Particle geometric reconstruction and morphological parameter automatic evaluation method based on three-dimensional image

Abstract

The invention discloses a particle geometric reconstruction and morphological parameter automatic evaluation method based on a three-dimensional image, and the method comprises the steps: firstly obtaining a three-dimensional gray image of a particle sample, carrying out the Gaussian noise reduction, binaryzation and watershed segmentation preprocessing, and achieving the precise segmentation of independent particles based on an established particle segmentation defect repair algorithm; and then, according to the segmented and marked three-dimensional particle image, an initial contour surface is reconstructed by using a marching cube algorithm to generate a particle surface grid, noise is suppressed by using a Taubin smooth optimization algorithm, a topological structure is automatically calibrated, and a digital model conforming to real particle surface features is constructed. And finally, importing each particle model into the created particle morphological parameter automatic evaluation software to realize quantitative analysis of particle morphological characteristics.

Description

Technical Field

[0001] The present invention relates to a particle geometry reconstruction and morphological parameter automatic evaluation method based on three-dimensional images, and belongs to computer vision and image processing. Background Art

[0002] In the study of granular material mechanics, the three-dimensional morphology of particles serves as a key parameter at the mesoscopic scale, directly influencing macroscopic mechanical responses, such as shear strength, stiffness anisotropy, and multi-physics field coupling mechanisms. Therefore, accurately and quantitatively characterizing particle morphology is a core component of particle mechanics research.

[0003] However, existing methods for real particle reconstruction and morphology characterization face significant technical defects. First, in terms of real particle geometric reconstruction, two-dimensional images are difficult to restore the three-dimensional topological structure of particles; affected by factors such as particle volume effect, the image segmentation process is prone to over-segmentation and under-segmentation defects, resulting in the inability to accurately extract independent particles from the sample, thereby reducing the reliability of particle morphology characterization.

[0004] Second, in terms of the quantitative characterization of particle morphological features, relying on manual morphological quantification methods (such as the caliper method to measure the major / minor axis) and empirical indicators (such as Wadell sphericity) makes it difficult to achieve automated and accurate characterization of the surface curvature distribution and contour non-convexity of irregular particles, and lacks systematic and comprehensive quantification technology.

[0005] Therefore, there is an urgent need to develop innovative methods for 3D image-based particle geometry reconstruction and automated morphological parameter assessment to overcome the fundamental shortcomings of existing technologies. This present invention is dedicated to this goal, aiming to provide an integrated, automated, and intelligent solution for real particle geometry reconstruction and shape parameter quantification. Summary of the Invention

[0006] The present invention aims to solve the above problems in the prior art and provides a method for automatically evaluating particle geometry reconstruction and morphological parameters based on three-dimensional images.

[0007] The technical solutions adopted in the present invention are:

[0008] A method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images comprises the following steps:

[0009] (1) Perform 360° scanning imaging on real three-dimensional particle samples to obtain three-dimensional grayscale image data containing particle morphology information and preprocess the image;

[0010] "Morphological information" refers to various characteristics of a particle's shape and structure. Specifically, this includes its geometry (e.g., cuboid, spherical, etc.), size (specific values ​​such as length, width, and height), surface texture (details such as smoothness and roughness), and curvature distribution (a quantitative description of the degree of surface curvature). This information, taken together, comprehensively describes a particle's appearance and structural characteristics in three-dimensional space, providing foundational data for subsequent particle geometry reconstruction and morphological parameter assessment.

[0011] (2) Based on the preprocessed 3D image, the marching cubes algorithm is used to extract the initial isosurface to generate the surface mesh. The Taubin smoothing optimization algorithm is used for noise suppression and the topology structure is automatically calibrated to generate a digital model consistent with the surface features of the real particles. The geometric reconstruction of the particles is achieved and a single 3D particle model file is output in .stl file format.

[0012] (3) Import the .stl file exported in step (2) into the software for automated quantitative evaluation of particle morphology parameters, and perform the following operations in the software:

[0013] a. Read the vertex and facet data of 3D particles from the imported .stl format file;

[0014] b. Determine the main direction of the particles using the PCA algorithm;

[0015] c. Calculating three-dimensional particle shape parameters from the particle image;

[0016] d. extracting the orthogonal projection surface of the three-dimensional image and forming a two-dimensional image for analyzing the three-dimensional particles, and calculating the two-dimensional particle shape parameters of the orthogonal projection surface of the three-dimensional image;

[0017] e. integrating the calculated three-dimensional image shape parameters of the particles and the two-dimensional image shape parameters of the particles;

[0018] f. Display the integrated parameter data through the visualization module;

[0019] g. Store the integrated parameter data, individual three-dimensional and two-dimensional particle shape parameters, and visualization results from step e in a separate folder in a specified format;

[0020] (4) Automated quantitative evaluation of particle morphology is achieved through software processing as described in step (3).

[0021] Furthermore, in step (1), the pre-processing includes:

[0022] Gaussian filter noise reduction processing;

[0023] Image binarization based on Otsu threshold algorithm;

[0024] The watershed algorithm is used to segment the binary image to mark and distinguish the independent particles in the three-dimensional particle sample.

[0025] Furthermore, when using the watershed algorithm to segment the binarized image, different methods are used to repair the "over-segmentation" and "under-segmentation" defects that appear after particle image segmentation;

[0026] The “over-segmentation” mentioned here refers to the phenomenon that a real particle is mistakenly divided into multiple independent particles during the image segmentation process;

[0027] "Under-segmentation" refers to the phenomenon that multiple real particles are mistakenly merged into a whole during the image segmentation process.

[0028] Furthermore, the repair method for the "over-segmentation" defect is as follows:

[0029] Calculate the over-segmentation coefficient of each particle and its adjacent label set, generate the over-segmentation coefficient histogram and spatial distribution visualization slice;

[0030] Set a threshold to filter out target particle labels with excessive over-segmentation coefficients;

[0031] Traverse the target label set, for each target particle, extract the 3D region between it and the adjacent over-segmented particles, and replace the adjacent label ID with the target label ID to achieve particle merging;

[0032] By dynamically merging over-segmented particles, visual monitoring, and updating label mapping, the particle fragmentation problem caused by the segmentation algorithm is effectively solved.

[0033] Furthermore, the repair method for the "under-segmentation" defect is as follows:

[0034] Calculate the under-segmentation coefficient matrix and filter out target particles whose under-segmentation coefficient exceeds a preset threshold;

[0035] The local 3D region of the target particle with a 5-pixel boundary is intercepted, and the particle core area is extracted as the watershed seed point through a 0.75 grayscale threshold. Secondary watershed segmentation based on seed markers is performed.

[0036] After verifying that the under-segmentation coefficients of all new particles are better than those of the original particles, the original particle labels are removed and continuous IDs are assigned to the new particles;

[0037] Accurate separation of adhered particles is achieved through dynamic re-segmentation, quantitative verification, and label update operations.

[0038] Furthermore, in step (2), the automatic calibration of the topological structure is performed by statistical analysis of the local voxel gradient distribution within the range of ±15% of the baseline threshold.

[0039] Furthermore, in step c, the shape parameters of each three-dimensional image particle are calculated, including: elongation index, flatness index, comprehensive three-dimensional aspect ratio, three-dimensional sphericity, three-dimensional concavity and convexity, and three-dimensional roundness.

[0040] Furthermore, the local coordinates of each particle are projected onto the three principal component directions using the PCA algorithm to obtain new coordinates. The coordinate range in each principal component direction is calculated as the dimension in that direction. The three dimensions are arranged in ascending order to obtain the minimum principal axis length, the intermediate principal axis length, and the maximum principal axis length. The shape parameters of each particle are calculated using the following formula:

[0041] Elongation index EI = intermediate spindle length / maximum spindle length;

[0042] Flatness index FI = minimum main axis length / intermediate main axis length;

[0043] Comprehensive 3D aspect ratio ;

[0044] The surface area SA of the closed surface formed by the triangular mesh on each particle is calculated by the triangular mesh volume and area calculation algorithm. The particle volume V is the product of the triangular mesh area, the z coordinate at the center of gravity, and the z component of the unit normal vector, and the three-dimensional particle shape parameter sphericity S is obtained. 3D for:

[0045] ;

[0046] Calculate the convex hull volume V using the convex hull algorithm con , combined with particle volume V, concavity C 3D for:

[0047] .

[0048] Furthermore, the curvature algorithm is used to calculate the curvature of four triangular mesh patches: minimum curvature k1, maximum curvature k2, average curvature and Gaussian curvature ,

[0049] Combined with the reference sphere curvature corresponding to each particle , and the surface area A of each triangular mesh in the closed surface on each particle i , obtain the roundness R based on the four curvatures 3D for:

[0050] , , , ,

[0051] Among them, R 3D1 Based on the minimum principal curvature, the smoothness of the flat area of ​​each particle is reflected. 3D2 The maximum principal curvature is used to characterize the roundness of each particle's angular region, R 3Dm Based on the average curvature, R 3Dg Quantifying local geometric complexity based on Gaussian curvature.

[0052] Furthermore, in step d, the orthogonal projection planes of the three-dimensional image include: XY plane, XZ plane and YZ plane; the particle shape parameters of the two-dimensional image include two-dimensional slenderness, two-dimensional sphericity, two-dimensional roundness and two-dimensional concavity and convexity.

[0053] Furthermore, it is characterized in that:

[0054] The particle ellipse of the two-dimensional projection is fitted based on the least squares method, and the length of the major axis of the ellipse is taken as D max , using Feret diameter method to calculate the minimum width D min , obtain the two-dimensional particle shape parameter slenderness ratio Ar of the three-dimensional particle projection surface 2D for:

[0055] ;

[0056] Calculate the minimum distance from the center of mass to the boundary point of the particle's two-dimensional projection as the maximum inscribed circle radius r max-in , the maximum distance from the center of mass to the boundary point is the minimum circumscribed circle radius r min-cir , obtain the two-dimensional particle shape parameter sphericity S of the three-dimensional particle projection surface 2D for:

[0057] ;

[0058] The average value of the radius of the curvature circle corresponding to the boundary points of the two-dimensional projection of the particle is , combined with the maximum inscribed circle radius r max-in , obtain the two-dimensional particle shape parameter roundness R of the three-dimensional particle projection surface 2D for:

[0059] ;

[0060] Calculate the total number of pixels belonging to the particle in the two-dimensional projection binary image as the actual area A 2D The total number of pixels inside the smallest convex polygon containing the particle is the convex hull area A 2Dcon , obtain the two-dimensional particle shape parameter concavity C of the three-dimensional particle projection surface2D for:

[0061] .

[0062] The present invention has the following beneficial effects:

[0063] Through the three-dimensional particle sample image, the independent particles in the sample can be accurately segmented and extracted, the geometric shape of the particles can be truly reconstructed, and the full-dimensional morphological parameter evaluation can be automatically performed to provide more complete quantitative information on the particle shape. The innovative integration of the three-dimensional and two-dimensional shape parameter systems facilitates the user to comprehensively evaluate the particle morphological characteristics. The system can automatically identify and process image data, perform intelligent and automated analysis, and simultaneously complete the calculation of the three-dimensional and two-dimensional shape parameters of the particles.

[0064] Figure 1 It is a flow chart of the present invention.

[0065] Figure 2 This is a slice image after grayscale, binarization, segmentation and repair marking after image preprocessing and optimization processing of the present invention.

[0066] Figure 3 Schematic diagram of the present invention before and after the "over-segmentation" and "under-segmentation" defect repair.

[0067] Figure 4 This is the three-dimensional particle sample image reconstructed by the present invention.

[0068] Figure 5 This is a single real three-dimensional particle model image extracted by the present invention.

[0069] Figure 6 Schematic diagram for the calculation of two-dimensional particle shape parameter indicators.

[0070] Figure 7 Schematic diagram of the interface of the software for automated quantitative evaluation of particle morphology parameters created for the present invention. DETAILED DESCRIPTION

[0071] The present invention will be further described below with reference to the accompanying drawings.

[0072] like Figures 1 to 7 The present invention provides a method for automatically evaluating particle geometry reconstruction and morphological parameters based on three-dimensional images, comprising the following steps:

[0073] (1) Three-dimensional image acquisition:

[0074] An X-ray micro-tomography (μCT) device was used in conjunction with a continuous rotating stage to perform 360° scanning imaging of real three-dimensional particle samples to obtain three-dimensional grayscale image data containing particle morphology information.

[0075] "Morphological information" refers to various characteristics of a particle's shape and structure. Specifically, this includes its geometry (e.g., cuboid, spherical, etc.), size (specific values ​​such as length, width, and height), surface texture (details such as smoothness and roughness), and curvature distribution (a quantitative description of the degree of surface curvature). This information, taken together, comprehensively describes a particle's appearance and structural characteristics in three-dimensional space, providing foundational data for subsequent particle geometry reconstruction and morphological parameter assessment.

[0076] (2) Image preprocessing:

[0077] Perform Gaussian filtering to reduce noise on the three-dimensional grayscale image; apply the Otsu threshold algorithm to binarize the image; use the watershed algorithm to segment the image and mark all independent particles contained in the sample;

[0078] (3) Image optimization processing:

[0079] When using the watershed algorithm to segment the binary image, different methods are used to repair the "over-segmentation" and "under-segmentation" defects that appear after particle image segmentation;

[0080] The “over-segmentation” mentioned here refers to the phenomenon that a real particle is mistakenly divided into multiple independent particles during the image segmentation process;

[0081] "Under-segmentation" refers to the phenomenon that multiple real particles are mistakenly merged into a whole during the image segmentation process.

[0082] The main algorithm process for repairing the "over-segmentation" defect includes: calculating the over-segmentation coefficient and adjacent label set of each particle, generating an over-segmentation coefficient histogram and spatial distribution visualization slices; setting a threshold to filter out target particle labels with excessive over-segmentation coefficients; traversing the target label set, extracting the three-dimensional area between each target particle and its adjacent over-segmented particles, and uniformly replacing the adjacent label IDs with the target label ID to achieve particle merging; through dynamic merging of over-segmented particles, visual monitoring and label mapping updates, the particle fragmentation problem caused by the segmentation algorithm is effectively solved.

[0083] The main algorithmic process for repairing the "under-segmentation" defect includes: calculating the under-segmentation coefficient matrix and screening target particles whose under-segmentation coefficient exceeds a preset threshold; intercepting the local three-dimensional region extending the 5-pixel boundary of the target particle, extracting the particle core area using a 0.75 grayscale threshold as the watershed seed point, and performing secondary watershed segmentation based on seed labeling; after verifying that the under-segmentation coefficients of all new particles are better than those of the original particles, removing the original particle labels and assigning continuous IDs to the new particles; and achieving accurate separation of adhered particles through dynamic re-segmentation, quantitative verification, and label update operations.

[0084] (4) Particle geometry reconstruction:

[0085] Based on the optimized 3D image, the marching cubes algorithm is used to extract the initial isosurface and generate the surface mesh. The Taubin smoothing optimization algorithm is used to suppress noise and automatically calibrate the topology to generate a digital model consistent with the actual particle surface features, achieving true geometric reconstruction of the particles and exporting a single 3D particle model file.

[0086] (5) Particle morphology characterization:

[0087] Start the created automated calculation program and create an interface. The user clicks "Select STL File" to select a single 3D particle model file. The user clicks "Start Analysis" and the program performs background analysis, mainly calculating various shape characteristic parameters of the 3D reconstructed particles and displaying them on the interface. The specific calculation process is as follows:

[0088] a. Read the vertex and facet data of 3D particles from the imported .stl format file;

[0089] b. Determine the main direction of the particles using the PCA algorithm;

[0090] c. Calculating three-dimensional particle shape parameters from the particle image;

[0091] d. extracting the orthogonal projection surface of the three-dimensional image and forming a two-dimensional image for analyzing the three-dimensional particles, and calculating the two-dimensional particle shape parameters of the orthogonal projection surface of the three-dimensional image;

[0092] e. integrating the calculated three-dimensional image shape parameters of the particles and the two-dimensional image shape parameters of the particles;

[0093] f. Display the integrated parameter data through the visualization module;

[0094] g. Store the integrated parameter data, individual three-dimensional and two-dimensional particle shape parameters, and visualization results from step e in a separate folder in a specified format;

[0095] Through the above-mentioned automated analysis and calculation steps, a comprehensive quantitative assessment of particle morphology is achieved. The user can select "Exit Program" at any time through the interface to end the calculation and result display.

[0096] The main contents of the background automatic analysis of particle morphology characteristics include:

[0097] The PCA algorithm is used to project the local coordinates of each particle onto the three principal component directions to obtain new coordinates. The coordinate range in each principal component direction is calculated as the dimension in that direction. The three dimensions are arranged in ascending order to obtain the minimum principal axis length, the intermediate principal axis length, and the maximum principal axis length. The three-dimensional shape parameters of each particle are calculated using the following formula:

[0098] Elongation Index EI = Intermediate spindle length / Maximum spindle length (1)

[0099] Flatness index FI = minimum spindle length / intermediate spindle length (2)

[0100] Comprehensive 3D aspect ratio (3)

[0101] For a three-dimensional particle represented by a closed triangular mesh, the sum of the areas of all triangular mesh patches is calculated as the three-dimensional particle surface area SA, and the particle volume V is the weighted sum of the area of ​​each triangular mesh patch and the projection of its center of gravity, thereby obtaining the three-dimensional particle shape parameter sphericity S 3D for:

[0102] (4)

[0103] Calculate the convex hull volume V using the convex hull algorithm con , combined with the particle volume V, the three-dimensional particle shape parameter concavity C 3D for:

[0104] (5)

[0105] Use the curvature algorithm to calculate the curvature of four triangular mesh surfaces: minimum curvature k1, maximum curvature k2, and average curvature and Gaussian curvature .

[0106] Combined with the reference sphere curvature corresponding to each particle , and the surface area A of each triangular mesh in the closed surface on each particle i , obtain the three-dimensional particle shape parameter roundness R based on four curvatures 3D for:

[0107] , , , (6)

[0108] Among them, R 3D1 Based on the minimum principal curvature, the smoothness of the flat area of ​​each particle is reflected. 3D2 The maximum principal curvature is used to characterize the roundness of each particle's angular region, R 3DmBased on the average curvature, R 3Dg Quantifying local geometric complexity based on Gaussian curvature.

[0109] The two-dimensional particle shape parameters of the orthogonal projection planes (XY plane, XZ plane, and YZ plane) of the three-dimensional image are calculated by the following formula: The particle shape parameters of the two-dimensional image include two-dimensional slenderness, two-dimensional sphericity, two-dimensional roundness, and two-dimensional concavity and convexity.

[0110] The particle ellipse of the two-dimensional projection is fitted based on the least squares method, and the length of the major axis of the ellipse is taken as D max , using Feret diameter method to calculate the minimum width D min , obtain the two-dimensional particle shape parameter slenderness ratio Ar of the three-dimensional particle projection surface 2D for:

[0111] (7)

[0112] Calculate the minimum distance from the center of mass to the boundary point of the particle's two-dimensional projection as the maximum inscribed circle radius r max-in , the maximum distance from the center of mass to the boundary point is the minimum circumscribed circle radius r min-cir , obtain the two-dimensional particle shape parameter sphericity S of the three-dimensional particle projection surface 2D for:

[0113] (8)

[0114] The average value of the radius of the curvature circle corresponding to the boundary points of the two-dimensional projection of the particle is , combined with the maximum inscribed circle radius r max-in , obtain the two-dimensional particle shape parameter roundness R of the three-dimensional particle projection surface 2D for:

[0115] (9)

[0116] Calculate the total number of pixels belonging to the particle in the two-dimensional projection binary image as the actual area A 2D The total number of pixels inside the smallest convex polygon containing the particle is the convex hull area A 2Dcon , obtain the two-dimensional particle shape parameter concavity C of the three-dimensional particle projection surface 2D for:

[0117] (10)

[0118] The above description is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principles of the present invention. These improvements should also be regarded as the scope of protection of the present invention.

Claims

1. A method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images, characterized by: The steps include: (1) Perform 360° scanning imaging on real three-dimensional particle samples to obtain three-dimensional grayscale image data containing particle morphology information and preprocess the image; (2) Based on the preprocessed 3D image, the marching cubes algorithm is used to extract the initial isosurface to generate the surface mesh. The Taubin smoothing optimization algorithm is used for noise suppression and the topology structure is automatically calibrated to generate a digital model consistent with the surface features of the real particles. The geometric reconstruction of the particles is achieved and a single 3D particle model file is output in .stl file format. (3) Import the .stl file exported in step (2) into the software for automated quantitative evaluation of particle morphology parameters, and perform the following operations in the software: a. Read the vertex and facet data of 3D particles from the imported .stl format file; b. Determine the main direction of the particles using the PCA algorithm; c. Calculating three-dimensional particle shape parameters from the particle image; d. extracting the orthogonal projection surface of the three-dimensional image and forming a two-dimensional image for analyzing the three-dimensional particles, and calculating the two-dimensional particle shape parameters of the orthogonal projection surface of the three-dimensional image; e. integrating the calculated three-dimensional image shape parameters of the particles and the two-dimensional image shape parameters of the particles; f. Display the integrated parameter data through the visualization module; g. Store the integrated parameter data, individual three-dimensional and two-dimensional particle shape parameters, and visualization results from step e in a separate folder in a specified format; (4) Automated quantitative evaluation of particle morphology is achieved through software processing as described in step (3).

2. The method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images according to claim 1, wherein: In step (1), the pretreatment includes: Gaussian filter noise reduction processing; Image binarization based on Otsu threshold algorithm; The watershed algorithm is used to segment the binary image to mark and distinguish the independent particles in the three-dimensional particle sample.

3. The method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images according to claim 2, wherein: When using the watershed algorithm to segment the binary image, different methods are used to repair the "over-segmentation" and "under-segmentation" defects that appear after particle image segmentation; The "over-segmentation" mentioned here refers to the phenomenon that a real particle is mistakenly divided into multiple independent particles during the image segmentation process; "Under-segmentation" refers to the phenomenon that multiple real particles are mistakenly merged into a whole during the image segmentation process.

4. The method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images according to claim 3, wherein: The specific repair methods for the "over-segmentation" defect are: Calculate the over-segmentation coefficient of each particle and its adjacent label set, generate the over-segmentation coefficient histogram and spatial distribution visualization slice; Set a threshold to filter out target particle labels with excessive over-segmentation coefficients; Traverse the target label set, for each target particle, extract the three-dimensional region between it and the adjacent over-segmented particles, and replace the adjacent label ID with the target label ID to achieve particle merging.

5. The method for automatically evaluating particle geometry reconstruction and morphological parameters based on three-dimensional images according to claim 3, wherein: The specific repair method for the "under-segmentation" defect is: Calculate the under-segmentation coefficient matrix and filter out target particles whose under-segmentation coefficient exceeds a preset threshold; The local 3D region of the target particle with a 5-pixel boundary is intercepted, and the particle core area is extracted as the watershed seed point through a 0.75 grayscale threshold. Secondary watershed segmentation based on seed markers is performed. After verifying that the under-segmentation coefficients of all new particles are better than those of the original particles, the original particle labels are removed and continuous IDs are assigned to the new particles.

6. The method for automatic particle geometry reconstruction and morphological parameter evaluation based on three-dimensional images according to claim 1, characterized in that: In step c, the shape parameters of each three-dimensional image particle are calculated, including: elongation index, flatness index, comprehensive three-dimensional aspect ratio, three-dimensional sphericity, three-dimensional concavity and convexity, and three-dimensional roundness.

7. The method for particle geometry reconstruction and morphological parameter automated evaluation based on three-dimensional images according to claim 6, characterized in that: The PCA algorithm is used to project the local coordinates of each particle onto the three principal component directions to obtain new coordinates. The coordinate range in each principal component direction is calculated as the dimension in that direction. The three dimensions are arranged in ascending order to obtain the minimum principal axis length, the intermediate principal axis length, and the maximum principal axis length. The shape parameters of each particle are calculated using the following formula: Elongation index EI = intermediate spindle length / maximum spindle length; Flatness index FI = minimum main axis length / intermediate main axis length; Comprehensive 3D aspect ratio ; The surface area SA of the closed surface formed by the triangular mesh on each particle is calculated by the triangular mesh volume and area calculation algorithm. The particle volume V is the product of the triangular mesh area, the z coordinate at the center of gravity, and the z component of the unit normal vector, and the three-dimensional particle shape parameter sphericity S is obtained. 3D for: ； Calculate the convex hull volume V using the convex hull algorithm con , combined with particle volume V, concavity C 3D for: 。 8. The method for automatically evaluating particle geometry and morphological parameters based on three-dimensional images according to claim 7, wherein: Use the curvature algorithm to calculate the curvature of four triangular mesh surfaces: minimum curvature k1, maximum curvature k2, and average curvature and Gaussian curvature , Combined with the reference sphere curvature corresponding to each particle , and the surface area A of each triangular mesh in the closed surface on each particle i , obtain the roundness R based on the four curvatures 3D for: ， ， ， ， Among them, R 3D1 Based on the minimum principal curvature, the smoothness of the flat area of ​​each particle is reflected. 3D2 The maximum principal curvature is used to characterize the roundness of each particle's angular region, R 3Dm Based on the average curvature, R 3Dg Quantifying local geometric complexity based on Gaussian curvature.

9. The method for particle geometry reconstruction and automatic morphological parameter evaluation based on three-dimensional images according to claim 1, characterized in that: In step d, the orthogonal projection planes of the three-dimensional image include: XY plane, XZ plane and YZ plane; the particle shape parameters of the two-dimensional image include two-dimensional slenderness, two-dimensional sphericity, two-dimensional roundness and two-dimensional concavity and convexity.

10. The method for automatic particle geometry reconstruction and morphological parameter evaluation based on three-dimensional images according to claim 9, characterized in that: The particle ellipse of the two-dimensional projection is fitted based on the least squares method, and the length of the major axis of the ellipse is taken as D max , using Feret diameter method to calculate the minimum width D min , obtain the two-dimensional particle shape parameter slenderness ratio Ar of the three-dimensional particle projection surface 2D for: ； Calculate the minimum distance from the center of mass to the boundary point of the particle's two-dimensional projection as the maximum inscribed circle radius r max-in , the maximum distance from the center of mass to the boundary point is the minimum circumscribed circle radius r min-cir , obtain the two-dimensional particle shape parameter sphericity S of the three-dimensional particle projection surface 2D for: ； The average value of the radius of the curvature circle corresponding to the boundary points of the two-dimensional projection of the particle is , combined with the maximum inscribed circle radius r max-in , obtain the two-dimensional particle shape parameter roundness R of the three-dimensional particle projection surface 2D for: ； Calculate the total number of pixels belonging to the particle in the two-dimensional projection binary image as the actual area A 2D The total number of pixels inside the smallest convex polygon containing the particle is the convex hull area A 2Dcon , obtain the two-dimensional particle shape parameter concavity C of the three-dimensional particle projection surface 2D for: 。

Images