Ore body refined three-dimensional modeling method based on offset point adaptive optimization

Abstract

The invention discloses an ore body refined three-dimensional modeling method based on offset point adaptive optimization. The method comprises the steps that control points of an ore body boundary are extracted according to basic exploration data, and then normal vectors of the control points are obtained through a multi-scale weighted PCA method; according to the control points and the normal vectors thereof, adopting an HRBF function to construct an ore body initial model; self-adaptive identification is carried out on a form distortion area in the ore body initial model, and a deviation point of distortion area deviation constraint is generated; and according to the offset point, the control point and the normal vector of the control point, solving an implicit function, and carrying out refined correction on the ore body initial model. According to the method, an index for describing model form deviation is designed, the rationality of the initial implicit model is evaluated, then the initial ore body model of the ore body is finely optimized, the form malformation problem occurring in the complex ore body implicit modeling process is effectively solved, and the accuracy and reliability of ore body three-dimensional modeling are greatly improved.

Description

Technical Field

[0001] The invention relates to an ore body modeling method, in particular to an ore body refined three-dimensional modeling method based on offset point adaptive optimization, and belongs to the field of ore body three-dimensional modeling. Background Art

[0002] Three-dimensional modeling of ore bodies is the three-dimensional mapping of the geometric shape and property distribution of mineral resources with economic value. It is of great significance to the study of mineralization laws, resource estimation, mine production and exploitation decisions, and prospecting prediction.

[0003] The concept of ore body is an economic concept, which is mainly based on the comprehensive delineation of ore quality, mineral processing methods, etc., so the shape is usually more complex. Traditional ore body three-dimensional modeling mainly constructs a three-dimensional ore body model through manual interaction and real-time preview, that is, explicit modeling. The explicit modeling method requires digitizing the outline of the ore body by comparing the drilling data of continuous sections, and then generating a three-dimensional solid model. Although this method can establish a relatively satisfactory ore body model, the modeling process has a high degree of manual participation, strong reliance on expert experience, high time and energy costs, and a low degree of automation, which limits its application in mines.

[0004] The implicit modeling method obtains the implicit function representing the three-dimensional geological body model by solving the interpolation equation that satisfies the geological constraints, and then obtains the three-dimensional geological body model by polygonizing the implicit function representing the geological body model. At present, it is still very challenging to carry out implicit modeling of geological objects with complex geometric features (such as ore bodies and rock bodies). The core problem is the lack of implicit modeling constraint rules. By adding more artificial constraints to control the implicit function, the three-dimensional model can be optimized to a certain extent, but it will greatly affect the degree of automation of the implicit modeling method. In addition, when the ore body structure is relatively complex, the establishment of artificial interpolation constraints usually has problems such as uneven distribution and large uncertainty. Therefore, the implicit modeling results of ore bodies under complex geological conditions and data conditions are often significantly different from the experts' understanding, which greatly affects the development of automated ore body modeling and restricts the development of digital mines and smart mines. Summary of the invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a method for fine three-dimensional modeling of ore bodies based on adaptive optimization of offset points. The method extracts control point information based on the original drilling data, uses the multi-scale weighted PCA method to more accurately calculate the normal vector information of the control points, and reasonably evaluates the morphological deviation of the model. On this basis, a method for optimizing offset points is designed to finely correct the initial model, thereby realizing the automatic construction of a complex ore body surface model. Compared with the prior art, this method not only ensures the continuity of the ore body model, but also avoids unreasonable connections caused by excessive extrapolation of the ore body model, thereby effectively ensuring the integrity and coherence of the complex ore body surface, which is more in line with actual geological conditions.

[0006] In order to achieve the above technical objectives, the present invention provides a method for fine 3D modeling of an ore body based on adaptive optimization of offset points, which is characterized by comprising:

[0007] Step S1, extracting control points of the ore body boundary according to basic exploration data;

[0008] Step S2, using a multi-scale weighted PCA method to obtain the normal vector of the control point in step S1;

[0009] Step S3, constructing an initial model of the ore body using the HRBF function according to the control points and their normal vectors;

[0010] Step S4, adaptively identifying the morphologically distorted area in the initial model of the ore body in step S3, and generating offset points of the offset constraints of the distorted area;

[0011] Step S5, solving the implicit function according to the above offset points, control points and their normal vectors, and making a fine correction to the initial model of the ore body.

[0012] The method provided by the present invention, based on the evaluation of the rationality of the initial implicit model, combines the structural ore control information and the drilling data, adaptively introduces the offset point as the inequality constraint, realizes the optimization of the initial ore body model, and effectively solves the technical problem of the unreasonable extrapolation phenomenon of the model.

[0013] As a preferred solution, the basic exploration data includes borehole bending measurement data, borehole opening coordinate position data and sample data.

[0014] As a preferred solution, the borehole bending measurement data includes the borehole name, inclination angle DIP, azimuth angle BRG and borehole depth.

[0015] As a preferred solution, the drilling hole coordinate position data includes the drilling hole name, the drilling hole opening ordinate X0, the drilling hole opening abscissa Y0, the drilling hole opening elevation Z0 and the drilling depth.

[0016] As a preferred solution, the sample data includes the borehole name, the starting depth and ending depth of the sampling point, and the grade value of the sampling point.

[0017] As a preferred solution, the process of extracting the control points of the ore body boundary is as follows:

[0018] Step S1-1, taking the average of the well start depth and the well end depth of the sampling point as the well depth of the sampling point, and determining the underground three-dimensional position of the well depth of the sampling point by measuring the depth H, the inclination angle DIP and the azimuth angle BRG;

[0019] Step S1-2, by performing linear interpolation on the inclinometer table data corresponding to the depth of the sampling point, the inclinometer information corresponding to each sampling point is obtained to obtain the coordinates of the sampling point;

[0020] Step S1-3, according to the above steps, the coordinates of all sampling points of the ore body are obtained, the sample segment is determined, and then all continuous sample segments are merged to extract the shallow points on the upper surface and the deep points on the lower surface.

[0021] As a preferred solution, the coordinate calculation process of the sampling point is:

[0022] Formula 1: ;

[0023] Formula 2: ;

[0024] Formula 3: ;

[0025] In formulas 1 to 3, X0, Y0 and Z0 are the coordinates of the drilling points, H is the depth of the drilling sampling point, is the inclination of the sampling point calculated by linear interpolation, The azimuth of the sampling point calculated by linear interpolation.

[0026] As a preferred solution, the calculation process of the normal vector in step S2 is:

[0027] Step S2-1: For any control point, calculate the neighborhood point set at three scales , and the centroid of the corresponding neighborhood and use Represents the number of points in the point set, and its calculation process is:

[0028] Formula 4: , ;

[0029] Step S2-2: The centroid obtained in step S2-1 Calculate the centralized neighborhood points And the covariance matrix , the calculation process is:

[0030] Formula 5: ;

[0031] Formula 6: ;

[0032] Step S2-3, obtain the covariance matrix The eigenvector and eigenvalue of , select the eigenvector corresponding to its minimum eigenvalue as the ore body normal vector of the neighborhood ;

[0033] Step S2-4: Calculate the corresponding scale The weight under , and perform weighted summation on the normal vectors to get the normal vector estimation result of the control point , the calculation process is:

[0034] Formula 7: .

[0035] As a preferred solution, the construction process of the initial model of the ore body is: according to the three-dimensional coordinates of the control point and its normal vector, an implicit equation is constructed through the HRBF implicit function, and then the implicit equation is solved to obtain the surface coefficient and the surface equation of the ore body, and then the surface equation is visualized by the moving cube method.

[0036] As a preferred solution, the adaptive identification process of the morphologically distorted area in the initial model of the ore body is as follows: Step S4-1, obtaining all control points and triangular facet data in the space where the model is located, determining its spatial boundary, and using the spatial boundary as the initial domain;

[0037] Step S4-2: Divide the initial domain into multiple subdomains by multi-scale spatial iteration, calculate the distortion index value in each subdomain, standardize the distortion index value, and mark the distorted area according to its absolute value.

[0038] As a preferred solution, the distortion index value in each sub-domain is The calculation process is:

[0039] Formula 8: ;

[0040] Formula 9: ;

[0041] The calculation process of the distortion index value standardization process is as follows:

[0042] Formula 10: ;

[0043] In formula 8~10: for The number of triangles in the subdomain extracted by the algorithm, is the number of control points in the subdomain, For subdomain The mean of For subdomain The standard deviation of .

[0044] As a preferred solution, the offset point generation process is as follows: after the subdomain marked as abnormal is divided into voxels according to the grid size, the center of the voxel is taken as the interpolation point, and the inverse distance weighted interpolation method is used to perform interpolation calculation on the center of the voxel, and the calculation result is compared with the set interpolation threshold to generate the offset point.

[0045] In order to evaluate the rationality of the initial ore body model and identify abnormal areas, the present invention designs an index that describes the conformity of the implicit model with known information, which is calculated by the ratio of the surface area of ​​the implicit model to the number of control points on the ore body surface; this metric can reflect the area where the model is over-extrapolated, because in the area where the index has a large abnormal value, the implicit model tends to show shapes such as bubbles or open sections that do not represent the characteristics of the real ore body, and there may be significant morphological deviations. Considering the efficiency of the model rationality evaluation, the present invention adopts an iterative segmentation strategy, which iteratively identifies the model parts with abnormal ratios by multi-scale spatial division of the model range.

[0046] As a preferred solution, the process of refining and correcting the initial ore body model is: constructing implicit function constraints of offset points, control points and their normal vectors, solving the implicit functions through GRBF, and then extracting the final implicit model through the marching cube algorithm.

[0047] As a preferred solution, the implicit function constraint is:

[0048] Formula 11: ;

[0049] Formula 12: ;

[0050] Formula 13: .

[0051] After obtaining all subdomains marked as abnormal, the present invention uses structural ore-control information to generate a local anisotropic field, divides voxels according to a certain grid size for each subdomain, takes the voxel center as the interpolation point, interpolates the voxel center based on the anisotropic IDW interpolation method, and uses the set interpolation threshold (idw_threshhold) to screen out qualified voxel centers as offset points according to the interpolation results. The probability of these offset points being located on the surface of the ore body is low, and the extended area of ​​the ore body model can be offset constrained to avoid unreasonable boundary prediction, making the model extension more stable and reliable.

[0052] Compared with the prior art, the beneficial technical effects of the technical solution of the present invention are:

[0053] 1) The modeling method provided by the present invention extracts control point information based on original drilling data, uses the multi-scale weighted PCA method to more accurately calculate the normal vector information of the control point, and reasonably evaluates the morphological deviation of the model. On this basis, a method for optimizing offset points is designed to finely correct the initial model, thereby realizing the automatic construction of a complex ore body surface model. Compared with the prior art, this method not only ensures the continuity of the ore body model, but also avoids unreasonable connections caused by excessive extrapolation of the ore body model, thereby effectively ensuring the integrity and coherence of the complex ore body surface, which is more in line with actual geological conditions.

[0054] 2) The indicator describing the conformity of the implicit model with the known information in the technical solution provided by the present invention is calculated by the ratio of the surface area of ​​the implicit model to the number of control points on the surface of the ore body, which can directly reflect the area where the model is over-extrapolated. Combined with the iterative segmentation strategy, multi-scale spatial division is performed within each subdomain of the model to iteratively identify model parts with abnormal ratios. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A flowchart of a three-dimensional modeling method provided in Example 1 of the present invention;

[0056] Figure 2 A schematic diagram of multi-scale weighted PCA normal vector estimation provided in Example 1 of the present invention;

[0057] in, Figure 2 (a) is the neighborhood point of scale k1, Figure 2 (b) is the neighborhood point of k2 scale, Figure 2 (c) is the neighborhood point of scale k3;

[0058] Figure 3 This is a schematic diagram of the initial model of the ore body in Example 1 of the present invention;

[0059] Figure 4 This is a schematic diagram of identifying abnormal areas based on the initial model in Example 1 of the present invention;

[0060] Figure 5 Schematic diagram of local anisotropic IDW interpolation in Example 1 of the present invention;

[0061] in, Figure 5 (a) is the local anisotropy field, Figure 5 (b) is the isotropic Euclidean distance and the elliptical distance after anisotropic attenuation;

[0062] Figure 6This is a schematic diagram of selecting offset points based on abnormal areas in Example 1 of the present invention;

[0063] Figure 7 This is the result of the refined correction of the ore body model based on the original control points, normal vectors and offset point constraints in Example 1 of the present invention. DETAILED DESCRIPTION

[0064] In order to facilitate the understanding of the present invention, the present invention will be described in more comprehensive and detailed manner in conjunction with the accompanying drawings and preferred embodiments of the specification. It should be noted that the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Example 1

[0066] This embodiment provides a method for fine 3D modeling of an ore body based on adaptive optimization of offset points, the steps of which are as follows:

[0067] (1) Extract ore body boundary control points based on original drilling data

[0068] For the original data in the geological database: the borehole bending measurement data table (survey table), the borehole opening coordinate position table (collar table) and the sample table (sample table). Among them, the data in the borehole bending measurement data table include the borehole name, inclination (DIP), azimuth (BRG), and borehole depth. The data in the borehole opening coordinate position table include the borehole name, the borehole orifice ordinate X0, the borehole orifice horizontal coordinate Y0, the borehole orifice elevation Z0, and the borehole depth. The data in the sample table include the borehole name, the well start depth and well end depth of the sampling point, and the grade value of the sampling point.

[0069] First, take the average of the well start depth and well end depth of the sampling point as the well depth of the sampling point. The position of the sampling point well depth in the underground three-dimensional space is determined by three parameters: the measuring point depth H along the borehole axis, the inclination DIP, and the azimuth BRG. The inclination DIP is the angle between the borehole axis and the horizontal plane; the azimuth BRG is the angle between the projection of the borehole axis on the horizontal plane and the magnetic north direction. The inclination information corresponding to each sampling point is obtained by linearly interpolating the inclinometer table data corresponding to the depth of the sampling point well depth, and the coordinates of the sampling point are calculated.

[0070] Assume that (X0, Y0, Z0) are the coordinates of the drilling opening point, H is the depth of the drilling sampling point, α is the inclination of the sampling point calculated by linear interpolation, and β is the azimuth of the sampling point calculated by linear interpolation. Calculate the coordinates of the sampling point (X, Y, Z):

[0071] Formula 1: ;

[0072] Formula 2: ;

[0073] Formula 3: ;

[0074] Similarly, the coordinates of each sampling point on the borehole can be obtained. For adjacent sampling points, if their grade values ​​are greater than or equal to the preset threshold and the depth difference is less than or equal to the threshold of the sample segment length, the segment between the two sampling points is considered as the sample segment. Finally, all continuous sample segments are merged to complete the extraction of sample segments. Based on the sample segments, the shallow points on the upper surface and the deep points on the lower surface are extracted as the key control points of the ore body boundary.

[0075] (2) Multi-scale weighted PCA normal vector estimation

[0076] like Figure 2 As shown, to determine each control point The normal vector , first calculate three scales for each control point The neighboring point set ,use Represents the number of points in the point set and calculates the centroid of the corresponding neighborhood , and then calculate the centralized neighborhood points , based on which the covariance matrix is ​​calculated , by performing eigenvalue decomposition on the covariance matrix, we can obtain the eigenvectors and eigenvalues ​​of the covariance matrix, and select the minimum eigenvalue The corresponding eigenvector As the normal vector of the neighborhood Then, according to the position of the control point at the boundary of the ore body, it is determined whether the direction of the normal vector needs to be adjusted in the opposite direction to ensure the consistency of the direction of the normal vector with the geometric shape of the ore body.

[0077] In order to integrate the normal vector estimation results at different scales, different weights are used to perform weighted summation on the normal vector estimation results. are the weights corresponding to the three scales, which are determined by the noise of the neighboring points at the corresponding scale:

[0078] Formula 4: ;

[0079] in, is the average amplitude of point cloud noise, Corresponding to the weight parameters at the first, second and third scales respectively, the weights in the above formula ensure that when the noise amplitude is small, The weight of is greater, and The weight will be smaller, and the algorithm will retain more estimation information under small radius; on the contrary, when the noise amplitude is large, The weight will be greater than , The algorithm will fully utilize the advantage of large radius and high robustness to noise and outliers. Therefore, the algorithm can ensure the accuracy of normal vector estimation under high levels of noise and outliers.

[0080] For each control point , at different neighborhood scales , , The normal vectors calculated below are . These normal vectors are divided according to predefined weights Perform weighted fusion to obtain the normal vector estimation result of each control point :

[0081] Formula 5: ;

[0082] (3) Preliminary modeling of ore bodies

[0083] When the three-dimensional coordinates of the ore body boundary control points are obtained , and its normal vector When , the implicit equation is constructed using the HRBF implicit function:

[0084] Formula 6: ;

[0085] Formula 7: ;

[0086] The surface coefficients are obtained by solving the implicit equations, and the surface equations of the complex ore body are obtained according to the surface coefficients. Finally, the surface equations are visualized using the marching cube method to complete the construction of the initial model of the ore body. Figure 3 shown.

[0087] (4) Adaptive identification of morphologically distorted regions

[0088] The rationality of the model is evaluated. According to the consistency between the implicit model and the known information, the abnormal area is adaptively identified and the offset points are added. Specifically, the morphological deformity parts in the model are identified, and new offset control points are automatically generated to constrain the initial model.

[0089] First, obtain the control points and triangular patch data of the model space, and calculate the spatial boundary of the entire geological model, and use the determined spatial boundary of the geological model as the initial domain. The initial domain is recursively subdivided into 8 subdomains in each iteration to improve the spatial resolution and identify potential abnormal features. Repeat the following steps until the maximum number of iterations is reached or there are no more subdomains to be split:

[0090] 4-1. Recursive subdivision: recursively subdivide the current block into 8 subdomains to improve spatial resolution;

[0091] 4-2. Statistical analysis: For each subdomain, count the number of control points and the number of valid triangles. Calculate the logarithmic transformation of the number of triangles to reduce its impact:

[0092] Formula 8: ;

[0093] 4-3. Calculate the index value in the subdomain:

[0094] Formula 9: ;

[0095] 4-4. Standardize the comparison values:

[0096] Formula 10: ;

[0097] 4-5. Abnormal subdomain marking: According to The absolute value of marks the abnormal subdomain. Figure 4 As shown, a larger ratio may indicate a larger morphological deviation of the model, reflecting a significant difference between the mineralization characteristics of the area and the surrounding geological bodies. When the maximum number of iterations is reached or there are no more subdomains to be segmented, the output result is: the output is the area marked as abnormal.

[0098] (5) Generation of offset points

[0099] Taking into account the anisotropy problem existing in geological space, this paper establishes a local anisotropy field based on key occurrence information such as the inclination and strike of the fault zone and the ore-controlling characteristics of the geological structure. Through the anisotropic IDW interpolation method, the structural ore-controlling characteristics are integrated into the interpolation calculation, thereby improving the accuracy of the offset point estimation. Figure 5 shown.

[0100] First, the present invention uses the three-dimensional wireframe model (TIN model) of the fracture surface as the data source, calculates the inclination and strike information of the fracture zone, and interpolates the mineralization space to construct a detailed local anisotropy field of structural control of ore. After obtaining the strike and inclination information of the local anisotropy field, for each identified abnormal subdomain, it is divided into a number of individual elements (such as small cubes or cuboid units) within the subdomain according to the set grid size (grid_size), and the center of the element is used as the target point for interpolation calculation.

[0101] For each voxel center position, the anisotropic interpolation parameters of the neighborhood space are calculated, and on this basis, the typical Euclidean distance is expanded to an anisotropic "elliptical" distance. Specifically, principal component analysis (PCA) is first used to calculate the dominant strike and dip directions in the neighborhood space, and then orthogonalized, and then the "vertical direction" perpendicular to the strike and dip directions is calculated, and finally three mutually orthogonal unit vectors are obtained. , respectively the main direction, secondary direction and vertical direction. These directions correspond to different ore control strengths. , construct the rotation matrix :

[0102] Formula 11: ;

[0103] After obtaining the rotation matrix, set reasonable distance attenuation parameters for strike, dip and vertical directions based on geological experience. , By adjusting the distance between the two points The distance in the direction is used to realize anisotropic interpolation. , then between two points The distance in the direction will be farther, that is, the spatial proximity will be weakened; on the contrary, when , then between two points The distance in the direction will be closer, that is, the spatial proximity of the two points will be enhanced. In the case of IDW interpolation, by setting a smaller attenuation parameter in the Euclidean distance in the main mineralization direction and a relatively large attenuation parameter in the non-main mineralization direction, the influence of structural occurrence changes on mineralization can be taken into account and more reliable offset points can be generated.

[0104] After obtaining the anisotropic interpolation parameters at the center of the voxel and After that, first use the rotation matrix Transform the coordinates of the voxel center and the neighborhood points:

[0105] Formula 12: ;

[0106] Then calculate the anisotropic distance from the voxel to the control points in the neighborhood:

[0107] Formula 13: ;

[0108] Finally, perform anisotropic IDW interpolation:

[0109] Formula 14: ;

[0110] According to the result of interpolation calculation, the set interpolation threshold (idw_threshold) is used to filter out the qualified voxel centers as offset points, such as Figure 6 As shown, these locations are considered to have a low probability of being located on the boundary surface of the ore body in space, which can offset the expansion area of ​​the ore body model, avoid unreasonable boundary prediction, and make the model expansion more stable and reliable.

[0111] (6) Refined correction of ore body model

[0112] Based on the generated offset points and the original control points and normal vectors, the following constraints can be constructed for the implicit function:

[0113] Formula 15: ;

[0114] Formula 16: ;

[0115] Formula 17: ;

[0116] Based on the constraints, GRBF is used to solve the implicit function parameters. After solving the implicit function, the marching cube algorithm is used to extract the final implicit model. The result is as follows: Figure 7 shown.

Claims

1. A refined three-dimensional modeling method for an ore body based on adaptive optimization of offset points, characterized in that: include: Step S1, extracting control points of the ore body boundary according to basic exploration data; Step S2, using a multi-scale weighted PCA method to obtain the normal vector of the control point in step S1; Step S3, constructing an initial model of the ore body using the HRBF function according to the control points and their normal vectors; Step S4, adaptively identifying the morphologically distorted area in the initial model of the ore body in step S3, and generating offset points of the offset constraints of the distorted area; Step S5, solving the implicit function according to the above offset points, control points and their normal vectors, and making a fine correction to the initial model of the ore body.

2. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 1, characterized in that: The basic exploration data include borehole bending measurement data, borehole opening coordinate position data and sample data; the borehole bending measurement data include borehole name, inclination DIP, azimuth BRG and borehole depth; the borehole opening coordinate position data include borehole name, borehole mouth ordinate X0, borehole mouth horizontal coordinate Y0, borehole mouth elevation Z0 and borehole depth; the sample data include borehole name, sampling point starting depth, well end depth and sampling point grade value.

3. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 2 is characterized in that: The process of extracting the control points of the ore body boundary is as follows: Step S1-1, taking the average of the well start depth and the well end depth of the sampling point as the well depth of the sampling point, and determining the underground three-dimensional position of the well depth of the sampling point by measuring the depth H, the inclination angle DIP and the azimuth angle BRG; Step S1-2, by performing linear interpolation on the inclinometer table data corresponding to the depth of the sampling point, the inclinometer information corresponding to each sampling point is obtained to obtain the coordinates of the sampling point; Step S1-3, according to the above steps, the coordinates of all sampling points of the ore body are obtained, the sample segment is determined, and then all continuous sample segments are merged to extract the shallow points on the upper surface and the deep points on the lower surface.

4. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 3 is characterized in that: The coordinate calculation process of the sampling point is: Formula 1: ; Formula 2: ; Formula 3: ; In formulas 1 to 3, X0, Y0 and Z0 are the coordinates of the drilling points, H is the depth of the drilling sampling point, is the inclination of the sampling point calculated by linear interpolation, The azimuth of the sampling point calculated by linear interpolation.

5. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 3 is characterized in that: The calculation process of the normal vector in step S2 is: Step S2-1: For any control point, calculate the neighborhood point set at three scales and the centroid of the corresponding neighborhood and use Represents the number of points in the point set, and its calculation process is: Formula 4: , ; Step S2-2: The centroid obtained in step S2-1 Calculate the centralized neighborhood points And the covariance matrix , the calculation process is: Formula 5: ; Formula 6: ; Step S2-3, obtain the covariance matrix The eigenvector and eigenvalue of , select the eigenvector corresponding to its minimum eigenvalue as the ore body normal vector of the neighborhood ; Step S2-4: Calculate the corresponding scale The weight under , and perform weighted summation on the normal vectors to get the normal vector estimation result of the control point , the calculation process is: Formula 7: .

6. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 1, characterized in that: The construction process of the initial ore body model is as follows: according to the three-dimensional coordinates of the control points and their normal vectors, implicit equations are constructed through HRBF implicit functions, and then the implicit equations are solved to obtain the surface coefficients and the surface equations of the ore body, and then the surface equations are visualized using the moving cube method.

7. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 1, characterized in that: The adaptive identification process of the morphologically distorted area in the initial model of the ore body is as follows: Step S4-1, obtaining all control points and triangular facet data in the space where the model is located, determining its spatial boundary, and using the spatial boundary as the initial domain; Step S4-2: Divide the initial domain into multiple subdomains by multi-scale spatial iteration, calculate the distortion index value in each subdomain, standardize the distortion index value, and mark the distorted area according to its absolute value.

8. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 7, characterized in that: The distortion index value in each sub-domain The calculation process is: Formula 8: ; Formula 9: ; The calculation process of the distortion index value standardization process is as follows: Formula 10: ; In formula 8~10: for The number of triangles in the subdomain extracted by the algorithm, is the number of control points in the subdomain, For subdomain The mean of For subdomain The standard deviation of .

9. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 7, characterized in that: The generation process of the offset point is as follows: after the subdomain marked as abnormal in the distorted area is divided into voxels according to the grid size, the center of the voxel is taken as the interpolation point, and the inverse distance weighted interpolation method is used to perform interpolation calculation on the center of the voxel, and the calculation result is compared with the set interpolation threshold to generate the offset point.

10. The method for refined three-dimensional modeling of an ore body based on adaptive optimization of offset points according to claim 1, characterized in that: The process of refining and correcting the initial model of the ore body is as follows: constructing implicit function constraints of offset points, control points and their normal vectors, solving the implicit functions through GRBF, and then extracting the final implicit model through the marching cube algorithm, that is, obtaining; The implicit function constraint is: Formula 11: ; Formula 12: ; Formula 13: .

Images