Metal ore body identification method based on remote sensing interpretation

Abstract

The invention discloses a metal ore body identification method based on remote sensing interpretation, relates to the technical field of mineral resource exploration, and aims to solve the problems of low identification precision, neglect of geological laws and poor generalization caused by dependence on single remote sensing data in the prior art. After standardization processing, spectrum and texture features are fused to generate a composite vector; extracting a multi-scale mineralization signal by using a convolutional neural network, and interpreting and constructing an ore control area in combination with a terrain gradient and a curvature; alteration zoning is identified through spectrum angle matching, and a favorable mineralization area is determined by combining lithology combination analysis; and dynamically weighting geological elements by adopting a deep learning fusion network, outputting an ore body position according to a comprehensive score, and finally optimizing a boundary through spatial clustering and morphological filtering. The method significantly improves the automation degree and reliability of ore body identification in a complex environment, and is suitable for metal mineral exploration target area delineation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ore body exploration, and particularly relates to a metal ore body identification method based on remote sensing interpretation. BACKGROUND

[0002] The current mineral resource exploration field mainly adopts a spectral feature matching method based on a single remote sensing data source to identify metal ore bodies. This method has significant defects in practical application: in vegetation-covered areas or surface water distribution areas, single spectral data is easily disturbed by environmental factors, causing mineral spectral features to be masked or distorted, resulting in a large number of misjudgment results; in cross-regional geological exploration scenarios, a supervised learning model relies on labeled samples in a specific area, and when migrating from a dry area to a humid area, the model recognition accuracy sharply decreases due to differences in lithological combination and alteration types, and samples need to be collected again to train the model; for the identification of concealed ore bodies, the existing technology ignores three-dimensional spatial geological knowledge such as alteration zoning patterns and structure ore-controlling rules, and cannot establish a correlation between surface remote sensing information and deep mineralization mechanism, for example, a copper ore body covered by sedimentary layers is missed due to a lack of deep structure feature interpretation, and the missing detection rate is significantly increased.

[0003] The root cause of the above defects lies in the double bottlenecks in the technical field: first, there is a lack of effective fusion mechanism when multi-source remote sensing data (including spectral data, radar data and thermal infrared data) have information redundancy and response conflicts, for example, the high emissivity feature of iron staining alteration minerals in the thermal infrared band and the absorption feature of spectral data are contradictory, causing the mineralization indication information to be incorrectly selected; second, key ore-forming rules in geological expert knowledge (such as fracture zone controlling alteration zoning spatial distribution, lithological combination limiting mineralization type, etc.) are difficult to be converted into quantifiable calculation models, and traditional algorithms cannot incorporate three-dimensional spatial knowledge such as structure-alteration coupling relationship and lithology-mineralization combination features into the decision-making process. Due to insufficient information fusion and expression barriers of geological knowledge, the existing technology forces the ore body identification process to rely heavily on artificial experience interpretation, which not only is low in efficiency, but also makes it difficult to realize standardized and automated exploration processes, seriously restricting the mineral resource guarantee capability. SUMMARY

[0004] To solve the above technical problems, the application provides a metal ore body identification method based on remote sensing interpretation to solve the problems existing in the prior art.

[0005] In the first aspect, to achieve the above object, the application provides a metal ore body identification method based on remote sensing interpretation, comprising the following steps:

[0006] Obtain multispectral remote sensing images, hyperspectral data, digital elevation models and geological structure maps, and after geometric correction and radiation calibration, unify the spatial resolution and coordinate system through a registration algorithm to generate a standardized multi-source remote sensing data set;

[0007] Based on the data set, the spectral feature vector is extracted by principal component analysis, and the spatial texture feature is obtained by calculating the gray level co-occurrence matrix parameter through the texture analysis operator, and when the correlation coefficient of the spectrum and the texture feature is greater than a preset threshold, a composite feature vector is formed by fusion;

[0008] For the composite feature vector, a multi-scale feature extraction model is constructed by a convolutional neural network, local and global spatial features are extracted by using different convolution kernel sizes, and when the feature response intensity is more than three times the standard deviation of the background noise, it is determined as an effective mineralization signal;

[0009] According to the distribution of the effective mineralization signal, the terrain gradient and curvature information are extracted in combination with the digital elevation model, the fracture zone and fold structure are identified through the geological structure interpretation algorithm, and when the spatial overlap degree of the mineralization signal and the structure control factor is greater than a set proportion, the structure ore-controlling area is determined;

[0010] The spectral angle matching algorithm is used to identify the altered minerals in the structure ore-controlling area, the altered zoning information of hydroxyl alteration, iron staining alteration and clay alteration is extracted by comparing with the standard mineral spectral library, and the alteration intensity distribution map is obtained;

[0011] Based on the alteration intensity distribution map, the lithology spectral feature database is combined, the support vector machine classifier is used to identify the lithology unit, the favorable ore-forming lithology combination type is determined through lithology combination analysis, and the lithology control factor is obtained;

[0012] The structure control factor, alteration zoning information and lithology control factor are input into the deep learning fusion network, the geological element weight is allocated through the attention mechanism, and when the comprehensive geological constraint score exceeds a preset ore-forming threshold, the potential ore body position prediction result is output;

[0013] The spatial clustering analysis is carried out on the potential ore body position, the density clustering algorithm is used to remove isolated abnormal points, the ore body boundary is optimized through morphological filtering, and the spatial distribution range of the metal ore body is determined.

[0014] In the second aspect, the application also provides a metal ore body identification system based on remote sensing interpretation, which is used to implement a metal ore body identification method based on remote sensing interpretation, and the system comprises:

[0015] A data preprocessing module is used to acquire multispectral remote sensing images, hyperspectral data, digital elevation models and geological structure maps, and after geometric correction and radiation calibration, the spatial resolution and coordinate system are unified to generate a standardized multi-source remote sensing data set;

[0016] A feature fusion module is used to extract spectral feature vectors and spatial texture features based on the standardized data set, and when the correlation coefficient is greater than a preset threshold, a composite feature vector is generated by fusion;

[0017] A mineralization extraction module is configured to process the composite feature vector through a multi-scale feature extraction model of a convolutional neural network, and output an effective mineralization signal when a feature response intensity exceeds three times a standard deviation of background noise;

[0018] A structure control module is configured to extract terrain gradient and curvature information in combination with a digital elevation model, identify a fault zone and a fold structure, and determine a structure-controlled mineralization area when a spatial overlap degree of the mineralization signal and the structure control factor is greater than a set proportion;

[0019] An alteration recognition module is configured to compare a standard mineral spectrum library by using a spectral angle matching algorithm, extract hydroxyl alteration, iron staining alteration and clay alteration zoning information in the structure-controlled mineralization area, and generate an alteration intensity distribution map;

[0020] A lithology analysis module is configured to identify a lithology unit in combination with a lithology spectral feature database, determine a favorable ore-forming lithology combination type through lithology combination analysis, and output a lithology control factor;

[0021] A decision output module is configured to input the structure control factor, the alteration zoning information and the lithology control factor into a deep learning fusion network, assign weights through an attention mechanism, and output a potential ore body position when a comprehensive geological constraint score exceeds a mineralization threshold value;

[0022] A result optimization module is configured to perform spatial clustering analysis on the potential ore body position, eliminate isolated abnormal points and optimize an ore body boundary, and determine a spatial distribution range of the metal ore body.

[0023] In a third aspect, the present application further provides a computer terminal device, comprising:

[0024] one or more processors;

[0025] a memory coupled to the processor, for storing one or more programs;

[0026] When the one or more programs are executed by the one or more processors, the one or more processors implement a metal ore body identification method based on remote sensing interpretation.

[0027] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement a metal ore body identification method based on remote sensing interpretation.

[0028] Compared with the prior art, the present application has the following advantages and technical effects:

[0029] The present invention provides a metal ore body identification method based on remote sensing interpretation. By combining multi-source remote sensing data fusion processing with a deep learning mechanism, the method effectively weakens the influence of surface cover and environmental interference on the identification results, and significantly improves the extraction accuracy of mineralization signals. A quantifiable constraint model constructed based on geological laws converts mineralization factors such as structural control of ore, alteration zoning and lithologic combination into calculable parameters, so that the identification results conform to the logic of geological theory. The automated processing flow replaces the manual interpretation link, which reduces reliance on subjective experience while ensuring the reliability of identification in complex geological environments, and ultimately achieves efficient and accurate delineation of the spatial position of metal ore bodies. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0031] Figure 1 Schematic diagram of the overall process of the metal ore body identification method based on multi-source remote sensing data fusion according to an embodiment of the present invention;

[0032] Figure 2 This is a flow chart of the spatial coupling relationship between alteration zoning and structural ore-controlling regions according to an embodiment of the present invention;

[0033] Figure 3 Schematic diagram of the process of identifying the spatial distribution of potential ore bodies in an exploration area according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] Example 1

[0037] like Figure 1 As shown, this embodiment provides a metal ore body identification method based on remote sensing interpretation, including:

[0038] Acquire multispectral remote sensing images, hyperspectral data, digital elevation models, and geological structure maps, and after geometric correction and radiometric calibration, use registration algorithms to unify spatial resolution and coordinate systems to generate standardized multi-source remote sensing datasets.

[0039] Based on the data set, the spectral feature vector is extracted by principal component analysis, and the spatial texture feature is obtained by calculating the gray level co-occurrence matrix parameter through the texture analysis operator; when the correlation coefficient of the spectrum and the texture feature is greater than a preset threshold, a composite feature vector is formed by fusion;

[0040] For the composite feature vector, a multi-scale feature extraction model is constructed by a convolutional neural network, local and global spatial features are extracted by using different convolution kernel sizes, and when the feature response intensity exceeds three times the standard deviation of the background noise, it is determined as an effective mineralization signal;

[0041] According to the distribution of the effective mineralization signal, the terrain gradient and curvature information are extracted in combination with the digital elevation model, the fracture zone and fold structure are identified through the geological structure interpretation algorithm, and when the spatial overlap degree of the mineralization signal and the structure control factor is greater than a set proportion, the structure ore-controlling area is determined;

[0042] The spectral angle matching algorithm is used to identify the altered minerals in the structure ore-controlling area, the alteration zoning information of hydroxyl alteration, iron staining alteration and clay alteration is extracted by comparing with the standard mineral spectral library, and the alteration intensity distribution map is obtained;

[0043] Based on the alteration intensity distribution map, the lithology spectral feature database is combined, the lithology unit is identified by using the support vector machine classifier, the favorable ore-forming lithology combination type is determined through lithology combination analysis, and the lithology control factor is obtained;

[0044] The structure control factor, alteration zoning information and lithology control factor are input into the deep learning fusion network, the geological element weight is allocated through the attention mechanism, and when the comprehensive geological constraint score exceeds a preset ore-forming threshold, the potential ore body position prediction result is output;

[0045] The spatial clustering analysis is performed on the potential ore body position, the density clustering algorithm is used to remove isolated abnormal points, the ore body boundary is optimized through morphological filtering, and the spatial distribution range of the metal ore body is determined.

[0046] As an embodiment in the embodiment, the process of generating the standardized multi-source remote sensing data set includes:

[0047] The position deviation of the original data set is corrected by a geometric transformation tool;

[0048] Radiometric calibration is performed by using a radiometric calibration tool, and when the radiometric feature value exceeds a preset threshold, linear interpolation is used for calibration;

[0049] The spatial resolution is unified by using a resampling method through a spatial registration tool, and the reference coordinate system is aligned by using a coordinate conversion tool;

[0050] The multispectral image and hyperspectral data are integrated by using an image fusion tool. When the integrated features are inconsistent with the digital elevation model or the geological structure map, a local adjustment algorithm is used for optimization.

[0051] Specifically, in S101, multispectral remote sensing images, hyperspectral data, digital elevation models, geological structure maps, and other multi-source remote sensing data are obtained. After geometric correction and radiation calibration, a registration algorithm is used to unify the spatial resolution and coordinate system of the multi-source data, and a standardized multi-source remote sensing data set is obtained.

[0052] According to the source of the multi-source remote sensing data, the original data set is obtained from the multispectral remote sensing image, the hyperspectral data, the digital elevation model, and the geological structure map. The position deviation of the original data set is corrected by using a geometric transformation tool, and a preliminary corrected multi-source data set is obtained. The radiation characteristic value is extracted from the preliminary corrected multi-source data set by using a radiation calibration tool. If the radiation characteristic value exceeds the preset threshold, the linear interpolation method is used for calibration, and a radiation consistent multi-source corrected data set is obtained. The radiation consistent multi-source corrected data set is processed by using a spatial registration tool. The spatial resolution of the multi-source corrected data set is unified by using a resampling method, and the reference coordinate system is aligned by using a coordinate conversion tool, and a spatial standardized multi-source remote sensing data set is obtained. The multispectral image and hyperspectral data in the spatial standardized multi-source remote sensing data set are integrated by using an image fusion tool. If the integrated features are inconsistent with the spatial features of the digital elevation model and the geological structure map, a local adjustment algorithm is used for optimization, and a final standardized multi-source remote sensing data set is determined.

[0053] For example, the processing of multi-source remote sensing data involves multiple technical links from obtaining the final standardized data set from multi-source data, covering geometric correction, radiation calibration, spatial registration, image fusion, and optimization adjustment. The implementation method and technical effect of each link are analyzed below through specific examples, focusing on a single business scenario of remote sensing image processing, and focusing on geological exploration applications.

[0054] As an optional implementation, in geological exploration, the original data set may include Landsat-8 multispectral images, 30-meter resolution hyperspectral data, SRTM digital elevation models, and regional geological structure maps. The geometric transformation tool is used to correct the position deviation. Assuming that the Landsat-8 image is deviated due to the inclination of the sensor, the image can be aligned with the high-precision base map by using the control point matching method, and the position deviation after correction is controlled within 1 meter. This method ensures that the geographical positions of data from different sources are consistent, which is convenient for subsequent superposition analysis and significantly improves the accuracy of geological feature recognition.

[0055] In a possible implementation, the radiation calibration tool extracts the radiation characteristic value from the preliminary corrected data.

[0056] As an optional implementation, the reflectivity of the red light band of Landsat-8 may be higher than the preset threshold 0.3 due to atmospheric scattering. A linear interpolation method is used in combination with the ground measured reflectivity data to calibrate the image pixel value and generate a radiation consistent data set. This calibration eliminates the atmospheric influence, makes the radiation values of the multispectral and hyperspectral data comparable, enhances the extraction effect of the mineral spectral characteristics, and provides a reliable basis for geological classification.

[0057] As an optional implementation, the image fusion tool integrates multispectral images and hyperspectral data, extracts spectral features using principal component analysis, and generates a fused image. If the eigenvalue of the fused image is inconsistent with the spatial features of the digital elevation model or the geological structure map, such as the deviation of the mineral distribution indicated by the hyperspectral from the fault line position of the geological map, local adjustment algorithm can be used for optimization.

[0058] Optionally, the local registration based on the least squares method can adjust the deviation area to make the fused image consistent with the spatial features of the geological structure map. This optimization enhances the spatial consistency of the data and helps to accurately identify the distribution of the ore deposit.

[0059] It can be understood that the above processing link forms a standardized multi-source remote sensing data set, which significantly improves the identification accuracy of mineral distribution and structural features in geological exploration.

[0060] As an implementation in this embodiment, the process of fusing to form a composite feature vector includes:

[0061] Spectral analysis tools are used to extract spectral feature data, and a gray matrix parameter is generated through a texture calculation tool;

[0062] A correlation analysis tool is used to calculate the correlation coefficient of spectral and texture features;

[0063] When the correlation coefficient is greater than a preset threshold, a feature fusion tool is used to integrate into a composite vector;

[0064] A data integration tool is used for spatial feature comparison, and when the spatial consistency parameter meets the preset standard, a spatial calibrated composite vector is output.

[0065] Specifically, S102, according to the standardized multi-source remote sensing data set, a principal component analysis method is used to extract a spectral feature vector, and a gray level co-occurrence matrix parameter is calculated through a texture analysis operator to obtain a spatial texture feature. If the correlation coefficient of the spectral feature and the texture feature is greater than a preset threshold, a composite feature vector is fused.

[0066] According to the standardized multi-source remote sensing dataset, spectral feature data is extracted by a spectral analysis tool, and a gray matrix parameter is generated by a texture calculation tool to obtain a preliminary spectral feature and texture feature dataset. For the preliminary spectral feature and texture feature dataset, a correlation analysis tool is used to calculate the correlation coefficient between the spectral feature and the texture feature. If the correlation coefficient is greater than a preset threshold, an initial composite vector dataset is integrated by a feature fusion tool. Spatial feature comparison is performed on the initial composite vector dataset by a data integration tool to obtain a spatial consistency parameter related to the image data. If the parameter meets the preset standard, a spatially calibrated composite vector dataset is obtained. The spatially calibrated composite vector dataset is adjusted in detail by a parameter optimization tool, and the deviation of the spectral feature and the texture feature is locally corrected to determine a final composite feature vector dataset.

[0067] As an optional implementation, in the field of geological exploration, based on the processing of the standardized multi-source remote sensing dataset, the spectral analysis tool is used to extract spectral feature data, which is a key link. Spectral features reflect the spectral reflectance characteristics of surface materials and are commonly used to identify mineral types. Assuming that when processing multispectral images of a certain area, the tool analyzes specific bands such as near-infrared bands and extracts spectral absorption features related to iron oxides. The reflectance of this feature in certain bands may show a significant low value trend, which is preliminarily judged to be related to the distribution of iron-containing minerals.

[0068] As an optional implementation, the process of the texture calculation tool generating the gray matrix parameter is mainly to capture the spatial structure information of the image. In geological exploration, texture features can help identify the roughness or fracture distribution of surface rocks. Assuming that for image data of the same area, the tool calculates the contrast parameter in the gray level co-occurrence matrix and finds that the contrast value is high in some areas, which may indicate that there are obvious fracture zones or rock body boundaries on the surface, providing spatial clues for subsequent analysis.

[0069] As an optional implementation, for the correlation analysis of the spectral feature and texture feature dataset, a correlation analysis tool is used to calculate the correlation coefficient between the two. If the preset threshold is 0.7 and the calculation result shows that the correlation coefficient between the spectral feature and the texture feature is 0.8, it indicates that there is a strong correlation between the two. At this time, the feature fusion tool is used to integrate the two into an initial composite vector dataset. This fusion can integrate spectral and texture information to form a more comprehensive geological feature description.

[0070] As an optional implementation, the data integration tool performs spatial feature comparisons primarily to ensure spatial consistency between the composite vector dataset and the image data. If, during the comparison process, the spatial consistency parameters for certain areas are found to be low, failing to meet a pre-defined standard (e.g., a deviation of less than 2 meters), further adjustments to the data registration method are necessary until the parameters meet the standard, resulting in a spatially calibrated composite vector dataset. This calibration ensures the spatial reliability of the data.

[0071] As an optional implementation, the parameter optimization tool can locally correct for deviations in spectral and textural features when making detailed adjustments to the spatially calibrated dataset. For example, if in certain areas, the mineral distribution indicated by spectral features slightly deviates from the location of fault zones indicated by textural features, the tool can use a local weighted adjustment method to correct for these deviations and ultimately determine a composite feature vector dataset. This correction improves the overall consistency of the data and provides a more accurate basis for mineral distribution and structural analysis in geological exploration.

[0072] As an optional implementation, the above processing flow can also be adapted for data of varying resolutions. For example, if the original imagery has a lower resolution in some areas, resulting in inadequate texture feature extraction, the optimization tool can interpolate these areas with surrounding high-resolution data to provide additional detail. This approach further enhances data integrity and helps reveal more comprehensive geological features.

[0073] It's important to note that the entire process is closely centered around geological exploration applications, with each step designed to enhance the data's ability to represent geological features. The fusion of spectral and textural features, spatial calibration, and detail optimization all contribute to the construction of a high-quality composite feature vector dataset, laying a solid foundation for subsequent geological analysis.

[0074] S103. For the composite feature vector, a multi-scale feature extraction model is constructed through a convolutional neural network. Local and global spatial features are extracted using different convolution kernel sizes to obtain multi-level mineralization indication features. If the characteristic response intensity exceeds three times the standard deviation of the background noise, it is judged to be a valid mineralization signal.

[0075] According to the composite feature vector, an image processing tool is used to perform hierarchical processing on the data, different sizes of convolution kernel tools are applied to different levels of the data respectively, corresponding local spatial features and global spatial features are obtained, and a hierarchical spatial feature set is obtained. Through a feature integration tool, the hierarchical spatial feature set is fused, the local spatial features and the global spatial features are weighted and combined, and a comprehensive feature set with multi-scale characteristics is obtained. For the comprehensive feature set, a statistical calculation tool is used to extract feature response intensity data, and background noise standard deviation data is obtained from a pre-established background noise database. If the feature response intensity exceeds three times the background noise standard deviation, it is determined to be a preliminary mineralization indicating feature. Through a signal screening tool, the preliminary mineralization indicating feature is verified again, and the results of the spatial feature fusion are compared with the preset signal judgment standard to determine the final effective mineralization signal set.

[0076] As an optional implementation, in the field of geological exploration, hierarchical processing based on composite feature vectors is a key link to improve data analysis accuracy. For composite feature vectors generated from multi-source remote sensing data, an image processing tool can extract spatial information of different scales through hierarchical processing. Local spatial features usually capture subtle changes in images, such as the distribution of surface fissures, while global spatial features reflect larger-scale geological structures.

[0077] Optionally, when processing multispectral images of a certain area, the tool may apply a 3x3 pixel convolution kernel to low-resolution areas to extract local fissure textures, and use a 7x7 pixel convolution kernel to high-resolution areas to obtain the overall profile of the stratum boundary. This hierarchical processing ensures the integrity of features of different scales, providing diverse spatial information for subsequent analysis.

[0078] As an optional implementation, when the feature integration tool fuses the hierarchical spatial feature set, a weighted combination method is usually used. Local features may be given a higher weight, such as 0.6, due to their high-resolution details, while global features may be given a weight of 0.4 due to their overall nature.

[0079] In a possible implementation, for a certain iron ore exploration area, the fused comprehensive feature set can reflect both the details of surface fissures and the trends of regional geological structures. This multi-scale characteristic makes the analysis result more robust, suitable for feature extraction in complex geological environments.

[0080] Specifically, statistical calculation tools are used to extract the characteristic response intensity data from the comprehensive feature set and compare it with the standard deviation data from the background noise database. For example, if the background noise standard deviation is 0.1 and the characteristic response intensity in a certain area reaches 0.35, exceeding three times the standard deviation, it is preliminarily judged to be a mineralization indicator. This method quantifies the significance of the features and screens for reliable signals.

[0081] In one embodiment, for the preliminary mineralization indication features, the tool may combine the distribution pattern of the spatial features to check their consistency with the known mineralization pattern.

[0082] As an optional implementation, if the characteristics of a region show a linear distribution associated with copper mineralization and are consistent with the standard model, it is determined to be a valid mineralization signal. This verification process increases the confidence level of the signal and provides a precise basis for geological exploration decision-making.

[0083] As an optional implementation, the tool may use interpolation to supplement detailed information when processing low-resolution imagery. For example, if the image resolution of a certain area is 30 meters, and the texture features extracted are fuzzy, the tool can interpolate using surrounding high-resolution data to generate more detailed spatial features. This method enhances data continuity and helps reveal the distribution characteristics of concealed ore bodies.

[0084] It is understandable that the above process is closely centered around geological exploration needs. From layered processing to feature fusion and signal screening, each step is aimed at improving the accuracy and reliability of feature extraction and providing solid support for the identification of mineralization characteristics.

[0085] As an implementation method in this embodiment, the process of determining the structural ore-controlling area includes:

[0086] Process terrain height data through gradient calculation tools to generate terrain gradient sets;

[0087] Use curvature calculation tools to analyze terrain height data to generate terrain curvature sets;

[0088] Identify fault zones and fold structures by integrating terrain gradient and curvature sets through geological structure interpretation tools;

[0089] When the spatial overlap between the structural control factor and the effective mineralization signal is greater than the preset threshold ratio, the ore-controlling area is determined by spatial analysis tools.

[0090] Specifically, S104, based on the distribution of the effective mineralization signal, combined with the digital elevation model, the terrain gradient and curvature information are extracted, the fault zone and fold structure are identified through the geological structure interpretation algorithm, and the structural control factor is obtained. If the spatial overlap between the mineralization signal and the structural control factor is greater than the set ratio, the structurally controlled mineral area is determined.

[0091] According to the effective mineralization signal distribution, terrain height data is extracted by using a digital elevation model processing tool, and the terrain height data is processed by using a gradient calculation tool to obtain a terrain gradient set. According to the terrain height data, the terrain height data is analyzed by using a curvature calculation tool to obtain a terrain curvature set. The terrain gradient set and the terrain curvature set are comprehensively processed by using a geological structure interpretation tool to identify a fault zone and a fold structure to obtain a structure control factor set. If the spatial overlap degree of the structure control factor set and the effective mineralization signal distribution is greater than a preset threshold proportion, the overlapping region is processed by using a spatial analysis tool to determine a ore-controlling area set.

[0092] As an optional implementation, in geological exploration, the ore-controlling area identification based on the effective mineralization signal distribution needs to be combined with terrain feature analysis. When the digital elevation model processing tool extracts terrain height data, the terrain information is usually stored in the form of grid data, and the precision can reach 1 meter.

[0093] In an embodiment, for a certain copper exploration area, the tool generates an elevation data set from satellite remote sensing data, with a resolution of 5 meters and covering an area of 100 square kilometers. The elevation data is processed by using a gradient calculation tool to generate a terrain gradient set, reflecting the change of terrain slope.

[0094] As an optional implementation, the slope of a certain area increases from 5 degrees to 20 degrees, indicating that there may be a fault zone. Gradient calculation is usually based on the elevation difference of adjacent grids to generate a slope distribution map, which is convenient for identifying the terrain mutation area.

[0095] It can be understood that the curvature calculation tool analyzes the terrain curvature set to capture the concave-convex change of the terrain. Curvature reflects the shape of the ground surface, such as positive curvature indicating convex terrain and negative curvature indicating concave. In a certain gold mine area, curvature analysis shows a negative curvature area, which implies the presence of a sedimentary basin, which is consistent with the mineralization signal distribution. The tool generates a curvature distribution map by using quadratic curvature calculation, with the same resolution as the elevation data. The geological structure interpretation tool comprehensively processes the gradient and curvature sets to identify fault zones and fold structures.

[0096] As an optional implementation, the mineralization signal of a certain area is concentrated near the fault zone, and a 3-square-kilometer ore-controlling area is delineated by spatial analysis, which is preferentially arranged for drilling verification. This method significantly improves the exploration efficiency through comprehensive analysis of multi-dimensional terrain and mineralization signal, and provides a basis for accurately positioning the ore body.

[0097] As an implementation of the embodiment, the process of obtaining the alteration intensity distribution map includes:

[0098] The spectral angle matching tool is used to compare the standard mineral spectrum library to identify hydroxyl alteration, iron staining alteration, and clay alteration minerals;

[0099] The spatial distribution analysis tool is used to divide the distribution range of the alteration types, and extract alteration zonation information;

[0100] When the spatial distribution of at least one alteration type overlaps with the structural ore-controlling area by more than a preset threshold, the spectral feature extraction tool is used to perform weighted processing on the spectral data of the overlapping area.

[0101] Specifically, in S105, a spectral angle matching algorithm is used to identify alteration minerals in the structural ore-controlling area, and through comparison and analysis with a standard mineral spectral library, alteration zonation information such as hydroxyl alteration, iron staining alteration, and clay alteration is extracted, and an alteration intensity distribution map is obtained.

[0102] As shown in Figure 2 The spectral angle matching tool is used to process the geological remote sensing data, and in combination with the pre-established standard mineral spectral library, the similarity of spectral features is compared to identify hydroxyl alteration, iron staining alteration, and clay alteration minerals in the structural ore-controlling area, and obtain a mineral species distribution set. According to the mineral species distribution set, the spatial distribution analysis tool is used to divide the distribution range of the hydroxyl alteration, iron staining alteration, and clay alteration, and extract the spatial zonation characteristics of each alteration type, and obtain an alteration zonation information set. If the spatial distribution of at least one alteration type in the alteration zonation information set overlaps with the structural ore-controlling area by more than a preset threshold, the spectral feature extraction tool is used to perform weighted processing on the spectral data of the overlapping area, and obtain an alteration intensity distribution map.

[0103] As an optional implementation, in the processing of geological remote sensing data, the spectral angle matching tool is used to analyze the data, which is an important means to identify mineral species. The principle of spectral angle matching is to compare the angle difference between the spectral curve of each pixel in the remote sensing data and the spectral curve of the known minerals in the standard mineral spectral library, to determine the similarity, and then determine the mineral type.

[0104] In a possible implementation, for a certain copper exploration area, a technical person uses hyperspectral remote sensing data in combination with a pre-established standard spectral library to identify hydroxyl alteration minerals such as kaolinite, iron staining alteration minerals such as hematite, and clay alteration minerals such as montmorillonite, forming a mineral species distribution set. The data covers an area of 50 square kilometers with a resolution of 10 meters, ensuring the identification accuracy.

[0105] As an optional implementation, for the spatial distribution analysis of the mineral species distribution set, the spatial distribution analysis tool can effectively divide the distribution range of the hydroxyl alteration, iron staining alteration, and clay alteration, and extract the spatial zonation characteristics.

[0106] Specifically, in the above copper mine area, the analysis tool found that the hydroxyl alteration was mainly distributed near the fault zone, with a range of about 8 square kilometers, while the iron alteration was concentrated on both sides of the fault zone, with a distribution width of about 2 kilometers. Through spatial zoning analysis, the alteration zoning information set was obtained, showing the transition characteristics of alteration types from the center to the periphery, providing a spatial basis for subsequent research.

[0107] As an optional implementation, when judging the overlap degree of alteration zoning and structure ore-controlling area, the preset threshold is usually set to 50%. In this copper mine area, the spatial distribution of hydroxyl alteration has an overlap degree of 65% with the identified structure ore-controlling area, which exceeds the threshold, indicating that this alteration type is closely related to structure control. For this overlapping area, the spectral feature extraction tool performs weighted processing on the spectral data to highlight the characteristic values of specific wavebands, generating an alteration intensity distribution map. The figure shows that the alteration intensity is higher in the central area and gradually decreases in the periphery, with the intensity value decreasing from 80% in the core area to 30% in the edge, providing a key area guide for further exploration.

[0108] As an optional implementation, the weighted processing process of the spectral feature extraction tool mainly enhances the characteristic signals related to alteration minerals by assigning different weights to the spectral reflectance of different wavebands. In this copper mine case, the tool particularly enhances the 2.2 micrometer waveband signal related to hydroxyl alteration, and combined with the spatial distribution data, the generated alteration intensity distribution map clearly reflects the mineralization potential area. This method helps to accurately delineate the target area and improve the targeting of exploration.

[0109] S106、According to the alteration intensity distribution map, combined with the lithology spectral feature database, a support vector machine classifier is used to identify different lithology units, and a favorable ore-forming lithology combination type is determined through lithology combination analysis to obtain a lithology control factor.

[0110] According to the alteration intensity distribution map, combined with the pre-established lithology spectral feature database, a data comparison tool is used to extract features from the spectral data in the distribution map to obtain a preliminary lithology unit classification set. Through a classification identification tool, the preliminary lithology unit classification set is subjected to secondary screening, and if the spectral features of a region in the classification set match at least one lithology unit in the database by more than a preset threshold, it is classified as the corresponding lithology unit to determine the final lithology unit distribution range. A spatial distribution analysis tool is used to analyze the combination of the final lithology unit distribution range, and the spatial features of different lithology units in the distribution range are superimposed and compared to determine the potential favorable ore-forming combination type. According to the favorable ore-forming combination type, combined with the extraction tool of the lithology control factor, the key lithology features in the combination type are subjected to weighted processing to obtain the lithology control factor related to mineralization, and the final target factor set is determined.

[0111] As an optional implementation, in the geological remote sensing data processing, the lithology unit classification based on the alteration intensity distribution map is an important link of the ore prediction. The lithology spectral feature database usually contains spectral curves of various rock types, such as granite, basalt, andesite, etc. Through comparison with the spectral data in the distribution map, the lithology unit can be preliminarily identified. The principle of the data comparison tool is to calculate the similarity of the remote sensing spectral curve and the lithology spectral curve in the database, and to judge the matching degree by using the cosine similarity or the correlation coefficient method. In a certain copper exploration area, the technical personnel use the hyperspectral data with a resolution of 10 meters and covering an area of 50 square kilometers, extract the spectral features, and preliminarily obtain the classification set of granite, andesite and sedimentary rock. The spectral feature of granite has a clear absorption peak at 2.1 microns, and the matching degree reaches 85%. The andesite shows unique reflection characteristics at 1.4 microns.

[0112] Specifically, the secondary screening further improves the classification accuracy through the classification identification tool. The preset threshold is usually 80%. If the spectral feature of a certain area has a matching degree of 90% with the silicified cataclastic rock in the database, it is classified as a silicified altered rock unit. In this gold mine case, after screening, it is determined that the silicified cataclastic rock is distributed in the western part of the mining area with an area of about 8 square kilometers, and the K-feldspar granite is concentrated in the eastern part with an area of about 6 square kilometers. The spatial distribution analysis tool performs superposition analysis on these lithology units to identify the favorable ore-forming assemblage.

[0113] As an optional implementation, the ductile shear zone of the contact zone between the silicified cataclastic rock and the K-feldspar granite is often accompanied by gold mineralization. The analysis shows that the contact area between the two is about 3 square kilometers, showing high ore-forming potential.

[0114] In the lithology control factor extraction process, the tool highlights the gold-related features through weighted processing:

[0115] The 2.2 micron band pyritization absorption feature is given a high weight (factor value 0.78), reflecting the enrichment degree of gold-bearing pyrite in the silicified cataclastic rock;

[0116] The 1.4 micron band K-feldspar reflection feature is given a secondary weight (factor value 0.62), indicating the hydrothermal alteration intensity of the granite.

[0117] The spatial distribution of these factors shows that the pyritization factor value of the western silicified cataclastic rock area is 0.75, and the K-feldspar factor value of the eastern granite area is 0.58. Combined with the alteration intensity map, a 2 square kilometer high potential target area is delineated in the intersection area of the shear zone, where the pyritization factor weighting value is increased to 0.81, which verifies the control of the structure on the mineralization.

[0118] As an embodiment in the present embodiment, the process of outputting the potential ore body position prediction result comprises:

[0119] The input geological data is standardized and converted by a data preprocessing tool, and abnormal values are removed;

[0120] The tectonic control factor, alteration zoning information and lithological control factor are spatially superimposed by a data fusion tool;

[0121] The weight of the geological element is dynamically adjusted by an attention mechanism tool, and the priority is improved when the weight value exceeds the preset threshold;

[0122] When the comprehensive geological constraint score exceeds the preset ore-forming threshold, the ore body position is output by a position prediction tool.

[0123] Specifically, S107, the tectonic control factor, alteration zoning information and lithological control factor are input into a deep learning fusion network, and different geological element weights are allocated by an attention mechanism. If the comprehensive geological constraint score exceeds the preset ore-forming threshold, the potential ore body position prediction result is output.

[0124] The geological data carrying the tectonic control factor, alteration zoning information and lithological control factor is received, the geological data is standardized and converted by a data preprocessing tool, and abnormal values are removed by a data cleaning tool to obtain a normalized geological data set. According to the normalized geological data set, the spatial distribution characteristics of the tectonic control factor, alteration zoning information and lithological control factor are superimposed and processed by a data fusion tool to determine the fused geological information set. For the fused geological information set, the geological element weight is dynamically adjusted by an attention mechanism tool. If the weight value of a certain element exceeds the preset threshold, its priority is improved, and the comprehensive geological constraint score result is obtained. According to the comprehensive geological constraint score result, a result comparison tool is used to match the preset ore-forming threshold. If the score result exceeds the preset ore-forming threshold, the potential ore body position is determined by a position prediction tool to obtain the prediction result output.

[0125] As an optional implementation, in the processing of geological remote sensing data, when receiving geological data containing tectonic control factors, alteration zoning information and lithological control factors, the standardized conversion of the data preprocessing tool is crucial. The standardized conversion eliminates the dimensional differences by normalizing the factor data of different dimensions to a unified range, such as 0 to 1. Assuming that in a lead-zinc mine exploration area, the fracture density value of the tectonic control factor is 0.1 to 5, and the mineralization intensity value of the alteration zoning information is 50 to 200, the preprocessing tool maps these data to the 0 to 1 interval, facilitating subsequent analysis. The data cleaning tool identifies outliers such as fracture density values exceeding 10 to eliminate their impact and ensure the reliability of the data set. After cleaning, the data set covers 100 square kilometers and contains 5000 sampling points.

[0126] Specifically, the data fusion tool superimposes the spatial distribution of tectonic control factors, alteration zoning information and lithological control factors. The fusion process is based on spatial interpolation methods to superimpose grid data of the three types of factors to generate a comprehensive distribution map.

[0127] As an optional implementation, the fracture density is high in the north of the lead-zinc mine area, the alteration zoning shows strong silicification characteristics, and the lithology is mainly carbonate rock. The generated geological information set highlights the high-potential area of 5 square kilometers in the north after fusion. The fusion tool uses a weighted average method to assign a weight of 0.4 to fracture density, 0.3 to alteration information, and 0.3 to lithological factors to form a comprehensive information layer.

[0128] In a possible implementation, the attention mechanism tool dynamically adjusts the weight of geological elements. The attention mechanism allocates weights according to the contribution of elements to mineralization. If the correlation between fracture density and mineralization intensity reaches 0.8, its weight is increased to 0.5. Assuming that the preset threshold is 0.4 and the weight of the fracture factor in the northern region is 0.6, the priority is increased, and a comprehensive geological constraint score is generated. The scoring result shows that the score of the northern region is 0.85 and the score of the southern region is 0.6, reflecting that the northern region has higher mineralization potential.

[0129] As an optional implementation, the result comparison tool matches the comprehensive score with the preset mineralization threshold of 0.7. The northern region score of 0.85 exceeds the threshold, indicating that it has mineralization conditions. The location prediction tool further analyzes the spatial characteristics of the region, combines the fracture zone trend and alteration zoning boundary, and predicts the location of potential ore bodies. The prediction result shows that the fracture intersection area of 2 square kilometers in the north is the location of high-potential ore bodies. The tool uses a spatial clustering algorithm to delineate the center coordinates of the ore body and outputs the prediction result, providing accurate guidance for drilling.

[0130] It can be understood that the above method has the advantages of comprehensively capturing the complex relationship of ore-forming elements through multi-factor fusion and dynamic weight adjustment. Standardization and cleaning ensure data quality, fusion and attention mechanism highlight key features, and prediction tools accurately locate target areas. These steps are closely linked to form a complete chain from data processing to ore body prediction, providing efficient support for geological exploration.

[0131] In S108, spatial clustering analysis is performed on the potential ore body location prediction result, density clustering algorithm is used to remove isolated abnormal points, and morphological filtering is used to optimize the ore body boundary, and finally the metal ore body recognition result and its spatial distribution range are determined.

[0132] As shown in Figure 3 According to the potential ore body location prediction result, the spatial distribution analysis tool is used to group process the data, the geological features in the grouped area are analyzed, the preliminary spatial distribution set is obtained, and the distribution data group for subsequent classification is obtained. For the distribution data group, the density clustering tool is used to classify the data points in each region, if the density of the data point is lower than the preset threshold, it is marked as an isolated point and removed, the distribution data set after removing abnormal points is obtained, and the clean data group for boundary processing is determined. Through the clean data group, the morphological filtering tool is used to smooth the boundary of each region, the irregular shape in the boundary region is adjusted, the smoothed boundary data is obtained, and the boundary range set for final recognition is obtained. According to the boundary range set, the result verification tool is used to compare the geological features in each region, if the features in the region meet the preset ore body recognition standard, it is marked as a target area, and the spatial distribution range of the metal ore body is determined.

[0133] As an optional implementation, in the scene of metal ore body exploration, the application of spatial distribution analysis tool is crucial for subsequent processing of potential ore body location prediction result. Spatial distribution analysis tool divides the large area into several sub-regions by grouping the data in the prediction result, so as to analyze the geological features more carefully. Assuming that in a lead-zinc mine exploration project, the prediction result covers an area of 200 square kilometers, the tool divides it into 10 sub-regions, each about 20 square kilometers, and analyzes its geological features such as fault distribution and mineralization intensity, and finally forms a preliminary spatial distribution set, providing data support for subsequent classification.

[0134] For the classification of distributed data sets, the use of density clustering tools is a key step. Density clustering tools classify based on the spatial density of data points. If the density of a data point is lower than a preset threshold, such as less than 5 sampling points per square kilometer, it will be marked as an isolated point and excluded. Assuming that in a sub-region of the above lead-zinc mine area, there are a total of 1000 sampling points, and the density of 20 points is lower than the threshold, these points may be due to sampling errors or complex terrain, resulting in abnormal data. After excluding, a clean data set is obtained, ensuring the reliability of subsequent analysis.

[0135] In the boundary processing stage, the morphological filtering tool is particularly important for smoothing the region boundary. The morphological filtering tool adjusts the irregular shape within the boundary region to make the boundary line more continuous and natural. In a sub-region of the lead-zinc mine, the original boundary presents a jagged shape due to surface undulations. The tool adjusts the boundary to a gentle curve through smoothing, reducing noise interference and forming a clear boundary range set. This processing helps to more accurately delineate the target region.

[0136] In the final recognition stage, the result verification tool compares the geological features within the boundary range to determine whether they meet the ore body recognition criteria. Assuming that the preset criteria are mineralization intensity greater than 100 and fracture density greater than 0.5, in a sub-region of the lead-zinc mine, the detection result shows that the mineralization intensity is 120 and the fracture density is 0.7, which meets the criteria, so it is marked as a target region. In this way, the tool finally determines the spatial distribution range of the metal ore body, providing a basis for subsequent exploration planning. This multi-step process, from grouping to classification, to boundary adjustment and feature comparison, forms a complete chain, ensuring the accuracy and practicality of the results.

[0137] Based on this, the metal ore body recognition method based on remote sensing interpretation provided by the embodiment of the present application has the technical effects that through the combination of multi-source remote sensing data fusion processing and deep learning mechanism, the influence of surface cover and environmental interference on the recognition result is effectively weakened, and the extraction accuracy of mineralization signal is significantly improved; the quantifiable constraint model constructed based on geological rules converts ore-forming elements such as tectonic ore control, alteration zoning, and lithological combination into calculable parameters, so that the recognition result conforms to the logical of geological theory; the automatic processing process replaces the manual interpretation link, reduces the dependence on subjective experience, ensures the recognition reliability in complex geological environment, and finally realizes the efficient and accurate delineation of the spatial position of the metal ore body.

[0138] Embodiment Two

[0139] In this embodiment, a computer terminal device is provided, comprising:

[0140] one or more processors;

[0141] a memory, coupled to the processor, storing one or more programs;

[0142] When the one or more programs are executed by the one or more processors, the one or more processors implement the method in the above embodiments.

[0143] In the embodiment, a computer readable storage medium is also provided, which stores a computer program, and the computer program is executed by a processor to implement the method in the above embodiments.

[0144] In the embodiment, an electronic device is also provided, which includes a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to execute the method in the above embodiments.

[0145] The above program can be run in the processor, or can also be stored in the memory (or called computer readable medium), the computer readable medium includes permanent and non-permanent, removable and non-removable media, which can be realized by any method or technology to store information. The information can be computer readable instructions, data structure, program module or other data. Examples of computer storage medium include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.

[0146] These computer programs can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate computer implemented processing, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks, and the steps corresponding to different steps can be realized by different modules.

[0147] In the embodiment, such a device or system is provided. The system is called metal ore body identification system based on remote sensing interpretation, which includes:

[0148] The data preprocessing module is used to obtain multispectral remote sensing image, hyperspectral data, digital elevation model and geological structure map, and after geometric correction and radiation calibration, the spatial resolution and coordinate system are unified to generate standardized multi-source remote sensing data set;

[0149] a feature fusion module configured to extract a spectral feature vector and a spatial texture feature based on a standardized dataset, and to generate a composite feature vector by fusing the two when a correlation coefficient therebetween is greater than a preset threshold value;

[0150] a mineralization extraction module configured to process the composite feature vector through a multi-scale feature extraction model of a convolutional neural network, and to output an effective mineralization signal when a feature response intensity exceeds three times a standard deviation of background noise;

[0151] a structure control module configured to extract terrain gradient and curvature information in combination with a digital elevation model, to identify a fault zone and a fold structure, and to determine a structure-controlled mineralization area when a spatial overlap degree of the mineralization signal and the structure control factor is greater than a set proportion;

[0152] an alteration recognition module configured to compare a standard mineral spectrum library by using a spectral angle matching algorithm, to extract hydroxyl alteration, iron staining alteration and clay alteration zoning information in the structure-controlled mineralization area, and to generate an alteration intensity distribution map;

[0153] a lithology analysis module configured to identify a lithology unit in combination with a lithology spectral feature database, to determine a favorable ore-forming lithology combination type by lithology combination analysis, and to output a lithology control factor;

[0154] a decision output module configured to input the structure control factor, the alteration zoning information and the lithology control factor into a deep learning fusion network, to assign weights by an attention mechanism, and to output a potential ore body location when a comprehensive geological constraint score exceeds a mineralization threshold value;

[0155] a result optimization module configured to perform spatial clustering analysis on the potential ore body location, to eliminate isolated abnormal points and optimize an ore body boundary, and to determine a spatial distribution range of a metal ore body.

[0156] As an embodiment in the present embodiment, the data preprocessing module comprises:

[0157] a geometric correction unit configured to perform position deviation correction on the original dataset by a geometric transformation tool;

[0158] a radiation calibration unit configured to extract a radiation feature value by using a radiation scaling tool, and to perform linear interpolation calibration when the radiation feature value exceeds a preset threshold value;

[0159] a spatial registration unit configured to unify a spatial resolution by a resampling method, and to align to a reference coordinate system by using a coordinate conversion tool;

[0160] a feature integration unit configured to perform feature fusion on multispectral images and hyperspectral data, and to perform local adjustment optimization when inconsistent with a digital elevation model or a geological structure map.

[0161] As an implementation form in the embodiment, the feature fusion module comprises:

[0162] a spectrum extraction unit configured to extract spectral feature data by using a spectrum analysis tool;

[0163] a texture calculation unit configured to generate a gray level co-occurrence matrix parameter by using a texture calculation tool;

[0164] a correlation determination unit configured to calculate a correlation coefficient of the spectral and texture features;

[0165] a vector generation unit configured to integrate and generate a composite feature vector when the correlation coefficient is greater than a preset threshold.

[0166] As an implementation form in the embodiment, the structure ore-controlling module comprises:

[0167] a gradient analysis unit configured to generate a terrain gradient set by using a gradient calculation tool;

[0168] a curvature analysis unit configured to generate a terrain curvature set by using a curvature calculation tool;

[0169] a structure interpretation unit configured to identify a fault zone and a fold structure by comprehensively analyzing the terrain gradient set and the terrain curvature set;

[0170] an ore-controlling determination unit configured to determine an ore-controlling area when an overlap degree of the structure control factor and the mineralization signal is greater than a threshold proportion.

[0171] As an implementation form in the embodiment, the alteration recognition module comprises:

[0172] a mineral matching unit configured to identify a hydroxyl alteration, an iron staining alteration and a clay alteration by comparing a standard mineral spectrum library by using a spectrum angle matching tool;

[0173] a zoning extraction unit configured to divide an alteration type distribution range by using a spatial distribution analysis tool;

[0174] an intensity generation unit configured to generate an alteration intensity distribution map by weighting the spectral data when an overlap degree of the alteration distribution and the ore-controlling area is greater than a threshold.

[0175] As an implementation form in the embodiment, the decision output module comprises:

[0176] a data normalization unit configured to perform standardization conversion on input geological data and remove abnormal values;

[0177] a factor fusion unit configured to perform spatial superposition on the structure control factor, the alteration zoning information and the lithology control factor;

[0178] a weight distribution unit configured to dynamically adjust weights of geological elements by using an attention mechanism.

[0179] a position prediction unit configured to output a position of the ore body when the integrated geological constraint score exceeds a mineralization threshold.

[0180] As an implementation in the embodiment, the result optimization module comprises:

[0181] a spatial clustering unit configured to remove isolated abnormal points by using a density clustering algorithm;

[0182] a boundary optimization unit configured to process a shape of the ore body boundary by morphological filtering;

[0183] a range determination unit configured to generate a spatial distribution range of the metal ore body.

[0184] The system or device is used to realize the functions of the method in the above-mentioned embodiments. Each module in the system or device corresponds to each step in the method, which has been described in the method and will not be repeated here.

[0185] Through the above-mentioned embodiments, the problem of metal ore body identification based on remote sensing interpretation in the related art is solved, thereby being able to ensure solving the problems in the prior art.

[0186] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for identifying metal ore bodies based on remote sensing interpretation, characterized in that: The following steps are involved: Acquire multispectral remote sensing images, hyperspectral data, digital elevation models, and geological structure maps, and after geometric correction and radiometric calibration, use registration algorithms to unify spatial resolution and coordinate systems to generate standardized multi-source remote sensing datasets. Based on this data set, principal component analysis is used to extract spectral feature vectors. At the same time, the gray-level co-occurrence matrix parameters are calculated through the texture analysis operator to obtain spatial texture features. When the correlation coefficient between spectral and texture features is greater than the preset threshold, they are fused to form a composite feature vector. For the composite feature vector, a multi-scale feature extraction model is constructed through a convolutional neural network. Different convolution kernel sizes are used to extract local and global spatial features. When the characteristic response intensity exceeds three times the standard deviation of the background noise, it is determined to be a valid mineralization signal. Based on the distribution of effective mineralization signals, combined with digital elevation models, terrain gradient and curvature information are extracted, and fault zones and fold structures are identified through geological structure interpretation algorithms. When the spatial overlap between mineralization signals and structural control factors is greater than a set ratio, the structurally controlled ore area is determined; A spectral angle matching algorithm is used to identify altered minerals in structurally controlled ore-bearing areas. By comparing with a standard mineral spectral library, alteration zoning information for hydroxyl alteration, iron-stained alteration, and clay alteration is extracted to obtain an alteration intensity distribution map. Based on the alteration intensity distribution map and the lithologic spectral feature database, the support vector machine classifier is used to identify lithologic units. The favorable mineralization lithologic combination type is determined through lithologic combination analysis, and the lithologic control factors are obtained. The structural control factors, alteration zoning information, and lithologic control factors are input into the deep learning fusion network. The geological factor weights are assigned through the attention mechanism. When the comprehensive geological constraint score exceeds the preset mineralization threshold, the potential ore body location prediction result is output. A spatial cluster analysis is conducted on the potential ore body locations, and a density clustering algorithm is used to eliminate isolated outliers. The ore body boundaries are optimized through morphological filtering to determine the spatial distribution range of the metal ore body.

2. The method according to claim 1, characterized in that The process of generating a standardized multi-source remote sensing dataset includes: Correct the position deviation of the original dataset using geometric transformation tools; A radiation calibration tool is used to extract radiation characteristic values, and when the radiation characteristic values ​​exceed the preset threshold, linear interpolation calibration is performed; The spatial resolution was unified by using the resampling method using the spatial registration tool, and aligned to the reference coordinate system using the coordinate conversion tool; Image fusion tools are used to integrate the features of multispectral images and hyperspectral data. When the integrated features are inconsistent with the digital elevation model or geological structure map, they are optimized through local adjustment algorithms.

3. The method according to claim 1, characterized in that The process of fusing to form a composite feature vector includes: Spectral feature data are extracted using spectral analysis tools, and grayscale matrix parameters are generated using texture calculation tools; The correlation coefficient between spectrum and texture features was calculated using correlation analysis tools; When the correlation coefficient is greater than the preset threshold, it is integrated into a composite vector through the feature fusion tool; Spatial feature comparison is performed through data integration tools, and when the spatial consistency parameters meet the preset standards, the spatially calibrated composite vector is output.

4. The method according to claim 1, wherein The process of determining the structural ore-controlling area includes: Process terrain height data through gradient calculation tools to generate terrain gradient sets; Use curvature calculation tools to analyze terrain height data to generate terrain curvature sets; Identify fault zones and fold structures by integrating terrain gradient and curvature sets through geological structure interpretation tools; When the spatial overlap between the structural control factor and the effective mineralization signal is greater than the preset threshold ratio, the ore-controlling area is determined by spatial analysis tools.

5. The method according to claim 1, wherein The process of obtaining the alteration intensity distribution map includes: Identify hydroxyl alteration, iron staining alteration and clay alteration minerals by comparing with the standard mineral spectrum library through spectral angle matching tools; Use spatial distribution analysis tools to divide the distribution range of alteration types and extract alteration zoning information; When the degree of overlap between the spatial distribution of at least one alteration type and the structurally controlled ore-bearing area is greater than a preset threshold, the spectral data of the overlapping area is weighted using a spectral feature extraction tool.

6. The method according to claim 1, characterized in that The process of outputting the potential ore body location prediction result includes: Use data preprocessing tools to standardize input geological data and remove outliers; Data fusion tools are used to spatially overlay structural control factors, alteration zoning information, and lithologic control factors; Dynamically adjust the weights of geological elements through the attention mechanism tool, and increase the priority when the weight value exceeds the preset threshold; When the comprehensive geological constraint score exceeds the preset mineralization threshold, the ore body location is output through the location prediction tool.

7. A metal ore body identification system based on remote sensing interpretation, characterized in that: The system comprises: The data preprocessing module is used to obtain multispectral remote sensing images, hyperspectral data, digital elevation models, and geological structure maps, unify the spatial resolution and coordinate system after geometric correction and radiometric calibration, and generate standardized multi-source remote sensing data sets; A feature fusion module is used to extract spectral feature vectors and spatial texture features based on a standardized data set, and fuse them to generate a composite feature vector when the correlation coefficient between the two is greater than a preset threshold; The mineralization extraction module is used to process the composite feature vector through the multi-scale feature extraction model of the convolutional neural network and output a valid mineralization signal when the characteristic response intensity exceeds three times the standard deviation of the background noise; The structural ore-control module is used to extract terrain gradient and curvature information in combination with the digital elevation model, identify fault zones and fold structures, and determine the structural ore-control area when the spatial overlap between the mineralization signal and the structural control factor is greater than the set ratio; The alteration identification module is used to compare the standard mineral spectrum library using a spectral angle matching algorithm to extract hydroxyl alteration, iron-stained alteration, and clay alteration zoning information within the structurally controlled ore-bearing area and generate an alteration intensity distribution map; The lithologic analysis module is used to identify lithologic units by combining the lithologic spectral feature database, determine favorable mineralization lithologic combination types through lithologic combination analysis, and output lithologic control factors; The decision output module is used to input structural control factors, alteration zoning information, and lithologic control factors into the deep learning fusion network, assign weights through the attention mechanism, and output the potential ore body location when the comprehensive geological constraint score exceeds the mineralization threshold; The result optimization module is used to perform spatial cluster analysis on potential ore body locations, eliminate isolated outliers, optimize ore body boundaries, and determine the spatial distribution range of metal ore bodies.

8. The system according to claim 7, characterized in that The data preprocessing module includes: A geometric correction unit, used to correct the position deviation of the original data set through a geometric transformation tool; a radiation calibration unit, configured to extract radiation characteristic values ​​using a radiation calibration tool, and perform linear interpolation calibration when the radiation characteristic values ​​exceed a preset threshold; A spatial registration unit, used to unify the spatial resolution through resampling methods and align to the reference coordinate system using coordinate transformation tools; The feature integration unit is used to fuse the features of multispectral images and hyperspectral data, and perform local adjustment and optimization when they are inconsistent with the digital elevation model or geological structure map.

9. A computer terminal device, characterized in that: include: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the metal ore body identification method based on remote sensing interpretation as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for identifying a metal ore body based on remote sensing interpretation as claimed in any one of claims 1 to 6 is implemented.

Images