Dynamic correction method for multi-point driving coupled physical field magma activity model

Abstract

The application provides a multi-point driving coupling physical field magma activity model dynamic correction method, relates to the technical field of geological modeling, and comprises the following steps: multi-source data linkage positioning is performed to obtain spatial discrete control points; a non-uniform gridding initial geological model is constructed; grid neighborhood lithology-property correlation modeling is performed for each control point to generate a correlation parameterized submodel; multi-source measured data binding is performed to obtain spatial discrete property true value data; spatial mapping multi-source data constraint residual diagnosis is performed to output a spatial discrete residual distribution field; grid reconstruction and correlation parameterized submodel parameter correction iteration of a residual anomaly area are triggered until global residual converges to a preset residual threshold value, and a three-dimensional dynamic geological model is output. The technical problem that remote sensing interpretation and geophysical interpretation are independent in the prior art, magma activity information on the ground and underground cannot be effectively combined, and accurate analysis of magma activity mechanism and evolution rules is hindered is solved.

Description

Technical Field

[0001] This invention relates to the field of geological modeling technology, and in particular to a method for dynamic correction of magmatic activity models with multi-point driven coupled physical fields. Background Technology

[0002] Existing technologies have significant limitations in interpreting magmatic activity. Specifically, remote sensing methods can only obtain two-dimensional planar information such as the extent of surface magmatic rock masses and the distribution of dikes, and cannot penetrate the surface to detect underground structures. Although geophysical methods can infer the distribution of underground rock masses, due to the multiple interpretations of geophysical inversion, it is difficult to accurately determine the rock mass type and its correlation with surface structures.

[0003] The two technological achievements are independent of each other. The remote sensing interpretation results lack underground extension verification, and the geophysical interpretation lacks surface geological constraints, resulting in a disconnect between surface and underground magmatic activity information.

[0004] This fragmentation makes it impossible to establish a three-dimensional understanding of magmatic systems, a coherent understanding of lithological unity and thermodynamic behavior, and hinders the accurate analysis of magmatic activity mechanisms and evolutionary laws.

[0005] It should be noted that the information disclosed in this background section is intended only to enhance the understanding of the overall background of the present invention, and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a dynamic correction method for magmatic activity models based on multi-point driven coupled physical fields. This method solves the technical problem that existing technologies, which separate remote sensing interpretation from geophysical exploration interpretation, fail to effectively combine surface and subsurface magmatic activity information, thus hindering the accurate analysis of magmatic activity mechanisms and evolutionary patterns.

[0007] The specific technical solution is as follows:

[0008] This invention provides a method for dynamic correction of a magmatic activity model driven by multiple points coupled with physical fields, the method comprising:

[0009] Using the geographical boundary of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatial discrete control points. These spatial discrete control points are then used as topological nodes to construct a non-uniform gridded initial geological model. By performing lithology-physical property correlation modeling on each control point in the non-uniform gridded initial geological model, a correlated parameterized sub-model is generated. Multi-source measured data are bound to the spatial discrete control points to obtain spatial discrete physical property ground truth data. The spatial discrete physical property ground truth data is used to perform multi-source data constraint residual diagnosis on the correlated parameterized sub-model, outputting a spatial discrete residual distribution field. Based on the spatial discrete residual distribution field, iterative grid reconstruction of the non-uniform gridded initial geological model and parameter correction of the correlated parameterized sub-model in residual anomaly areas are triggered until the global residual converges to a preset residual threshold, outputting a three-dimensional dynamic geological model.

[0010] In one implementation, based on the spatial discrete residual distribution field, the mesh reconstruction of the non-uniformly meshed initial geological model and the parameter correction iteration of the associated parameterized sub-model in the residual abnormal region are triggered until the global residual converges to a preset residual threshold, outputting a three-dimensional dynamic geological model, including:

[0011] S1: Based on the residual abnormal areas of the spatial discrete residual distribution field, the local mesh refinement reconstruction of the non-uniform meshed initial geological model with associated control points is performed to obtain an adaptive refined mesh; S2: Based on the adaptive refined mesh, the lithology-physical property parameters of the associated parameterized sub-model are relaxed and coordinated to obtain the associated parameterized correction model; S3: The spatially mapped multi-source data constraint residual diagnosis of the associated parameterized correction model is performed using the spatial discrete physical property true data, and the spatially corrected residual distribution field is output; the mesh reconstruction and parameter correction iterations of steps S1 to S3 are iteratively executed until the global residual converges to the preset residual threshold, and the three-dimensional dynamic geological model is output.

[0012] In one implementation, using the geographical boundary of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatially discrete control points, including:

[0013] Using the geographical boundary of the magmatic rock mass distribution area as the interpretation benchmark, remote sensing geological data is extracted using remote sensing data. By interpreting the remote sensing geological data, discrete geological structure control points on the surface are located, including the boundary inflection points of discrete surface rock masses and the intersection points of discrete surface dikes. Anomaly gradient fields are obtained by calculating the geophysical gradient field of the magmatic rock mass distribution area. Extreme value focusing inversion is performed on the anomaly gradient field to locate the subsurface discrete anomaly gradient abrupt change points. The discrete geological structure control points on the surface and the subsurface discrete anomaly gradient abrupt change points are spatially aggregated to obtain the spatial discrete control points.

[0014] In one implementation, it further includes:

[0015] The Canny operator is used to extract the blurred boundary of the rock mass from the remote sensing geological data, generating an initial boundary vector line. After overlaying DEM data to correct the terrain shadow interference of the initial boundary vector line, directional gradient filtering is used to enhance the geometric discontinuity of the boundary, outputting a corrected boundary vector line. Curvature calculation is performed on the corrected boundary vector line to obtain a curvature value sequence, and multiple candidate boundary turning points are traversed and screened using a preset curvature threshold. Regional geological maps and field survey data are retrieved according to the geographical boundaries of the magmatic rock mass distribution area. By projecting the multiple candidate boundary turning points onto the regional geological map and comparing them with the field survey data, pseudo-turning point identification and elimination based on the contradiction analysis of the survey results are performed, outputting the boundary turning points of the discrete rock mass on the surface.

[0016] In one implementation, it further includes:

[0017] Linear structural enhancement is performed on the remote sensing geological data based on directional filtering, outputting a multi-directional linear structural feature response field. After binarizing the multi-directional linear structural feature response field using a preset structural response intensity threshold, morphological skeleton extraction is performed, and skeleton line intersection nodes are detected. By performing multi-scale spatial consistency verification on the skeleton line intersection nodes to eliminate isolated noise points, a candidate intersection point set is output. Field verification points are collected in the magmatic rock mass distribution area to perform spatial geological rationality matching of the candidate intersection point set, and the discrete surface dike intersection points are output.

[0018] In one implementation, a non-uniformly gridded initial geological model is constructed using the spatially discrete control points as topological nodes, including:

[0019] A three-dimensional topological connection relationship is established between the spatial discrete control points to generate a three-dimensional constrained topological framework. Density calculations are performed on the spatial discrete control points to obtain a control point density distribution field. Spatial discrete density attributes are assigned to the spatial discrete control points based on the control point density distribution field. Based on the spatial discrete density attributes, grid cells are allocated to the spatial discrete control points to obtain a set of spatial discrete unstructured grid cells. After injecting geological interface constraints into the spatial discrete unstructured grid cell set within the three-dimensional constrained topological framework, inverse distance weighted interpolation is performed based on the geological attributes of the control points to output the non-uniform gridded initial geological model.

[0020] In one implementation, a Delaunay triangulation is used to connect the discrete surface geological structure control points, a constrained tetrahedral partition is used to connect the discrete surface geological structure control points and the discrete underground anomaly gradient abrupt change points, and a forced adjacency constraint is used to connect the discrete underground anomaly gradient abrupt change points.

[0021] In one implementation, a correlated parameterized sub-model is generated by performing grid neighborhood lithology-physical property correlation modeling point-by-point on the non-uniformly gridded initial geological model, including:

[0022] Centered on the spatial discrete control points, a set of spatial discrete neighborhood grid cells is radially extracted from the non-uniformly gridded initial geological model. Lithological distribution patterns are identified within the set of spatial discrete neighborhood grid cells to obtain a set of dominant lithological types and a set of spatial discrete lithological contact relationships. Based on the set of dominant lithological types and spatial discrete lithological contact relationships, a spatial discrete property parameter transfer function is constructed for the spatial discrete control points. Based on the spatial discrete property parameter transfer function, a discrete neighborhood property field is calculated for the set of spatial discrete neighborhood grid cells. The discrete neighborhood property field is used to quantify discrete uncertainty parameters. The set of dominant lithological types and spatial discrete lithological contact relationships are fused to generate a spatial discrete lithological pattern. The discrete neighborhood property field, spatial discrete lithological pattern, and discrete uncertainty parameters are structurally encapsulated in the non-uniformly gridded initial geological model to generate the associated parameterized sub-model.

[0023] In one implementation, multi-source data constraint residual diagnosis is performed by spatially mapping the associated parameterized sub-model using the spatially discrete physical property truth data, outputting a spatially discrete residual distribution field, including:

[0024] Based on the spatial distribution of grid cells, topological matching is performed on the spatial discrete physical property ground truth data and the associated parameterized sub-model to obtain spatial discrete ground truth points. Each spatial ground truth point includes measured physical property values, measured lithology types, and measured uncertainty values. Based on the spatial discrete ground truth points, the spatial mapping residual elements of the associated parameterized sub-model are calculated to obtain spatial discrete residual elements. Each spatial residual element includes absolute physical property residuals, lithology consistency residuals, and uncertainty deviation residuals. Based on the heterogeneous geological scene of the spatial discrete control points, the spatial discrete residual elements are geologically constrained and weighted to output spatial discrete weighted residuals. Controlled kriging interpolation is constructed to diffuse the spatial field of the spatial discrete weighted residuals, generating the spatial discrete residual distribution field.

[0025] Beneficial effects of the embodiments of the present invention:

[0026] By using the geographical boundaries of the magmatic rock mass distribution area as the interpretation benchmark and combining multi-source data linkage positioning, spatial discrete control points can be accurately obtained. This process ensures the accuracy of subsequent model building. Furthermore, the joint positioning of multi-source data can reduce data bias and improve positioning accuracy. By using spatial discrete control points as topological nodes, a non-uniform gridded initial geological model is constructed. Non-uniform gridding ensures that areas with high-density control points have finer grids, which can more accurately capture local geological features, while large grid cells are used in low-density areas to save computational resources. This allows the model to adopt appropriate accuracy for the complexity of different regions, ensuring a balance between overall computational efficiency and accuracy. By performing lithology-physical property correlation modeling at each control point in the non-uniform gridded initial geological model, a correlated parameterized sub-model is generated. This process enables the model to accurately reflect the physical property changes between different lithologies and capture local lithology-physical property relationships in spatial distribution. By binding multi-source measured data to spatially discrete control points, spatially discrete physical property ground truth data is obtained. This process combines measured and predicted data, enabling the model to more accurately reflect real-world physical property characteristics and calibrate the model's initial results. Residual diagnosis is performed using the spatially discrete physical property ground truth data, outputting a spatially discrete residual distribution field. By calculating the difference between the model output and the measured data, the model's deviation in different regions is identified, providing necessary information for model correction. By triggering grid reconstruction in residual abnormal areas and parameter correction iterations of the correlated parameterized sub-model until the global residual converges to a preset threshold, the model can gradually reduce errors through multiple rounds of optimization, ultimately outputting a high-precision three-dimensional dynamic geological model. Through this dynamic correction, the model can continuously adapt to new data and changes, thus providing more accurate predictions.

[0027] Of course, implementing any product or method of the present invention does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This invention illustrates a flowchart of a dynamic correction method for a magma activity model based on a multi-point driven coupled physical field, provided by the present invention.

[0030] Figure 2This paper illustrates a schematic diagram of the multi-source data linkage positioning process in the dynamic correction method for the magma activity model of multi-point driven coupled physical field provided by the present invention. Detailed Implementation

[0031] To facilitate understanding of the present invention, a more complete description of the invention will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein; rather, these embodiments are provided so that the disclosure of the invention will be more thorough and complete.

[0032] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0033] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0034] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0035] The present invention provides a dynamic correction method for magmatic activity models based on multi-point driven coupled physical fields, which addresses the technical problem that existing technologies have independent remote sensing interpretation and geophysical exploration interpretation, resulting in the ineffective integration of magmatic activity information from the surface and subsurface, thus hindering the accurate analysis of magmatic activity mechanisms and evolution laws.

[0036] Example 1: See Figure 1 The present invention provides a method for dynamic correction of a magmatic activity model based on a multi-point driven coupled physical field, the method comprising:

[0037] Y100: Using the geographical boundary of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatial discrete control points.

[0038] The geographical boundaries of igneous rock mass distribution areas are used as interpretation benchmarks. These boundaries are known geological limits, topographic changes, or other significant geological features, and are established through geological maps, remote sensing imagery, and survey data. Multi-source data integration is performed, including: remote sensing geological data, which uses satellite remote sensing and aerial photography to obtain information on the distribution of surface rock masses; and geophysical data, which uses geophysical gradient field calculations to obtain information on subsurface anomalies, helping to identify the distribution characteristics of subsurface rock masses.

[0039] By using remote sensing data analysis and geological surveys, the boundary inflection points and dike intersections of discrete rock masses on the Earth's surface are identified. These control points help define the distribution of surface rock masses. Geophysical gradient field inversion technology is used to locate anomalous gradient abrupt change points underground, which help reveal the structure and properties of subsurface rock masses. These surface and subsurface control points are then spatially aggregated to obtain spatially discrete control points, which form the basis for subsequent model construction.

[0040] Y200: Using the spatial discrete control points as topological nodes, construct a non-uniform gridded initial geological model.

[0041] Spatial discrete control points are used as topological nodes, and their connections are defined to form a three-dimensional topological framework. This framework helps define the grid boundaries and structure, providing a foundation for subsequent meshing. Based on the location, density, and geological characteristics of the spatial discrete control points, a non-uniform grid is constructed. Unlike a uniform grid, the non-uniform grid's cell size can be dynamically adjusted according to actual geological features. For example, areas with igneous rock masses require a denser grid, while other areas can use larger cells. Based on the spatial density, geological characteristics, and topological relationships between control points, the grid cells are rationally allocated. Each grid cell represents a small region in space and possesses corresponding physical properties. Finally, through the meshing process, a non-uniform meshed initial geological model with geological characteristics is formed, serving as the basis for subsequent modeling and analysis.

[0042] Y300: Generates a parametric sub-model by performing grid neighborhood lithology-physical property correlation modeling at each control point in the non-uniform gridded initial geological model.

[0043] Centered on each spatial discrete control point, grid cells surrounding that point are selected one by one. These cells constitute the neighborhood, enabling localized modeling of the geological features around each control point. Lithological patterns are then identified within the selected grid neighborhood. By analyzing geological data (such as lithology type and contact relationships) of the area, the main lithological distribution patterns are identified, involving rock types, structural characteristics, and other relevant factors.

[0044] Based on lithological pattern identification, the relationship between lithology and physical properties (such as magnetic susceptibility, density, resistivity, and radioactivity) is established. Through these parametric transfer functions, physical property data can be inferred and mapped across different regions. Based on the physical property transfer functions, the physical property field within the grid neighborhood is calculated; for example, the physical properties of each grid cell are calculated, forming a complete physical property field. Considering the limitations of geological data and measurement errors, uncertainty analysis is performed on the physical property values ​​of each grid cell. This process generates a confidence interval for the physical property data by quantifying uncertainty parameters. By fusing the discrete neighborhood physical property field, uncertainty parameters, and lithological patterns, a correlated parameterized sub-model is constructed. This model can more accurately describe the lithology-physical property relationship within the region and support subsequent geological analysis.

[0045] Y400: Bind multi-source measured data at the spatial discrete control points to obtain spatial discrete physical property true data.

[0046] Physical property data related to spatially discrete control points are collected through field surveys, sensor measurements, geophysical exploration, or other data acquisition methods. This data includes: lithological types, i.e., classifications of different rocks, such as granite and basalt; and physical properties, such as magnetic susceptibility, density, resistivity, and radioactivity. The collected multi-source measured data are then bound to the previously determined spatially discrete control points. Specifically, corresponding physical property data is assigned to each control point; for example, lithology, magnetic susceptibility, and other physical property data are bound to the control point. Appropriate data preprocessing and verification are performed to ensure the accuracy and consistency of the bound data. After binding at each control point, spatially discrete physical property true value data is generated, containing the true values ​​of lithology, physical properties, and other relevant measurements at the control point.

[0047] Y500: Multi-source data constraint residual diagnosis is performed on the associated parameterized sub-model using the spatial discrete physical property true data, and the spatial discrete residual distribution field is output.

[0048] Mapping spatial discrete physical property ground truth data to the associated parameterized sub-model involves comparing the ground truth data of each control point with the model's predictions, thus mapping the real data to the corresponding spatial location in the model. Multi-source measured data are applied to model constraints; each spatial discrete control point in the model is subject to constraints from different data sources, allowing for more accurate model calibration. Residual diagnosis is then performed. Specifically, for each spatial discrete control point, the difference between the model output and the actual measured value is calculated, termed the residual. Residuals include physical property residuals (such as density and hardness), lithological consistency residuals, and uncertainty deviation residuals. By statistically analyzing the residuals of all control points, a spatial discrete residual distribution field is generated. This field represents the error distribution of the model at various spatial locations, presenting a spatial field that shows areas of significant error.

[0049] Y600: Based on the spatial discrete residual distribution field, trigger the mesh reconstruction of the non-uniform meshed initial geological model and the parameter correction iteration of the associated parameterized sub-model in the residual abnormal area until the global residual converges to the preset residual threshold, and output the three-dimensional dynamic geological model.

[0050] Based on the spatial discrete residual distribution field, regions where the error exceeds a preset threshold are identified. These regions are called residual abnormal regions. The larger error in these regions indicates that the model's predictions in these regions are inaccurate and require optimization and adjustment.

[0051] For residual anomaly areas, the initial non-uniformly meshed geological model undergoes mesh reconstruction. Specific steps include: in residual anomaly areas, the mesh is refined and optimized to improve the mesh resolution in these regions. This helps to simulate magmatic activity and its impacts more precisely. After mesh refinement, details and variations are better captured, improving the model's accuracy in these areas. After mesh reconstruction in the anomaly areas, the model's parameters are corrected iteratively based on the differences between real data and model predictions. Parameters are gradually adjusted to optimize lithology, physical properties, and other attributes. Specifically, this includes: correcting the physical property parameters of mesh nodes to make the model's output more closely match the real data; and continuously adjusting parameters such as the lithological model and physical property transfer function according to the corrected model.

[0052] After completing mesh reconstruction and parameter correction, the residuals are recalculated, and a new residual distribution field is generated. If the new residual values ​​still do not meet the preset threshold, mesh reconstruction and parameter correction continue until the global residuals of the model converge to the set preset residual threshold. Finally, when the model's residuals meet the convergence condition, an optimized three-dimensional dynamic geological model is output. This model can more accurately reflect the spatial distribution and physical characteristics of magmatic activity and is suitable for further analysis and prediction.

[0053] In one implementation, based on the spatial discrete residual distribution field, the mesh reconstruction of the non-uniformly meshed initial geological model and the parameter correction iteration of the associated parameterized sub-model in the residual abnormal region are triggered until the global residual converges to a preset residual threshold, outputting a three-dimensional dynamic geological model, including:

[0054] S1: Based on the residual abnormal areas of the spatial discrete residual distribution field, the local mesh refinement reconstruction of the non-uniform meshed initial geological model with associated control points is performed to obtain an adaptive refined mesh; S2: Based on the adaptive refined mesh, the lithology-physical property parameters of the associated parameterized sub-model are relaxed and coordinated to obtain the associated parameterized corrected model; S3: The spatially mapped multi-source data constraint residual diagnosis of the associated parameterized corrected model is performed using the spatial discrete physical property true data, and the spatial corrected residual distribution field is output; Y610: The mesh reconstruction and parameter correction iterations of steps S1 to S3 are iteratively executed until the global residual converges to the preset residual threshold, and the three-dimensional dynamic geological model is output.

[0055] By discretizing the spatial residual distribution field, regions with significant model errors are identified; these regions are residual anomaly areas and are key areas for model optimization, requiring more detailed modeling. For these identified residual anomaly areas, local mesh refinement is performed, i.e., reducing the size of mesh cells in these regions to increase spatial resolution. Mesh refinement allows for more detailed simulation of property changes and the dynamic processes of magmatic activity in these regions. This local mesh refinement is adaptive, meaning the degree of refinement adjusts according to the size and distribution of the residuals. The refined mesh provides higher accuracy in the problem area, forming an adaptively refined mesh that better captures the details of magmatic activity.

[0056] Based on the adaptively refined mesh, relaxation and coordination correction of lithology and physical properties are performed. Relaxation and coordination correction is an iterative optimization technique that adjusts lithology and physical property parameters to make the model more consistent with actual geological data and physical laws. Lithology correction adjusts the lithology distribution in the model according to the new mesh distribution and physical property data to ensure that the lithology types in the model are more consistent with reality. Physical property correction adjusts physical property parameters, such as magnetic susceptibility, density, resistivity, and radioactivity, to ensure they match actual measured values. Through relaxation and coordination correction, the parameters are adjusted to ensure a more reasonable relationship between lithology and physical properties. Finally, a correlated parameterized correction model is generated, which can more accurately reflect the true characteristics of magmatic activity.

[0057] Mapping spatially discrete physical property ground truth data to an associated parameterized correction model involves comparing and calibrating the actual physical property data with the corrected model. Real physical property data from different data sources, such as geological exploration data and geophysical data, are used to constrain the model, allowing for more accurate model correction. The residuals between the model output and actual measurements are calculated to obtain the spatially corrected residual distribution field. This distribution field reflects the difference between the corrected model and the actual data, showing which areas have larger prediction errors and require further optimization.

[0058] In each iteration, steps S1 to S3 are executed, specifically including: refining and reconstructing the mesh for residual anomalies; correcting lithological and physical property parameters based on the refined mesh; performing residual diagnosis using measured physical property data and the corrected model; and generating a new residual distribution field. After each iteration, it is checked whether the global residuals have converged to a preset residual threshold. If the residual values ​​have not converged, the next iteration continues until the global residuals meet the preset convergence criteria. When the global residuals converge to the preset threshold, the final three-dimensional dynamic geological model is output. This model accurately reflects the spatial distribution, lithological changes, and physical properties of magmatic activity, and is suitable for further analysis and prediction.

[0059] In one implementation, see Figure 2 Using the geographical boundaries of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatially discrete control points, including:

[0060] Y110: Using the geographical boundary of the magmatic rock mass distribution area as the interpretation benchmark, remote sensing geological data is extracted using remote sensing data; Y120: By interpreting the remote sensing geological data, discrete geological structure control points on the surface are located, wherein the discrete geological structure control points on the surface include the boundary turning points of discrete rock masses and the intersection points of discrete dikes on the surface; Y130: Anomaly gradient field is obtained by performing geophysical gradient field calculation on the magmatic rock mass distribution area; Y140: Extreme value focusing inversion is performed on the anomaly gradient field to locate the subsurface discrete anomaly gradient abrupt change points; Y150: The discrete geological structure control points on the surface and the subsurface discrete anomaly gradient abrupt change points are spatially aggregated to obtain the spatial discrete control points.

[0061] The geographical boundaries of the magmatic rock mass distribution area are used as the interpretation benchmark. Geographical boundaries are defined based on historical geological surveys, maps or other geological research data. They can be the natural boundaries of geological units, such as faults, fold belts or other obvious geological features.

[0062] Remote sensing data is used to extract information related to the distribution of magmatic bodies within the geographical boundaries of the magmatic body distribution area, thus obtaining the remote sensing geological data. This remote sensing geological data excludes interference data from non-magmatic body distribution areas. The specific data composition includes: satellite remote sensing data, which effectively identifies surface lithology and extracts geological information such as mineral composition; airborne remote sensing data, which obtains higher-resolution remote sensing data from aerial photography or lidar scanning, allowing for better identification of surface lithology and extraction of mineral composition; and infrared and thermal imaging, which help identify different rock types and geological structures, especially magmatic activity.

[0063] By interpreting geological remote sensing data, we can identify the geological structural features of the Earth's surface, especially discrete control points associated with magmatic rock masses. The boundary inflection points of discrete rock masses on the surface indicate the locations where the boundary position or direction of the magmatic rock mass changes. For example, the distribution, orientation, and density of dikes, and the contact interface between the rock mass and the surrounding rock. In remote sensing images, these points may appear as obvious texture or color changes. The intersection points of discrete dikes on the surface indicate the locations where different rock masses and dikes, or dikes, meet. These are usually areas with relatively complex geological structures. In remote sensing images, the intersection points appear as areas with intersecting structures or overlapping boundaries.

[0064] In areas where igneous rock masses are distributed, geophysical techniques, such as electromagnetic exploration, seismic wave detection, and gravity field measurement, can be used to calculate the underground gradient field. Geophysical techniques can infer the underground structure and lithological distribution by measuring the physical properties of the underground medium, such as conductivity, density, and wave velocity. The gradient field reflects the spatial variation of these physical properties, especially in areas where igneous rock masses are distributed, where the gradient field exhibits significant variations.

[0065] By calculating the gradient field, anomalous regions in the subsurface can be identified. These anomalous regions represent traces of magmatic activity. For example, magmatic rocks cause abnormal changes in the electrical conductivity and density of the subsurface medium. Therefore, changes or abrupt changes in the gradient field become important indicators for identifying these rock masses. Anomaly gradient fields refer to regions where gradient values ​​deviate from the normal background values. These anomalous gradient regions correspond to potential magmatic activity areas and can provide necessary subsurface structural information for subsequent geological modeling.

[0066] The extreme value focusing inversion technique is used to analyze anomalous gradient fields. The core idea of ​​this method is to identify the regions in the gradient field where changes are most significant—that is, the points where gradient values ​​change most drastically. These points correspond to underground geological anomalies or areas of magmatic activity. Extreme value focusing inversion uses mathematical methods, such as backpropagation and optimization algorithms, to convert anomalous gradient values ​​into a speculative model of underground structures. By inverting the gradient values, the specific location of underground anomalies can be obtained. By analyzing the extreme values ​​in the gradient field, i.e., abrupt change points, anomalous gradient abrupt change points are identified. These points indicate where the properties of underground materials change abruptly, indicating the boundaries of magmatic bodies or other geological anomalies. Abrupt change points are usually located at local maxima or minima of the gradient field, marking the boundaries between different underground lithologies or physical properties.

[0067] The extracted discrete surface geological structure control points and subsurface anomalous gradient abrupt change points are spatially aggregated. Using spatial computation methods, these control points are aggregated based on their spatial location, ensuring their interrelationship in three-dimensional space. The aggregated spatial discrete control points integrate surface and subsurface information, enabling a more comprehensive representation of the spatial distribution of geological structures.

[0068] One implementation also includes:

[0069] Y121: The Canny operator is used to extract the blurred boundary of the rock mass from the geological remote sensing data, generating an initial boundary vector line; Y122: After overlaying DEM data to correct the terrain shadow interference of the initial boundary vector line, the boundary geometric discontinuity is enhanced by directional gradient filtering, and the corrected boundary vector line is output; Y123: After calculating the curvature of the corrected boundary vector line to obtain a curvature value sequence, multiple candidate boundary turning points are traversed and filtered using a preset curvature threshold; Y124: Regional geological maps and field survey data are retrieved according to the geographical boundary of the magmatic rock mass distribution area; Y125: By projecting the multiple candidate boundary turning points onto the regional geological map and comparing them with the field survey data, pseudo-turning point identification and elimination based on the contradiction analysis of the survey results are performed, and the boundary turning points of the discrete rock mass on the surface are output.

[0070] The Canny operator is an edge detection algorithm that accurately extracts edge information from images through steps such as filtering, gradient calculation, and non-maximum suppression. In geological remote sensing data, the Canny operator is used to extract rock mass boundaries because these boundaries typically appear as high-contrast areas in the image. Processing remote sensing images with the Canny operator extracts blurred rock mass boundaries. These boundaries may not be entirely accurate due to image quality, noise, or other factors, but further processing can yield more accurate boundary information. The extracted boundary information is then converted into initial boundary vector lines, representing the approximate location of the rock mass boundaries. These vectorized boundary lines can be conveniently used for subsequent geological modeling and analysis.

[0071] Overlaying DEM (Digital Elevation Model) data with the initial boundary vector lines provides terrain height information, helping to eliminate the interference of terrain shadows on boundary extraction, especially in complex terrain or mountainous areas, thus improving the accuracy of boundary extraction. Using DEM data for terrain shadow correction removes boundary deviations caused by terrain factors such as ridges and valleys. This step improves the accuracy of boundary extraction, ensuring more realistic rock mass boundaries.

[0072] The initial boundary vector lines are further enhanced using directional gradient filtering. This filtering helps to enhance the geometric discontinuities of the boundary, making it more prominent in the image. Through this filtering, inflection points and discontinuous areas of the rock mass boundary can be effectively identified. The final corrected boundary vector line is then output, representing more accurate rock mass boundary information and providing more precise data for subsequent geological analysis and modeling.

[0073] Curvature calculation is performed on the positive boundary vector line. Curvature is an index that represents the degree of curvature of a curve. By calculating the curvature at each point, a curvature value sequence is obtained. The curvature value sequence reveals the degree of curvature of the boundary vector line at various locations. In the rock mass boundary, the turning point is manifested as a significant change in curvature value.

[0074] A preset curvature threshold is set to filter out points with larger curvature values, representing inflection or change points at the boundary. This threshold can be adjusted based on geological characteristics. Typically, curvature changes are larger at rock mass boundaries or abrupt structural changes, making these points more likely to be boundary inflection points. The curvature value sequence is iterated through, and points with curvature exceeding the preset threshold are selected as candidate boundary inflection points. These candidate points represent possible rock mass boundary inflection locations and form the basis for subsequent analysis.

[0075] Based on the geographical boundaries of the magmatic rock mass distribution area, geological maps and field survey data for the corresponding region are selected. These geological maps and data contain important information such as the region's geological structure, rock mass distribution, mineral composition, and stratigraphic strike. Regional geological maps and field survey data can be retrieved from geological exploration, historical geological surveys, or databases, providing geological background and structural information to help analyze whether candidate turning points conform to the actual geological structure.

[0076] Multiple candidate boundary inflection points are projected onto regional geological maps and field survey data. The projection process maps the candidate points onto the actual geological maps and field survey data for comparison with the geological background. Based on the contradiction analysis of the survey results, the candidate inflection points are checked to see if they match the distribution and structural characteristics of rock masses and dikes in the regional geological maps and field survey data. If a candidate inflection point is inconsistent with the distribution of rock masses and dikes, it is likely a false inflection point, i.e., an erroneous point caused by image noise or error, and does not represent the true geological structure. Through contradiction analysis, false inflection points that do not match the stratigraphic strike are identified and eliminated. After eliminating false inflection points, the remaining inflection points are the verified boundary inflection points of discrete rock masses on the surface. These inflection points represent the actual changing positions of the rock mass boundaries and are key data in geological modeling and analysis.

[0077] One implementation also includes:

[0078] Y126: Linear structural enhancement is performed on the geological remote sensing data based on directional filtering, and a multi-directional linear structural feature response field is output; Y127: After binarizing the multi-directional linear structural feature response field using a preset structural response intensity threshold, morphological skeleton extraction is performed, and skeleton line intersection nodes are detected; Y128: Multi-scale spatial consistency verification is performed on the skeleton line intersection nodes to eliminate isolated noise points and output a candidate intersection point set; Y129: Field verification points are collected in the magmatic rock mass distribution area to perform spatial geological rationality matching of the candidate intersection point set and output the discrete surface dike intersection points.

[0079] Directional filtering is an image processing technique that enhances structural features in an image in a specific direction. For linear structures in geological remote sensing data, such as faults, folds, and dikes, directional filtering can highlight these features and reduce noise unrelated to linear structures. This filtering technique employs different filter kernels, such as Gabor filters and Sobel filters, which enhance directional features in the image. Multiple directional filters are used to process remote sensing data in different directions, such as horizontal, vertical, and diagonal, enhancing linear structural features in the image. The output is a multi-directional linear structural feature response field, which displays the linear structural features in the image, helping to identify surface and subsurface geological structures.

[0080] By setting a threshold for structural response intensity, the multi-directional linear structural feature response field is binarized. This step separates the strong response regions in the image—areas with obvious geological structural features—from the background, forming a binary image. The purpose of this is to highlight those strong linear structural regions and filter out other irrelevant information. Morphological operations, such as erosion, dilation, opening, and closing operations, are then used to further process the binarized image, extracting skeleton lines. These skeleton lines represent the main linear structures extracted from the remote sensing data, such as dikes and faults, which are crucial information for geological analysis. Intersection points are detected within the extracted skeleton lines. These intersections represent the convergence areas of geological structures, such as dike intersections and fault intersections. Image analysis algorithms are used to accurately locate these intersection points, providing key geological features for subsequent analysis.

[0081] By performing multi-scale analysis on the intersection nodes of the skeleton lines, the consistency of these nodes across different scales is examined. This means that nodes should have similar spatial locations and structural features at different resolutions or scales. Through multi-scale analysis, stable nodes appearing at multiple scales can be identified, and these nodes are generally more likely to be actual geological junctions. Spatial consistency verification of the intersection nodes ensures that their spatial locations do not change significantly across different scales. Only intersection nodes existing at at least two scales are considered reliable geological features. This method can eliminate some isolated noise points, which typically only appear at a certain scale and do not represent actual geological structures. After screening and verification, the final set of candidate intersection points is output, representing the locations of dikes, faults, and other geological junctions.

[0082] In areas where magmatic rock masses are distributed, field verification points are collected. These verification points are obtained through on-site geological surveys, field measurements, or existing geological data. Field verification points can be dike intersections, fault intersections, etc., providing geologically verifiable evidence for candidate intersections. The set of candidate intersections is then matched with the collected field verification points for spatial geological plausibility. By comparing the consistency between the candidate intersections and actual geological characteristics, the rationality of these intersections is verified. During the matching process, the spatial distribution of dikes, lithological characteristics, and other geological structural information are considered. After verification, intersections that conform to actual geological characteristics are confirmed, and the final discrete surface dike intersections are output.

[0083] In one implementation, a non-uniformly gridded initial geological model is constructed using the spatially discrete control points as topological nodes, including:

[0084] Y210: Establish the three-dimensional topological connection relationship between the spatial discrete control points to generate a three-dimensional constrained topological framework; Y220: Calculate the density of the spatial discrete control points to obtain the control point density distribution field; Y230: Assign spatial discrete density attributes to the spatial discrete control points based on the control point density distribution field; Y240: Assign grid cells to the spatial discrete control points based on the spatial discrete density attributes to obtain a set of spatial discrete unstructured grid cells; Y250: After injecting geological interface constraints into the set of spatial discrete unstructured grid cells in the three-dimensional constrained topological framework, perform inverse distance weighted interpolation based on the geological attributes of the control points to output the non-uniform gridded initial geological model.

[0085] Three-dimensional topological connectivity refers to the connection methods between control points. It is defined based on the physical distance, geological relationships, or spatial adjacency between control points. By calculating the relative positions and topological connections between control points, adjacency relationships are established—that is, which control points are adjacent—and the connection patterns in three-dimensional space are defined accordingly. Based on these three-dimensional topological connectivity relationships, a three-dimensional constrained topological framework is generated. This framework describes the connection methods and spatial structure between control points.

[0086] Density calculations are performed on discrete control points in space to understand their distribution in three-dimensional space. By calculating the distances and relative positions between control points, the control point density within each region can be obtained. Density calculations employ methods such as kernel density estimation or neighborhood density estimation, which can assess the distribution of control points within each region. The resulting density distribution field represents the degree of clustering of control points in various regions of space, providing the basis for the subsequent meshing process.

[0087] Based on the control point density distribution field, a density attribute is assigned to each spatial discrete control point. This attribute reflects the control point density in the region where the control point is located. Control points in high-density regions have higher density attributes, while control points in low-density regions have lower density attributes.

[0088] Based on the density attributes of the control points, corresponding grid cells are assigned to each control point. The size of the grid cells is adjusted according to the density attributes: smaller grid cells are used in high-density areas to capture more detail, while larger grid cells are used in low-density areas to improve computational efficiency. The allocation of grid cells is unstructured, meaning that the shape and size of the grid can be freely adjusted to adapt to the needs of different regions. This grid is not a regular rectangle or cube, but can be adaptively distributed in space. After the grid cell allocation is completed, a spatially discrete unstructured grid cell set is obtained, which contains information from all grid cells and serves as the basis for the geological model.

[0089] Geological interface constraint injection refers to injecting actual interface information from the geological model, such as stratigraphic interfaces and faults, into the three-dimensional topological framework. This step ensures that the model adheres to the actual geological structure and interface constraints during construction. These constraints, derived from geological surveys, remote sensing data, or geophysical data, are used to determine the boundaries of different geological units.

[0090] Inverse distance weighted interpolation (IRW) is a spatial interpolation method that estimates the attribute values ​​of unknown points by calculating a weighted average of control points in space. In this method, the weight of a control point is determined inversely proportional to its distance from the interpolation point; that is, control points closer to the interpolation point contribute more to the interpolation result, while those farther away contribute less. This interpolation method allows the generation of geological attributes for each grid cell based on the geological attributes of the control points. Based on the ILDW results, a non-uniformly meshed initial geological model is generated. This model can flexibly reflect the geological characteristics of different regions. The non-uniform mesh improves the accuracy of high-density areas, while using larger grid cells for low-density areas, thus saving computational resources.

[0091] In one implementation, a Delaunay triangulation is used to connect the discrete surface geological structure control points, a constrained tetrahedral partition is used to connect the discrete surface geological structure control points and the discrete underground anomaly gradient mutation points, and a forced adjacency constraint is used to connect the discrete underground anomaly gradient mutation points.

[0092] Delaunay triangulation is a meshing method suitable for meshing irregular point sets. This method ensures that the circumcircle of each triangle does not contain other points by connecting the edges between points, thus avoiding inappropriate meshing. Delaunay triangulation is used to connect discrete geological control points on the Earth's surface, generating a structurally sound triangular mesh that guarantees reasonable and non-overlapping connections between control points.

[0093] Tetrahedral meshing is used to divide a three-dimensional space into multiple tetrahedra, each consisting of four points (vertices). Similar to Delaunay triangulation, tetrahedral meshing aims to ensure a reasonable mesh geometry. Constrained tetrahedral meshing adds constraints to the traditional tetrahedral meshing, such as ensuring the reasonableness of connections between tetrahedra and guaranteeing the accurate representation of geological interfaces. Constrained tetrahedral meshing is used to connect discrete surface geological structural control points and discrete subsurface anomalous gradient abrupt change points, which correspond to subsurface magmatic body boundaries or other geological anomalies.

[0094] Forced adjacency constraints are used to ensure that the connections between discrete anomalous gradient mutation points in the subsurface are reasonable and conform to the actual geological structure. By setting adjacency constraints, the mesh cells in the model are kept consistent and physically reasonable when connected. This constraint helps to avoid the generation of connected regions that do not conform to the actual geological structure when building the mesh, and ensures that each tetrahedron and each triangle of the mesh can accurately reflect the structure of the subsurface rock mass.

[0095] In one implementation, a correlated parameterized sub-model is generated by performing grid neighborhood lithology-physical property correlation modeling point-by-point on the non-uniformly gridded initial geological model, including:

[0096] Y310: Using the spatial discrete control point as the center, radially extract the spatial discrete neighborhood grid cell set from the non-uniform gridded initial geological model; Y320: Perform lithological distribution pattern identification on the spatial discrete neighborhood grid cell set to obtain the spatial discrete neighborhood dominant lithological type set and the spatial discrete lithological contact relationship set; Y330: Construct the spatial discrete physical property parameter transfer function of the spatial discrete control point based on the spatial discrete neighborhood dominant lithological type set and the spatial discrete lithological contact relationship set; Y340: Calculate the discrete neighborhood physical property field of the spatial discrete neighborhood grid cell set based on the spatial discrete physical property parameter transfer function; Y350: Quantify the discrete uncertainty parameter using the discrete neighborhood physical property field; Y360: Fuse the spatial discrete neighborhood dominant lithological type set and the spatial discrete lithological contact relationship set to generate the spatial discrete lithological pattern; Y370: Encapsulate the discrete neighborhood physical property field, the spatial discrete lithological pattern, and the discrete uncertainty parameter in the non-uniform gridded initial geological model mapping structure to generate the associated parameterized sub-model.

[0097] In a non-uniformly gridded geological model, based on the coordinates of a spatially discrete control point, a set of neighboring grid cells is extracted around the point with a certain radius or distance. These neighboring grid cell sets include grid cells within a certain range around the control point, have similar geological properties, and can better represent the local geological features of the control point.

[0098] Within the extracted spatially discrete neighborhood grid cell set, lithological characteristics such as rock type, mineral composition, magnetic susceptibility, and density are analyzed to identify lithological distribution patterns. This process can be based on known geological data, remote sensing imagery, or geophysical results. The most dominant lithological types within the neighborhood grid cell set are identified; these lithological types are the main components of the geological body. By statistically analyzing the lithological distribution of the grid cells, a dominant lithological type set can be obtained, containing grid cells related to the main lithological types of the region. Within the neighborhood grid cell set, contact relationships between lithologies, such as contact surfaces, fault planes, and dike intersections, determine the transition and mutual influence between different lithologies. By analyzing these contact relationships, a lithological contact relationship set is obtained, which describes the spatial contact and interaction between different lithological types.

[0099] Based on the extracted set of dominant lithological types and lithological contact relationships within a spatially discrete neighborhood, a property transfer function is constructed. This function describes the relationship between lithological types and property parameters, and can infer variations in property parameters based on the distribution of different lithological types. The property transfer function can be used to map known physical property data, such as measured values ​​obtained through well logging, drilling, or remote sensing, to neighboring grid cells, thereby inferring the physical parameters of each grid cell. This process helps to accurately describe the physical characteristics of the area near spatially discrete control points and effectively combines the physical properties in the geological model with actual observational data.

[0100] Based on the constructed property transfer function, known lithological types and lithological contact relationships are mapped to spatially discrete neighborhood grid cells. The property transfer function helps infer the physical properties associated with each neighborhood grid cell, such as density, porosity, and permeability. These parameters vary according to changes in lithological type and contact relationships within the neighborhood. Using the property transfer function, physical property calculations are performed on the spatially discrete neighborhood grid cell set. Through this calculation, a physical property value is generated for each grid cell, and the physical property distribution within the entire neighborhood is constructed. Ultimately, the generated discrete neighborhood physical property field describes the changes in physical properties within the region and can accurately reflect local geological characteristics.

[0101] In geological modeling, physical property parameters often exhibit uncertainty due to data incompleteness, measurement errors, and model assumptions. Quantifying discrete uncertainty parameters allows for the assessment and description of the degree of uncertainty in physical property parameters. For example, measurement errors in physical property data, assumption errors in the model, and other uncertainties can be quantified using statistical methods and sensitivity analysis. Uncertainty parameters provide a metric for subsequent model optimization and result interpretation, ensuring the reliability of the model results.

[0102] The spatial discrete neighborhood dominant lithology type set describes the main lithology types in the area surrounding the control point, while the lithology contact relationship set provides the contact and transition relationships between different lithology types. By integrating these two, a comprehensive spatial discrete lithology model is created that describes the distribution and contact relationships between different lithology types.

[0103] The previously generated discrete neighborhood physical property field, spatial discrete lithological model, and discrete uncertainty parameters are mapped to the non-uniformly meshed initial geological model. This process requires ensuring the correct distribution of physical property parameters, lithological models, and uncertainty parameters within the mesh. Structured encapsulation means rationally combining these physical properties and uncertainty parameters according to the structure and topological relationships of the mesh to form a complete geological model. Finally, a correlated parameterized sub-model is generated, which contains lithological, physical property, and uncertainty information and can accurately describe the spatial distribution of geological bodies.

[0104] In one implementation, multi-source data constraint residual diagnosis is performed by spatially mapping the spatially discrete physical property truth data to the associated parameterized sub-model, outputting a spatially discrete residual distribution field, including:

[0105] Y510: Based on the spatial distribution of grid cells, perform topological matching on the spatial discrete physical property ground truth data and the associated parameterized sub-model to obtain spatial discrete ground truth points. Each spatial ground truth point includes measured physical property values, measured lithology types, and measured uncertainty values. Y520: Calculate the spatial mapping residual elements of the associated parameterized sub-model based on the spatial discrete ground truth points to obtain spatial discrete residual elements. Each spatial residual element includes absolute physical property residuals, lithology consistency residuals, and uncertainty deviation residuals. Y530: Based on the heterogeneous geological scene of the spatial discrete control points, perform geological constraint weighting on the spatial discrete residual elements to output spatial discrete weighted residuals. Y540: Construct controlled kriging interpolation to diffuse the spatial field of the spatial discrete weighted residuals, generating the spatial discrete residual distribution field.

[0106] Based on the spatial distribution structure of the grid cells, such as the three-dimensional coordinates of the grid nodes, topological matching is performed between the spatially discrete physical property ground truth data and the associated parameterized sub-model. Here, the grid cells represent local regions of the model, and each grid cell has corresponding physical properties and spatial location. For each grid cell, topological matching compares the actually measured physical property data with the model-predicted physical property data. Topological matching ensures that the physical property data corresponding to each control point can be correctly associated with the physical properties of its location.

[0107] Spatial discrete true value points are the actual measurement points associated with each grid cell. Each true value point consists of three parts: measured physical property values, which are physical property data obtained through actual measurements, such as magnetic susceptibility, density, resistivity, and radioactivity; measured lithology types, which are lithology types obtained through geological exploration; and measured uncertainty values, which are uncertainty data introduced due to measurement errors, model assumptions, and other factors. These true value points provide benchmark data for subsequent residual calculations and model calibration.

[0108] The measured physical property values, lithological types, and uncertainty information of spatially discrete true value points are mapped into the associated parameterized sub-model. This mapping allows for a comparison between the actual data and the model's predictions. For each spatially discrete true value point, its residual elements are calculated. These residual elements represent the differences between the model's predictions and the actual measurements, including: absolute physical property residuals (the absolute difference between the model's predicted physical property values ​​and the actual measurements, reflecting the error in the physical property parameters); lithological consistency residuals (the difference between the model's predicted lithological types and the measured lithological types, reflecting the accuracy of the lithological predictions); and uncertainty deviation residuals (the difference between the uncertainty parameters in the model and the actual measured uncertainty values, reflecting the model's ability to handle data uncertainty). By calculating these residual elements, a discrete residual element set for the entire spatial region is generated, providing data support for subsequent model adjustments and optimizations.

[0109] Heterogeneous geological scenarios refer to geologically diverse regions with significant differences in geological characteristics. This heterogeneity leads to variations in physical properties and lithological distribution. Geological constraint weighting involves weighting residual elements based on the geological characteristics of each region, such as lithology and structure. Typically, geological regions with higher density (such as veins or igneous bodies) require more weighting, while regions with relatively lower density have lighter weights. This weighting ensures that the error correction for different geological regions during model optimization reflects their actual geological characteristics. The weighted residuals generate spatially discrete weighted residuals, which provide more refined error correction data for subsequent model adjustments, ensuring that errors from different geological regions are appropriately handled.

[0110] Kriging interpolation is an interpolation method based on the spatial autocorrelation of geological data, used to estimate the attribute values ​​of unknown points. Here, Kriging interpolation is used to estimate the residual field of the entire region based on weighted residual data. Controlling Kriging interpolation means considering geological constraints and weighted residuals during interpolation to ensure that the interpolation results more closely match the actual geological structure. Kriging interpolation spreads the known weighted residual data across the entire space, generating a spatial residual distribution field. This distribution field shows the error distribution in different regions, thus providing data support for subsequent model correction and optimization. The final generated spatial discrete residual distribution field reveals the error distribution in various regions of the model, providing visual information for adjusting and optimizing the model. This residual information can be used to further correct and improve the geological model, ensuring that the model more accurately reflects the actual geological environment.

[0111] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0112] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A dynamic correction method for a magmatic activity model driven by multiple points coupled with physical fields, characterized in that, The method includes: Using the geographical boundaries of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatial discrete control points; Using the spatially discrete control points as topological nodes, a non-uniformly gridded initial geological model is constructed; By performing grid neighborhood lithology-physical property correlation modeling at each control point in the non-uniform gridded initial geological model, a correlation parameterized sub-model is generated. Multi-source measured data are bound together at the spatial discrete control points to obtain spatial discrete physical property true data; Multi-source data constraint residual diagnosis is performed on the associated parameterized sub-model using the spatial discrete physical property true data to perform spatial mapping, and the spatial discrete residual distribution field is output. Based on the spatial discrete residual distribution field, the grid reconstruction of the non-uniform gridded initial geological model and the parameter correction iteration of the associated parameterized sub-model in the residual abnormal area are triggered until the global residual converges to the preset residual threshold, and a three-dimensional dynamic geological model is output. Specifically, based on the spatial discrete residual distribution field, the non-uniformly meshed initial geological model of the residual abnormal region is triggered to undergo mesh reconstruction and parameter correction iteration of the associated parameterized sub-model until the global residual converges to a preset residual threshold, outputting a three-dimensional dynamic geological model, including: S1: Based on the residual abnormal area of ​​the spatial discrete residual distribution field, the non-uniform gridded initial geological model is reconstructed by local grid densification of associated control points to obtain an adaptive refined grid; S2: Based on the adaptive refined mesh, perform relaxation and coordination correction of lithological and physical property parameters on the associated parameterized sub-model to obtain the associated parameterized correction model; S3: Use the spatial discrete physical property true value data to perform spatial mapping on the associated parameterized correction model for multi-source data constraint residual diagnosis, and output the spatial correction residual distribution field; The mesh reconstruction and parameter correction iterations in steps S1 to S3 are performed iteratively until the global residual converges to the preset residual threshold, and the three-dimensional dynamic geological model is output.

2. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 1, characterized in that, Using the geographical boundaries of the magmatic rock mass distribution area as the interpretation benchmark, multi-source data linkage positioning is performed to obtain spatially discrete control points, including: Using the geographical boundaries of the magmatic rock mass distribution area as the interpretation benchmark, remote sensing geological data is extracted using remote sensing data; By interpreting the remote sensing geological data, control points of discrete geological structures on the surface are located, including the boundary inflection points of discrete rock masses and the intersection points of discrete rock veins on the surface. Anomaly gradient fields were obtained by performing geophysical gradient field calculations on the distribution area of ​​the magmatic rock mass. Extreme value focusing inversion is performed on the abnormal gradient field to locate the abrupt change points of the underground discrete abnormal gradient; The discrete control points are obtained by spatially aggregating the discrete geological structure control points on the surface and the discrete anomalous gradient change points in the underground.

3. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 2, characterized in that, Also includes: The Canny operator is used to extract the rock mass boundary from the remote sensing geological data and generate initial boundary vector lines. After overlaying DEM data to correct terrain shadow interference on the initial boundary vector line, the boundary geometric discontinuity is enhanced by directional gradient filtering, and the corrected boundary vector line is output. After calculating the curvature of the correction boundary vector line to obtain a curvature value sequence, multiple candidate boundary inflection points are traversed and filtered using a preset curvature threshold. Based on the geographical boundaries of the magmatic rock mass distribution area, retrieve regional geological maps and field survey data; By projecting the multiple candidate boundary turning points onto the regional geological map and comparing them with the field survey data, false turning points are identified and eliminated based on the contradiction analysis of the survey results, and the boundary turning points of the discrete rock mass on the surface are output.

4. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 2, characterized in that, Also includes: The remote sensing geological data is linearly structurally enhanced based on directional filtering, and a multi-directional linear structural feature response field is output. After binarizing the multi-directional linear structural feature response field using a preset structural response intensity threshold, morphological skeleton extraction is performed, and skeleton line intersection nodes are detected. By performing multi-scale intersection space consistency verification on the skeleton line intersection nodes, isolated noise points are eliminated, and a candidate intersection point set is output. Field verification points are collected in the magmatic rock mass distribution area to perform spatial geological rationality matching of the candidate intersection point set, and the discrete surface dike intersection points are output.

5. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 2, characterized in that, Using the spatially discrete control points as topological nodes, a non-uniformly gridded initial geological model is constructed, including: Establish the three-dimensional topological connection relationship between the spatial discrete control points to generate a three-dimensional constrained topological framework; Density calculations are performed on the spatially discrete control points to obtain the control point density distribution field; The spatial discrete density attribute is assigned to the spatial discrete control points based on the control point density distribution field. Based on the spatial discrete density attribute, the spatial discrete control points are assigned grid cells to obtain a set of spatial discrete unstructured grid cells; After injecting geological interface constraints into the spatial discrete unstructured grid cell set within the three-dimensional constrained topology framework, inverse distance weighted interpolation is performed based on the geological attributes of the control points to output the non-uniform gridded initial geological model.

6. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 5, characterized in that, Delaunay triangulation is used to connect the discrete surface geological structure control points. Constrained tetrahedral partitioning is used to connect the discrete surface geological structure control points and the discrete underground anomaly gradient abrupt change points. Forced adjacency constraints are used to connect the discrete underground anomaly gradient abrupt change points.

7. The dynamic correction method for a magma activity model with multi-point driven coupled physical fields as described in claim 5, characterized in that, By performing grid neighborhood lithology-physical property correlation modeling at each control point in the non-uniform gridded initial geological model, a correlated parameterized sub-model is generated, including: Centered on the spatial discrete control points, a set of spatial discrete neighborhood grid cells is radially extracted from the non-uniformly gridded initial geological model; Lithology distribution pattern identification is performed on the spatial discrete neighborhood grid cell set to obtain the spatial discrete neighborhood dominant lithology type set and the spatial discrete lithology contact relationship set; Based on the mapping of the spatial discrete neighborhood dominant lithology type set and the spatial discrete lithology contact relationship set, the spatial discrete property parameter transfer function of the spatial discrete control point is constructed. The discrete neighborhood property field is calculated based on the spatial discrete property parameter transfer function for the spatial discrete neighborhood grid cell set; The discrete uncertainty parameters are quantified using the discrete neighborhood property field. By fusing the set of dominant lithological types in the spatial discrete neighborhood with the set of contact relationships between spatial discrete lithologies, a spatial discrete lithology model is generated. The discrete neighborhood physical field, spatial discrete lithological pattern, and discrete uncertainty parameters are encapsulated in the non-uniform gridded initial geological model mapping structure to generate the associated parameterized sub-model.

8. The method for dynamic correction of a magmatic activity model with multi-point driven coupled physical fields as described in claim 1, characterized in that, Multi-source data constraint residual diagnosis is performed on the associated parameterized sub-model using the spatial discrete physical property truth data, and the output spatial discrete residual distribution field is included: Based on the spatial distribution of grid cells, topological matching is performed on the spatial discrete physical property true value data and the associated parameterized sub-model to obtain spatial discrete true value points. Each spatial true value point includes the measured physical property value, the measured lithology type, and the measured uncertainty value. Based on the spatial discrete truth points, the residual elements of the spatially mapped parametric sub-model are calculated to obtain spatial discrete residual elements, wherein each spatial residual element includes physical property absolute residual, lithological consistency residual and uncertainty deviation residual. Based on the heterogeneous geological scene of the spatial discrete control points, the spatial discrete residual elements are subjected to geological constraint weighting, and the spatial discrete weighted residuals are output. The spatial field diffusion of the spatial discrete weighted residuals is generated by constructing controlled Kriging interpolation to produce the spatial discrete residual distribution field.

Images