Hexahedral mesh generation method based on image feature point local encryption

Abstract

The invention discloses a hexahedral mesh generation method based on image feature point local encryption, and the method comprises the steps: S1) obtaining an industrial CT image, and carrying out the preprocessing of the industrial CT image; s2) carrying out feature extraction on the preprocessed industrial CT image, and determining an area containing grid features in the industrial CT image; s3) performing grid division on the industrial CT image; s4) carrying out local encryption on the grid containing the grid feature part; and S5) generating a boundary unit grid by using a shortest distance method to obtain a local encrypted grid of the workpiece. According to the method, the space complexity and the calculation time in the grid generation process are reduced, the generated grid can adaptively realize local refinement in a workpiece curvature feature obvious region, and the problem that features are difficult to maintain in a direct generation method is solved.

Description

Technical Field

[0001] The invention relates to the field of imaging of internal and external structures of workpieces, and in particular to a method for generating hexahedral meshes based on local encryption of image feature points. Background Art

[0002] Industrial computed tomography (CT) technology can obtain the internal and external structure of workpieces under non-destructive conditions and is widely used in defect detection, reverse engineering, etc. Finite element analysis can understand the overall stress and strain distribution of the workpiece and is an effective numerical simulation method for solving practical engineering problems. Three-dimensional finite element meshes include tetrahedron, hexahedron, prism and pyramid meshes.

[0003] However, the existing three-dimensional finite element analysis method is computationally complex and cannot locally refine areas with obvious curvature features of the workpiece, and is prone to losing information on subtle features of the workpiece. Summary of the invention

[0004] The object of the present invention is to provide a method for generating a hexahedral mesh based on local encryption of image feature points, comprising the following steps:

[0005] S1) acquiring an industrial CT image and preprocessing the industrial CT image;

[0006] S2) performing feature extraction on the preprocessed industrial CT image to determine an area containing grid features in the industrial CT image;

[0007] S3) gridding the industrial CT image;

[0008] S4) using an octree algorithm in combination with the maximum and minimum grid sizes and the information of the characteristic points contained in the grid, locally encrypting the grid containing the characteristic parts of the grid, and obtaining a non-uniform hexahedral grid with characteristic local encryption;

[0009] S5) generating a boundary unit grid using the shortest distance method to obtain a local encrypted grid of the workpiece;

[0010] Furthermore, the preprocessing includes noise reduction and image binarization processing.

[0011] Further, the steps of noise reduction processing include:

[0012] S1.1) performing wavelet transform processing on the industrial CT image, thereby decomposing the industrial CT image into several frequency components;

[0013] S1.2) De-noising all frequency components, removing the wavelet coefficients of the noise signal and retaining the wavelet coefficients of the effective image signal;

[0014] S1.3) Use the soft threshold function to apply a threshold to the retained wavelet coefficients to obtain the wavelet coefficients ω after applying the threshold λ ,Right now:

[0015]

[0016] Where: ω is the wavelet coefficient; λ is the threshold; N is the signal length, σ is the noise standard deviation;

[0017] S1.4) The high-frequency wavelet coefficients representing the image boundary and detail feature information after applying the threshold are processed using the threshold shrinkage method, and the low-frequency wavelet coefficients after applying the threshold are processed using the guided filtering method; then the processed wavelet coefficients are reconstructed using the inverse wavelet transform to obtain a denoised image.

[0018] Furthermore, the method of image binarization processing includes an adaptive threshold segmentation method;

[0019] The threshold T used by the adaptive threshold segmentation method is as follows:

[0020]

[0021] Where: G max is the maximum gray value in the window, G min Indicates the minimum grayscale value within the window.

[0022] Furthermore, the step of extracting features from the preprocessed industrial CT image includes:

[0023] Based on the Shi-Tomasi algorithm, a sliding detection cube is established in three-dimensional space;

[0024] The preprocessed industrial CT image is subjected to sliding detection using the sliding detection cube, and the grayscale value changes E(x, y, z) of the pixels at the same position in the sliding detection cube before and after sliding are compared. If the average corner point intensity of the cube is greater than the corresponding corner point response function R, it is determined that there are feature points in the industrial CT image area within the current sliding detection cube.

[0025] Furthermore, the gray value change of the pixel point E(x,y,z) is as follows:

[0026]

[0027] Where ω(x, y, z) is the sliding cube function of the current position, which represents the weight of the grayscale value of each pixel in the sliding cube, I(x, y, z) is the grayscale value of the pixel at the current position, and I(x+u, y+v, z+w) is the grayscale value of the pixel at the new position.

[0028] Furthermore, the corresponding corner point response function R is as follows:

[0029] R=min(λ1 2 ,λ2 2 ,λ3 2 )(4)

[0030] Where λ1, λ2, and λ3 represent the magnitude of the vertical gradients in the x, y, and z directions.

[0031] Furthermore, the steps of meshing the industrial CT image include:

[0032] S3.1) Determine the size of the octree root node in, It means rounding up, col×row is the pixel size of the input industrial CT image;

[0033] S3.2) Setting the recursive termination conditions of the adaptive octree, including whether the node contains feature point information and the maximum size maxL and the minimum size minL that control the overall grid size; 1≤minL≤maxL≤s;

[0034] S3.3) recursively partitioning the hexahedral mesh using an octree algorithm;

[0035] Further, the step of locally encrypting the mesh containing the mesh feature part includes:

[0036] S4.1) Find the non-uniform grid transition region using the hanging node transition chain;

[0037] S4.2) Use the 2-refinement template to encrypt the mesh of the non-uniform mesh hanging nodes, thereby solving the hanging node problem caused by dividing the non-uniform hexahedral mesh.

[0038] Further, the steps of generating the boundary cell mesh using the shortest distance method include:

[0039] S5.1) performing a layer deletion operation on the outer grids whose distance from the generated core grid to the image is less than a minimum unit size, and retaining the outer grids whose distance from the edge of the image is greater than or equal to a minimum unit size;

[0040] S5.2) projecting four vertices of the outer surface of the core mesh along the direction of the shortest distance from the edge of the image to obtain four projection points;

[0041] S5.3) Connect the four vertices of the outer surface of the core mesh and the four points obtained by projection in sequence according to the order of mesh node construction to obtain a fitting mesh unit.

[0042] The technical effect of the present invention is unquestionable. The present invention focuses on the extraction of three-dimensional feature points of industrial CT sequence images, adaptively divides the octree, performs adaptive division according to feature point information and the set maximum and minimum grid sizes, uses the hanging node transition chain search template to find the hanging nodes in the initial core grid, uses the template method on the hanging node transition chain to eliminate the hanging nodes, generates a core grid with a smooth transition, deletes the layer deletion operation close to the edge of the image, uses the shortest distance method to perform surface fitting on the core grid, and generates boundary units that meet the boundary consistency requirements. It not only reduces the spatial complexity and calculation time in the grid generation process, but also enables the generated grid to adaptively achieve local refinement in the area with obvious curvature characteristics of the workpiece, and solves the problem of difficulty in maintaining features in the direct generation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 The generation process of the local encrypted grid based on image feature points of the present invention is shown.

[0044] Figure 2 A schematic diagram of feature point detection for industrial CT sequence images is shown.

[0045] Figure 3 The local mesh encryption detection process of the local encrypted hexahedral mesh generation method based on image feature points of the present invention is shown.

[0046] Figure 4 Schematic diagram showing the core mesh of the local encrypted mesh generated from the industrial CT image of the gear workpiece

[0047] Figure 5 A schematic diagram of the hanging node transition chain search is shown.

[0048] Figure 6 A schematic diagram of 2-refined template processing hanging nodes is shown.

[0049] Figure 7 The locally encrypted three-dimensional hexahedral mesh after projection using the shortest distance method is shown. DETAILED DESCRIPTION

[0050] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

[0051] Embodiment 1:

[0052] See also Figures 1 to 7 , a hexahedral mesh generation method based on local encryption of image feature points, comprising the following steps:

[0053] S1) acquiring an industrial CT image of a workpiece (such as a gear) and preprocessing the industrial CT image;

[0054] S2) performing feature extraction on the preprocessed industrial CT image to determine an area containing grid features in the industrial CT image;

[0055] S3) gridding the industrial CT image;

[0056] S4) using an octree algorithm in combination with the maximum and minimum grid sizes and the information of the characteristic points contained in the grid, locally encrypting the grid containing the characteristic parts of the grid, and obtaining a non-uniform hexahedral grid with characteristic local encryption;

[0057] S5) generating a boundary unit grid using the shortest distance method to obtain a local encrypted grid of the workpiece;

[0058] The preprocessing includes noise reduction and image binarization.

[0059] The steps of noise reduction processing include:

[0060] S1.1) performing wavelet transform processing on the industrial CT image, thereby decomposing the industrial CT image into several frequency components;

[0061] S1.2) De-noising all frequency components, removing the wavelet coefficients of the noise signal and retaining the wavelet coefficients of the effective image signal;

[0062] S1.3) Use the soft threshold function to apply a threshold to the retained wavelet coefficients to obtain the wavelet coefficients ω after applying the threshold λ ,Right now:

[0063]

[0064] Where: ω is the wavelet coefficient; λ is the threshold; N is the signal length, σ is the noise standard deviation;

[0065] S1.4) The high-frequency wavelet coefficients representing the image boundary and detail feature information after applying the threshold are processed using the threshold shrinkage method, and the low-frequency wavelet coefficients after applying the threshold are processed using the guided filtering method; then the processed wavelet coefficients are reconstructed using the inverse wavelet transform to obtain a denoised image. The high-frequency wavelet coefficients represent the detail information such as the edge and texture in the image. In the wavelet transform process, the high-frequency components are obtained by high-pass filtering the image, and the details in the horizontal, vertical and diagonal directions can be retained. Processing high-frequency coefficients can be used to enhance the details of the image and denoise at the same time.

[0066] The methods of image binarization processing include adaptive threshold segmentation method;

[0067] The threshold T used by the adaptive threshold segmentation method is as follows:

[0068]

[0069] Where: G max is the maximum gray value in the window, G min Indicates the minimum grayscale value within the window.

[0070] The steps of feature extraction for preprocessed industrial CT images include:

[0071] Based on the Shi-Tomasi algorithm, a sliding detection cube is established in three-dimensional space;

[0072] The preprocessed industrial CT image is subjected to sliding detection using the sliding detection cube, and the grayscale value changes E(x, y, z) of the pixels at the same position in the sliding detection cube before and after sliding are compared. If the average corner point intensity of the cube is greater than the corresponding corner point response function R, it is determined that there are feature points in the industrial CT image area within the current sliding detection cube.

[0073] The change in pixel grayscale value E(x,y,z) is as follows:

[0074]

[0075] Where ω(x, y, z) is the sliding cube function of the current position, which represents the weight of the grayscale value of each pixel in the sliding cube, I(x, y, z) is the grayscale value of the pixel at the current position, and I(x+u, y+v, z+w) is the grayscale value of the pixel at the new position. λ1,λ2,λ3 represent the magnitude of the vertical gradient in the x, y, and z directions. u, v, and w are the coordinate differences between the pixel at the new position and the pixel at the current position; Ix, Iy, and Iz represent the grayscale values ​​of the pixel at the current position in the x, y, and z directions;

[0076] The corresponding corner point response function R is as follows:

[0077] R=min(λ1 2 ,λ2 2 ,λ3 2 )(4)

[0078] The steps of meshing industrial CT images include:

[0079] S3.1) Determine the size of the octree root node in, It means rounding up, col×row is the pixel size of the input industrial CT image;

[0080] S3.2) Setting the recursive termination conditions of the adaptive octree, including whether the node contains feature point information and the maximum size maxL and the minimum size minL that control the overall grid size; 1≤minL≤maxL≤s;

[0081] S3.3) recursively partitioning the hexahedral mesh using an octree algorithm;

[0082] The steps of locally encrypting the mesh containing the mesh features include:

[0083] S4.1) Find the non-uniform grid transition region using the hanging node transition chain;

[0084] S4.2) Use the 2-refinement template to encrypt the mesh of the non-uniform mesh hanging nodes, thereby solving the hanging node problem caused by dividing the non-uniform hexahedral mesh.

[0085] The steps to generate a boundary cell mesh using the shortest distance method include:

[0086] S5.1) performing a layer deletion operation on the outer grids whose distance from the generated core grid to the image is less than a minimum unit size, and retaining the outer grids whose distance from the edge of the image is greater than or equal to a minimum unit size;

[0087] S5.2) projecting four vertices of the outer surface of the core mesh along the direction of the shortest distance from the edge of the image to obtain four projection points;

[0088] S5.3) Connect the four vertices of the outer surface of the core mesh and the four points obtained by projection in sequence according to the order of mesh node construction to obtain a fitting mesh unit.

[0089] Embodiment 2:

[0090] A method for generating a hexahedral mesh based on local encryption of image feature points comprises the following steps:

[0091] S1) acquiring an industrial CT image and preprocessing the industrial CT image;

[0092] S2) performing feature extraction on the preprocessed industrial CT image to determine an area containing grid features in the industrial CT image;

[0093] S3) gridding the industrial CT image;

[0094] S4) using an octree algorithm in combination with the maximum and minimum grid sizes and the information of the characteristic points contained in the grid, locally encrypting the grid containing the characteristic parts of the grid, and obtaining a non-uniform hexahedral grid with characteristic local encryption;

[0095] S5) generating a boundary unit grid using the shortest distance method to obtain a local encrypted grid of the workpiece;

[0096] Embodiment 3:

[0097] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as that of Example 2, and further, the preprocessing includes noise reduction and image binarization processing.

[0098] Embodiment 4:

[0099] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-3, and further, the step of noise reduction processing comprises:

[0100] S1.1) performing wavelet transform processing on the industrial CT image, thereby decomposing the industrial CT image into several frequency components;

[0101] S1.2) De-noising all frequency components, removing the wavelet coefficients of the noise signal and retaining the wavelet coefficients of the effective image signal;

[0102] S1.3) Use the soft threshold function to apply a threshold to the retained wavelet coefficients to obtain the wavelet coefficients ω after applying the threshold λ ,Right now:

[0103]

[0104] Where: ω is the wavelet coefficient; λ is the threshold; N is the signal length, σ is the noise standard deviation;

[0105] S1.4) The high-frequency wavelet coefficients representing the image boundary and detail feature information after applying the threshold are processed using the threshold shrinkage method, and the low-frequency wavelet coefficients after applying the threshold are processed using the guided filtering method; then the processed wavelet coefficients are reconstructed using the inverse wavelet transform to obtain a denoised image.

[0106] Embodiment 5:

[0107] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-4, and further, the method for image binarization processing includes an adaptive threshold segmentation method;

[0108] The threshold T used by the adaptive threshold segmentation method is as follows:

[0109]

[0110] Where: G max is the maximum gray value in the window, G min Indicates the minimum grayscale value within the window.

[0111] Embodiment 6:

[0112] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-5, and further, the step of extracting features from the preprocessed industrial CT image comprises:

[0113] Based on the Shi-Tomasi algorithm, a sliding detection cube is established in three-dimensional space;

[0114] The preprocessed industrial CT image is subjected to sliding detection using the sliding detection cube, and the grayscale value changes E(x, y, z) of the pixels at the same position in the sliding detection cube before and after sliding are compared. If the average corner point intensity of the cube is greater than the corresponding corner point response function R, it is determined that there are feature points in the industrial CT image area within the current sliding detection cube.

[0115] Embodiment 7:

[0116] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-6, and further, the gray value variation E(x, y, z) of the pixel point is as follows:

[0117]

[0118] Where ω(x, y, z) is the sliding cube function at the current position, which represents the weight of the grayscale value of each pixel in the sliding cube, I(x, y, z) is the grayscale value of the pixel at the current position, and I(x+u, y+v, z+w) is the grayscale value of the pixel at the new position. λ1, λ2, λ3 represent the vertical gradients in the three directions of x, y, and z.

[0119] To maximize the value of E(x, y, z), the three values ​​of λ1, λ2, and λ3 must meet the requirements at the same time. To achieve this goal, the corresponding corner point response function R is proposed as follows:

[0120] R=min(λ1 2 ,λ2 2 ,λ3 2 )(2)

[0121] The R value indicates the strength of the corner point. If it is greater than this value, it is considered that the sliding cube contains points with high corner point strength, that is, image feature points. The smaller the R value is, the more feature points are detected, and the lower the average corner point strength of the feature points is, and vice versa. Usually, the R value is 0.016.

[0122] Embodiment 8:

[0123] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-7, and further, the step of meshing the industrial CT image comprises:

[0124] S3.1) Determine the size of the octree root node in, It means rounding up, col×row is the pixel size of the input industrial CT image;

[0125] S3.2) Setting the recursive termination conditions of the adaptive octree, including whether the node contains feature point information and the maximum size maxL and the minimum size minL that control the overall grid size; 1≤minL≤maxL≤s;

[0126] S3.3) recursively partitioning the hexahedral mesh using an octree algorithm;

[0127] Embodiment 9:

[0128] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of embodiments 2-8, and further, the step of locally encrypting the mesh containing the mesh feature parts comprises:

[0129] S4.1) Find the non-uniform grid transition region using the hanging node transition chain;

[0130] S4.2) Use the 2-refinement template to encrypt the mesh of the non-uniform mesh hanging nodes, thereby solving the hanging node problem caused by dividing the non-uniform hexahedral mesh.

[0131] Embodiment 10:

[0132] A method for generating a hexahedral mesh based on local encryption of image feature points, the technical content of which is the same as any one of Embodiments 2-9, and further, the step of generating a boundary unit mesh using the shortest distance method comprises:

[0133] S5.1) performing a layer deletion operation on the outer grids whose distance from the generated core grid to the image is less than a minimum unit size, and retaining the outer grids whose distance from the edge of the image is greater than or equal to a minimum unit size;

[0134] S5.2) projecting four vertices of the outer surface of the core mesh along the direction of the shortest distance from the edge of the image to obtain four projection points;

[0135] S5.3) Connect the four vertices of the outer surface of the core mesh and the four points obtained by projection in sequence according to the order of mesh node construction to obtain a fitting mesh unit.

[0136] Embodiment 11:

[0137] A method for generating a hexahedral mesh based on local encryption of image feature points, the steps are as follows:

[0138] S1, binarization processing is performed on the industrial CT image of the gear workpiece after denoising;

[0139] Specifically, step S1 includes, S1.1, decomposing the image into different frequency components by wavelet transform;

[0140] S1.2, remove the wavelet coefficients of noise and retain the wavelet coefficients of effective image signals;

[0141] S1.3, use the soft threshold function to determine the threshold, which is expressed as:

[0142]

[0143] Where: ω is the size of the wavelet coefficient, ω λ is the size of the wavelet coefficient after applying the threshold, and λ is the threshold;

[0144] S1.4, using a fixed threshold to select λ, its expression is:

[0145] Where λ is the threshold, N is the signal length, and σ is the noise standard deviation.

[0146] S1.5, using threshold shrinkage method to process high frequency coefficients, using guided filtering to process low frequency coefficients; using inverse wavelet transform to reconstruct the processed wavelet coefficients, thereby obtaining a denoised image;

[0147] S1.6, use the adaptive threshold segmentation technique to perform global binary segmentation on the image, and the threshold T can be expressed as:

[0148]

[0149] Where: G max is the maximum gray value in the window, G min is the minimum gray value in the window;

[0150] S2, extracting features of industrial CT sequence images in the three-dimensional target area.

[0151] Specifically, Figure 2 As shown, step S2 includes: S2.1, extending the Shi-Tomasi algorithm to apply in three-dimensional space, establishing a sliding detection cube in the three-dimensional space, and comparing the gray value change of pixels at the same position in the cube before and after sliding. If there is a large gray value change, it can be considered that there is a feature point in the cube at this time.

[0152]

[0153] E(x,y,z) is the change in the grayscale value of the pixel at the position (x,y,z), ω(x,y,z) is the sliding cube function of the current position, which represents the weight of the grayscale value of each pixel in the sliding cube, I(x,y,z) is the grayscale value of the pixel at the current position, and I(x+u,y+v,z+w) is the grayscale value of the pixel at the new position.

[0154] S2.2, rewritten into matrix form as follows:

[0155]

[0156] S2.3, this patent proposes the corresponding corner point response function:

[0157] R=min(λ1 2 ,λ2 2 ,λ3 2 )When R is set to 0.016, better corner detection effect can be achieved.

[0158] S3, recursively divide the non-uniform grid according to the characteristics of the industrial CT sequence images, and locally encrypt the grid feature parts.

[0159] Specifically, Figure 3 As shown, step S3 includes: S3.1, first determine the size of the octree root node, the root node size can be expressed by Get, where It means rounding up, col×row is the pixel size of the input industrial CT image;

[0160] S3.2, set the recursive termination conditions of the adaptive octree, including whether the node contains feature point information and the maximum size maxL and minimum size minL of the overall grid size. The overall grid size can be controlled according to demand, and the maximum and minimum grid sizes need to be set, and the relationship 1≤minL≤maxL≤s must be satisfied when setting;

[0161] S3.3, recursively divide the hexahedral mesh using the octree algorithm;

[0162] S4, using the hanging node transition chain search template method to find the hanging node transition chain, and using 2 refinement templates on the hanging node transition chain to eliminate the hanging nodes;

[0163] Specifically, in Figure 4 Find the hanging node transition chain on Figure 5 , using 2 refinement templates to obtain Figure 6 ,Step S4 includes: 4.1, proposing a hanging node transition chain to find the non-uniform grid transition area;

[0164] S4.2, use the 2-refine template to refine the mesh of the non-uniform mesh hanging nodes.

[0165] S5, using the shortest distance method to generate the surface unit mesh, and finally obtain the local encrypted mesh of the workpiece.

[0166] Specifically, Figure 7 As shown, step S5 includes: S5.1, performing a layer deletion operation on the outer grids whose distance from the generated core grid to the image is less than a minimum unit size, and retaining the outer grids whose distance from the edge of the image is greater than or equal to a minimum unit size;

[0167] S5.2, projecting four vertices of the outer surface of the core mesh along the direction of the shortest distance from the edge of the image to obtain four projection points;

[0168] S5.3, according to the order of mesh node construction, the four vertices of the outer surface of the core mesh and the four points obtained by projection are connected in sequence to obtain a fitting mesh.

Claims

1. A hexahedral mesh generation method based on local encryption of image feature points, It is characterized in that The following steps are involved: S1) Acquire industrial CT images and preprocess the industrial CT images. S2) performing feature extraction on the preprocessed industrial CT image to determine an area containing grid features in the industrial CT image; S3) gridding the industrial CT image; S4) using an octree algorithm in combination with the maximum and minimum grid sizes and the information of the characteristic points contained in the grid, locally encrypting the grid containing the characteristic parts of the grid, and obtaining a non-uniform hexahedral grid with characteristic local encryption; S5) Generate a boundary unit grid using the shortest distance method to obtain a local encrypted grid of the workpiece.

2. According to claim 1, a hexahedral mesh generation method based on local encryption of industrial CT image feature points, It is characterized in that The preprocessing includes noise reduction and image binarization.

3. According to claim 2, a hexahedral mesh generation method based on local encryption of industrial CT image feature points, It is characterized in that The steps of noise reduction processing include: S1.1) performing wavelet transform processing on the industrial CT image, thereby decomposing the industrial CT image into several frequency components; S1.2) De-noising all frequency components, removing the wavelet coefficients of the noise signal and retaining the wavelet coefficients of the effective image signal; S1.3) Use the soft threshold function to apply a threshold to the retained wavelet coefficients to obtain the wavelet coefficients ω after applying the threshold λ ,Right now: Where: ω is the wavelet coefficient; λ is the threshold; N is the signal length, σ is the noise standard deviation; S1.4) The high-frequency wavelet coefficients representing the image boundary and detail feature information after applying the threshold are processed using the threshold shrinkage method, and the low-frequency wavelet coefficients after applying the threshold are processed using the guided filtering method; then the processed wavelet coefficients are reconstructed using the inverse wavelet transform to obtain a denoised image.

4. According to claim 2, a hexahedral mesh generation method based on local encryption of industrial CT image feature points, It is characterized in that The methods of image binarization processing include adaptive threshold segmentation method; The threshold T used by the adaptive threshold segmentation method is as follows: Where: G max is the maximum gray value in the window, G min Indicates the minimum grayscale value within the window.

5. According to claim 1, a hexahedral mesh generation method based on local encryption of industrial CT image feature points, It is characterized in that The steps of feature extraction for preprocessed industrial CT images include: Based on the Shi-Tomasi algorithm, a sliding detection cube is established in three-dimensional space; The preprocessed industrial CT image is subjected to sliding detection using the sliding detection cube, and the grayscale value changes E(x, y, z) of the pixels at the same position in the sliding detection cube before and after sliding are compared. If the average corner point intensity of the cube is greater than the corresponding corner point response function R, it is determined that there are feature points in the industrial CT image area within the current sliding detection cube.

6. A hexahedral mesh generation method based on local encryption of industrial CT image feature points according to claim 5, It is characterized in that The change in pixel grayscale value E(x,y,z) is as follows: Where ω(x, y, z) is the sliding cube function of the current position, which represents the weight of the grayscale value of each pixel in the sliding cube, I(x, y, z) is the grayscale value of the pixel at the current position, and I(x+u, y+v, z+w) is the grayscale value of the pixel at the new position.

7. The method for generating a hexahedral mesh based on local encryption of industrial CT image feature points according to claim 5, It is characterized in that The corresponding corner point response function R is as follows: In the formula, λ 1 ,λ 2 ,λ 3 Indicates the magnitude of the vertical gradient in the x, y, and z directions.

8. The method for generating a hexahedral mesh based on local encryption of industrial CT image feature points according to claim 1, It is characterized in that The steps of meshing industrial CT images include: S3.1) Determine the size of the octree root node in, It means rounding up, col×row is the pixel size of the input industrial CT image; S3.2) Setting the recursive termination conditions of the adaptive octree, including whether the node contains feature point information and the maximum size maxL and the minimum size minL that control the overall grid size; 1≤minL≤maxL≤s; S3.3) Recursively divide the hexahedral mesh using the octree algorithm.

9. The method for generating a hexahedral mesh based on local encryption of industrial CT image feature points according to claim 1, It is characterized in that The steps of locally encrypting the mesh containing the mesh features include: S4.1) Find the non-uniform grid transition region using the hanging node transition chain; S4.2) Use the 2-refinement template to encrypt the mesh of the non-uniform mesh hanging nodes, thereby solving the hanging node problem caused by dividing the non-uniform hexahedral mesh.

10. The method for generating a hexahedral mesh based on local encryption of industrial CT image feature points according to claim 1, It is characterized in that The steps to generate a boundary cell mesh using the shortest distance method include: S5.1) performing a layer deletion operation on the outer grids whose distance from the generated core grid to the image is less than a minimum unit size, and retaining the outer grids whose distance from the edge of the image is greater than or equal to a minimum unit size; S5.2) projecting four vertices of the outer surface of the core mesh along the direction of the shortest distance from the edge of the image to obtain four projection points; S5.3) Connect the four vertices of the outer surface of the core mesh and the four points obtained by projection in sequence according to the order of mesh node construction to obtain a fitting mesh unit.

Images