A method and system for identifying and predicting weak information of concealed ore

Abstract

This invention relates to the field of mineral resource exploration and discloses a method and system for identifying and predicting weak information in concealed mineralization. The method acquires and spatially registers and normalizes multi-source exploration data, including geological structures, geochemistry, and mineralogical data, to extract various features related to mineralization and construct a multi-dimensional feature set. A reference distribution is constructed based on geological background information and compared with the multi-dimensional feature set to separate responses indicating mineralization. Mineralization potential is then assessed, and the results are presented in a three-dimensional format, supporting interactive analysis and verification. This invention effectively identifies weak mineralization information obscured by surface background through multi-source data fusion, refined background stripping, and adaptive threshold adjustment, significantly improving the accuracy and efficiency of concealed mineralization exploration. It solves the problems of traditional methods relying on experience, difficulty in weak signal extraction, and lack of intuitive three-dimensional visualization support.

Description

Technical Field

[0001] This application relates to the field of mineral resource exploration, and more specifically, to a method and system for identifying and predicting hidden mineral weaknesses. Background Technology

[0002] In the field of mineral resource exploration, especially in identifying and predicting deep, concealed ore bodies, has always been a challenging task. With the continuous advancement of geological exploration technology, we are able to collect increasingly abundant exploration data from various sources, including geological structures, geochemistry, and mineralogical data. This data provides a valuable foundation for comprehensively analyzing subsurface conditions and predicting mineral deposit locations. However, traditional exploration methods often rely excessively on the personal experience of geologists, and when processing this complex and diverse information, they often focus only on a single type of data or isolated indicators. This makes it difficult to effectively capture those weak signals indicating deep ore bodies that are strongly obscured by the surface background from the massive amount of information.

[0003] For example, in the field of mineral resource exploration, the identification and prediction of concealed mineral deposits are crucial for discovering deep mineral deposits and ensuring resource supply. Traditionally, geologists rely on this data for empirical judgment, but information on concealed mineral deposits is often masked by surface background noise, making it difficult to extract weak signals. There is an urgent need for efficient and intelligent methods to integrate multi-source information to improve exploration success rates. However, existing methods for predicting concealed mineral deposits have significant technical shortcomings. Traditional methods often rely on single data types or indicators, such as isolated geochemical anomalies or tectonic analyses, resulting in limited identification capabilities and difficulty in effectively integrating multi-source data to capture weak information about deep mineralization. Data processing is highly dependent on human experience, with low automation, low efficiency, and strong subjectivity. Furthermore, prediction results are usually presented in two-dimensional form, lacking intuitive three-dimensional visualization support, making it difficult for geologists to make quick decisions and verify their findings. These problems restrict the accuracy and reliability of concealed mineral exploration and hinder the effective development of mineral resources. Summary of the Invention

[0004] The purpose of this invention is to propose a method for identifying and predicting weak information in concealed mineral deposits, which aims to solve the problems of traditional exploration methods such as over-reliance on experience, difficulty in effectively integrating multi-source data, difficulty in extracting weak signals, and lack of intuitive three-dimensional visualization support for prediction results.

[0005] Specifically, the present invention provides a method for identifying and predicting hidden mineral weaknesses, which includes the following steps: We acquire exploration data from various sources, including geological structure, geochemistry, and mineralogical data, from the exploration area, and perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. From the processed exploration data, various features related to mineralization are extracted and integrated to form a multi-dimensional feature set that reflects potential mineralization information. Based on the geological background information of the exploration area, a reference distribution reflecting the changes in the regional geological background is constructed; the multidimensional feature set is compared with the reference distribution to separate the responses indicating mineralization. Based on the response of the isolated indicator mineralization, the mineralization potential of the exploration area is assessed to obtain the mineralization potential assessment results. The mineralization potential assessment results are presented in a three-dimensional format, supporting geologists to conduct interactive analysis and verification.

[0006] A system for identifying and predicting hidden mineral weaknesses, comprising: The data acquisition and processing module is used to acquire exploration data from various sources, including geological structure, geochemistry, and mineralogy, from the exploration area, and to perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. The feature extraction and integration module is used to extract various features related to mineralization from the processed exploration data and integrate these features to form a multi-dimensional feature set that reflects potential mineralization information. The background stripping module is used to construct a reference distribution reflecting changes in the geological background of the exploration area based on the geological background information of the area; it compares the multidimensional feature set with the reference distribution to separate the responses indicating mineralization; the background stripping module includes: a latent geological texture recognition unit, used to infer latent geological textures by combining remote sensing spectral data and high-resolution geological images to correct the background field; an adaptive threshold adjustment unit, used to dynamically adjust the recognition threshold of weak mineralization signals according to the background field gradient; and a weak signal discrimination unit, used to discriminate weak mineralization clues based on spatial continuity. The mineralization potential assessment module is used to assess the mineralization probability of the exploration area based on the response of the isolated indicator mineralization, and obtain the mineralization potential assessment results; The 3D visualization and interaction module is used to present the mineralization potential assessment results in a 3D form and support geologists to conduct interactive analysis and verification.

[0007] The beneficial effects provided by this invention are as follows: The method for identifying and predicting weak information in concealed mineralization disclosed in this application effectively solves the limitations of traditional methods based on single data types or indicators by acquiring and processing multi-source exploration data, extracting various features related to mineralization, and constructing a multi-dimensional feature set. By constructing a reference distribution reflecting changes in regional geological background and comparing it with the multi-dimensional feature set, responses indicating mineralization can be separated, thereby overcoming the problem of surface background noise masking weak mineralization signals. Furthermore, this application presents the mineralization potential assessment results in a three-dimensional form and supports interactive analysis and verification by geologists, greatly improving the intuitiveness and operability of the prediction results and overcoming the shortcomings of traditional two-dimensional presentation methods. In summary, this application effectively solves the problems of difficulty in extracting weak information in concealed mineralization, low prediction accuracy, insufficient automation, and lack of intuitive visualization support in existing technologies by integrating multi-source information, refined background stripping, and three-dimensional visualization interaction, significantly improving the success rate and efficiency of concealed mineralization exploration. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0010] Before formally describing the present invention, a general description of the solution of the present invention will be given first to facilitate understanding.

[0011] Example 1 Please refer to Figure 1 The present invention provides a method for identifying and predicting hidden mineral weaknesses, comprising the following steps: We acquire exploration data from various sources, including geological structure, geochemistry, and mineralogical data, from the exploration area, and perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. From the processed exploration data, various features related to mineralization are extracted and integrated to form a multi-dimensional feature set that reflects potential mineralization information. Based on the geological background information of the exploration area, a reference distribution reflecting the changes in the regional geological background is constructed; the multidimensional feature set is compared with the reference distribution to separate the responses indicating mineralization. Based on the response of the isolated indicator mineralization, the mineralization potential of the exploration area is assessed to obtain the mineralization potential assessment results. The mineralization potential assessment results are presented in a three-dimensional format, supporting geologists to conduct interactive analysis and verification.

[0012] This application integrates and refines multi-source exploration data to provide a more comprehensive reflection of subsurface geological conditions. By constructing a refined background field and performing background stripping, it effectively identifies and separates weak signals indicating mineralization, overcoming the limitations of traditional methods that are susceptible to background noise interference. Furthermore, this application significantly improves geologists' understanding and verification efficiency of mineralization potential assessment results through three-dimensional visualization and interactive analysis, thereby enhancing the accuracy and reliability of concealed mineralization identification and prediction.

[0013] The “exploration data” mentioned in this application refers to raw information obtained through various means during mineral exploration. These data come from diverse sources, including but not limited to geological structural data (such as faults, folds, and lithological boundaries), geochemical data (such as elemental content in soils, rocks, and stream sediments), and mineralogical data (such as mineral composition, crystal structure, and distribution of altered minerals). This data forms the basis for mineral resource assessment.

[0014] Spatial registration refers to unifying data from different sources and coordinate systems into a single spatial reference system to ensure that all data can be accurately mapped geographically. Normalization refers to converting data with different dimensions and numerical ranges into a unified numerical interval to eliminate the impact of dimensional differences on subsequent analysis. For example, the content data of geochemical elements can be normalized to the range of 0-1.

[0015] "Multidimensional feature set" refers to the integration of multiple processed features related to mineralization into a comprehensive dataset containing information in multiple dimensions, such as geochemical anomaly intensity, tectonic density, alteration mineral index, etc., to comprehensively reflect potential mineralization information.

[0016] "Reference distribution" refers to a benchmark model constructed based on the macro-geological background information of the exploration area, which reflects the normal range of variation of regional geological background elements or characteristics. It is used to compare with actual exploration data to identify anomalies.

[0017] "Response indicating mineralization" refers to an abnormal signal or feature that clearly indicates the existence of mineralization and is separated from a multidimensional feature set through methods such as background stripping.

[0018] "Mineralization potential assessment results" refer to the quantitative assessment of the probability of mineral resource occurrence in the exploration area based on the response of indicated mineralization, and are usually presented in the form of probability values, potential levels or resource estimates.

[0019] The specific implementation method for identifying and predicting hidden mineral weaknesses in this application is as follows: First, it is necessary to acquire exploration data from various sources, including geological structure, geochemistry, and mineralogical data, from the exploration area. This data can be obtained in several ways. For example, geological structure data can be obtained through field geological surveys, borehole core analysis, and geophysical exploration (such as seismic exploration, gravity exploration, and magnetic exploration). Geochemical data can be obtained through laboratory analysis of soil, rock, and stream sediment samples. Mineralogical data can be obtained through thin section identification, X-ray diffraction (XRD), and scanning electron microscopy (SEM). After acquiring these raw data, they need to be spatially registered and normalized. Spatial registration can be performed using Geographic Information System (GIS) software to project data from different sources onto a unified coordinate system; for example, unifying all data to the WGS84 coordinate system. Normalization can be achieved using methods such as min-max normalization and Z-score normalization to convert data of different dimensions to a unified numerical range. For example, the content data of geochemical elements can be normalized to between 0 and 1 to eliminate the influence of dimensional differences on subsequent analysis and ensure that the data are consistent in spatial location and numerical range.

[0020] Secondly, from the processed exploration data, it is necessary to extract various features related to mineralization and integrate these features to form a multidimensional feature set reflecting potential mineralization information. Feature extraction can be performed according to different data types and mineralization models. For example, for geochemical data, features such as indicator element anomalies, element combination anomalies, anomaly intensity, and anomaly range can be extracted. For geological structural data, features such as fault density, structural line density, and fold intensity can be extracted. For mineralogical data, features such as alteration mineral assemblages, mineral content, and mineral crystal structure characteristics can be extracted. These features can be extracted using specialized geochemical software, geological modeling software, or custom algorithms. After extracting these features, they need to be integrated to form a multidimensional feature set. The integration method can be simply superimposing all features, or multivariate statistical methods such as principal component analysis (PCA) and factor analysis can be used for dimensionality reduction and feature fusion to form a comprehensive multidimensional dataset reflecting potential mineralization information. For example, geochemical anomaly intensity, fault density, and alteration index can be used as different dimensions to construct a multidimensional feature vector.

[0021] Secondly, based on the geological background information of the exploration area, a reference distribution reflecting changes in the regional geological background is constructed. The multidimensional feature set is then compared with the reference distribution to separate responses indicating mineralization. There are several methods for constructing the reference distribution. One approach is based on regional geological maps, dividing the exploration area into different lithological units and statistically analyzing the average content or distribution characteristics of indicator elements in each lithological unit as the background value for that unit. Another approach uses statistical methods, such as moving averages and kriging interpolation, to smooth the geochemical data within the exploration area, obtaining a continuous background field. After constructing the reference distribution, the actual measured values ​​in the multidimensional feature set are compared with the reference distribution. The comparison method can be a simple difference calculation, subtracting the background value from the actual measured value to obtain the residual signal. More complex statistical methods, such as anomaly detection algorithms, can be used to identify anomalous points or regions that significantly deviate from the background distribution. For example, if the geochemical element content of a region is much higher than the average background value of its lithological unit, then mineralization anomalies may exist in that region.

[0022] Next, based on the responses of the isolated indicative mineralizations, the mineralization probability of the exploration area is assessed, yielding a mineralization potential assessment result. After isolating the responses of the indicative mineralizations, these responses need further analysis and evaluation to determine their reliability and potential for indicative mineralization. Assessment methods may include: classifying anomalous signals, for example, classifying anomalous signals into three levels—weak, medium, and strong—based on their intensity; assigning different weights to different types of anomalous signals based on the experience of geological experts; and using machine learning models, such as Support Vector Machines (SVM) and Random Forests, to classify and predict anomalous signals to assess mineralization probability. For example, the mineralization potential can be comprehensively judged based on the intensity, extent, morphology, and similarity to known mineral deposits of geochemical anomalies. Finally, the assessment results can be presented in the form of mineralization potential maps, mineralization probability maps, or resource estimation reports.

[0023] Finally, the mineralization potential assessment results are presented in three-dimensional form, supporting interactive analysis and verification by geologists. Three-dimensional visualization is key to improving the intuitiveness and usability of the mineralization potential assessment results. The results can be imported into professional 3D geological modeling software, such as GOCAD and Petrel, or displayed using custom 3D visualization tools. 3D presentation can include: displaying mineralization potential values ​​in three-dimensional space as volume rendering, isosurfaces, or slice maps; overlaying geological structural models, borehole data, geophysical anomalies, and other geological information to form a comprehensive 3D geological model. Interactive analysis and verification functions can include: allowing geologists to rotate, scale, and translate the 3D model at any angle; supporting cross-sectioning and borehole trajectory simulation of specific areas; providing a query function, allowing users to view detailed mineralization potential assessment data and related characteristic information by clicking on any point in the model; and supporting geologists to correct or mark the assessment results based on their experience and judgment, and record the verification process. For example, geologists can observe the spatial relationship between a geochemical anomaly and a deep fault zone in the 3D model, thereby determining the mineralization potential of the anomaly.

[0024] The method for identifying and predicting weak information in concealed mineral deposits in this application systematically integrates multi-source exploration data and performs refined spatial registration and normalization processing, ensuring the consistency of the data in both space and numerical terms, thus laying a solid foundation for subsequent feature extraction and analysis. Compared with traditional methods, this application fully considers the diversity and complexity of data during the data processing stage, avoiding the limitations of a single data source.

[0025] In some embodiments described above in this application, a reference distribution is constructed based on the geological background information of the exploration area, and a multidimensional feature set is compared with the reference distribution to separate responses indicating mineralization. However, in practical applications, the complexity and diversity of the geological background may cause the initially constructed reference distribution to fail to fully and accurately reflect the subtle changes in the regional geological background. Especially in the identification of weak information on concealed mineralization, weak mineralization signals are easily masked by complex background noise, thus affecting the accuracy of mineralization response separation.

[0026] In this regard, this application further proposes the steps of constructing a reference distribution reflecting changes in the regional geological background based on the geological background information of the exploration area; and comparing the multidimensional feature set with the reference distribution to separate the responses indicating mineralization, including: Based on macro-geological maps, the lithological units of the exploration area are divided, and the content of indicator elements in each lithological unit is statistically analyzed to construct a preliminary background field; The initial residual signal is obtained by calculating the difference between the geochemical measurements in the multidimensional feature set and the preliminary background field values. Spatial pattern recognition was performed on the initial residual signal, and latent geological textures were inferred by combining remote sensing spectral data and high-resolution geological images; The background correction amount is calculated based on the implicit geological texture and superimposed on the initial background field to construct a refined background field; The refined residual signal is obtained by calculating the difference between the geochemical measurements in the multidimensional feature set and the refined background field values.

[0027] Specifically, constructing a preliminary background field involves analyzing macroscopic geological maps of the exploration area and dividing the entire region into lithological units with similar geological characteristics. Subsequently, for each divided lithological unit, the average or typical content of indicator elements related to mineralization is statistically analyzed, and this is used as the background value for that lithological unit. These background values ​​collectively constitute the preliminary background field, the purpose of which is to provide an initial geological background reference based on macroscopic geological units.

[0028] The initial residual signal is obtained by calculating the difference between the actual geochemical measurements in the multidimensional feature set and the background values ​​at the corresponding locations in the aforementioned preliminary background field. This initial residual signal initially reflects anomalous information that may exist outside the macroscopic geological background, but it may still contain background noise that has not been completely stripped away by the preliminary background field, as well as local background variations caused by latent geological textures.

[0029] Furthermore, spatial pattern recognition is performed on the initial residual signal, and latent geological textures are inferred by combining remote sensing spectral data and high-resolution geological imagery. Spatial pattern recognition aims to identify anomalous patterns with specific spatial distribution characteristics from the initial residual signal; these patterns may be associated with concealed geological bodies or structures. Simultaneously, remote sensing spectral data and high-resolution geological imagery are used to provide detailed information about the surface or near-surface, such as surface lithology, alteration information, and linear structures. This information helps infer latent geological textures that are difficult to directly identify on macroscopic geological maps, such as concealed faults, lithological contact zones, and faint alteration halos. Although these latent geological textures do not directly manifest as mineralization anomalies, they influence the distribution of the local background field.

[0030] Therefore, based on the inferred latent geological textures, a background correction amount can be calculated. This background correction amount quantifies the specific impact of these latent geological textures on the local geological background value. Subsequently, this background correction amount is superimposed on the preliminary background field to construct a refined background field. Compared to the preliminary background field, the refined background field can more accurately and meticulously reflect the geological background changes in the exploration area, especially in areas affected by latent geological textures.

[0031] Finally, the difference between the geochemical measurements in the multidimensional feature set and the refined background field values ​​is calculated to obtain the refined residual signal. This refined residual signal is obtained by removing background interference from a more accurate background field, thus it can more clearly and accurately indicate potential mineralization information, especially those weak and easily obscured hidden mineralization information.

[0032] In some preferred embodiments, it is assumed that in a certain exploration area, preliminary geological surveys indicate the presence of two main lithological units: granite and gneiss. Based on macro-geological maps, the area is first divided into granite and gneiss units, and the average content of indicator elements (e.g., Cu, Au) is calculated for each unit to construct a preliminary background field. For example, the Cu background value for the granite unit is 50 ppm, and for the gneiss unit it is 80 ppm. Subsequently, the difference between the actual collected geochemical measurements and this preliminary background field is calculated to obtain the initial residual signal. Analysis of the initial residual signal reveals weak linear anomalies in certain areas, but their geochemical values ​​are not significantly higher than the preliminary background field. At this point, combined with high-resolution remote sensing imagery, these linear anomalies can be identified as actually corresponding to concealed fault structures or lithological contact zones. Although these latent geological textures are not directly ore bodies, they affect the migration and enrichment of local elements, thus causing changes in local background values. For example, the Cu background value near a fault zone may be slightly higher than that of the surrounding granite. By using a pre-defined texture-background value influence coefficient database, the correction amount of these latent geological textures to the background values ​​is calculated and superimposed onto the initial background field, thus constructing a more refined background field. For example, near fault zones, the Cu background value of granite units may be corrected to 55 ppm. Finally, the difference between geochemical measurements and the refined background field is calculated again. The resulting refined residual signal will more accurately reflect the true mineralization anomalies. For example, near fault zones, if the Cu measurement value reaches 100 ppm, it may be considered a weak anomaly under the initial background field, but under the refined background field, because the background value is corrected to 55 ppm, its residual signal (45 ppm) will more clearly indicate potential mineralization. Through this refined background stripping process, even weak mineralization information can be effectively identified and separated.

[0033] In some embodiments described above in this application, spatial pattern recognition is performed on the initial residual signal, and latent geological textures are inferred by combining remote sensing spectral data and high-resolution geological images to correct the background field. However, in practical applications, the influence of latent geological textures on the background values ​​of indicator elements may be multi-scale and complex. Simply inferring their existence and making corrections may not completely eliminate the background interference they cause, thereby affecting the accuracy of the refined background field and the recognition effect of weak mineralization signals.

[0034] In response, this application further proposes a process for spatial pattern recognition of the initial residual signal and inference of latent geological textures by combining remote sensing spectral data and high-resolution geological images, including: The initial residual signal is decomposed into multiple scales to separate anomalous components at different spatial scales; For each isolated anomalous component, by combining remote sensing spectral data and high-resolution geological images, latent geological texture features related to specific lithological units or structures are identified. These latent geological texture features include concealed faults, lithological contact zones, and geological bodies that exhibit weak linear or ring-like features in surface images. Based on the preset texture-background value influence coefficient database, query the contribution weight of each latent geological texture feature to the background value of the indicator element; The initial residual signal is corrected based on the contribution weight to remove background interference caused by hidden geological textures.

[0035] Specifically, multi-scale decomposition of the initial residual signal refers to breaking down the original initial residual signal into multiple signal components that are representative of different spatial scales. The aim is to reveal the contribution of different geological processes or structures to geochemical anomalies at different spatial scales. For example, regional background field variations may correspond to large-scale components, while local mineralization anomalies may correspond to small-scale components. This decomposition can be achieved through methods such as wavelet transform, empirical mode decomposition, or multi-scale morphological filtering.

[0036] For each isolated anomalous component, identifying latent geological texture features by combining remote sensing spectral data and high-resolution geological imagery can be understood as using multi-source remote sensing information to refine the interpretation of the decomposed signals. Remote sensing spectral data can provide information on surface material composition and alteration, while high-resolution geological imagery can reveal weak linear, ring-shaped, or speckled features on the surface. These features are often indirect manifestations of concealed faults, lithological contact zones, or alteration halos on the surface. For example, concealed faults may appear as weak linear structures or vegetation anomalies on the surface; lithological contact zones may appear as abrupt changes in surface rock color or texture; and alteration halos may appear as specific mineral assemblages or ring-shaped anomalies in surface morphology. By comprehensively analyzing this information, these latent geological bodies that are difficult to identify directly with the naked eye can be more accurately inferred.

[0037] In practical applications, the pre-set texture-background value influence coefficient database specifically stores quantitative data on the influence of different types of latent geological texture features (such as faults of different scales, contact zones of different types, and alteration halos of different intensities) on the background values ​​of different indicator elements. This database can be constructed and updated using historical exploration data, geological model simulations, or expert experience. Querying the contribution weight of each latent geological texture feature to the background value of indicator elements aims to quantify the degree of disturbance of these latent geological bodies on the local geochemical background, thereby providing accurate correction parameters for subsequent background stripping.

[0038] Therefore, correcting the initial residual signal based on contribution weights to remove background interference caused by hidden geological textures means subtracting the quantified influence of hidden geological textures from the initial residual signal. For example, if a concealed fault exists in a region and this fault has a positive contribution weight to the background value of copper, then during correction, this contribution weight will be used to adjust the initial residual signal of copper in that region to eliminate the uplift effect of the fault itself on the background value of copper, thereby making the corrected signal more realistically reflect the mineralization anomaly.

[0039] In some preferred embodiments, a specific example is given below. Suppose that in an exploration area, preliminary geochemical measurements show some weak anomalies, but these anomalies may be affected by concealed geological structures. First, these initial residual signals are decomposed at multiple scales, for example, using wavelet transform to decompose them into high-frequency, mid-frequency, and low-frequency components, representing anomaly information at local, intermediate, and regional scales, respectively.

[0040] For the high-frequency components, by combining high-resolution UAV imagery and remote sensing spectral data (such as shortwave infrared bands), weak linear structures on the surface were identified, which may indicate concealed faults. Meanwhile, in the mid-frequency components, by analyzing the texture features and spectral anomalies of remote sensing images, some annular or irregular alteration halos were identified, which may be related to concealed rock masses or hydrothermal activity.

[0041] Subsequently, based on a pre-defined texture-background value influence coefficient database, the contribution weights of these identified latent fractures and alteration halos to the background values ​​of indicator elements (e.g., copper, lead, zinc) are queried. For example, the database may show that a latent fracture with a specific orientation has a 0.1 ppm increase in the background value of copper, while an alteration halo with a specific spectral feature has a 0.05 ppm increase in the background value of lead.

[0042] Finally, based on these retrieved contribution weights, the initial residual signal was corrected point by point. Specifically, in areas with concealed faults, the copper value in the initial residual signal was reduced by 0.1 ppm; in areas with alteration halos, the lead value was reduced by 0.05 ppm. In this way, background interference caused by hidden geological textures was accurately removed, allowing the corrected residual signal to more accurately reflect potential mineralization anomalies, thus providing a more reliable data foundation for subsequent mineralization potential assessment.

[0043] In some embodiments described above in this application, a refined background field is constructed and a refined residual signal is calculated to isolate responses indicative of mineralization. However, in actual concealed mineral exploration, mineralization information often manifests as weak geochemical anomalies, which may be masked by complex geological background variations, or their signal intensity may be close to background noise. Relying solely on a single residual signal analysis and using a fixed identification threshold may lead to the underidentification of weak mineralization signals or misjudgments in areas with drastic background field changes, thereby affecting the accuracy and reliability of mineralization potential assessment.

[0044] In response, this application further proposes a method for identifying and predicting weak information in concealed mineral deposits, which further includes: The recognition threshold for weak mineralization signals is dynamically adjusted based on the refined background field value. The refined residual signal is compared with the identification threshold to identify weak mineralization clues; Based on weak mineralization cues, combined with fracture density and other elemental combination anomalies in the multidimensional feature set, a mineralization potential score is calculated to separate responses indicating mineralization.

[0045] Specifically, in the above method, the identification threshold for weak mineralization signals is first dynamically adjusted based on the refined background field values. This means that the identification threshold is no longer fixed but is adaptively adjusted according to the complexity and changing trends of the local geological background. For example, in areas with gentle background field changes, a lower identification threshold can be used to improve the sensitivity to weak mineralization signals; while in areas with steep gradients or discontinuities in the background field, the identification threshold can be appropriately increased to avoid misidentifying background noise as mineralization signals. This dynamic adjustment mechanism aims to make the identification process more flexible and accurate.

[0046] Furthermore, the refined residual signal is compared with the dynamically adjusted identification threshold described above to identify weak mineralization cues. This step aims to screen out weak anomalies from a complex geochemical background that, while having low signal intensity, may indicate potential mineralization. These weak mineralization cues may not be significant individually, but their spatial distribution and combination characteristics may be of significant indicative importance.

[0047] Based on this, and using identified weak mineralization clues, combined with fracture density and other elemental anomalies from the multidimensional feature set, a mineralization potential score is calculated to ultimately separate the responses indicating mineralization. Fracture density refers to the density of fracture structures within the exploration area. Fracture structures are typically favorable channels for fluid migration and mineral deposition; therefore, areas with high fracture density are often closely related to mineralization. Other elemental anomalies refer to anomalous patterns formed by combinations of multiple associated or indicator elements in addition to the main indicator element. These anomalies provide more comprehensive mineralization information. By integrating this multi-source information, the mineralization potential of the area can be assessed more comprehensively and accurately, thus reliably separating the responses indicating mineralization.

[0048] In some preferred embodiments, a specific example is given below. Assume an exploration area with a complex geological background, containing various lithological units and fault structures. First, a refined background field is constructed using the method described above. When analyzing the refined residual signals, the system dynamically adjusts the identification threshold for weak mineralization signals based on the local gradient of the refined background field values ​​and the type of geological unit. For example, within a granite body, the background field changes gently, and the identification threshold is set to a lower value to capture weak copper and molybdenum anomalies; while near lithological contact zones or fault zones, the background field gradient is larger, and the identification threshold is correspondingly increased to avoid misinterpreting natural changes in the background field as mineralization signals. Subsequently, the refined residual signals are compared with these dynamically adjusted thresholds to identify multiple spatially continuous but low-intensity weak mineralization clues. For example, a series of weak lead-zinc anomalies are found in a certain area. Although individual anomalies are not significant, they exhibit a certain linear distribution in space, consistent with the direction of known secondary faults; these are identified as weak mineralization clues. Finally, combining these weak mineralization clues, the system further integrates fault density data for the region (e.g., fault distribution and density maps obtained through remote sensing image interpretation and geophysical data inversion) and other elemental combination anomalies (e.g., gold-silver associated elemental combination anomalies, polymetallic elemental combination anomalies). By constructing a multi-level feature fusion model, this information is weighted and fused to calculate the final mineralization potential score. For example, if a region simultaneously possesses weak mineralization clues, high fault density, and favorable elemental combination anomalies, its mineralization potential score will be significantly improved, thus identifying it as a high-potential concealed mineralization area. In this way, even extremely weak mineralization information can be effectively identified and predicted in complex contexts.

[0049] In some embodiments described above in this application, a threshold for identifying weak mineralization signals is dynamically adjusted based on refined background field values. However, in practical applications, the complexity and heterogeneity of the geological background may make it difficult for simple dynamic adjustment methods to accurately capture all weak mineralization information. If the identification threshold fails to adequately adapt to rapid changes in the local geological background or differences in mineralization favorability among different geological units, the identification accuracy of weak mineralization signals may decrease, or even result in misjudgment or missed judgment.

[0050] In response, this application further proposes a step of dynamically adjusting the identification threshold of weak mineralization signals based on refined background field values, including: Gradient calculations are performed on local regions of the refined background field values ​​to identify areas where the background field has steep gradients or discontinuities. Based on the identified steep gradients or discontinuous regions, a threshold adjustment function T(x) = T0 + k·| B(x)|, where T0 is the basic threshold, | B(x)| represents the gradient intensity of the background field, and k is an adjustment coefficient. In regions where the background field changes gently, a lower recognition threshold is used. In regions where the background field has a steep gradient or is discontinuous, the recognition threshold is increased according to the intensity of the background field gradient. Based on the local geological unit types and known metallogenic models of the region, the adjustment coefficient k is weighted and corrected according to the metallogenic favorability level corresponding to the local geological unit type.

[0051] Specifically, calculating the gradient of a local region in a refined background field involves using mathematical methods (such as finite difference, Sobel operator, or Prewitt operator) to calculate the rate of change of the background field at each spatial location. The aim is to quantify the intensity and direction of local changes in the background field, thereby identifying areas with drastic changes in the background field value that may contain geological boundaries or anomalies. This involves constructing a threshold adjustment function T(x) = T0 + k·| B(x)| can be understood as an adaptive threshold setting mechanism. T0 is the base threshold, representing the default recognition standard in the stable region of the background field. B(x)| represents the background field gradient intensity, reflecting the drastic degree of local changes in the background field. k is an adjustment coefficient used to control the influence of the gradient intensity on the threshold adjustment. When the background field changes gently, the gradient intensity is low, and the identification threshold is close to the base threshold, maintaining high sensitivity to capture weak signals. When the background field has a steep gradient or is discontinuous, the gradient intensity is high, and the identification threshold is increased accordingly to avoid misjudging natural changes in the background field as mineralization signals. In practical applications, combining the local geological unit type and known mineralization model, the adjustment coefficient k is weighted and corrected according to the mineralization favorability level corresponding to the local geological unit type. This means assigning different mineralization favorability levels to different lithologies, structures, or alteration units based on the experience of geological experts and regional mineralization patterns. For example, in areas known to be mineralization-favorable contact zones or fault zones, even with a high background field gradient, it may be necessary to maintain a relatively low identification threshold or assign a high adjustment coefficient k to avoid missing genuine weak mineralization signals; while in areas with low mineralization potential, the value of k can be appropriately adjusted to make the threshold more sensitive to background changes and reduce false anomalies. The purpose is to ensure that threshold adjustment not only considers the numerical changes in the background field, but also incorporates geological expertise, thereby improving the geological rationality of threshold setting.

[0052] In some preferred embodiments, it is assumed that in a certain exploration area, refined background field values ​​show a significant steep gradient near a known concealed fault zone. If a fixed identification threshold is used, the high background values ​​in this fault zone area may be misjudged as weak mineralization anomalies. However, according to the scheme of this application, a local gradient calculation is first performed on the refined background field values ​​to identify the steep gradient in the fault zone area. Subsequently, based on the gradient intensity, a threshold adjustment function T(x) = T0 + k·| B(x)|, dynamically increasing the identification threshold for this region. Simultaneously, since this fault zone is judged by geological experts to be a favorable structure for mineralization, a relatively high weight is assigned when weighting the adjustment coefficient k to ensure that even with an increased threshold, it does not completely block out potentially weak mineralization signals within this favorable structural zone. For example, in the core region of the fault zone, T0 can be set to 0.5, and the gradient intensity | With B(x)| equal to 2.0 and k value corrected to 0.3, the identification threshold T(x) for this region is 0.5 + 0.3 * 2.0 = 1.1. However, in gentle rock mass areas far from fault zones, the gradient intensity may be only 0.1, and k value corrected to 0.1, resulting in an identification threshold T(x) of 0.5 + 0.1 * 0.1 = 0.51. In this way, the system can intelligently adapt to different geological environments, maintaining appropriate sensitivity in favorable mineralization areas with drastic background changes, while maintaining high sensitivity in areas with gentle background changes, thus more accurately identifying genuine weak mineralization clues.

[0053] In some embodiments described above in this application, a refined residual signal is compared with an identification threshold to identify weak mineralization clues. However, in actual exploration, relying solely on a single threshold comparison method may be insufficient to accurately capture concealed mineralization anomalies with low signal intensity but good spatial continuity, or may easily misjudge local background fluctuations as mineralization clues, thereby affecting the accuracy and reliability of weak mineralization information identification.

[0054] In response, this application further proposes a step of comparing refined residual signals with an identification threshold to determine weak mineralization clues, specifically including: A preliminary threshold comparison is performed on the refined residual signal values ​​of each spatial grid cell. Specifically, this preliminary threshold comparison aims to quickly filter out grid cells with signal values ​​higher or lower than a certain range as potential anomalous regions.

[0055] When the refined residual signal value is less than the recognition threshold, the local continuity index C = Σ (the reciprocal of the difference between signal values ​​of adjacent grid cells) in three-dimensional space is calculated, taking into account the local geological unit type of the region. The local continuity index C quantifies the smoothness of signal variation between adjacent spatial grid cells; a larger value indicates better signal continuity in the local area. The reciprocal of the difference between signal values ​​of adjacent grid cells highlights the continuity of areas with smooth signal variation. Incorporating the local geological unit type means considering the lithology, structure, and other geological background of the grid cell when calculating the local continuity index, as different geological units may exhibit different responses and continuity characteristics to mineralization signals.

[0056] The potential mineralization continuity index P is calculated based on the local continuity index C, the proximity of the refined residual signal value to the identification threshold, and the favorable nature of the local geological unit type. The formula is: P = α·C + β·(S / T) + γ·G, where S is the refined residual signal value, T is the identification threshold, G is the geological favorableness coefficient, and α, β, and γ are preset weighting coefficients. Specifically, the potential mineralization continuity index P is a comprehensive indicator that considers not only the spatial continuity of the signal (C), but also the relative relationship between signal intensity and the identification threshold (S / T), and the favorable nature of the regional geological background for mineralization (G). The S / T term reflects the proximity of the signal intensity to the threshold; the closer the signal value S is to the threshold T, the closer this ratio is to 1, indicating a greater likelihood of it being a weak mineralization clue. The geological favorableness coefficient G is assigned a value based on the local geological unit type, combined with known mineralization patterns and expert experience. For example, in known favorable lithological or tectonic zones for mineralization, the G value will be relatively high. Preset weighting coefficients α, β, and γ are used to balance the contributions of these three factors in the calculation of the potential mineralization continuity index P.

[0057] The potential mineralization continuity index P is compared with a preset continuity discrimination threshold. This continuity discrimination threshold is a pre-set critical value used to distinguish genuine weak mineralization cues from background noise.

[0058] Based on the comparison results, weak mineralization clues are identified. When the potential mineralization continuity index P is greater than or equal to the continuity discrimination threshold, the spatial grid cell is identified as a weak mineralization clue.

[0059] In some preferred embodiments, it is assumed that refined residual signal data has been obtained through geochemical measurements in an exploration area. In a certain local area, the refined residual signal values ​​of some grid cells are slightly lower than a preset identification threshold. If only traditional threshold comparison methods are used, these areas may be directly excluded. However, with the scheme of this application, these signal values ​​are first subjected to preliminary threshold comparison. For those grid cells with signal values ​​lower than the identification threshold, the system further calculates their local continuity index C in three-dimensional space. For example, if these grid cells have low signal values ​​but the signal value differences between their adjacent grid cells are small, a higher local continuity index C will be obtained. At the same time, combined with the local geological unit type of the area, such as the area being identified as a favorable fault zone or lithological contact zone for mineralization, a higher geological favorability coefficient G will be assigned. Subsequently, based on the local continuity index C, the closeness (S / T) between the refined residual signal value and the identification threshold, and the geological favorability coefficient G, a potential mineralization continuity index P is calculated. If the calculated P value exceeds the preset continuity discrimination threshold, even if the original signal value is lower than the identification threshold, the area will still be identified as a weak mineralization clue. For example, in a region with a signal value S of 0.8T (where T is the identification threshold), if its C value is high (e.g., 0.9) and G value is also high (e.g., 0.8), the calculated P value, using appropriate weighting coefficients α, β, and γ, may reach 0.7. If the preset continuity discrimination threshold is 0.6, then the region is successfully identified as a weak mineralization clue. This method enables the effective identification of concealed mineralization anomalies with weak signals but good spatial distribution and favorable geological background, avoiding potential missed detections caused by traditional methods.

[0060] In some embodiments described above, this application proposes a scheme for calculating mineralization potential scores based on weak mineralization clues, fracture density, and other elemental combination anomalies. However, in practical applications, these indicative mineralization characteristics often exhibit different patterns and intensities at different spatial scales. Simply performing combined calculations may not fully capture their inherent mineralization information, affecting the accuracy and reliability of mineralization potential assessment. If the above problems are not addressed, important weak mineralization information may be missed, or non-mineralized backgrounds may be misjudged as mineralization anomalies, thus affecting the identification and prediction of concealed mineralization. Therefore, this application further proposes a more refined and robust method for calculating mineralization potential scores, using multi-scale analysis and hierarchical fusion to more comprehensively separate the responses of indicative mineralizations.

[0061] Based on weak mineralization cues, and combined with fracture density and other elemental combination anomalies in a multidimensional feature set, the steps for calculating a mineralization potential score to isolate responses indicative of mineralization include: Spatial scale decomposition was performed on weak mineralization clues, fracture density, and elemental combination anomalies to separate characteristic components at different scales. For each separated scale component, local structural features are identified and extracted; A multi-level feature fusion mechanism is constructed to fuse local structural features at different scales step by step; Based on the experience of geological experts and the laws of mineralization, weights are assigned to local structural features at different scales; Mineralization potential scores are calculated based on the fused multi-level features.

[0062] Specifically, spatial scale decomposition of weak mineralization clues, fracture density, and elemental assemblage anomalies refers to using techniques such as wavelet transform, multi-scale morphological filtering, or Gaussian pyramids to decompose the original weak mineralization clue, fracture density map, and elemental assemblage anomaly map into a series of components reflecting characteristics at different spatial resolutions or frequencies. The aim is to reveal the spatial distribution characteristics of mineralization information at different scales; for example, large-scale decomposition may reflect regional tectonic control of ore deposits, while small-scale decomposition may indicate the occurrence of local ore bodies.

[0063] For each separated scale component, identifying and extracting local structural features can be understood as identifying local anomalies with specific geometric shapes, spatial arrangements, or statistical significance at each specific spatial scale through image processing, pattern recognition, or spatial statistical methods. These local structural features may include linear anomalies (such as fault zones), ring anomalies (such as magmatic bodies or alteration halos), and planar anomalies (such as mineralization enrichment zones within specific lithological units), etc. The purpose is to accurately locate and quantify spatial patterns directly related to mineralization from a complex background.

[0064] In practical applications, constructing a multi-level feature fusion mechanism involves progressively fusing local structural features at different scales. Specifically, this means designing a fusion framework, such as using hierarchical clustering, neural networks, or evidence-based reasoning methods, to gradually integrate local structural features extracted from different scales according to their geological significance and spatial correlation. For example, small-scale features can be fused as sub-components of large-scale features, or features from different scales can be combined through weighted summation, logical operations, etc. The aim is to construct a multi-dimensional, multi-level feature representation that comprehensively reflects mineralization information, avoiding the limitations of single-scale analysis.

[0065] Furthermore, based on the experience of geological experts and metallogenic regularities, weights are assigned to local structural features at different scales. This involves combining known regional metallogenic models, deposit types, and geological experts' understanding of the ore-controlling effects of different geological elements to assign corresponding contribution or importance coefficients to each scale component and its extracted local structural features. For example, in porphyry deposit exploration, mesoscale anomalies associated with ring structures may be given higher weights; in hydrothermal vein deposits, local features associated with linear faults may be more emphasized. The aim is to integrate prior geological knowledge into the feature fusion process, improving the accuracy and geological rationality of mineralization potential assessment.

[0066] Therefore, calculating the mineralization potential score based on the fused multi-level features involves taking the comprehensive features, after multi-scale decomposition, local feature extraction, multi-level fusion, and weighting, as input, and then using a mathematical model (e.g., based on regression analysis, machine learning classifiers, or fuzzy comprehensive evaluation models) to calculate the mineralization potential value of each spatial unit. This score quantifies the mineralization probability of that spatial unit, aiming to provide an intuitive and comparable indicator to guide subsequent exploration decisions.

[0067] In some preferred embodiments, this application is implemented as follows: Suppose that in a certain exploration area, it is necessary to assess the mineralization potential based on geochemical anomalies, fault structure distribution, and remote sensing alteration information.

[0068] First, spatial scale decomposition is performed on geochemical anomaly data (such as the content of indicator elements like Cu, Au, and Mo), fracture density maps, and remote sensing alteration index maps. For example, wavelet transform can be used to decompose these data into high-frequency (corresponding to small-scale local anomalies) and low-frequency (corresponding to large-scale regional background) components.

[0069] For each scale component, local structural features are identified and extracted. For example, in high-frequency components, linear geochemical anomaly zones or high-density fault intersections can be identified; in low-frequency components, regional alteration halos or large rock mass boundaries can be identified. These local structural features are extracted and their boundaries optimized through morphological processing and region growing algorithms.

[0070] Subsequently, a multi-level feature fusion mechanism was constructed. For example, small-scale (high-frequency) linear geochemical anomalies can be fused with mesoscale (medium-frequency) fault zones to identify fault-controlled local mineralization enrichment areas. At the same time, large-scale (low-frequency) regional alteration halos are integrated with these local mineralization enrichment areas to form a comprehensive mineralization potential feature map.

[0071] In this process, based on the known metallogenic regularities of porphyry deposits in the region, geologists might argue that mesoscale alteration anomalies associated with ring structures and localized high-density fault features associated with the intersection of large faults contribute more significantly to mineralization. Therefore, these features will be given higher weight during the fusion process.

[0072] Finally, based on the fused multi-level features, a mineralization potential score for each spatial grid cell is calculated using a support vector machine (SVM) model or a logistic regression model. This score is represented by a value between 0 and 1, with higher values ​​indicating a greater likelihood of mineralization, thus providing precise target area guidance for subsequent drilling verification.

[0073] Specifically, the steps for identifying and extracting local structural features described above can be performed in the following manner.

[0074] For each isolated scale component, the steps for identifying and extracting local structural features include: Each separated scale component undergoes spatial morphological processing to smooth feature boundaries, remove noise, and connect broken regions; Based on the scale components after morphological processing, the texture features of local regions are calculated to quantify the uniformity and complexity within the region. By combining the geological background information corresponding to the scale components, spatial constraints and corrections are applied to the morphological processing results and texture analysis results; Based on the corrected morphological and textural features, local structural features with independent spatial morphology and geological significance are extracted through region growing or segmentation methods. The extracted features are optimized for boundaries to reduce feature cross-contamination.

[0075] The spatial morphological processing of each separated scale component refers to using mathematical morphology methods, such as erosion, dilation, opening, and closing operations, to process the scale components. The aim is to eliminate random noise in the data, smooth feature edges, and connect potential mineralization regions that are broken due to data sparsity or insufficient sampling, thereby making local structural features more complete and continuous. For example, closing operations can connect broken mineralization anomaly zones, and opening operations can remove isolated noise points.

[0076] Furthermore, based on the scale components after morphological processing, the texture features of local regions are calculated. This can be understood as quantifying the statistical characteristics of local regions, such as uniformity, contrast, correlation, and energy, through methods such as gray-level co-occurrence matrix (GLCM), local binary pattern (LBP), or wavelet transform. The aim is to describe the degree of aggregation, directionality, and complexity of mineralization anomalies in terms of spatial distribution, thereby more comprehensively reflecting the spatial structure of potential mineralization information.

[0077] Furthermore, spatial constraints and corrections are applied to the morphological processing and texture analysis results by incorporating geological background information corresponding to scale components. This involves overlaying the processed features with known geological structures (such as faults and folds), lithological distribution, alteration zones, and other information. The aim is to ensure that the extracted local structural features conform to geological laws, avoid extracting geologically meaningless false anomalies, and reasonably adjust the morphology and extent of the features. For example, if the morphological processing results show that a linear anomaly is consistent with the strike of a known fault zone, the confidence level of that anomaly can be increased.

[0078] Specifically, based on the corrected morphological and textural features, local structural features with independent spatial morphology and geological significance are extracted using region growing or segmentation methods. Region growing methods can start from one or more seed points and expand to the surrounding neighborhood according to preset similarity criteria (e.g., signal strength, texture similarity) until a stopping condition is met, thereby extracting complete mineralization anomaly areas. Segmentation methods can divide the entire exploration area into several sub-regions with similar characteristics, each sub-region representing a local structural feature. The aim is to aggregate scattered weak mineralization signals into entities with clear boundaries and geological meaning.

[0079] Finally, boundary optimization of the extracted features is performed to reduce feature cross-contamination. This refers to the fine-tuning of the boundaries of these features after the local structural features have been extracted. For example, edge detection algorithms, active contour models (Snake models), or graph cut-based methods can be used to make the feature boundaries more accurately fit the edges of the actual geological body and avoid confusion between different features due to blurred boundaries, thereby improving the accuracy of subsequent mineralization potential assessment.

[0080] In some of the embodiments described above in this application, although a multi-level feature fusion mechanism is proposed to integrate local structural features at different scales, in practical applications, if the correlation between each feature and the actual mineralization point, as well as the experience of geological experts, are not fully considered, the fusion weights may be set unreasonably, thus affecting the accuracy of the mineralization potential score. To address this, this application further proposes a method for constructing a multi-level feature fusion mechanism, which, by introducing known mineralization point information and expert experience, achieves dynamic optimization of feature weights and adaptive adjustment of the fusion strategy.

[0081] The steps for constructing a multi-level feature fusion mechanism to progressively fuse local structural features at different scales include: Obtain the spatial distribution information of known mineralization points, calculate the spatial correlation strength between each local structural feature and the known mineralization points, and use it as the initial weight; Based on the experience of geological experts, the initial weights were corrected to obtain the comprehensive weights; During the fusion process, the information content of the fused features and their correlation with known mineralization points are monitored, and the fusion weights or fusion strategies are adjusted based on the monitoring results.

[0082] Specifically, obtaining spatial distribution information of known mineralization points refers to collecting and organizing precise spatial coordinate data of discovered mineralization points (e.g., proven ore bodies, mineralization anomalies) within the exploration area through various means such as geological exploration reports, borehole data, and tunnel exposure. This data is typically stored in three-dimensional coordinates, reflecting the true spatial occurrence of mineralization. Calculating the spatial correlation strength between each local structural feature and known mineralization points can be understood as quantifying the spatial proximity or statistical correlation between each local structural feature (e.g., geochemical anomaly clumps, fault zones, alteration halos, etc.) and known mineralization points. For example, methods such as Euclidean distance, spatial autocorrelation analysis, and kernel density estimation can be used for calculation. Features closer to known mineralization points or with distribution patterns more similar to known mineralization points have higher spatial correlation strengths. Using this correlation strength as an initial weight aims to provide a data-driven, objective weighting basis for subsequent feature fusion, ensuring that features more relevant to known mineralization occupy a more significant position in the fusion process.

[0083] The modification of initial weights based on the experience of geological experts involves incorporating their knowledge and judgment into the initial weights. Geological experts can adjust the initial weights based on their in-depth understanding of regional metallogenic regularities, specific deposit type characteristics, and the controlling role of geological structures. For example, experts might consider certain features with weak correlation to the data to be crucial indicators of mineralization from a geological perspective, thus increasing their weights; conversely, the opposite would also be true. This modification can be achieved through weighted adjustments, multiplicative factor adjustments, or direct assignment. The resulting comprehensive weights aim to combine the objectivity of data-driven analysis with the subjective insight of expert experience, forming a more comprehensive and accurate feature weighting system to better reflect the true contribution of different features to mineralization indicators.

[0084] In practical applications, monitoring the information content of the fused features and their correlation with known mineralization points during the fusion process can be understood as a real-time or phased evaluation of the fusion results. Information content can be measured using statistical indicators such as information entropy and variance, reflecting the effective information contained in the fused features. Correlation with known mineralization points can be evaluated using indicators such as prediction accuracy and correlation coefficient, reflecting the predictive ability of the fusion results for known mineralization points. Adjusting the fusion weights or fusion strategy based on the monitoring results aims to achieve adaptive optimization of the fusion process. For example, if monitoring reveals redundancy in the information content of the fused features or a decrease in their correlation with known mineralization points, the fusion weights of each feature can be adjusted (e.g., reducing the weights of features with lower contribution or higher information overlap), or the parameters of the fusion algorithm can be adjusted (e.g., changing the type of fusion model or its internal parameters) to optimize the final mineralization potential score.

[0085] This application's scheme first utilizes the spatial distribution information of known mineralization points to objectively quantify the intrinsic relationship between each local structural feature and actual mineralization, thus providing data-driven initial weights for feature fusion. This avoids weight bias caused by subjective assumptions or lack of experience. Furthermore, by introducing the experience of geological experts to correct the initial weights, the fusion mechanism can incorporate the region's unique metallogenic regularities and the experts' profound understanding of complex geological phenomena, compensating for the limitations of purely data-driven methods. In addition, during the fusion process, the information content of the fused features and their correlation with known mineralization points are monitored in real time, and the fusion weights or fusion strategies are dynamically adjusted based on the monitoring results, forming a closed-loop adaptive optimization process. This dynamic adjustment mechanism ensures that the fusion process can be iteratively optimized according to actual results, thereby making the final mineralization potential score more accurately reflect the true occurrence state of concealed ore.

[0086] Through the above technical solutions, this application can significantly improve the scientificity and accuracy of multi-level feature fusion. By combining known mineralization point information, the feature weight setting becomes more objective; by incorporating the experience of geological experts, it ensures that the fusion mechanism conforms to actual geological laws; and through dynamic monitoring and adjustment, the fusion process can adaptively optimize, thereby effectively avoiding the blindness and rigidity of weight setting in traditional fusion methods. This greatly improves the reliability and accuracy of identifying and predicting hidden mineral weakness information, providing more accurate decision support for geological exploration work.

[0087] In some preferred embodiments, it is assumed that multiple local structural features have been identified in a certain exploration area, such as geochemical anomalies at different scales, fault zones, and alteration halos. First, the spatial coordinates of known mineralization points (such as the locations of ore bodies revealed by boreholes) discovered within the area are obtained. For each local structural feature, the reciprocal of its Euclidean distance to the nearest known mineralization point, or its statistical correlation with the distribution of known mineralization points (such as the spatial autocorrelation coefficient), can be calculated as the initial weight for that feature. For example, features closer to known mineralization points or with stronger correlations receive higher initial weights. Subsequently, geological experts can adjust these initial weights based on the metallogenic model of the area and their own experience. For example, if experts believe that fault structures within a specific lithological unit have particularly important indicative significance for the metallogenesis of the area, their weights can be appropriately increased based on expert experience, even if the initial weight is low. During feature fusion, machine learning models (such as random forests or support vector machines) can be used for fusion, and the prediction accuracy of the model on known mineralization points and the entropy value of the fused features can be monitored in real time. If a decrease in prediction accuracy or redundancy in information is detected, the fusion weights of each feature can be adjusted. For example, the weights of features with lower contribution or higher information overlap can be reduced, or the parameters of the fusion model can be adjusted to optimize the final mineralization potential score.

[0088] Example 2: This application also discloses a system for identifying and predicting hidden mineral weaknesses, the system comprising: The data acquisition and processing module is used to acquire exploration data from various sources, including geological structure, geochemistry, and mineralogy, from the exploration area, and to perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. The feature extraction and integration module is used to extract various features related to mineralization from the processed exploration data and integrate these features to form a multi-dimensional feature set that reflects potential mineralization information. The background stripping module is used to construct a reference distribution reflecting changes in the geological background of the exploration area based on the geological background information of the area; it compares the multidimensional feature set with the reference distribution to separate the responses indicating mineralization; the background stripping module includes: a latent geological texture recognition unit, used to infer latent geological textures by combining remote sensing spectral data and high-resolution geological images to correct the background field; an adaptive threshold adjustment unit, used to dynamically adjust the recognition threshold of weak mineralization signals according to the background field gradient; and a weak signal discrimination unit, used to discriminate weak mineralization clues based on spatial continuity. The mineralization potential assessment module is used to assess the mineralization probability of the exploration area based on the response of the isolated indicator mineralization, and obtain the mineralization potential assessment results; The 3D visualization and interaction module is used to present the mineralization potential assessment results in a 3D form and support geologists to conduct interactive analysis and verification.

[0089] This application presents a system for identifying and predicting weak information in concealed mineralization, aiming to address the challenges of traditional exploration methods in effectively capturing weak mineralization signals obscured by surface background when processing complex multi-source data, and the lack of intuitive 3D visualization support for prediction results. Through a modular design, the system achieves full automation and intelligence across the entire process, from acquiring, processing, extracting and integrating features from multi-source exploration data, to refined background stripping, mineralization potential assessment, and finally, 3D visualization and interactive analysis. The collaborative work of each module ensures comprehensive data processing, accurate weak signal identification, and intuitive assessment results, thereby significantly improving the efficiency and reliability of concealed mineralization identification and prediction.

[0090] The specific implementation method of the concealed mineral weakness information identification and prediction system of this application is as follows: The data acquisition and processing module is used to acquire exploration data from various sources, including geological structure, geochemistry, and mineralogical data, from the exploration area. It then performs spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. The methods for acquiring exploration data, spatial registration, and normalization have already been described in the above embodiments and will not be repeated here. It is important to emphasize that this module can be configured to connect to external data sources through various interfaces. For example, it can receive data from geological databases or remote sensing satellites via a network interface, or read locally stored exploration data via a file interface. Data processing functions can be implemented by integrating various data processing algorithm libraries. For example, the Python-based GDAL library can be used for spatial registration, or the Scikit-learn library can be used for data normalization.

[0091] The feature extraction and integration module is used to extract multiple features related to mineralization from the processed exploration data and integrate these features to form a multi-dimensional feature set reflecting potential mineralization information. The specific methods for feature extraction and integration have been described in the above embodiments and will not be repeated here. It is important to emphasize that this module can be configured to support multiple feature extraction algorithms. For example, statistical methods (such as outlier analysis) can be used to extract geochemical anomaly features, or image processing techniques (such as edge detection) can be used to extract structural features. Feature integration can be achieved through simple feature stitching or through advanced feature fusion using machine learning models (such as neural networks).

[0092] The background stripping module is used to construct a reference distribution reflecting changes in the geological background of the exploration area based on the geological background information of the area. It compares the multidimensional feature set with the reference distribution to separate responses indicating mineralization. This module is crucial for the system to identify weak mineralization signals. It can be configured to construct the reference distribution in various ways; for example, it can be based on a preset geological unit background value table or generate a background field using spatial interpolation algorithms (such as inverse distance weighting). Comparison between the multidimensional feature set and the reference distribution can be achieved through simple difference calculations or by using statistical methods (such as Z-score analysis) to identify anomalies.

[0093] The background stripping module further includes a latent geological texture recognition unit, used to infer latent geological textures by combining remote sensing spectral data and high-resolution geological imagery to correct the background field. This unit can be configured to perform texture analysis on remote sensing spectral data and high-resolution geological imagery using image processing algorithms (such as Fourier transform and wavelet analysis) to identify weak linear or ring-shaped features on the surface. For example, preliminary identification can be performed through manual visual interpretation combined with geographic information system tools, or preliminary inference of texture features can be made through a rule-based expert system.

[0094] The adaptive threshold adjustment unit dynamically adjusts the identification threshold of weakly mineralized signals based on the background field gradient. This unit can be configured to calculate the gradient by performing numerical differentiation on the background field data and then make simple linear or non-linear adjustments based on the gradient value and a preset fixed threshold. For example, a global fixed threshold can be set, and then the threshold can be manually adjusted according to the changing trend of the background field.

[0095] The weak signal discrimination unit is used to identify weak mineralization clues based on spatial continuity. This unit can be configured to perform a simple neighborhood analysis on the residual signal, for example, counting the number of outliers within a fixed radius around a signal point, and comparing this number with a preset empirical value to determine whether it constitutes a weak mineralization clue.

[0096] The mineralization potential assessment module is used to evaluate the mineralization probability of the exploration area based on the response of the isolated indicator mineralization, thus obtaining the mineralization potential assessment result. The specific methods for mineralization potential assessment have already been described in the above embodiments and will not be repeated here. It is important to emphasize that this module can be configured to support multiple assessment models; for example, it can employ a scoring model based on expert experience or a prediction model based on statistical regression. The assessment results can be output in the form of probability maps, potential level maps, etc.

[0097] The 3D visualization and interaction module is used to present the mineralization potential assessment results in 3D form and support geologists in interactive analysis and verification. The specific methods for 3D visualization and interaction have already been described in the above embodiments and will not be repeated here. It is important to emphasize that this module can be configured to achieve 3D visualization through various graphics rendering engines. For example, a custom 3D display interface can be developed based on OpenGL or DirectX, or an application programming interface (API) of existing 3D geological modeling software can be integrated. Interactive functions can be implemented through input devices such as a mouse and keyboard, supporting rotation, zoom, and panning of the viewpoint, as well as querying and annotating model elements.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying and predicting hidden mineral weaknesses, characterized in that: Includes the following steps: We acquire exploration data from various sources, including geological structure, geochemistry, and mineralogical data, from the exploration area, and perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. From the processed exploration data, various features related to mineralization are extracted and integrated to form a multi-dimensional feature set that reflects potential mineralization information. Based on the geological background information of the exploration area, a reference distribution reflecting the changes in the regional geological background is constructed; the multidimensional feature set is compared with the reference distribution to separate the responses indicating mineralization. Based on the response of the isolated indicator mineralization, the mineralization potential of the exploration area is assessed to obtain the mineralization potential assessment results. The mineralization potential assessment results are presented in a three-dimensional format, supporting geologists to conduct interactive analysis and verification.

2. The method for identifying and predicting hidden mineral weaknesses as described in claim 1, characterized in that: The steps of comparing a multidimensional feature set with a reference distribution to separate responses indicating mineralization include: Based on macro-geological maps, the lithological units of the exploration area are divided, and the content of indicator elements in each lithological unit is statistically analyzed to construct a preliminary background field; The initial residual signal is obtained by calculating the difference between the geochemical measurements in the multidimensional feature set and the preliminary background field values. Spatial pattern recognition was performed on the initial residual signal, and latent geological textures were inferred by combining remote sensing spectral data and high-resolution geological images; The background correction amount is calculated based on the implicit geological texture and superimposed on the initial background field to construct a refined background field; The refined residual signal is obtained by calculating the difference between the geochemical measurements in the multidimensional feature set and the refined background field values.

3. The method for identifying and predicting hidden mineral weaknesses as described in claim 2, characterized in that: The steps for spatial pattern recognition of the initial residual signal and inference of latent geological texture by combining remote sensing spectral data and high-resolution geological images include: The initial residual signal is decomposed into multiple scales to separate anomalous components at different spatial scales; For each isolated anomalous component, by combining remote sensing spectral data and high-resolution geological images, latent geological texture features related to specific lithological units or structures are identified. These latent geological texture features include concealed faults, lithological contact zones, and geological bodies that exhibit weak linear or ring-like features in surface images. Based on the preset texture-background value influence coefficient database, query the contribution weight of each latent geological texture feature to the background value of the indicator element; The initial residual signal is corrected based on the contribution weight to remove background interference caused by hidden geological textures.

4. The method for identifying and predicting hidden mineral weaknesses as described in claim 2, characterized in that: The method further includes: The recognition threshold for weak mineralization signals is dynamically adjusted based on the refined background field value. The refined residual signal is compared with the identification threshold to identify weak mineralization clues; Based on weak mineralization cues, combined with fracture density and other elemental combination anomalies in the multidimensional feature set, a mineralization potential score is calculated to separate the responses indicating mineralization.

5. The method for identifying and predicting hidden mineral weaknesses as described in claim 4, characterized in that: The steps for dynamically adjusting the identification threshold of weak mineralization signals based on refined background field values ​​include: Gradient calculations are performed on local regions of the refined background field values ​​to identify areas where the background field has steep gradients or discontinuities. Based on the identified steep gradients or discontinuous regions, a threshold adjustment function T(x) = T0 + k·| B(x)|, where T0 is the basic threshold, | B(x)| represents the gradient intensity of the background field, and k is an adjustment coefficient. In regions where the background field changes gently, a lower recognition threshold is used. In regions where the background field has a steep gradient or is discontinuous, the recognition threshold is increased according to the intensity of the background field gradient. Based on the local geological unit types and known metallogenic models of the region, the adjustment coefficient k is weighted and corrected according to the metallogenic favorability level corresponding to the local geological unit type.

6. The method for identifying and predicting hidden mineral weaknesses as described in claim 4, characterized in that, The steps for comparing refined residual signals with identification thresholds to determine weak mineralization cues include: A preliminary threshold comparison is performed on the refined residual signal values ​​of each spatial grid cell; When the refined residual signal value is less than the identification threshold, the local continuity index of the signal in three-dimensional space is calculated as C = Σ (the reciprocal of the difference between the signal values ​​of adjacent grid cells), and combined with the local geological unit type of the region; Based on the local continuity index C, the proximity of the refined residual signal value to the identification threshold, and the favorableness of the local geological unit type, the potential mineralization continuity index P = α·C + β·(S / T) + γ·G is calculated, where S is the refined residual signal value, T is the identification threshold, G is the geological favorableness coefficient, and α, β, and γ are preset weight coefficients. The potential mineralization continuity index P is compared with a preset continuity discrimination threshold; Based on the comparison results, weak mineralization clues were identified.

7. The method for identifying and predicting hidden mineral weaknesses as described in claim 4, characterized in that: Based on weak mineralization cues, and combined with fracture density and other elemental combination anomalies in a multidimensional feature set, the steps for calculating a mineralization potential score to isolate responses indicative of mineralization include: Spatial scale decomposition was performed on weak mineralization clues, fracture density, and elemental combination anomalies to separate characteristic components at different scales. For each separated scale component, local structural features are identified and extracted; A multi-level feature fusion mechanism is constructed to fuse local structural features at different scales step by step; Based on the experience of geological experts and the laws of mineralization, weights are assigned to local structural features at different scales; Mineralization potential scores are calculated based on the fused multi-level features.

8. The method for identifying and predicting hidden mineral weaknesses according to claim 7, characterized in that, For each isolated scale component, the steps for identifying and extracting local structural features include: Each separated scale component undergoes spatial morphological processing to smooth feature boundaries, remove noise, and connect broken regions; Based on the scale components after morphological processing, the texture features of local regions are calculated to quantify the uniformity and complexity within the region. By combining the geological background information corresponding to the scale components, spatial constraints and corrections are applied to the morphological processing results and texture analysis results; Based on the corrected morphological and textural features, local structural features with independent spatial morphology and geological significance are extracted through region growing or segmentation methods. The extracted features are optimized for boundaries to reduce feature cross-contamination.

9. The method for identifying and predicting hidden mineral weaknesses according to claim 7, characterized in that, The steps for constructing a multi-level feature fusion mechanism to progressively fuse local structural features at different scales include: Obtain the spatial distribution information of known mineralization points, calculate the spatial correlation strength between each local structural feature and the known mineralization points, and use it as the initial weight; Based on the experience of geological experts, the initial weights were corrected to obtain the comprehensive weights; During the fusion process, the information content of the fused features and their correlation with known mineralization points are monitored, and the fusion weights or fusion strategies are adjusted based on the monitoring results.

10. A system for identifying and predicting hidden mineral weaknesses, characterized in that, The system includes: The data acquisition and processing module is used to acquire exploration data from various sources, including geological structure, geochemistry, and mineralogy, from the exploration area, and to perform spatial registration and normalization on the exploration data to ensure consistency in spatial location and numerical range. The feature extraction and integration module is used to extract various features related to mineralization from the processed exploration data and integrate these features to form a multi-dimensional feature set that reflects potential mineralization information. The background stripping module is used to construct a reference distribution reflecting changes in the geological background of the exploration area based on the geological background information of the exploration area; it compares the multidimensional feature set with the reference distribution to separate the responses indicating mineralization; the background stripping module includes: a latent geological texture recognition unit, used to infer latent geological textures by combining remote sensing spectral data and high-resolution geological images to correct the background field; an adaptive threshold adjustment unit, used to dynamically adjust the recognition threshold of weak mineralization signals according to the background field gradient; and a weak signal discrimination unit, used to discriminate weak mineralization clues based on spatial continuity. The mineralization potential assessment module is used to assess the mineralization probability of the exploration area based on the response of the isolated indicator mineralization, and obtain the mineralization potential assessment results; The 3D visualization and interaction module is used to present the mineralization potential assessment results in a 3D form and support geologists to conduct interactive analysis and verification.

Images