Rock pore three-dimensional reconstruction method and system based on adaptive hierarchical cube algorithm, and storage medium

Abstract

The invention discloses a rock pore three-dimensional reconstruction method and system based on an adaptive hierarchical cube algorithm, and a storage medium, relates to the technical field of rock structure measurement, and aims to solve the problems of low extraction execution efficiency, ambiguity and poor reconstruction precision of an existing moving cube algorithm. According to an image segmentation model based on RefineNet, the model adopts three 3 * 3 convolution cascading to replace 7 * 7 convolution, and a coordinate attention mechanism is introduced behind each block of an encoder to improve the segmentation precision and efficiency of a key area; dividing the cube into tetrahedrons by adopting a saddle point-adaptive layered downward subdivision method, solving intersection point coordinates and normal vectors, and constructing a triangular patch index; determining the contour surface shape penetrating through the cube by adopting a cube lookup table, and merging triangles of all tetrahedrons to form a complete contour surface; and creating a new triangular patch by using the vertex of the edge of the hole and the newly generated vertex, and repairing the hole to obtain the accurately reconstructed three-dimensional hole of the rock.

Description

Technical Field

[0001] The present invention relates to the technical field of rock structure measurement, and in particular to a method, system and storage medium for three-dimensional reconstruction of rock pores based on an adaptive layered cube algorithm. Background Art

[0002] In petrological research, the pore structure of rocks covers the size, shape, arrangement and interconnection of pores and throats. Different types of rocks have their own unique pore characteristics, which are mainly affected by the spatial layout of pores, the properties of cements and the rock sedimentation process, and are closely related to the permeability of reservoir rocks. As an indispensable part of rock composition, rock pores are of fundamental significance for understanding the migration, storage mechanism and controlling factors of oil and gas. The diversity of rock pore structure has a significant impact on the permeability of oil and gas. Therefore, three-dimensional reconstruction of rock pores is crucial for the effective development of oil and gas fields.

[0003] When it comes to rock pore-related measurements, the main methods include electron microscope scanning, nuclear magnetic resonance and CT scanning, but they are not able to reflect the situation of rock pores. Using neural network technology to process and analyze data helps to extract information about pore structure. Marching Cubes (MC) is a commonly used method for image-based 3D reconstruction. However, the traditional marching cube algorithm spends a lot of time on useless voxel traversal during operation, resulting in low efficiency of the entire reconstruction algorithm, ambiguity and poor reconstruction accuracy. Summary of the invention

[0004] The technical problems to be solved by the present invention are:

[0005] The existing marching cubes algorithms have the problems of low extraction execution efficiency, ambiguity and poor reconstruction accuracy.

[0006] The present invention adopts the following technical solutions to solve the above technical problems:

[0007] The present invention provides a method for three-dimensional reconstruction of rock pores based on an adaptive layered cube algorithm, comprising the following steps:

[0008] Step 1: Slice the rock sample, collect CT scan image data of the rock sample slices, divide the slice image of each rock sample into a group, and pre-process the image data;

[0009] Step 2: Build an image segmentation model based on RefineNet. The model uses three 3x3 convolution cascades to replace 7x7 convolutions. A coordinate attention mechanism is introduced after each block of the encoder to improve the segmentation accuracy and efficiency of key areas. The rock image is segmented based on the image segmentation model.

[0010] Step 3: traverse all voxels based on the marching cube algorithm, combine adjacent slices to create a logical cube, and compare the vertex values ​​of the cube with the preset threshold of the isosurface to determine the isosurface;

[0011] Step 4: Use the saddle point-adaptive hierarchical downward subdivision method to divide the cube into tetrahedrons, calculate the intersection coordinates and normal vectors, and construct the triangle patch index;

[0012] Step 5: Construct a cube lookup table, determine the shape of the isosurface passing through the cube through the lookup table, and merge the triangles of all tetrahedrons to form a complete isosurface;

[0013] Step 6: Use the vertices at the edge of the hole and the newly generated vertices to create new triangular faces, repair the holes, and obtain the reconstructed three-dimensional pores of the rock.

[0014] Furthermore, the preprocessing of the image data described in step 1 specifically includes: enhancing the contrast of the image, and expanding the data through random cropping, flipping, and mirror expansion operations.

[0015] Furthermore, the functional implementation process of the coordinate attention mechanism described in step 2 is:

[0016] a. Define coordinate encoding for each pixel position of the feature map, normalize the position information to generate a coordinate encoding matrix, perform pooling operations on the feature maps in the horizontal and vertical directions respectively, and generate a pair of feature maps with two-directional perception capabilities;

[0017] b. Connect the obtained feature maps with two-directional perception capabilities along the spatial dimension to form a fused feature map, generate an intermediate feature map through a nonlinear activation function, and generate attention weights through a Sigmoid function, including: horizontal attention weight A h and vertical attention weight A w ;

[0018] c. Set the horizontal attention weight A h and vertical attention weight A w Applied to the corresponding direction of the original feature map, the weighted feature map is finally output.

[0019] Furthermore, in the training process of the image segmentation model in step 2, a hybrid loss function is constructed by using BCE loss and Dice loss.

[0020] Furthermore, in step 5, linear interpolation is used to calculate the intersection point P of the isosurface and the edge of the cube, that is:

[0021]

[0022] Among them, V1 and V2 are the values ​​in the cube lookup table, iso is the threshold of the isosurface, and P1 and P2 are the positions of the vertices on the edge of the cube.

[0023] Furthermore, step six specifically includes the following process:

[0024] (1) Fine-tune the position of the newly generated vertices according to the weight rules and the geometry of the hole;

[0025] (2) Merge the original vertices and newly generated vertices at the edge of the hole into a vertex set;

[0026] (3) triangulate the vertex set to generate new triangular facets;

[0027] (4) Filter out the faces that are completely inside the hole from the faces generated by triangulation and use them for hole repair to obtain the reconstructed three-dimensional pores of the rock.

[0028] A three-dimensional reconstruction system for rock pores based on an adaptive layered cube algorithm. The system has a program module corresponding to the steps of the method described in any one of the above technical solutions, and executes the steps in the above three-dimensional reconstruction method for rock pores based on an adaptive layered cube algorithm when running.

[0029] A computer-readable storage medium stores a computer program, wherein the computer program is configured to implement the steps of the rock pore three-dimensional reconstruction method based on an adaptive layered cube algorithm described in any one of the above technical solutions when called by a processor.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention discloses a method for three-dimensional reconstruction of rock pores based on an adaptive hierarchical cube algorithm. It adopts multi-layer small convolution kernels to replace large kernel convolution, reduces parameters and deepens the network depth, thereby improving network capacity and complexity. In addition, a coordinate attention mechanism is introduced after each block of the encoder, so that the segmentation efficiency and accuracy are significantly improved. The present invention designs a saddle point-adaptive hierarchical downward subdivision method to divide the cube into tetrahedrons to solve the ambiguity problem. By performing detailed subdivision of the cube, the complex geometric features inside the cube can be more accurately captured and represented, especially when there is a saddle point. This method can effectively handle complex situations inside the cube and ensure the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of a method for three-dimensional reconstruction of rock pores based on an adaptive layered cube algorithm in an embodiment of the present invention;

[0033] Figure 2 It is a CT scan image of a rock slice in an embodiment of the present invention;

[0034] Figure 3 The image is an image after histogram equalization and labelme annotation in the embodiment of the present invention;

[0035] Figure 4 The image is obtained by segmenting the rock pores using the image segmentation model in the embodiment of the present invention;

[0036] Figure 5 A diagram showing a dissection situation when there is a body saddle point in the body in an embodiment of the present invention;

[0037] Figure 6 The diagram is a diagram showing the segmentation of a body with two body saddle points in the embodiment of the present invention.

[0038] Figure 7 is a flow chart of a cube lookup table strategy in an embodiment of the present invention;

[0039] Figure 8 A schematic diagram of triangulation in an embodiment of the present invention;

[0040] Fig. 9 Schematic diagram of 3D reconstruction of pores in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the scheme of the present invention, exemplary implementations or embodiments of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described implementations or embodiments are only implementations or embodiments of a part of the present invention, not all of them. Based on the implementations or embodiments of the present invention, all other implementations or embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0043] Specific implementation plan 1: Figure 1 As shown, the present invention provides a rock pore three-dimensional reconstruction method based on an adaptive layered cube algorithm, comprising the following steps:

[0044] Step 1: Slice the rock sample, collect CT scan image data of the rock sample slices, divide the slice image of each rock sample into a group, and pre-process the image data;

[0045] Step 2: Build an image segmentation model based on RefineNet. The model uses three 3x3 convolution cascades to replace 7x7 convolutions, and introduces a coordinate attention mechanism after each block of the encoder to improve the segmentation accuracy and efficiency of key areas.

[0046] Step 3: traverse all voxels based on the marching cube algorithm, combine adjacent slices to create a logical cube, and compare the vertex values ​​of the cube with the preset threshold of the isosurface to determine the isosurface;

[0047] Step 4: Use the saddle point-adaptive hierarchical downward subdivision method to divide the cube into tetrahedrons, calculate the intersection coordinates and normal vectors, and construct the triangle patch index;

[0048] Step 5: Construct a cube lookup table, determine the shape of the isosurface passing through the cube through the lookup table, and merge the triangles of all tetrahedrons to form a complete isosurface;

[0049] Step 6: Use the vertices at the edge of the hole and the newly generated vertices to create new triangular faces, repair the holes, and obtain the reconstructed three-dimensional pores of the rock.

[0050] Specific implementation scheme 2: The image data is preprocessed as described in step 1, specifically including: contrast enhancement of the image, and data expansion by random cropping, flipping, and mirror expansion operations. The rest of this implementation scheme is the same as the specific implementation scheme 1.

[0051] Specific implementation plan three: The functional implementation process of the coordinate attention mechanism described in step 2 is:

[0052] a. Define coordinate encoding for each pixel position of the feature map, normalize the position information to generate a coordinate encoding matrix, perform pooling operations on the feature maps in the horizontal and vertical directions respectively, and generate a pair of feature maps with two-directional perception capabilities;

[0053] b. Connect the obtained feature maps with two-directional perception capabilities along the spatial dimension to form a fused feature map, generate an intermediate feature map through a nonlinear activation function, and generate attention weights through a Sigmoid function, including: horizontal attention weight A h and vertical attention weight A w ;

[0054] c. Set the horizontal attention weight A h and vertical attention weight A wIt is applied to the corresponding direction of the original feature map, and finally outputs a weighted feature map. The rest of this implementation is the same as the first implementation.

[0055] Specific implementation scheme 4: In the training process of the image segmentation model in step 2, a hybrid loss function is constructed by using BCE loss and Dice loss. The rest of this implementation scheme is the same as that of specific implementation scheme 3.

[0056] Specific implementation plan 5: In step 3, the pores are rendered white and the background is black, so the isosurface threshold is 255. All voxels are traversed, and the Kth slice and the K+1th slice are combined to create a logical cube. The eight vertices of the cube are compared with the threshold. If one of the two vertices on a certain edge of the voxel is greater than the isosurface threshold and the other is less than the isosurface threshold, then this edge has one and only one point that is the same as the isosurface threshold, that is, the intersection position of the isosurface and this edge.

[0057] Compare the value of the voxel vertex with the threshold of the isosurface. If the vertex value is not less than the threshold, the vertex is determined to be outside the isosurface and marked as "0"; if the vertex value is lower than the threshold, the vertex is determined to be inside the isosurface and marked as "1". The rest of this embodiment is the same as the specific embodiment 1.

[0058] Specific implementation scheme six: In step four, the cube is divided into tetrahedrons using the saddle point-adaptive layered downward subdivision method. The specific process is as follows:

[0059] The coordinates of the four points in the plane of a certain slice are known to be (x1, y1), (x2, y2), (x3, y3), (x4, y4). Substitute the coordinate values ​​of these known points into the equation

[0060] F(x,y)=axy+bx+cy+d=0

[0061] Get the values ​​of coefficients a, b, c and d, and find their partial derivatives The value of the saddle point is obtained. For the data in three-dimensional space, it is known that eight P i The coordinate value of (i=0,1...7) uses the trilinear interpolation function as follows:

[0062] P(x,y,z)=a+bx+cy+dz+exy+fxz+gyz+hxyz

[0063] Substitute the coordinates of the eight known points into the above equation and solve it to get the coefficients of the equation. Substitute the coefficients into the equation, then find the partial derivative of the equation and set it to 0. For example, the system of equations As shown. Solve the system of equations to obtain the value of the body saddle point. If the system of equations has no solution, it means that there is no body saddle point in the data space.

[0064] (1) n-face saddle point - m-body saddle point (0≤n≤6,0≤m≤1)

[0065] If there is only one saddle point in the cube, the point will be connected to the 8 vertices of the cube. For each face, if there is no saddle point, it will be divided into two tetrahedrons by connecting the diagonals; if there is a saddle point, the saddle point will be connected to the four corner points and the body saddle point to form four tetrahedrons, such as Figure 5 B and Figure 5 As shown in C.

[0066] When there is no body saddle point in the cube, use its center as a substitute. Then split it as shown in the figure above. In case 1, 12+2n tetrahedrons can be formed, where n represents the face saddle point.

[0067] (2) n-face saddle point - 2-body saddle point (0≤n≤6)

[0068] The existence of two saddle points makes the segmentation more complicated. The vertices and saddle points of the cube are numbered as V i (i=0,1…7) and B j (j=1,2).

[0069] When there is no body saddle point (n = 0), first remove the three edges connected to vertex V0 and the three edges connected to V7 (here vertices V0 and V7 can be replaced by any pair of opposite vertices). Next, connect the two body saddle points B1 and B2, and connect them to the other six vertices except V0 and V7. Figure 6 As shown in B, this operation can generate six tetrahedrons, namely (B1B2V1V2), (B1B2V2V3), (B1B2V3V4), (B1B2V4V5), (B1B2V5V6), and (B1B2V6V7), a total of 6 tetrahedrons.

[0070] The vertices and saddle points of the cube are numbered V i (i=0,1…7) and B j (j = 1, 2), in the case of only two body saddle points, if B1 is closer to the three adjacent faces of V0, then B1 is connected to the corner points of these faces, and the structure formed is as follows Figure 6 C, then, according to Figure 6 The subdivision method of B processes these three faces separately. Similarly, after the vertices of the three adjacent faces of B2 and V7 are connected, they are subdivided according to the same subdivision principle (such as Figure 6 D) In ​​this way, 18 tetrahedrons will be generated.

[0071] In the case of a saddle point on the surface (and n is not equal to 0), the internal tetrahedron with edges from B1 to B2 remains unchanged. The surface containing the saddle point is divided according to Figure 6The partitioning method shown in C is used for processing. At this point, a total of 18+2n tetrahedrons can be partitioned, and for each tetrahedron, its vertex coordinates are determined. The vertex can be a vertex, a saddle point, or a body saddle point of a cube. The linear interpolation method is used to calculate the coordinates of the intersection points on each edge in the tetrahedron. For each calculated intersection point, its neighboring vertex in the original data grid is found, and the scalar value of the neighboring vertex is used to calculate the gradient of the point. At the intersection point, the opposite direction of the gradient is the normal vector. The calculated gradient vector is normalized to obtain a unit normal vector, and the intersection coordinates and the normal vector are used to construct a triangular face index, and the mesh data formed by the tetrahedron partitioning is converted into a renderable three-dimensional surface. The rest of this implementation plan is the same as the specific implementation plan one.

[0072] Specific implementation plan seven: (1) Figure 7 As shown, a queue Q, a marking array F and a result array Z are established. Among them, Q is responsible for storing the cubic voxels that intersect the isosurface; F is used to identify the processing status of the voxels in the queue Q, with unprocessed marked as "0" and processed marked as "1"; Z is used to save the triangular patch information generated by the processed voxels.

[0073] (2) After the algorithm starts, each voxel is checked one by one until the first voxel that intersects the isosurface is found. The voxel is then added to Q and marked as 1 in F.

[0074] (3) For each voxel, determine the triangle index by looking up the table based on the values ​​of its eight vertices, and find the voxels adjacent to it that intersect with the isosurface. Add these adjacent voxels to the queue Q and mark them as 0 in the array F. At the same time, record the triangle vertex and normal information of the voxel in the array T.

[0075] (4) After completing the processing of the first voxel in queue Q, continue to process the queue and perform the same operation on the next voxel. When there are no unprocessed voxels in the queue, the loop ends. At this point, the isosurface can be drawn and the algorithm repetition rate can be reduced, thereby improving the execution efficiency of the entire algorithm.

[0076] This implementation scheme establishes a cube lookup table to store cubes that have been confirmed to be non-empty, and records whether the cube has been operated, so as to reduce the inspection of empty cubes, prevent missed inspections, improve reconstruction efficiency, and solve the problem that a large amount of time is spent on useless voxel traversal during the operation of the MC algorithm, resulting in low efficiency of the entire reconstruction algorithm. The rest of this implementation scheme is the same as the specific implementation scheme six.

[0077] Specific implementation scheme eight: In step five, linear interpolation is used to calculate the intersection point P of the isosurface and the edge of the cube, that is:

[0078]

[0079] Wherein, V1 and V2 are the values ​​in the cube lookup table, iso is the threshold of the isosurface, and P1 and P2 are the positions of the vertices on the cube edge. The rest of this embodiment is the same as the first embodiment.

[0080] Specific implementation plan nine: Figure 8 As shown, step six specifically includes the following process:

[0081] (1) Fine-tune the position of the newly generated vertices according to the weight rules and the geometry of the hole;

[0082] (2) Merge the original vertices and newly generated vertices at the edge of the hole into a vertex set;

[0083] (3) triangulate the vertex set to generate new triangular facets;

[0084] (4) Filter out the facets completely located inside the hole from the facets generated by triangulation, and use them for hole repair to obtain a reconstructed three-dimensional rock pore. The rest of this embodiment is the same as the specific embodiment 1. This embodiment obtains a three-dimensional rendering that is more consistent with the real pores, so as to achieve stitching of the mesh model and optimization of hole repair.

[0085] Specific implementation scheme ten: The present invention provides a rock pore three-dimensional reconstruction method (algorithm) based on an adaptive layered cube algorithm, which is the underlying technical core of the present invention, and various products can be derived based on the algorithm.

[0086] Based on the method proposed in the present invention, a rock pore three-dimensional reconstruction system based on an adaptive layered cube algorithm is developed using a programming language. The system has program modules corresponding to the steps of the above-mentioned technical solution, and executes the steps in the above-mentioned rock pore three-dimensional reconstruction method based on the adaptive layered cube algorithm during operation.

[0087] The computer program of the developed system (software) is stored on a computer-readable storage medium, and the computer program is configured to implement the steps of the above-mentioned rock pore three-dimensional reconstruction method based on the adaptive layered cube algorithm when called by a processor. That is, the present invention is materialized on a carrier to become a computer program product.

[0088] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0089] The computer programs (also referred to as programs, software, software applications, or codes) of the present invention include machine instructions for programmable processors, and these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or device (e.g., disk, optical disk, memory, programmable logic device PLD) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0090] Example 1

[0091] Image acquisition: Sandstone samples of appropriate size were selected, and the cores were scanned using a GE Healthcare LightSpeed ​​Plus CT scanner to collect rock slice image data;

[0092] Dataset creation: The rock image size obtained by CT scanning is 753x753. A total of 1074 slices were scanned for a rock column. The results are as follows: Figure 2 As shown in the figure, the histogram equalization algorithm is used to enhance the image contrast of all images, and labelme is used to annotate the pores of the 100 preprocessed rock images to divide the rock images into pores and background. The results are shown in Figure 3 As shown in the figure, the image was randomly cropped, flipped, and mirrored to 1000 images, and the training set and test set were randomly generated at a ratio of 4:1. The training set was used to train the image segmentation model based on RefineNet constructed by the present invention, and the test set was used to test the model performance. The trained image segmentation model was used to segment the rock image, and the results were shown in the figure. Figure 4 shown.

[0093] An improved MC model was constructed, and the saddle point-adaptive hierarchical downward subdivision method was used to subdivide the cube into tetrahedrons. A cube lookup table was established to store those cubes that have been confirmed to be non-empty, and to record whether the cube has been operated. Finally, a set of triangular facets was drawn to obtain a three-dimensional model. At the same time, in order to obtain a three-dimensional rendering that is more consistent with the actual pores, the mesh model was stitched and optimized for hole repair. The results show that the number of triangular facets reconstructed by the MC model before the improvement increased by nearly 34.2%, and the average drawing time was further accelerated by nearly 18.5 seconds. Fig. 9 As shown in Figure 3, the obtained 3D pore map is visually smoother and retains more detailed features, and its details are better than those of the model before optimization.

[0094] Table 1

[0095] Detection Methods Reconstruction patches Reconstruction time MC algorithm before improvement 45938330 46s Improved MC algorithm 61670212 27.5s

[0096] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A method for three-dimensional reconstruction of rock pores based on an adaptive layered cube algorithm, characterized in that: The steps include: Step 1: Slice the rock sample, collect CT scan image data of the rock sample slices, divide the slice image of each rock sample into a group, and pre-process the image data; Step 2: Build an image segmentation model based on RefineNet. The model uses three 3x3 convolution cascades to replace 7x7 convolutions. A coordinate attention mechanism is introduced after each block of the encoder to improve the segmentation accuracy and efficiency of key areas. The rock image is segmented based on the image segmentation model. Step 3: traverse all voxels based on the marching cube algorithm, combine adjacent slices to create a logical cube, and compare the vertex values ​​of the cube with the preset threshold of the isosurface to determine the isosurface; Step 4: Use the saddle point-adaptive hierarchical downward subdivision method to divide the cube into tetrahedrons, calculate the intersection coordinates and normal vectors, and construct the triangle patch index; Step 5: Construct a cube lookup table, determine the shape of the isosurface passing through the cube through the lookup table, and merge the triangles of all tetrahedrons to form a complete isosurface; Step 6: Use the vertices at the edge of the hole and the newly generated vertices to create new triangular faces, repair the holes, and obtain the reconstructed three-dimensional pores of the rock.

2. The method for three-dimensional reconstruction of rock pores based on the adaptive layered cube algorithm according to claim 1, characterized in that: The image data is preprocessed as described in step 1, specifically including: enhancing the contrast of the image, and expanding the data through random cropping, flipping, and mirror expansion operations.

3. The method for three-dimensional reconstruction of rock pores based on the adaptive layered cube algorithm according to claim 1, characterized in that: The functional implementation process of the coordinate attention mechanism described in step 2 is: a. Define coordinate encoding for each pixel position of the feature map, normalize the position information to generate a coordinate encoding matrix, perform pooling operations on the feature maps in the horizontal and vertical directions respectively, and generate a pair of feature maps with two-directional perception capabilities; b. Connect the obtained feature maps with two-directional perception capabilities along the spatial dimension to form a fused feature map, generate an intermediate feature map through a nonlinear activation function, and generate attention weights through a Sigmoid function, including: horizontal attention weight A h and vertical attention weight A w ; c. Set the horizontal attention weight A h and vertical attention weight A w Applied to the corresponding direction of the original feature map, the weighted feature map is finally output.

4. The method for three-dimensional reconstruction of rock pores based on the adaptive layered cube algorithm according to claim 3, characterized in that: In the training process of the image segmentation model described in step 2, BCE loss and Dice loss are used to construct a hybrid loss function.

5. The method for three-dimensional reconstruction of rock pores based on the adaptive layered cube algorithm according to claim 1, characterized in that: In step 5, linear interpolation is used to calculate the intersection point P of the isosurface and the edge of the cube, that is: Among them, V1 and V2 are the values ​​in the cube lookup table, iso is the threshold of the isosurface, and P1 and P2 are the positions of the vertices on the edge of the cube.

6. The method for three-dimensional reconstruction of rock pores based on the adaptive layered cube algorithm according to claim 1, characterized in that: Step 6 specifically includes the following process: (1) Fine-tune the position of the newly generated vertices according to the weight rules and the geometry of the hole; (2) Merge the original vertices and newly generated vertices at the edge of the hole into a vertex set; (3) triangulate the vertex set to generate new triangular facets; (4) Filter out the faces that are completely inside the hole from the faces generated by triangulation and use them for hole repair to obtain the reconstructed three-dimensional pores of the rock.

7. A rock pore 3D reconstruction system based on an adaptive layered cube algorithm, characterized in that: The system has a program module corresponding to the steps of the method described in any one of claims 1 to 6, and executes the steps of the rock pore three-dimensional reconstruction method based on the adaptive layered cube algorithm when running.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps in the rock pore three-dimensional reconstruction method based on the adaptive layered cube algorithm described in any one of claims 1 to 6 when called by a processor.

Images