-scale feature extraction backbone network... The PPA module is embedded at multiple levels (such as 8x, 16x, and 32x downsampling layers) to form a context and detail co-enhancement module. During the forward pass, the multi-branch feature flow of the C3k2 module provides multi-scale input to the PPA module. During the backward pass, the attention weights generated by the PPA module are fed back to the C3k2 module, guiding it to focus on key target features, thus achieving bidirectional co-optimization between the two. The cross-scale feature fusion network is a BiFPN (Bidirectional Feature Pyramid Network), which performs bidirectional cross-scale fusion of the multi-level feature maps enhanced by the PPA module from top to bottom and bottom to top, outputting the potential uranium source region delineation results.

[0028] Furthermore, the method for determining the key layers of the multi-scale feature extraction backbone network is as follows: in the C3k2 module stacking structure of the multi-scale feature extraction backbone network, the output layers of the C3k2 modules with downsampling ratios of 8x, 16x and 32x are determined as the key layers for embedding the PPA module.

[0029] Step S22: In the potential uranium source region, mineral inversion is performed using hyperspectral images, and the mineral inversion results are output to obtain mineral composition information of the potential uranium source region.

[0030] Specifically, the process involves constructing a diffusion model-guided hyperspectral mineral inversion network, PMNet, which includes a multi-scale depth-separable convolutional branch, a Vision Transformer branch, a frequency domain feature interaction module, and a Vision Mamba sequence modeling module. Hyperspectral images of potential uranium source regions are input into PMNet. Local spectral detail features are extracted through the multi-scale depth-separable convolutional branch, and global spatial context features are extracted through the Vision Transformer branch. The frequency domain feature interaction module performs cross-domain fusion of the output features from the multi-scale depth-separable convolutional branch and the Vision Transformer branch. Finally, the Vision Mamba sequence modeling module performs selective state updates and sequence modeling to output the mineral inversion results.

[0031] Step S23: Based on airborne gamma-ray spectroscopy data and combined with spatial clustering algorithms, delineate uranium anomaly regions in potential uranium source areas; within the uranium anomaly regions, combine mineral inversion results with ground verification to achieve high-precision prediction of the source rock distribution in the uranium anomaly regions.

[0032] In traditional airborne gamma-ray anomaly identification, uranium content is easily affected by multiple factors such as surface conditions, observation accuracy, and background field fluctuations, leading to a risk of misjudgment with single uranium anomaly identification methods. Furthermore, while the Th / U ratio reflects the relative enrichment of uranium, its numerical variation is also affected by thorium background interference, often resulting in missed anomalies, especially in igneous rocks or high-Th background areas. To improve the stability and accuracy of uranium source information extraction, this application uses airborne gamma-ray energy spectrum data combined with spatial clustering algorithms to accurately delineate uranium-rich anomaly blocks and provides prior constraints for source rock type identification. Step S23 specifically involves: Step S231: Denoising and correction are performed on the airborne gamma spectrum data. The singular value decomposition of the airborne gamma spectrum data is performed using the NASVD method. Principal component components with a cumulative variance contribution rate of not less than 95% are selected for signal reconstruction to obtain the reconstructed airborne gamma spectrum data, effectively filtering out random noise. The reconstructed airborne gamma spectrum data is sorted according to the signal-to-noise ratio using the MNF method to obtain the airborne gamma reconstruction data.

[0033] Step S232: By extracting background samples from potential uranium source areas, the mean and standard deviation of uranium content and Th / U ratio are calculated respectively to construct a dual-threshold model for uranium content and Th / U ratio; based on airborne gamma reconstruction data, the K-means clustering algorithm is used to spatially cluster the raster cells that meet the dual-threshold model for uranium content and Th / U ratio in potential uranium source areas in order to extract uranium anomaly regions with good spatial continuity and clear morphological structure.

[0034] Preferably, in the dual-threshold model of uranium content and Th / U ratio, the uranium content threshold is defined as: ,in, Indicates the uranium content threshold. This represents the average uranium content. The standard deviation of uranium content is represented; the threshold for the Th / U ratio is defined as: ,in, Indicates the threshold value of the Th / U ratio. This represents the mean Th / U ratio. This represents the standard deviation of the Th / U ratio. By jointly screening a dual-threshold model of uranium content and Th / U ratio, it can effectively correspond geologically to the distribution of source rocks or uranium-rich source rocks, significantly improving the ability to identify surface anomalies.

[0035] Preferably, the silhouette coefficient can be used to determine the optimal number of clusters during spatial clustering, thereby ensuring the objectivity and geological rationality of uranium anomaly region delineation. In actual aerial surveys, undulating terrain causes changes in the detector's altitude relative to the ground surface, affecting gamma-ray intensity. Furthermore, the high-energy background contribution from cosmic rays also has a significant impact on the total count. Therefore, in this embodiment, the following correction model is used for data preprocessing to ensure the accuracy of the identification results: in, This indicates the net uranium content or net count rate after radon, cosmic rays, and high correction, in units of ppm or cps; This indicates the uranium content or count rate measured by the airborne gamma spectrometer, including surface signals and various noise interferences, and is expressed in ppm or cps. This represents the radon correction factor, which is determined through calibration experiments or by field measurements over water bodies without uranium anomalies, and is a dimensionless empirical constant. The count rate, representing the contribution of radon background, is estimated by low-energy scattering from the detector on the uranium window or by measurement using a dedicated radon window, and is measured in cps. The count rate representing the contribution of cosmic rays to the background is estimated through high-altitude flight measurements or empirical models, and is measured in cps. Represents the high attenuation correction function. This indicates the actual flight altitude of the probe above the Earth's surface, usually obtained by a radar altimeter or laser rangefinder, and is measured in meters (m).

[0036] After data preprocessing through the calibration model, the airborne gamma reconstruction data after NSVD denoising and MNF transformation can be restored to a "net anomaly" that is closer to the real surface radioactivity distribution, laying the foundation for subsequent source rock optimization.

[0037] Step S233: In the preferred exposed area within the uranium anomaly region, the anomaly is verified by combining high-resolution surveys, geological mapping, and detailed evaluation to achieve high-precision prediction of the source rock distribution in the uranium anomaly region, thereby improving the scientific nature and accuracy of the mining area evaluation.

[0038] Furthermore, the high-resolution survey specifically involves using a high-resolution handheld gamma spectrometer for fine-grained surveys. Through dense sampling and high-frequency sampling, it enables quantitative monitoring of anomalies in U, Th, and K content, capturing details of radiation changes within a small-scale space, and providing basic data support for the precise location and morphological analysis of spatial anomalies.

[0039] Furthermore, the geological mapping specifically involves: conducting geological mapping at a scale of 1:10,000 to systematically clarify the spatial relationship between uranium anomalies and regional lithological characteristics, tectonic units, and alteration zones.

[0040] Furthermore, the detailed evaluation specifically involves: quantitatively analyzing the in-situ content, occurrence form, and mineralogical characteristics of uranium to reveal the formation of uranium. Mineral morphology (such as source rock type, mineralization intensity, and alteration degree) and its spatial distribution characteristics provide experimental data support for constructing a quantitative model between airborne anomalies and actual mineralization degree.

[0041] Step S3: Within the uranium anomaly region, surface water flow direction information is extracted based on topographic data from multi-source remote sensing data. The basement undulation morphology is inverted using gravity and magnetic data, and the tectonic stress field is inverted based on InSAR deformation data. A coupled analysis process of surface water flow, basement morphology, and tectonic stress field is constructed to analyze uranium migration channels and transport mechanisms. Specifically: Step S31: Based on high-resolution optical data and SAR remote sensing data, the surface water flow direction in the uranium anomaly area is analyzed by remote sensing identification of key terrain parameters and division of topographic and geomorphic units, extraction of valley networks and intelligent division of catchment basins, and estimation and verification of surface water flow direction.

[0042] Preferably, the key terrain parameter remote sensing identification and terrain and geomorphological unit division specifically involves: generating 12.5m resolution DEM data using sentinel SAR data and preprocessing the DEM data; remotely identifying key terrain parameters in the DEM data; based on the key terrain parameters, extracting feature lines such as valley lines and ridge lines from the DEM data using the geometric morphology method, and determining the slope threshold in the DEM data using the natural discontinuity classification method; based on the extracted ridge lines and valley lines, dividing the uranium anomaly area into linear feature areas and non-linear feature areas; delineating ridges, valleys, and saddles within the linear feature areas based on the feature line morphology and local slope transitions; and classifying steep slopes and gentle slopes within the non-linear feature areas based on the slope threshold, ultimately forming five types of terrain and geomorphological units: ridges, valleys, saddles, steep slopes, and gentle slopes.

[0043] Furthermore, the preprocessing of DEM data includes: performing Gaussian filtering on the DEM data to reduce the noise impact of low-rise buildings, vegetation, etc., in the uranium anomaly area; and performing depression filling on the DEM data after Gaussian filtering to avoid depressions and flat areas in the DEM data causing the raster to have no flow direction or unclear flow direction.

[0044] Furthermore, the key topographic parameters include basic topographic factors (slope, aspect, elevation, etc.), morphological parameters (planar curvature, profile curvature, surface roughness, topographic relief), and hydrological parameters (cumulative runoff, topographic humidity index, etc.).

[0045] The extraction of the gully network and the intelligent division of the catchment basins are specifically as follows: The Douglas-Peuker algorithm is used to obtain the upward-curving grid points in the DEM data to obtain the prototype of the water system network; Since the water system location extraction in the network prototype obtained by the Douglas-Peuker algorithm is relatively accurate, but the water system integrity is poor and some areas need to be refined, the D8 algorithm is also used to analyze and optimize the water system network prototype to extract the gully network; and the catchment basins are divided on the gully network.

[0046] Furthermore, the analysis and optimization of the water network prototype using the D8 algorithm includes: filling depressions, determining the direction of water flow, calculating the cumulative water flow matrix, and extracting the water network.

[0047] The determination of the water flow direction specifically includes: defining the water flow direction in the D8 algorithm as the direction of the line connecting the center of the processed grid cell to the center of the grid cell with the largest slope among its eight adjacent grid cells. Valid water flow directions are defined as east, northeast, north, northwest, west, southwest, south, and southeast, and are represented by eight valid feature codes: 1, 2, 3, 4, 5, 6, 7, and 8, respectively.

[0048] The formula for calculating the slope is as follows: in, Represents grid cells With grid cells The slope between; , These represent the serial numbers of the grid cells; Represents grid cells With grid cells The elevation between; Represents grid cells With grid cells Elevation between adjacent grid cells , ,and , Not both are 0; Indicates the width of the grid cell.

[0049] The calculation of the water flow accumulation matrix specifically includes: calculating the cumulative water flow value flowing into each grid cell along the water flow direction to obtain the water flow accumulation matrix. During calculation, each grid cell in the DEM data represents one unit of water volume, and the total number of all grid cells that directly or indirectly flow through a given grid cell constitutes the cumulative water flow value for that grid cell. In the calculation of the water flow accumulation matrix, only grid cells marked as 1 in the Douglas-Peuker algorithm are accumulated; the weights of other grid cells are assigned to 0. First, a two-dimensional water flow accumulation matrix (with the same number of rows and columns as the DEM data) is defined, and the water flow accumulation matrix value of each grid cell is initialized to 0. Then, the water flow direction is scanned sequentially, starting from the grid cells marked as 1 in the Douglas-Peuker algorithm, and tracing forward along the water flow direction until the watershed boundary is reached. Each time a grid cell is scanned, its water flow direction is determined. When the grid cell is marked as 1 in the Douglas-Peuker algorithm, the corresponding upstream accumulation cell pointing to the grid cell is increased by one unit. After all water flow directions have been scanned along the tracking line, the water flow accumulation matrix of all grid cells marked as 1 in the Douglas-Peuker algorithm can be obtained.

[0050] The water network extraction specifically involves: based on the water flow accumulation matrix, extracting grid cells whose water flow accumulation value is greater than the water catchment area threshold as river channels to obtain the valley network.

[0051] For each raster cell in the flow accumulation matrix, the accumulated flow value exceeds the catchment area threshold, and the raster cell is considered a river channel. The choice of the catchment area threshold varies depending on the method used, and it significantly impacts the gully network and its characteristics extracted from DEM data. A smaller catchment area threshold results in a denser river network, while a larger threshold leads to a sparser network. Therefore, a reasonable catchment area threshold must be set based on the actual watershed conditions.

[0052] This embodiment uses the Broscoe algorithm to calculate the catchment area threshold. This algorithm is based on the water level difference in topography. The catchment area threshold is selected as the minimum area threshold that can keep the average water level difference of a certain level of water system in the Strahler water system classification small from the average water level difference of the higher level water system.

[0053] The process of dividing catchment basins on the valley network specifically involves: using the preprocessed DEM data, the unidirectional D8 algorithm is used to calculate the flow direction of each grid cell in valleys, ridges, and steep slopes, while the multidirectional MFD-md algorithm is used to calculate the flow direction of each grid cell in saddles and gentle slopes; based on the flow direction of each grid cell, the number of grid cells flowing into each grid cell (i.e., the area of ​​the upstream catchment area) is calculated; the natural lowest point of the entire study area (e.g., the river mouth) is determined as the outlet point; potential river network nodes are generated based on the cumulative runoff threshold (e.g., the number of grid cells flowing into each grid cell > 1000 grid cells); and on the potential river network nodes, based on the outflow direction of the water flow in each grid cell and the outlet point, GIS hydrological tools are used to delineate all catchment areas upstream of each outlet point, i.e., catchment basins.

[0054] The surface water flow direction estimation and verification specifically involves: using the valley network as a flow direction constraint, the water flow direction is forced to flow along the center line of the valley, overcoming the problem of uncertain flow direction in DEM data in flat valley areas. Specifically, the gully network is input into the hydrological processing as a "burn line," and the flow direction is modified at the gully location to follow the direction of the gully extension. Based on the flow direction of each grid cell, starting from any point (or a specific point, such as a watershed), the flow path is traced to the outlet or stopping point to generate the confluence path. The water flow direction of each grid cell in the divided catchment basins, gully network, and each grid cell is spatially overlaid with the divided topographic and geomorphological units, optical image interpretation, and SAR features to check the degree of agreement between the catchment basin boundaries and ridge lines, the matching degree between the gully network and the actual river valley locations in the optical image, and the rationality of the flow direction at key topographic nodes such as saddles, thus verifying the accuracy of the surface runoff pattern and the consistency of geomorphological logic. The generated distribution of accumulated runoff is compared with the actual river network location to obtain the surface water flow direction within the uranium anomaly area.

[0055] Step S32: The basal undulation morphology of the uranium anomaly region is inverted using gravity and magnetic data to characterize the spatial distribution of the local uplift in the uranium anomaly region; the tectonic stress field inversion and activity intensity assessment based on InSAR deformation data are used to interpret the activity of the faults surrounding the local uplift in the uranium anomaly region and the direction of the tectonic stress field.

[0056] The method of using gravity and magnetic data to invert the basement undulation morphology of the uranium anomaly region specifically involves: acquiring land gravity data, aeromagnetic data, deep reflection seismic profiles, and magnetotelluric sounding data of the uranium anomaly region, and performing standardized preprocessing; performing joint gravity and magnetic inversion based on land gravity data and aeromagnetic data to obtain a three-dimensional density structure and magnetic distribution model of the uranium anomaly region; performing density interface inversion based on land gravity data, deep reflection seismic profiles, and magnetotelluric sounding data to obtain the burial depth distribution of the main density interfaces; and inverting the basement undulation morphology based on the three-dimensional density structure and magnetic distribution model of the uranium anomaly region and the burial depth distribution of the main density interfaces to characterize the spatial distribution of local uplifts in the uranium anomaly region.

[0057] Furthermore, the density of measurement points for acquiring land gravity data is ≥1 point / 10km², with the CG-5 relative gravimeter accuracy ±0.01mGal; the flight altitude for acquiring aeromagnetic data is <300m, the sampling interval is 200m, and the accuracy is ±1nT; the trace spacing for acquiring deep reflection seismic profiles is 25m; and the MT station spacing for acquiring magnetotelluric sounding data is 5km.

[0058] The standardized preprocessing includes: gravity data correction (terrain correction, Bouguer correction, equalization correction), magnetic data processing, and physical property parameter fusion.

[0059] The tectonic stress field inversion and activity intensity assessment based on InSAR deformation data specifically involves: acquiring Sentinel-1 satellite C-band SAR data of the uranium anomaly region as the core data source; systematically downloading SLC (single-look complex) data covering the uranium anomaly region through the European Space Agency's Copernicus Open Access Hub; standardizing the SAR and SLC data using the ISCE2 software stack; performing high-precision temporal deformation field calculations on the standardized SAR and SLC data using Small Baseline Sets (SBAS-InSAR) technology to ultimately obtain the vertical and east-west deformation components of the uranium anomaly region; and quantitatively analyzing the tectonic stress field and assessing the tectonic activity intensity based on the vertical and east-west deformation components, respectively. Furthermore, the standardized data processing specifically includes: applying precise orbital ephemeris (POD) to correct orbital errors; achieving sub-pixel-level registration of master-slave images through a cross-correlation algorithm (residual ≤ 0.125 pixels); generating interferograms using complex data conjugate multiplication, and introducing SRTM 30-meter DEM data to eliminate terrain effects based on a terrain phase model; recovering the true phase using a minimum cost flow algorithm (coherence threshold > 0.25); and constructing an elevation regression model by fusing ERA5 reanalysis meteorological data to effectively suppress atmospheric delay errors and ensure the accuracy of deformation signal extraction.

[0060] The quantitative analysis of the constructed stress field specifically involves: the strain energy field calculation begins with the displacement gradient tensor. The spatial differential of , where, Represents the components of the strain tensor and describes deformation; , They represent the displacement field at... direction and Component of direction; , They represent direction and Spatial coordinates of direction; partial derivatives Represents displacement components Along coordinates Rate of change of direction, partial derivative Represents displacement components Along coordinates The rate of change of direction, i.e., the displacement gradient. The central difference method is used to solve for it within a 1 km grid. Based on elasticity theory, the strain energy density is defined as... ,in, Represents strain energy density. Represents the stress tensor components. Represents the components of the strain tensor. , Let represent the spatial coordinate directions, and Einstein's summation convention be used, where the stress-strain relationship follows the generalized Hooke's law.

[0061] The tectonic stress field inversion employs a layered viscoelastic finite element model, establishing a three-dimensional mesh for the upper crust (0-20 km), middle crust (20-35 km), and lower crust (>35 km). Material parameters are assigned from geophysical survey results of the uranium anomaly region (in this embodiment, elastic modulus 80 / 60 / 100 GPa, Poisson's ratio 0.25 / 0.28 / 0.24, and viscosity coefficient 1×10²² / 5×10²¹ / 1×10²³ Pa·s). Boundary conditions are set based on a plate tectonics model, and the equilibrium equations are solved. Obtain the stress tensor, where, The divergence of the stress tensor represents the net stress per unit volume. Represents the stress tensor. This represents the body force vector; a value of 0 on the right-hand side of the formula indicates equilibrium. The principal stress directions are determined through eigenvalue decomposition. Inversion, in which, Represents the stress tensor. Represents the eigenvector (corresponding to the principal stress direction). The eigenvalues ​​(corresponding to the magnitudes of principal stresses) are represented by an improved particle swarm optimization (IPSO) algorithm to optimize the friction coefficient, thereby improving the model strain rate. Compared with InSAR observations Minimizing the L2 norm reveals the orientation of the principal stress axes.

[0062] The assessment of tectonic activity intensity specifically involves constructing a Comprehensive Activity Index (TAI) to quantify tectonic activity intensity. The formula for calculating the Comprehensive Activity Index (TAI) is as follows: in, It represents strain energy density, reflecting the degree of energy accumulation; This represents the maximum value of the strain energy density; Indicates the sliding rate, indicating fracture activity (>5 mm / year is considered an active fracture); This represents the maximum sliding speed. Indicates the change in Coulomb stress. Characterizes the intensity of fault activity. This indicates the change in shear stress at the fracture surface. Indicates the fault friction coefficient. This indicates the change in normal stress along the fault plane; This represents the maximum value of the change in Coulomb stress.

[0063] Based on the Comprehensive Activity Index (TAI) value, uranium anomaly regions are divided into four categories of active regions: stable region (<0.3), weakly active region (0.3-0.6), moderately active region (0.6-0.9), and strongly active region (>0.9).

[0064] Step S33 integrates the surface water flow direction and tectonic stress field direction in the uranium anomaly region, and conducts comparative analysis with Cenozoic tectonic events in the uranium anomaly region to determine the flow direction characteristics of ore-forming fluids in the Cenozoic ore-bearing strata. Based on this, by comprehensively considering the spatial distribution of local uplifts in the surface water flow direction and the activity of faults around the local uplifts, the constraints of local uplifts on the runoff of ore-forming fluids are analyzed, thereby reconstructing the migration path of ore-forming fluids in the Cenozoic ore-bearing strata.

[0065] Specifically, through multi-source remote sensing data fusion and geodynamic analysis, the migration path and control mechanism of ore-forming fluids were systematically analyzed. First, based on Sentinel-1 InSAR deformation field data, the direction of the regional tectonic stress field was inverted, clarifying the control law of the dominant direction of the maximum principal stress on the deep fluid transport channel—fluids preferentially migrate along tensional fractures or strike-slip faults at a 45° angle to them. Second, combining the surface water flow direction and runoff network extracted from DEM data (calculated based on the D8 algorithm to determine the cumulative flow direction), the patterns of paleochannels and catchment basins were identified; simultaneously, a three-dimensional morphological model of local uplifts was constructed using ALOS PALSAR topographic data to analyze the fault density and tectonic activity intensity around the uplifts. Furthermore, by coupling the three elements of tectonic stress field, topographic uplift, and activity intensity, a dynamic model of ore-forming fluid runoff was established: ① Tectonic stress drives the upwelling of deep thermal fluids along high-permeability faults; ② Surface runoff carries uranium-bearing water to converge towards low-potential areas around local uplifts; ③ Tectonic activity intensity controls fluid migration efficiency. Finally, the migration path of ore-forming fluids in the Cenozoic ore-bearing strata was reconstructed through hydrological numerical simulation (such as TOUGHREACT software). The reliability of the path was verified by combining the spatial distribution of uranium mineralization, hydrothermal alteration mineral assemblage, and tectonic lithofacies mapping. The three-stage migration model of "deep upwelling - surface convergence - margin sedimentation" was revealed, providing a dynamic basis for the prediction of concealed uranium deposits.

[0066] Step S4 focuses on the complex correlation and synergistic response mechanism between multi-source remote sensing data (including spectral, high-resolution imagery, InSAR, etc.) and key ore-controlling elements of sandstone-type uranium deposits, and constructs a sandstone-type uranium deposit mineralization prediction model based on remote sensing data characteristics.

[0067] The construction of a sandstone-type uranium mineralization model based on remote sensing data features specifically involves: defining the oxidation front by iron staining anomaly boundaries; identifying reduction barriers (e.g., organic matter, pyrite) using shortwave infrared / radar; accurately locating the oxidation-reduction transition zone; verifying the spatial configuration relationship between the uranium source area, migration path, and oxidation-reduction transition zone (e.g., whether the fault cuts through the sand body, whether the oxidation-reduction transition zone is located downstream of the migration path) using GIS spatial overlay and buffer analysis; and fusing and quantifying variables derived from multi-source remote sensing (e.g., alteration index, tectonic density, sand body permeability, oxidation-reduction gradient, etc.) using machine learning methods (e.g., random forest, XGBoost) to generate a mineralization favorability probability map.

[0068] See Figure 2 This is a uranium mineralization prediction heatmap obtained from a uranium mineralization potential area in the Erlian Basin, based on an intelligent mineralization prediction method that integrates multiple data sources and sink process analysis provided by an embodiment of the present invention. Figure 2As shown, the horizontal axis represents longitude, and the vertical axis represents latitude. The blue to red color blocks represent mineralization probabilities; higher red values ​​indicate greater mineralization potential, while lower blue values ​​indicate less mineralization potential. The uranium mineralization potential area of ​​the Erlian Basin is located at the intersection of the Ulanqab Depression and the Manite Depression. Several sandstone-type uranium deposits have been discovered in this area (such as...). Figure 2 (as shown by the pentagram in the middle), from Figure 2 As can be seen from the data, the Chuanjing Depression at the southwestern end of the Erlian Basin has significant mineralization potential. Therefore, it can be determined that the intelligent mineralization prediction method integrating multi-source data and sink process analysis provided in this application can identify areas with significant uranium mineralization potential in the Erlian Basin, providing a basis for sandstone-type uranium exploration.

[0069] The above description is merely a specific embodiment of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be included within the protection scope of this invention. The protection scope of this invention should be determined by the scope of the claims.

Claims

1. A smart mineralization prediction method integrating multiple data sources and sink process analysis, characterized in that, The method includes: Step S1: Acquire multi-source remote sensing data, airborne gamma spectral data, gravity and magnetic data, and geological data covering the target area; Step S2: Delineate potential uranium source regions in the target area using a bidirectional feature fusion network; obtain mineral composition information of potential uranium source regions by performing mineral inversion using hyperspectral images; and obtain uranium anomaly regions in potential uranium source regions through a collaborative framework of airborne gamma spectroscopy and ground verification. Step S3: Within the uranium anomaly region, surface water flow direction information is extracted based on topographic data from multi-source remote sensing data. The undulating morphology of the basement is inverted using gravity and magnetic data, and the tectonic stress field is inverted based on InSAR deformation data. A coupled analysis process of surface water flow-basement morphology-tectonic stress field is constructed to analyze the uranium migration channel and transport mechanism. Step S4 focuses on the complex correlation and synergistic response mechanism between multi-source remote sensing data and key ore-controlling elements of sandstone-type uranium deposits, and constructs a mineralization prediction model for sandstone-type uranium deposits based on remote sensing data characteristics.

2. The intelligent mineralization prediction method based on the integration of multiple data sources and the sinking process analysis as described in claim 1, characterized in that, Step S2 includes: Step S21: Preprocess the multi-source optical and SAR images of the target area, select the main information bands and stack them to form a multi-source remote sensing fusion feature map; input the multi-source remote sensing fusion feature map into a bidirectional feature fusion network, which includes a multi-scale feature extraction backbone network, a context and detail co-enhancement module and a cross-scale feature fusion network. The multi-scale feature extraction backbone network is composed of multiple stacked C3k2 modules. PPA modules are embedded in the key layers of the multi-scale feature extraction backbone network to form a context and detail co-enhancement module. The cross-scale feature fusion network is a BiFPN. The multi-level feature map enhanced by the PPA module is fused bidirectionally from top to bottom and from bottom to top to output the potential uranium source area delineation result. Step S22: In the potential uranium source region, mineral inversion is performed using hyperspectral images, and the mineral inversion results are output to obtain mineral composition information of the potential uranium source region; Step S23: Based on airborne gamma-ray spectroscopy data and combined with spatial clustering algorithms, delineate uranium anomaly regions in potential uranium source areas; within the uranium anomaly regions, combine mineral inversion results with ground verification to achieve high-precision prediction of the source rock distribution in the uranium anomaly regions.

3. The intelligent mineralization prediction method integrating multiple data sources and sink process analysis as described in claim 2, characterized in that, Step S22 specifically involves: constructing a diffusion model-guided hyperspectral mineral inversion network PMNet, which includes a multi-scale depth-separable convolutional branch, a Vision Transformer branch, a frequency domain feature interaction module, and a Vision Mamba sequence modeling module; inputting hyperspectral images of potential uranium source regions into the hyperspectral mineral inversion network PMNet; extracting local spectral detail features through the multi-scale depth-separable convolutional branch and extracting global spatial context features through the Vision Transformer branch; performing cross-domain fusion of the output features of the multi-scale depth-separable convolutional branch and the Vision Transformer branch through the frequency domain feature interaction module; and performing selective state updates and sequence modeling by the Vision Mamba sequence modeling module to output the mineral inversion results.

4. The intelligent mineralization prediction method integrating multiple data sources and sink process analysis as described in claim 2, characterized in that, Step S23 specifically involves: Step S231: Denoise and correct the airborne gamma spectrum data. Use the NASVD method to perform singular value decomposition on the airborne gamma spectrum data, select the principal component components with a cumulative variance contribution rate of not less than 95% for signal reconstruction, and obtain the reconstructed airborne gamma spectrum data. Use the MNF method to sort the reconstructed airborne gamma spectrum data according to the signal-to-noise ratio, and obtain the airborne gamma reconstruction data. Step S232: By extracting background samples from potential uranium source areas, the mean and standard deviation of uranium content and Th / U ratio are calculated respectively to construct a dual threshold model for uranium content and Th / U ratio; based on airborne gamma reconstruction data, the grid cells that satisfy the dual threshold model for uranium content and Th / U ratio in potential uranium source areas are spatially clustered using the K-means clustering algorithm to extract uranium anomaly regions. Step S233: Select exposed surface areas within the uranium anomaly region, and verify the anomaly by combining high-resolution surveys, geological mapping, and detailed evaluation to achieve high-precision prediction of the source rock distribution in the uranium anomaly region.

5. The intelligent mineralization prediction method integrating multiple data sources and sink process analysis according to claim 4, characterized in that, In the dual-threshold model of uranium content and Th / U ratio, the uranium content threshold is defined as: ,in, Indicates the uranium content threshold. This represents the average uranium content. The standard deviation of uranium content is represented; the threshold for the Th / U ratio is defined as: ,in, Indicates the threshold value of the Th / U ratio. This represents the mean Th / U ratio. This represents the standard deviation of the Th / U ratio.

6. The intelligent mineralization prediction method integrating multiple data sources and sink process analysis according to claim 4, characterized in that, The high-resolution survey specifically involves using a high-resolution handheld gamma spectrometer for detailed surveys, achieving quantitative monitoring of U, Th, and K anomalies through dense sampling and high-frequency sampling; the geological mapping specifically involves conducting geological mapping at a scale of 1:10,000; and the detailed evaluation specifically involves quantitatively analyzing the in-situ content, occurrence form, and mineralogical characteristics of uranium.

7. The intelligent mineralization prediction method integrating multiple data sources and sink process analysis according to claim 1, characterized in that, Step S3 specifically involves: Step S31: Based on high-resolution optical data and SAR remote sensing data, the surface water flow direction in the uranium anomaly area is analyzed by remote sensing identification of key topographic parameters and division of topographic and geomorphic units, extraction of valley network and intelligent division of catchment basins, and estimation and verification of surface water flow direction. Step S32: Use gravity and magnetic data to invert the basement undulation morphology of the uranium anomaly region and characterize the spatial distribution of the local uplift in the uranium anomaly region; based on the tectonic stress field inversion and activity intensity assessment of InSAR deformation data, interpret the activity of the faults around the local uplift in the uranium anomaly region and the direction of the tectonic stress field. Step S33: Integrate the surface water flow direction and tectonic stress field direction in the uranium anomaly region, and conduct comparative analysis with Cenozoic tectonic events in the uranium anomaly region to determine the flow direction characteristics of ore-forming fluids in the Cenozoic ore-bearing strata; comprehensively analyze the spatial distribution of local uplifts in the surface water flow direction and the activity of faults around the local uplifts to analyze the constraints of local uplifts on the runoff of ore-forming fluids and reconstruct the migration path of ore-forming fluids in the Cenozoic ore-bearing strata.

8. The intelligent mineralization prediction method based on the integration of multiple data sources and the sinking process analysis as described in claim 7, characterized in that, The key terrain parameter remote sensing identification and terrain and geomorphological unit division are specifically as follows: 12.5m resolution DEM data is generated using Sentinel SAR data, and the DEM data is preprocessed; key terrain parameters in the DEM data are identified by remote sensing; based on the key terrain parameters, feature lines such as valley lines and ridge lines in the DEM data are extracted using the geometric morphology method, and the slope threshold in the DEM data is determined using the natural discontinuity classification method; based on the extracted ridge lines and valley lines, the uranium anomaly area is divided into linear feature areas and non-linear feature areas. In the linear feature areas, ridges, valleys, and saddles are delineated according to the feature line morphology and local slope transitions. In the non-linear feature areas, steep slopes and gentle slopes are divided according to the slope threshold, ultimately forming five types of terrain and geomorphological units: ridges, valleys, saddles, steep slopes, and gentle slopes.

9. The intelligent mineralization prediction method based on the integration of multiple data sources and sink process analysis as described in claim 8, characterized in that, Preprocessing of DEM data includes: performing Gaussian filtering on the DEM data, and performing depression filling on the Gaussian-filtered DEM data.

10. The intelligent mineralization prediction method based on the integration of multiple data sources and sink process analysis as described in claim 1, characterized in that, Specifically, step S4 involves defining the oxidation front by the iron staining anomaly boundary, identifying the reduction barrier using shortwave infrared / radar, and accurately locating the oxidation-reduction transition zone; using GIS spatial overlay and buffer analysis to verify the spatial configuration relationship between the uranium source area, migration path, and oxidation-reduction transition zone; and using machine learning methods to fuse and quantify variables derived from multi-source remote sensing to generate a mineralization favorability probability map.

Images