Deep pillar-oriented digital geological strength index intelligent acquisition method

Abstract

A kind of digital geological strength index intelligent acquisition method for deep pillar, first obtains the digital image of the pillar surface and carries out pretreatment;Based on image automatic identification joint fissure, extract trace length, direction, spacing and opening degree and other geometric and structural parameters;At the same time, identify the apparent weathering degree of rock mass to determine the weathering grade;And build three-dimensional model to extract surface roughness parameter. Further, the above structure, roughness, opening degree and weathering information are integrated to construct the comprehensive index JSVI of apparent joint quality of the pillar. Finally, the JSVI is coupled with the rock mass structure grade parameter SR to establish the GSI calculation model, realizing the non-contact, objective and digital expression of the pillar rock mass quality. The method is simple in implementation process and low in implementation cost. Based on digital image and intelligent identification technology, the whole process automation, objectivity and quantification of deep hard rock pillar rock mass fissure structure parameters, structure surface state index and geological strength index are realized.

Description

Technical Field

[0001] This invention belongs to the field of rock mass engineering and intelligent identification technology, specifically relating to an intelligent method for acquiring digital geological strength indicators for deep ore pillars. Background Technology

[0002] The Geological Strength Index (GSI) is a core parameter in rock engineering that characterizes rock mass quality. It is widely used in key areas such as stability analysis of deep mine pillars, optimization of underground engineering design, and quantitative estimation of rock mass mechanical parameters, and has important guiding significance for engineering construction safety and design rationality.

[0003] In current engineering practice, obtaining GSI values ​​largely relies on technicians manually interpreting rock mass structural characteristics and structural plane development, referring to relevant experience charts. This traditional method has significant limitations: firstly, the evaluation results are highly dependent on the operator's engineering experience, exhibiting strong subjectivity. Differences in interpretation standards among different technicians can lead to significant dispersion in the evaluation results, making it difficult to guarantee objectivity and consistency. Secondly, in deep pillar operations or complex geological environments, on-site geological surveys are hampered by spatial constraints and construction disturbances, resulting in high operational difficulty and low efficiency. This makes it impossible to achieve rapid and continuous evaluation of rock mass quality, failing to meet the actual needs of efficient management in deep engineering projects.

[0004] With the continuous development of digital image processing technology and intelligent recognition algorithms, the automatic extraction of fracture and structural parameters from rock mass images has made automated GSI acquisition possible. However, existing technologies still have significant shortcomings: most methods focus only on the identification of rock mass fractures or the extraction of single structural parameters, lacking comprehensive analysis and fusion application of key parameters across multiple dimensions such as structural surface roughness, opening, and weathering degree. A complete automated GSI acquisition technology solution adapted to the geological conditions of deep hard rock pillars has not yet been formed, failing to effectively address the pain points of traditional manual evaluation methods.

[0005] To address the shortcomings of existing technologies, there is an urgent need to propose an automatic method for acquiring digital geological strength indicators for deep hard rock pillars, so as to achieve automated, quantitative, and continuous rock mass quality evaluation, and provide reliable technical support for the safety management and design optimization of deep rock mass engineering. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method for intelligent acquisition of digital geological strength indicators for deep rock pillars. This method is simple to implement and has low implementation cost. Based on digital image and intelligent recognition technology, it realizes the fully automated, objective, and quantitative acquisition of rock mass fracture structure parameters, structural surface state indicators, and geological strength indicators of deep hard rock pillars.

[0007] To achieve the above objectives, this invention provides a method for intelligent acquisition of digital geological strength indicators for deep ore pillars, comprising the following steps: S1: Digital image acquisition and preprocessing; Acquire digital images of the surface of deep ore pillars or rock outcrops, and preprocess the digital images; S2: Joint and fracture identification and structural parameter extraction; Joint and fracture identification is performed based on preprocessed digital images, fracture geometric information is extracted, and corresponding rock mass structural parameters are obtained. S3: Identification of the apparent weathering degree of rock mass; The apparent weathering degree of the rock mass is identified based on the digital image of the pillar, and the corresponding weathering level of the rock mass of the pillar is obtained; S4: 3D modeling and roughness extraction of ore pillar surface; A three-dimensional model of the ore pillar surface is constructed based on digital images, and its roughness parameters are obtained from the three-dimensional model. S5: Construction of comprehensive index for apparent joint quality of ore pillars; Based on structural parameters, roughness, opening parameters, and weathering grade, a comprehensive index (JSVI) for the apparent joint quality of a pillar is constructed. S6: Calculation of Geological Strength Index (GSI) and Expression of Pillar Rock Mass Quality; Based on the comprehensive structural surface quality index JSVI and the rock mass structural grade parameter SR, a calculation model for the rock mass geological strength index GSI is established, ultimately realizing the digital expression of the quality of the pillar rock mass.

[0008] As a preferred embodiment, the digital image acquisition and preprocessing process in S1 is as follows: S11: Acquire image data; Collect digital images of the surface of deep hard rock pillars; S12: Preprocessing; preprocessing the digital image; the preprocessing includes noise suppression, brightness and contrast normalization, geometric distortion correction, and scale calibration and unification.

[0009] In this technical solution, multiple preprocessing steps, including noise suppression, brightness and contrast normalization, geometric distortion correction, and scale calibration and unification, effectively eliminate interference factors during the acquisition process and establish a mapping between pixels and actual physical dimensions, providing high-quality and standardized input images for subsequent crack identification and parameter calculation.

[0010] To achieve automatic crack identification and objective, automated extraction of geometric and structural parameters, the process of joint crack identification and structural parameter extraction in S2 is as follows: S21: Semantic segmentation of fractures; The preprocessed digital image of the pillar is input into the joint trace recognition model, and pixel-level semantic segmentation is performed on the structural disease area. The fractures on the surface of the pillar are automatically identified, and the fracture area is used as the joint trace recognition result. S22: Crack skeleton extraction and coordinate acquisition; refine the joint trace recognition results, extract the crack trace skeleton, and obtain the crack skeleton structure; based on this, extract the continuous pixel coordinate information of each crack skeleton structure; S23: Geometric parameter calculation; Based on the crack trace skeleton and its pixel coordinate information, calculate the trace length, orientation angle distribution and opening information; S24: Calculation of structural parameters; Based on the spatial distribution relationship of the fracture trace skeleton, calculate the rock mass structural parameters, including trace spacing and fracture density.

[0011] In this technical solution, the crack region is automatically identified based on deep learning semantic segmentation, single-pixel precision traces are obtained through skeleton extraction, geometric parameters such as length, orientation angle, and opening are intelligently calculated, and the trace spacing and crack density are automatically solved by angular periodic clustering and orthogonal projection, which completely replaces manual interpretation and manual surveying, significantly improving efficiency, accuracy and repeatability.

[0012] To achieve automated and high-precision geometric calculations of crack length, orientation angle, and opening, the geometric parameter calculation process in S23 is as follows: S23-1: Crack trace length calculation; The crack trace length is obtained by accumulating the Euclidean distance between adjacent skeleton points to reduce the error caused by pixel discretization; S23-2: Calculation of fracture orientation angle; The least squares method is used to fit a straight line of the fracture skeleton point set, and the slope of the straight line is used to calculate the dominant orientation angle of the fracture, which is used for subsequent grouping of dominant fractures; S23-3: Calculation of crack opening; Based on the crack boundary identification results, the local width at the skeleton point is calculated using the distance field method, the average opening and the maximum opening are statistically analyzed, and mapped to the opening rating.

[0013] In this technical solution, the trace length is accurately calculated by accumulating Euclidean distance based on the coordinates of the skeleton points, effectively eliminating pixel discrete errors; the dominant direction angle of the crack is obtained by least squares straight line fitting, providing a reliable basis for grouping the dominant cracks; the local width is extracted from the crack boundary using the distance field method, and the average and maximum opening are statistically analyzed and automatically mapped to the rating, realizing the comprehensive, objective and repeatable extraction of crack geometric parameters.

[0014] To achieve automatic identification of dominant fracture groups and objective, automated calculation of trace spacing, the calculation process for trace spacing in S24 is as follows: S24-1: Clustering of dominant fracture groups; By statistically analyzing the fracture orientation angle, the Adap-EOKM adaptive clustering method considering the periodicity of the angle is adopted, with the fracture length as the weight, and a clustering objective function based on the circumferential distance is constructed according to formula (5). Fractures with similar orientations are grouped together to identify the dominant fracture groups on the surface of deep hard rock pillars. (5); In the formula, This is a function of the circumferential distance between direction angles. In this embodiment, As the weight, take the crack length. ; S24-2: Geometric reconstruction of the crack group; For each crack group, the centroid position is calculated based on its pixel set, and a covariance matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue is taken as the main direction vector of the group, thereby obtaining the geometric center and dominant extension direction of the crack group. S24-3: Orthogonal projection and coordinate sequence generation; using the principal direction vector and centroid of the fracture group, the fracture points are orthogonally projected along the principal direction vector to transform the fracture distribution in two-dimensional space into a one-dimensional projected coordinate sequence. S24-4: Survey line generation and spacing statistics; Survey lines are automatically generated in the normal direction of the projection direction, the projection distance between adjacent fractures is calculated based on the sorted projection coordinates, and the average value is obtained to obtain the trace spacing of the fracture group.

[0015] In this technical solution, angular periodic adaptive clustering (Adap-EOKM) based on the fracture orientation angle is used with fracture length as the weight to overcome the boundary problem of orientation periodicity and accurately identify the dominant fracture group. The main orientation of each group is reconstructed by the eigenvector of the covariance matrix, and the two-dimensional distribution is transformed into one-dimensional coordinates by orthogonal projection. Survey lines are automatically generated in the normal direction. The distance between adjacent fractures is calculated by sorting the projected coordinates. This completely eliminates the subjectivity of traditional manual survey line layout and realizes the automation, objectivity and high repeatability of distance statistics.

[0016] To achieve automated identification and quantitative rating of the weathering level of structural surfaces, the process for identifying the apparent weathering degree of rock mass in S3 is as follows: S31: Feature extraction; Extracting surface texture and color features of the ore pillar based on digital images of the ore pillar; S32: Weathering Classification; Based on the preset engineering classification standards and combined with the on-site characteristics of deep hard rock pillars, the weathering state of structural surfaces is divided into five levels: unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered. S33: Deep learning discrimination; using a deep learning model to discriminate digital images of ore pillars and obtain the corresponding weathering level; S34: Quantification of weathering grade; converting the apparent weathering grade of the rock mass into a standardized weathering degree rating (WDR) according to a preset mapping rule.

[0017] In this technical solution, the texture and color features of the structural surface are extracted based on the digital image of the ore pillar. The weathering state is subdivided into five levels according to the engineering grading standard. A deep learning classification model (such as RegNetY-400MF) is used to automatically determine the weathering level. The weathering level is converted into a standardized weathering degree rating (WDR) through a preset mapping rule. The qualitative weathering description is transformed into a calculable numerical index, avoiding the subjectivity of manual visual judgment and providing a unified and objective input for the subsequent comprehensive evaluation of the structural surface state.

[0018] To achieve non-contact, high-precision 3D quantization of structural surface roughness, the 3D modeling and roughness extraction process for the ore pillar surface in S4 is as follows: S41: 3D reconstruction; using multi-view stereo vision or structured light methods, a 3D point cloud model of the ore pillar surface is constructed based on multiple digital images; S42: Roughness extraction; Based on a 3D point cloud model, roughness parameters of the structural surface are extracted using surface fitting, slope analysis, or inversion methods based on joint roughness coefficients.

[0019] In this technical solution, a three-dimensional point cloud model of the ore pillar surface is reconstructed based on multiple digital images using multi-view stereo vision or structured light methods. This eliminates the need for contact measurement and adapts to complex underground environments. Roughness parameters are extracted based on the point cloud model using methods such as surface fitting, slope analysis, or JRC inversion. Compared with the traditional standard profile comparison method, the three-dimensional reconstruction can simultaneously capture macroscopic undulations and microscopic roughness, resulting in higher accuracy and smaller errors. This provides more realistic mechanical boundary conditions for assessing the risk of slippage and instability of the ore pillar under high stress.

[0020] To achieve unified quantification and comprehensive index fusion of multi-source structural surface parameters, the comprehensive index construction process for the apparent joint quality of the ore pillar in S5 is as follows: S51: Parameter rating; The extracted fracture features are quantified into fracture feature rating (LDR) according to the preset grading standard; the fracture opening parameter is quantified into opening rating (ADR) according to the degree of opening; the obtained structural surface roughness parameter is quantified into roughness rating (RDR) according to the degree of roughness; the obtained weathering rating (WDR) is directly used as input. S52: JSVI calculation; taking into account the characteristics of crack traces, the opening of the structural surface, the roughness of the structural surface and the degree of weathering of the structural surface, the comprehensive index of structural surface quality JSVI is obtained according to formula (12); (12); In the formula, This is the normalization coefficient.

[0021] In this technical solution, the characteristics of fractures (length, spacing, density), opening, roughness, and weathering degree are quantified according to preset grading standards. After unifying the dimensions, the comprehensive quality index JSVI of the structural surface is calculated by weighted summation and normalization coefficient k. Geometric features, mechanical morphology, and weathering degree are integrated into a single digital index, which solves the problem of fragmented input parameters in traditional GSI. At the same time, it follows the basic principles of rock mechanics, so that the evaluation results have both mathematical objectivity and physical and mechanical significance.

[0022] To achieve coupled calculation of rock mass structure grade and structural plane state, as well as continuous GSI value taking and quality classification, the calculation process of geological strength index GSI and the expression of pillar rock mass quality in S6 are as follows: S61: Determination of rock mass structure grade parameters: Calculation of volume joint number based on the average spacing of discontinuities in each group. And then according to The range of values ​​is calculated using a piecewise logarithmic function to determine the rock mass structure grade parameters. S, as shown in formula (14); (14); 62: GSI Calculation; Based on the rock mass structure grade parameter SR and the structural surface state index JSVI, a geological strength index calculation model suitable for deep hard rock pillars is established to obtain the geological strength index of deep hard rock pillars. As shown in formula (16); (16); S63: Digital representation of the quality of the ore pillar rock mass; the calculated... The values ​​are divided into five levels according to a preset threshold: excellent, good, average, poor, and very poor. The output values ​​and corresponding levels are then displayed.

[0023] In this technical solution, the volumetric joint number is calculated based on the average spacing of each group of discontinuities, and a piecewise logarithmic function is used to map it into continuous rock mass structural grade parameters. This is more refined and suitable for automated calculation than the discrete grading of traditional GSI charts. The SR (Sequential Joint Scale) is coupled with the structural surface condition index JSVI (Structural Surface Condition Index) with equal weights to obtain... The model comprehensively reflects the structural characteristics and structural surface state characteristics of the rock mass; it is simple and has clear physical meaning. Finally, the continuous... The numerical values ​​are divided into five quality levels: excellent, good, medium, poor, and very poor according to the threshold. The output values ​​and corresponding levels transform quantitative indicators into intuitive evaluation conclusions, providing a scientific basis for the stability analysis and support design of deep mine pillars.

[0024] This invention addresses the problems of existing methods for obtaining geological strength indicators (GSI), such as reliance on manual interpretation, high subjectivity, low evaluation efficiency, and difficulty in achieving large-scale, continuous, and rapid evaluation. It provides a digital and automated GSI acquisition solution for deep hard rock pillars, resolving the core pain points of traditional methods and providing reliable technical support for deep rock engineering. First, digital images of the pillar surface are acquired and preprocessed to provide high-quality, standardized input images for subsequent identification and parameter calculation. Second, automatic fracture identification is achieved based on a deep learning model, capturing micro-fractures and hidden joints that are difficult to detect with the naked eye, improving the completeness of structural surface statistics. This allows for accurate and efficient calculation of structural parameters such as fracture traces, direction, density, and spacing, precisely determining the rock mass structural grade parameters. Third, through image color space conversion and texture feature extraction, the degree of weathering can be converted into specific quantitative indicators or probability distributions. Finally, 3D reconstruction extracts the true 3D undulation morphology, and the calculated JRC (Geological Strength Ratio) is more accurate and has less error than the traditional comparative standard profile method. The digital model incorporates both macroscopic undulations and microscopic roughness, providing crucial mechanical boundary conditions for evaluating the risk of slippage and instability of ore pillars under deep, high-stress conditions. Furthermore, it cleverly integrates geometric features, weathering degree, and mechanical morphology into a single comprehensive index, resolving the fragmentation problem of traditional GSI input parameters. Simultaneously, the JSVI's construction logic adheres to fundamental principles of rock mechanics, ensuring that the final evaluation results possess both mathematical objectivity and physical and mechanical significance. Finally, through a continuous calculation model of JSVI and SR, continuous GSI values ​​are achieved, better reflecting the gradual variation of rock mass quality in nature and enabling more accurate acquisition of geological strength indicators.

[0025] Compared with traditional methods that rely on experience for interpretation, this invention has the following advantages: 1. Based on digital image and intelligent recognition technology, the parameters of rock mass fracture structure, surface roughness and geological strength index are automatically acquired, which reduces the subjective error caused by traditional manual interpretation and improves the objectivity and consistency of rock mass quality evaluation. 2. By constructing the Structural Surface State Index (JSVI), a multi-parameter comprehensive characterization of the structural surface state of deep hard rock pillars was achieved, improving the comprehensiveness and objectivity of the evaluation. 3. By establishing a geological strength index calculation model applicable to deep hard rock pillars, the geological strength index of deep hard rock pillars can be obtained quickly, automatically, and intelligently, which greatly improves the evaluation efficiency. 4. By organically combining crack identification, spacing calculation, weathering identification, roughness and opening analysis, and GSI model construction, a complete digital evaluation process is formed, which has good engineering application prospects and promotion value.

[0026] 5. By accurately mapping pixels to actual physical dimensions, the crack size parameters under different images and different acquisition distances are made comparable in a uniform manner, providing a metrological basis for multi-source data fusion.

[0027] 6. By adopting angular periodic clustering and orthogonal projection to automatically generate survey lines, the subjectivity of traditional manual survey line layout is completely eliminated, realizing the automation and repeatability of spacing statistics and significantly improving the reliability of spacing parameters.

[0028] This method is simple to implement and has low implementation costs. Based on digital image and intelligent recognition technology, it realizes the full-process automated, objective and quantitative acquisition of rock mass fracture structure parameters, structural surface state indicators and geological strength indicators of deep hard rock pillars. It overcomes the shortcomings of traditional methods, such as reliance on manual interpretation, strong subjectivity and low efficiency, and provides an efficient and reliable technical means for the quality evaluation of deep rock mass engineering. Attached Figure Description

[0029] Figure 1 This is a flowchart of the present invention; Figure 2 This is a diagram showing the identification results of structural defects in the pillars in this invention; Figure 3 This is a schematic diagram illustrating the calculation of pillar fracture spacing in this invention; Figure 4 This is a diagram showing the results of the apparent weathering degree determination of the ore pillar in this invention; Figure 5 This is a diagram showing the calculation of surface roughness and opening of the pillar structure in this invention; Figure 6 This is the GSI evaluation chart for hard rock pillars in this invention. Detailed Implementation

[0030] like Figures 1 to 6 As shown, this invention provides a method for intelligent acquisition of digital geological strength indicators for deep ore pillars, characterized by the following steps: S1: Digital image acquisition and preprocessing; acquire digital images of the surface of deep ore pillars or rock outcrops, and preprocess the digital images to provide high-quality input images for subsequent fracture identification and weathering discrimination; As a preferred embodiment, the digital image acquisition and preprocessing process in S1 is as follows: S11: Acquire image data; Collect digital images of the surface of deep hard rock pillars; In this embodiment, a digital image of the surface of a deep hard rock pillar is first acquired. The image can be obtained using an industrial camera, an underground mobile acquisition terminal, a drone platform, or other image acquisition equipment suitable for the underground mining environment. During the acquisition process, it is preferable to ensure that the image contains complete structural information of the pillar surface and to minimize the impact of uneven lighting, occlusion, and motion blur on subsequent identification. S12: Preprocessing; Preprocessing the digital image. The preprocessing includes noise suppression, brightness and contrast normalization, geometric distortion correction, and scale calibration and unification. Among them, noise suppression uses Gaussian filtering, median filtering or bilateral filtering to eliminate sensor noise and environmental interference; Brightness and contrast normalization employs adaptive histogram equalization or Retinex-based enhancement methods to suppress the effects of non-uniform downhole illumination and unify image grayscale distribution characteristics. Geometric distortion correction is based on pre-calibrated camera intrinsic parameters and lens distortion coefficients to correct radial and tangential distortion of the image; for perspective distortion caused by shooting angle, it is compensated by planar homography transformation or 3D reconstruction methods in the subsequent structural surface parameter extraction stage. The scaling and unification are based on reference objects with known physical dimensions in the image (such as the spacing between anchor bolt holes in the pillar, calibration ruler, etc.) or the object distance and focal length parameters recorded during acquisition. The proportional relationship between pixel size and actual physical size is established, and all images are resampled to a uniform spatial resolution.

[0031] S13: After preprocessing, a standardized digital image of the ore pillar is obtained, which serves as input data for subsequent fracture identification, weathering identification, and acquisition of structural surface parameters.

[0032] Through multiple preprocessing steps, including noise suppression, brightness and contrast normalization, geometric distortion correction, and scale calibration and unification, interference factors during the acquisition process are effectively eliminated and a mapping between pixels and actual physical dimensions is established, providing high-quality, standardized input images for subsequent crack identification and parameter calculation.

[0033] S2: Joint and fracture identification and structural parameter extraction; joint and fracture identification is performed based on preprocessed digital images, fracture geometric information is extracted, and corresponding rock mass structural parameters are obtained; To achieve automatic crack identification and objective, automated extraction of geometric and structural parameters, the process of joint crack identification and structural parameter extraction in S2 is as follows: S21: Semantic segmentation of fractures; The preprocessed digital image of the pillar is input into the joint trace recognition model, and pixel-level semantic segmentation is performed on the structural disease area. The fractures on the surface of the pillar are automatically identified, and the fracture area is used as the joint trace recognition result. The specific process is as follows: In this embodiment, intelligent identification of structural defects is performed based on the preprocessed digital image of the mine pillar (see attached). Figure 2 Preferably, a deep integrated network ECA-SegFormer-BiFPN is used to identify and detect the apparent deterioration areas of the ore pillar; wherein, the ECA module is used to enhance the channel feature representation capability, and BiFPN is used to realize multi-scale feature fusion, thereby improving the recognition accuracy of complex trace morphology, small and intersecting fracture traces; let the input image be x, the model output is a fracture probability distribution map of the same size as the input image. ,in These are the network parameters for the crack identification model.

[0034] To transform the probability distribution map into fracture region results that can be used for geometric analysis, a threshold is set. The probability map is binarized according to formula (1) to obtain the fracture binary map. ; (1); In the formula, This indicates that the pixel belongs to the crack region; This indicates that the pixel belongs to a non-cracked region; S22: Crack skeleton extraction and coordinate acquisition; the joint trace recognition results are refined to extract the crack trace skeleton and obtain the crack skeleton structure; based on this, the continuous pixel coordinate information of each crack skeleton structure is extracted; the specific process is as follows: Morphological refinement is performed on the binary crack image. An iterative skeleton extraction algorithm is preferred to remove redundant boundary pixels to obtain a crack skeleton structure with a single pixel width. The skeleton structure can better maintain the topological morphology, spatial connectivity and main direction information of the crack. After obtaining the fracture skeleton, connected component analysis is performed on the skeleton graph to separate different fractures into several independent fracture objects. For each fracture object, its skeleton pixel coordinate set is extracted and sorted according to pixel adjacency to construct a fracture path sequence. Let a certain fracture skeleton point sequence be... Based on this, the apparent structural parameters of hard rock pillars are extracted; S23: Geometric parameter calculation; Based on the fracture trace skeleton and its pixel coordinate information, calculate the trace length, orientation angle distribution and opening information to characterize the structural features of deep hard rock pillar rock mass; To achieve automated and high-precision geometric calculations of crack length, orientation angle, and opening, the specific process for calculating geometric parameters is as follows: S23-1: Crack trace length calculation; The crack trace length is obtained by accumulating the Euclidean distance between adjacent skeleton points to reduce the error caused by pixel discretization; Specifically, the length of the fracture trace is obtained according to formula (2). ; (2); In the formula, The coordinates of the skeleton points; The above methods can effectively reduce the length estimation deviation caused by pixel discreteness; S23-2: Calculation of fracture orientation angle; The least squares method is used to fit a straight line of the fracture skeleton point set, and the slope of the straight line is used to calculate the dominant orientation angle of the fracture, which is used for subsequent grouping of dominant fractures; Specifically, to obtain the dominant direction information of the fractures, a least-squares method is used to fit a straight line to the fracture point set. Let the expression of the fitted straight line be... Then, according to formula (3), the least squares objective function is obtained; (3); Slope obtained by fitting Afterwards, the direction angle of the fracture Represented as When the line is perpendicular, take fracture direction angle Used to characterize the orientation features of joint traces and as a basis for subsequent grouping of dominant traces and calculation of corresponding spacing; S23-3: Calculation of crack opening; Based on the crack boundary identification results, the local width at the skeleton point is calculated using the distance field method, the average opening and the maximum opening are statistically analyzed, and mapped to the opening rating.

[0035] Specifically, the crack opening is used to characterize the width of the crack and is obtained based on digital image measurement methods. The crack boundary is extracted based on the crack boundary recognition results, and the local crack opening is calculated using the distance field method.

[0036] Calculate any point according to formula (4) Distance field to the boundary ; (4); In the formula, Indicates the fracture boundary; For the center point of the fracture skeleton Its local width can be expressed as Based on the local width information corresponding to each skeleton point. The average width of all skeleton points of the crack is taken as the average opening. Alternatively, the maximum value can be taken as the maximum opening. Then, the opening rating is obtained by mapping the range of values ​​for the opening parameter. The trace length is accurately calculated by accumulating Euclidean distances based on the coordinates of the skeleton points, effectively eliminating pixel discrepancy errors. Least squares linear fitting is used to obtain the dominant fracture orientation angle, providing a reliable basis for grouping dominant fractures. The local width is extracted from the fracture boundary using a distance field method, and the average and maximum opening are statistically analyzed and automatically mapped to a rating, achieving comprehensive, objective, and repeatable extraction of fracture geometric parameters. Through these methods, the intelligent acquisition and expression of the opening of structural surfaces in deep hard rock pillars is realized.

[0037] S24: Structural parameter calculation; Based on the spatial distribution relationship of the fracture trace skeleton, calculate the rock mass structural parameters, including trace spacing and fracture density, to provide basic data for subsequent structural grade evaluation; Regarding trace spacing: After obtaining the geometric information of the fracture in the pillar, the corresponding spacing parameters are further calculated. Unlike the traditional method of manually laying out survey lines, this implementation method achieves intelligent calculation of trace spacing through trace grouping, geometric reconstruction, orthogonal projection, and spacing statistics (see appendix). Figure 3 This reduces the subjectivity caused by human selection of the survey line direction; To achieve automatic identification of dominant fracture groups and objective, automated calculation of trace spacing, the calculation process for trace spacing is as follows: S24-1: Clustering of dominant fracture groups; By statistically analyzing the fracture orientation angles, the Adap-EOKM adaptive clustering method, which considers the periodicity of angles, is adopted. A clustering objective function based on circumferential distance is constructed with fracture length as the weight. Fractures with similar orientations are grouped together to identify the dominant fracture groups on the surface of deep hard rock pillars. The specific process is as follows: Suppose there are N fractures in total, with orientation angles of respectively... The direction of the cluster center is The corresponding weight is The clustering objective function is constructed according to formula (5). : (5); In the formula, This is a function of the circumferential distance between direction angles. In this embodiment, the weights Acceptable crack length .

[0038] S24-2: Geometric reconstruction of the crack group; For each crack group, the centroid position is calculated based on its pixel set, and a covariance matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue is taken as the main direction vector of the group, thereby obtaining the geometric center and dominant extension direction of the crack group. The specific process is as follows: For the first Group of cracks, let its pixel set be... First, calculate the centroid position of the point set according to formula (6). Then, construct the covariance matrix according to formula (7). ; (6); (7); The eigenvector corresponding to the largest eigenvalue of the matrix is ​​taken as the principal direction vector of the trace group. ; The geometric reconstruction of the trace group is completed using the above method, and the direction vector and geometric center position that can characterize the dominant extension direction of the fracture are obtained.

[0039] S24-3: Orthogonal projection and coordinate sequence generation; using the principal direction vector and centroid of the fracture group, the fracture points are orthogonally projected along the principal direction vector to transform the fracture distribution in two-dimensional space into a one-dimensional projected coordinate sequence, thereby revealing the spatial arrangement of fractures in the principal direction. The specific process is as follows: For any point in the point set... The projected coordinates are obtained according to formula (8). ; (8); In the formula, ; For the first The coordinates of each pixel; The direction vector of the corresponding trace; After obtaining the projected coordinates of all pixels within the same crack, the median (or mean) of the projected coordinates of that crack is taken as the representative projected coordinates of that crack, thus compressing each crack into a one-dimensional point. Let the... The group has a total of A crack yields a representative coordinate sequence. Used for subsequent spacing calculations.

[0040] By using the above method, the interference of adjacent pixels within the same crack on the spacing statistics can be eliminated, thus truly reflecting the spatial interval between different cracks.

[0041] S24-4: Survey line generation and spacing statistics; Survey lines are automatically generated in the normal direction of the projection direction, the projection distance between adjacent fractures is calculated based on the sorted projection coordinates, and the average value is obtained to obtain the trace spacing of the fracture group.

[0042] The specific process is as follows: Let the sorted projected coordinates be... Then the distance between adjacent traces can be expressed as The average spacing is obtained according to formula (9). , which is the trace spacing; (9); In the formula, This represents the number of cracks within the group. Based on the fracture orientation angle, angular periodic adaptive clustering (Adap-EOKM) is used with fracture length as the weight to overcome the boundary problem of orientation periodicity and accurately identify the dominant fracture group. The main orientation of each group is reconstructed by the eigenvector of the covariance matrix, and the two-dimensional distribution is transformed into one-dimensional coordinates by orthogonal projection. Survey lines are automatically generated in the normal direction. The spacing between adjacent fractures is calculated by sorting the projected coordinates, completely eliminating the subjectivity of traditional manual survey line layout and realizing the automation, objectivity and high repeatability of spacing statistics.

[0043] The calculation process for fracture density is as follows: Fracture density reflects the degree of fracture development within a unit area. Let the actual area corresponding to the image be... The image was identified by the CCP If there are multiple cracks, then the crack density can be expressed as: When necessary, the number of cracks per unit area can also be used to characterize crack density.

[0044] Based on deep learning semantic segmentation, the system automatically identifies fracture regions. Single-pixel precision traces are obtained through skeleton extraction, and geometric parameters such as length, orientation angle, and opening are intelligently calculated. Angular periodic clustering and orthogonal projection are used to automatically solve for trace spacing and fracture density, completely replacing manual interpretation and surveying, significantly improving efficiency, accuracy, and repeatability. Through these steps, the geometric features of fractures on the surface of deep hard rock pillars are automatically extracted, providing a data foundation for subsequent construction of rock mass structural grade parameters and structural surface state indicators.

[0045] S3: Identification of the apparent weathering degree of rock mass; The apparent weathering degree of the rock mass is identified based on the digital image of the pillar, and the corresponding weathering level of the rock mass of the pillar is obtained; To achieve automated identification and quantitative rating of the weathering level of structural surfaces, the process for identifying the apparent weathering degree of rock mass in S3 is as follows: S31: Feature extraction; Extracting surface texture and color features of the ore pillar based on digital images of the ore pillar; S32: Weathering Classification; Based on the preset engineering classification standards and combined with the on-site characteristics of deep hard rock pillars, the weathering state of the structural surface is divided into five levels: unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered, in order to quantitatively characterize the weathering degree of the structural surface of deep hard rock pillars. In this embodiment, the degree of weathering of the structural surface is automatically identified using an image classification model (see attached image). Figure 4Specifically, a weathering classification dataset was constructed based on collected images of the structural surfaces of deep hard rock pillars, and manually labeled using field geological survey results. According to engineering geological grading standards, the weathering degree of the structural surfaces was divided into five levels: unweathered (UWT), slightly weathered (SWT), moderately weathered (MWT), strongly weathered (HWT), and completely weathered (CWT). S33: Deep learning discrimination; using a deep learning model to discriminate digital images of ore pillars and obtain the corresponding weathering level; In S33, the deep learning model adopts the RegNetY-400MF network structure to determine the weathering level of the structural surface image of the deep ore pillar rock mass. Preferably, the input structural surface image is set as follows: After obtaining the output features through the classification model, they are normalized using the Softmax function to obtain the probability distribution of each weathering level; let the weathering category be... The category with the highest probability is selected as the weathering degree identification result. In this embodiment, the weathering level result can be mapped using formula (10): (10); If necessary, this mapping relationship can be adjusted based on engineering experience or the grading system of different mining areas. Through the above process, the weathering degree of the structural surface of deep hard rock pillars can be automatically identified and quantitatively expressed.

[0046] S34: Quantification of Weathering Grade; First, the apparent weathering grade of the rock mass (such as unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered) obtained by deep learning is used as input. Then, according to the preset engineering mapping rules, it is converted into a weathering degree rating WDR score that can be used for quantitative calculation. Specifically, when using a 1-5 point system, unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered are mapped to 1 point, 2 points, 3 points, 4 points, and 5 points, respectively; or when using a 0-10 point system, they are mapped to 0 points, 2.5 points, 5 points, 7.5 points, and 10 points, respectively. The 1-5 point system is preferred. In this way, the qualitative description of weathering grade is transformed into a standardized numerical rating index, providing a unified and quantifiable input for the WDR parameters in the subsequent JSVI calculation.

[0047] Based on the extraction of structural surface texture and color features from digital images of ore pillars, the weathering state is subdivided into five levels according to engineering grading standards. A deep learning classification model (such as RegNetY-400MF) is used to automatically determine the weathering level, and the weathering level is converted into a standardized weathering degree rating (WDR) through preset mapping rules. This transforms the qualitative weathering description into a calculable numerical index, avoiding the subjectivity of manual visual judgment and providing a unified and objective input for the subsequent comprehensive evaluation of the structural surface state.

[0048] S4: 3D modeling and roughness extraction of ore pillar surface; constructing a 3D model of the ore pillar surface based on digital images and obtaining its roughness parameters from the 3D model; In this embodiment, in addition to fracture structure parameters, aperture parameters, and weathering grade, surface roughness parameters are further obtained to comprehensively characterize the geometric state of the structural surfaces of deep hard rock pillars (see appendix). Figure 5 ); To achieve non-contact, high-precision 3D quantization of structural surface roughness, the 3D modeling and roughness extraction process for the ore pillar surface in S4 is as follows: S41: 3D reconstruction; using multi-view stereo vision or structured light methods, a 3D point cloud model of the ore pillar surface is constructed based on multiple digital images; S42: Roughness extraction; Structural surface roughness parameters are obtained based on 3D reconstruction data; Based on the 3D point cloud model, surface fitting, slope analysis, or inversion methods based on JRC (joint roughness coefficient) are used to extract the roughness parameters (such as root mean square slope, fractal dimension, or JRC value) of the structural surface; The roughness parameters are obtained through 3D topographic profile analysis of the structural surface and are used to characterize the surface undulation characteristics of the structural surface; Among them, the roughness parameters, the opening parameters (based on the fracture boundary identification results, obtained through distance measurement methods to characterize the degree of fracture cracking), the fracture structure parameters (trace length, spacing, density), and the weathering grade together constitute the basic parameters for evaluating the structural surface condition of deep hard rock pillars; Specifically, based on the 3D point cloud model reconstructed from the digital image of the ore pillar, the structural surface is profiled. Profile curves of the structural surface can be extracted along multiple directions, and the profile height undulation characteristics can be calculated to characterize the roughness of the structural surface. Let the profile height value be... The roughness parameters are obtained according to formula (11). ; (11); Furthermore, second-order parameters of the profile can also be used. Characterizes roughness. Among them, For section length, joint roughness coefficient ; A 3D point cloud model of the ore pillar surface is reconstructed from multiple digital images using multi-view stereo vision or structured light methods. This method eliminates the need for contact measurement and is adaptable to complex underground environments. Roughness parameters are extracted from the point cloud model using methods such as surface fitting, slope analysis, or JRC inversion. Compared to traditional standard profile comparison methods, 3D reconstruction can simultaneously capture macroscopic undulations and microscopic roughness, resulting in higher accuracy and smaller errors. This provides more realistic mechanical boundary conditions for assessing the slippage and instability risk of the ore pillar under high stress. Through the above calculations, the root mean square slope of the structural surface is obtained. value or The value is used as the roughness parameter input for subsequent structural surface condition evaluation.

[0049] S5: Construction of a comprehensive index for the quality of apparent joints in ore pillars; Based on structural parameters, roughness, opening parameters and weathering grade, a comprehensive index for the quality of apparent joints in ore pillars, JSVI (Jointed Surface Visual Index), is constructed to digitally characterize the quality of deep hard rock pillars. After obtaining the structural parameters of the pillar trace, joint spacing, roughness, opening parameters and weathering degree, the structural surface state index corresponding to the pillar is constructed to comprehensively and quantitatively characterize the structural surface state of deep hard rock pillars.

[0050] In this embodiment, each parameter is first mapped according to a preset grading rule or engineering experience and converted into a unified scoring scale. On this basis, the structural surface state index JSVI is then calculated. To achieve unified quantification and comprehensive index fusion of multi-source structural surface parameters, the comprehensive index construction process for the apparent joint quality of the ore pillar in S5 is as follows: S51: Parameter Rating; The extracted fracture features (including trace length, spacing, and density) are quantified into a fracture feature rating (LDR) based on a preset grading standard. The specific grading standard is as follows: Grade 1 (Excellent) is given when the fracture length is less than 0.1m, the spacing is greater than 2m, and the density is less than 1 fracture per square meter; Grade 2 (Good) is given when the fracture length is 0.1 to 0.5m, the spacing is 0.6 to 2m, and the density is 1 to 3 fractures per square meter; Grade 3 (Medium) is given when the fracture length is 0.5 to 1m, the spacing is 0.2 to 0.6m, and the density is 3 to 5 fractures per square meter; Grade 4 (Poor) is given when the fracture length is 1 to 2m, the spacing is 0.06 to 0.2m, and the density is 5 to 10 fractures per square meter; Grade 5 (Very Poor) is given when the fracture length is greater than 2m, the spacing is less than 0.06m, and the density is greater than 10 fractures per square meter. The above grades 1 to 5 correspond to LDR scores of 1, 2, 3, 4, and 5 points, respectively. The crack opening parameter is quantified into an Opening Rating (ADR) based on the degree of opening. The specific grading standards are as follows: an opening less than 0.1 mm is rated as Grade 1 (closed), corresponding to an ADR score of 1 point; an opening of 0.1 to 1 mm is rated as Grade 2 (slightly open), corresponding to an ADR score of 2 points; an opening of 1 to 5 mm is rated as Grade 3 (medium open), corresponding to an ADR score of 3 points; an opening of 5 to 10 mm is rated as Grade 4 (open), corresponding to an ADR score of 4 points; and an opening greater than 10 mm is rated as Grade 5 (wide open), corresponding to an ADR score of 5 points. The obtained surface roughness parameters (such as JRC values ​​or root mean square slope) are quantified into Roughness Rating (RDR) according to the degree of roughness. The specific grading standards are as follows: a JRC value greater than 18 is rated as Level 1 (very rough), corresponding to an RDR score of 1 point; a JRC value of 14 to 18 is rated as Level 2 (rough), corresponding to an RDR score of 2 points; a JRC value of 10 to 14 is rated as Level 3 (medium rough), corresponding to an RDR score of 3 points; a JRC value of 6 to 10 is rated as Level 4 (smooth), corresponding to an RDR score of 4 points; and a JRC value less than 6 is rated as Level 5 (very smooth), corresponding to an RDR score of 5 points. The weathering rating WDR is directly used as input to ensure that the fracture feature rating LDR, aperture rating ADR, roughness rating RDR and weathering rating WDR adopt a uniform quantitative score (1 to 5), providing a standardized input for the subsequent weighted summation of JSVI. S52: JSVI calculation; taking into account the characteristics of crack traces, the opening of the structural surface, the roughness of the structural surface and the degree of weathering of the structural surface, the comprehensive index of structural surface quality JSVI is obtained according to formula (12); (12); In the formula, This is a normalization coefficient used to adjust the range of JSVI values ​​to a reasonable range (e.g., 0 to 100). Fracture characteristics (length, spacing, density), aperture, roughness, and weathering degree are quantified according to preset grading standards. After unifying the dimensions, the weighted summation and normalization coefficient k are used to calculate the comprehensive structural surface quality index (JSVI). This integrates geometric features, mechanical morphology, and weathering degree into a single digital index, solving the problem of fragmented input parameters in traditional GSI. Simultaneously, it adheres to the basic principles of rock mechanics, ensuring that the evaluation results possess both mathematical objectivity and physical and mechanical significance. Through this construction method, multi-source parameters of deep hard rock pillar structural surfaces can be uniformly mapped into a single structural surface state index, achieving a digital and quantitative expression of the structural surface state.

[0051] S6: Calculation of Geological Strength Index (GSI) and Expression of Pillar Rock Mass Quality; Based on the comprehensive structural surface quality index JSVI and the rock mass structural grade parameter SR, a calculation model for the rock mass geological strength index GSI is established, ultimately realizing the digital expression of the quality of the pillar rock mass. To achieve coupled calculation of rock mass structure grade and structural plane state, as well as continuous GSI value taking and quality classification, the calculation process of geological strength index GSI and the expression of pillar rock mass quality in S6 are as follows: S61: Determination of rock mass structure grade parameters: Calculation of volume joint number based on the average spacing of discontinuities in each group. And then according to The range of values ​​is calculated using a piecewise logarithmic function to determine the rock mass structure grade parameters. The specific process is as follows: First, calculate the volume joint number according to formula (13). ; (13); In the formula, For hard rock pillars The average spacing of the group of discontinuous surfaces. The number of discontinuous facets in the pillar; Then, based on the different ranges of volume joint numbers, the rock mass structure grade parameters are obtained using a piecewise function, as shown in formula (14). (14); The above segmented calculation method can more accurately characterize the structural features of deep hard rock pillars under different joint development conditions, providing a unified input basis for subsequent GSI calculations. S62: GSI Calculation; Based on the rock mass structure grade parameter SR and the structural surface state index JSVI, a geological strength index calculation model suitable for deep hard rock pillars is established to obtain the geological strength index of deep hard rock pillars. The specific process is as follows: First, the rock mass structure grade parameter SR and the comprehensive quality index JSVI of the structural surface are unified and normalized to the range of 0-100. The geological strength index is the geological strength index for constructing deep pillar rock masses. The geological strength index is obtained by coupling the rock mass structure grade parameter SR with the structural surface state index JSVI, and then calculated according to formula (15). ; (15); In the formula, , These are the weighting coefficients for the rock mass structure grade parameter SR and the comprehensive quality index JSVI of structural surfaces, respectively. ; The rock mass structure grade parameter SR and the comprehensive quality index of structural surfaces JSVI are normalized to the same dimension, so that , convert formula (15) into formula (16); (16); By coupling the structural characteristics of the rock mass with the state characteristics of the structural plane, the geological strength index of deep hard rock pillars can be automatically calculated and comprehensively evaluated.

[0052] S63: Digital representation of pillar rock mass quality; firstly, the calculated... The numerical value (ranging from 0 to 100) is used as input, and then, according to the preset rock mass quality grade mapping rules, it is... The values ​​are divided into five levels: 80-100 corresponds to excellent, 60-80 to good, 40-60 to average, 20-40 to poor, and 0-20 to very poor. The final output is a digital assessment result of the quality of the ore pillar rock mass, which includes... The numerical values ​​and their corresponding rock mass quality grades transform quantitative geological strength indicators into intuitive and readable rock mass quality evaluation conclusions, providing a scientific basis for the stability analysis and support design of deep mine pillars, such as... Figure 6 As shown.

[0053] The volumetric joint number is calculated based on the average spacing of discontinuities in each group, and a piecewise logarithmic function is used to map it to continuous rock mass structural grade parameters, which is more refined and suitable for automated calculation than the discrete grading of traditional GSI charts; the SR is coupled with the structural surface condition index JSVI with equal weights to obtain... The model comprehensively reflects the structural characteristics and structural surface state characteristics of the rock mass; it is simple and has clear physical meaning. Finally, the continuous... The numerical values ​​are divided into five quality levels: excellent, good, medium, poor, and very poor according to the threshold. The output values ​​and corresponding levels transform quantitative indicators into intuitive evaluation conclusions, providing a scientific basis for the stability analysis and support design of deep mine pillars.

[0054] This invention addresses the problems of existing methods for obtaining geological strength indicators (GSI), such as reliance on manual interpretation, high subjectivity, low evaluation efficiency, and difficulty in achieving large-scale, continuous, and rapid evaluation. It provides a digital and automated GSI acquisition solution for deep hard rock pillars, resolving the core pain points of traditional methods and providing reliable technical support for deep rock engineering. First, digital images of the pillar surface are acquired and preprocessed to provide high-quality, standardized input images for subsequent identification and parameter calculation. Second, automatic fracture identification is achieved based on a deep learning model, capturing micro-fractures and hidden joints that are difficult to detect with the naked eye, improving the completeness of structural surface statistics. This allows for accurate and efficient calculation of structural parameters such as fracture traces, direction, density, and spacing, precisely determining the rock mass structural grade parameters. Third, through image color space conversion and texture feature extraction, the degree of weathering can be converted into specific quantitative indicators or probability distributions. Finally, 3D reconstruction extracts the true 3D undulation morphology, and the calculated JRC (Geological Strength Ratio) is more accurate and has less error than the traditional comparative standard profile method. The digital model incorporates both macroscopic undulations and microscopic roughness, providing crucial mechanical boundary conditions for evaluating the risk of slippage and instability of ore pillars under deep, high-stress conditions. Furthermore, it cleverly integrates geometric features, weathering degree, and mechanical morphology into a single comprehensive index, resolving the fragmentation problem of traditional GSI input parameters. Simultaneously, the JSVI's construction logic adheres to fundamental principles of rock mechanics, ensuring that the final evaluation results possess both mathematical objectivity and physical and mechanical significance. Finally, through a continuous calculation model of JSVI and SR, continuous GSI values ​​are achieved, better reflecting the gradual variation of rock mass quality in nature and enabling more accurate acquisition of geological strength indicators.

[0055] Compared with traditional methods that rely on experience for interpretation, this invention has the following advantages: 1. Based on digital image and intelligent recognition technology, the parameters of rock mass fracture structure, surface roughness and geological strength index are automatically acquired, which reduces the subjective error caused by traditional manual interpretation and improves the objectivity and consistency of rock mass quality evaluation. 2. By constructing the Structural Surface State Index (JSVI), a multi-parameter comprehensive characterization of the structural surface state of deep hard rock pillars was achieved, improving the comprehensiveness and objectivity of the evaluation. 3. By establishing a geological strength index calculation model applicable to deep hard rock pillars, the geological strength index of deep hard rock pillars can be obtained quickly, automatically, and intelligently, which greatly improves the evaluation efficiency. 4. By organically combining crack identification, spacing calculation, weathering identification, roughness and opening analysis, and GSI model construction, a complete digital evaluation process is formed, which has good engineering application prospects and promotion value.

[0056] 5. By accurately mapping pixels to actual physical dimensions, the crack size parameters under different images and different acquisition distances are made comparable in a uniform manner, providing a metrological basis for multi-source data fusion.

[0057] 6. By adopting angular periodic clustering and orthogonal projection to automatically generate survey lines, the subjectivity of traditional manual survey line layout is completely eliminated, realizing the automation and repeatability of spacing statistics and significantly improving the reliability of spacing parameters.

[0058] This method is simple to implement and has low implementation costs. Based on digital image and intelligent recognition technology, it realizes the full-process automated, objective and quantitative acquisition of rock mass fracture structure parameters, structural surface state indicators and geological strength indicators of deep hard rock pillars. It overcomes the shortcomings of traditional methods, such as reliance on manual interpretation, strong subjectivity and low efficiency, and provides an efficient and reliable technical means for the quality evaluation of deep rock mass engineering.

Claims

1. A method for intelligent acquisition of digital geological strength indicators for deep ore pillars, characterized in that, Includes the following steps: S1: Digital image acquisition and preprocessing; Acquire digital images of the surface of deep ore pillars or rock outcrops, and preprocess the digital images; S2: Joint and fracture identification and structural parameter extraction; Joint and fracture identification is performed based on preprocessed digital images, fracture geometric information is extracted, and corresponding rock mass structural parameters are obtained. S3: Identification of the apparent weathering degree of rock mass; The apparent weathering degree of the rock mass is identified based on the digital image of the pillar, and the corresponding weathering level of the rock mass of the pillar is obtained; S4: 3D modeling and roughness extraction of ore pillar surface; A three-dimensional model of the ore pillar surface is constructed based on digital images, and its roughness parameters are obtained from the three-dimensional model. S5: Construction of comprehensive index for apparent joint quality of ore pillars; Based on structural parameters, roughness, opening parameters, and weathering grade, a comprehensive index (JSVI) for the apparent joint quality of a pillar is constructed. S6: Calculation of Geological Strength Index (GSI) and Expression of Pillar Rock Mass Quality; Based on the comprehensive structural surface quality index JSVI and the rock mass structural grade parameter SR, a calculation model for the rock mass geological strength index GSI is established, ultimately realizing the digital expression of the quality of the pillar rock mass.

2. The method for intelligent acquisition of digital geological strength indicators for deep ore pillars according to claim 1, characterized in that, In S1, the digital image acquisition and preprocessing process is as follows: S11: Acquire image data; Collect digital images of the surface of deep hard rock pillars; S12: Preprocessing; preprocessing the digital image; the preprocessing includes noise suppression, brightness and contrast normalization, geometric distortion correction, and scale calibration and unification.

3. The method for intelligent acquisition of digital geological strength indicators for deep ore pillars according to claim 1, characterized in that, In S2, the process of joint and fracture identification and structural parameter extraction is as follows: S21: Semantic segmentation of fractures; The preprocessed digital image of the pillar is input into the joint trace recognition model, and pixel-level semantic segmentation is performed on the structural disease area. The fractures on the surface of the pillar are automatically identified, and the fracture area is used as the joint trace recognition result. S22: Crack skeleton extraction and coordinate acquisition; refine the joint trace recognition results, extract the crack trace skeleton, and obtain the crack skeleton structure; based on this, extract the continuous pixel coordinate information of each crack skeleton structure; S23: Geometric parameter calculation; Based on the crack trace skeleton and its pixel coordinate information, calculate the trace length, orientation angle distribution and opening information; S24: Calculation of structural parameters; Based on the spatial distribution relationship of the fracture trace skeleton, calculate the rock mass structural parameters, including trace spacing and fracture density.

4. The intelligent acquisition method for digital geological strength indicators of deep ore pillars according to claim 3, characterized in that, In S23, the geometric parameters are calculated as follows: S23-1: Crack trace length calculation; The crack trace length is obtained by accumulating the Euclidean distance between adjacent skeleton points to reduce the error caused by pixel discretization; S23-2: Calculation of fracture orientation angle; The least squares method is used to fit a straight line of the fracture skeleton point set, and the slope of the straight line is used to calculate the dominant orientation angle of the fracture, which is used for subsequent grouping of dominant fractures; S23-3: Calculation of crack opening; Based on the crack boundary identification results, the local width at the skeleton point is calculated using the distance field method, the average opening and the maximum opening are statistically analyzed, and mapped to the opening rating.

5. The intelligent acquisition method for digital geological strength indicators of deep ore pillars according to claim 3, characterized in that, In S24, the calculation process for the trace spacing is as follows: S24-1: Clustering of dominant fracture groups; By statistically analyzing the fracture orientation angle, the Adap-EOKM adaptive clustering method considering the periodicity of the angle is adopted, with the fracture length as the weight, and a clustering objective function based on the circumferential distance is constructed according to formula (5). Fractures with similar orientations are grouped together to identify the dominant fracture groups on the surface of deep hard rock pillars. （5）； In the formula, This is a function of the circumferential distance between direction angles. ; In this embodiment, As the weight, take the crack length. ; S24-2: Geometric reconstruction of the crack group; For each crack group, the centroid position is calculated based on its pixel set, and a covariance matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue is taken as the main direction vector of the group, thereby obtaining the geometric center and dominant extension direction of the crack group. S24-3: Orthogonal projection and coordinate sequence generation; using the principal direction vector and centroid of the fracture group, the fracture points are orthogonally projected along the principal direction vector to transform the fracture distribution in two-dimensional space into a one-dimensional projected coordinate sequence. S24-4: Survey line generation and spacing statistics; Survey lines are automatically generated in the normal direction of the projection direction, the projection distance between adjacent fractures is calculated based on the sorted projection coordinates, and the average value is obtained to obtain the trace spacing of the fracture group.

6. The intelligent acquisition method for digital geological strength indicators of deep ore pillars according to claim 1, characterized in that, In S3, the process for identifying the apparent weathering degree of the rock mass is as follows: S31: Feature extraction; Extracting surface texture and color features of the ore pillar based on digital images of the ore pillar; S32: Weathering Classification; Based on the preset engineering classification standards and combined with the on-site characteristics of deep hard rock pillars, the weathering state of structural surfaces is divided into five levels: unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered. S33: Deep learning discrimination; using a deep learning model to discriminate digital images of ore pillars and obtain the corresponding weathering level; S34: Quantification of weathering grade; converting the apparent weathering grade of the rock mass into a standardized weathering degree rating (WDR) according to a preset mapping rule.

7. The method for intelligent acquisition of digital geological strength indicators for deep ore pillars according to claim 1, characterized in that, In S4, the process of 3D modeling and roughness extraction of the ore pillar surface is as follows: S41: 3D reconstruction; using multi-view stereo vision or structured light methods, a 3D point cloud model of the ore pillar surface is constructed based on multiple digital images; S42: Roughness extraction; Based on a 3D point cloud model, roughness parameters of the structural surface are extracted using surface fitting, slope analysis, or inversion methods based on joint roughness coefficients.

8. The intelligent acquisition method for digital geological strength indicators of deep ore pillars according to claim 1, characterized in that, In S5, the process of constructing the comprehensive index of apparent joint quality of the ore pillar is as follows: S51: Parameter rating; quantify the extracted crack features into crack feature rating (LDR) according to the preset grading standard; quantify the crack opening parameter into opening rating (ADR) according to the opening degree; quantify the obtained structural surface roughness parameter into roughness rating (RDR) according to the roughness degree. The weathering rating (WDR) will be used directly as input. S52: JSVI calculation; taking into account the characteristics of crack traces, the opening of the structural surface, the roughness of the structural surface and the degree of weathering of the structural surface, the comprehensive index of structural surface quality JSVI is obtained according to formula (12); （12）； In the formula, This is the normalization coefficient.

9. The intelligent acquisition method for digital geological strength indicators of deep ore pillars according to claim 1, characterized in that, In S6, the calculation process of the Geological Strength Index (GSI) and the expression of the quality of the ore pillar rock mass are as follows: S61: Determination of rock mass structure grade parameters: Calculation of volume joint number based on the average spacing of discontinuities in each group. And then according to The range of values ​​is calculated using a piecewise logarithmic function to determine the rock mass structure grade parameters. S, as shown in formula (14); （14）； 62: GSI Calculation; Based on the rock mass structure grade parameter SR and the structural surface state index JSVI, a geological strength index calculation model suitable for deep hard rock pillars is established to obtain the geological strength index of deep hard rock pillars. As shown in formula (16); （16）； S63: Digital representation of the quality of the ore pillar rock mass; the calculated... The values ​​are divided into five levels according to a preset threshold: excellent, good, average, poor, and very poor. The output values ​​and corresponding levels are then displayed.

Images