A bone model reconstruction method based on tomographic medical images

Through unidirectional linear interpolation, grayscale threshold method, edge detection and moving least squares method, the problem of ambiguity in bone three-dimensional reconstruction of bone in tomography images is solved, and high-precision three-dimensional bone model reconstruction is achieved and improve visual effects.

CN114549675BActive Publication Date: 2024-12-03TIANJIN POLYTECHNIC UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011333033.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-25
Publication Date
2024-12-03
Estimated Expiration
2040-11-25

AI Technical Summary

Technical Problem

The existing three-dimensional reconstruction method of tomography images has problems such as poor three-dimensional model clarity and limited grid editing. The surface grid-based method has ambiguity in topology, resulting in a step effect and affecting surface accuracy.

Method used

A three-dimensional data field is established by one-way linear interpolation method, combining grayscale threshold method and region growth method to segment the target bones, obtain the set of discrete points of the bone contour through edge detection and polygon approximation, and fit it with the cubic spline function using the moving least squares solution. The curves are connected with the shortest span method to create a high-precision three-dimensional bone model.

Benefits of technology

It improves the quality and visualization of the three-dimensional reconstruction of the skeleton model, reduces the impact of noise, is suitable for the fitting of complex surface models, and improves the efficiency of data processing and the application scope of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114549675B_ABST
    Figure CN114549675B_ABST
Patent Text Reader

Abstract

The technical scheme of the present invention is: firstly, a DICOM format tomographic medical image is obtained by CT scanning, and then the voxels between the torn slices are increased by unidirectional linear interpolation to establish a three-dimensional data field. Then, the target tissue is segmented by gray threshold method and regional growing method to separate it from other tissues, and then edge detection and polygon approximation are performed on each slice one by one to obtain a discrete point set of the target bone contour. Then, a moving least square solution is used, and a cubic spline function is selected as a compact support weighting function, and a discrete point set of the bone contour obtained by polygon approximation is fitted to a curve, and finally, the shortest span method is used as a connection criterion, and a triangular paving connection is performed between two adjacent contour lines to obtain a three-dimensional bone model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of modeling of tomographic medical images, and in particular to a bone model reconstruction method based on tomographic medical images. Background Art

[0002] Visualization and VR technology have been used more and more widely and deeply in medicine. Among them, three-dimensional reconstruction of bones is one of the important means of visualizing tomographic medical images. Through three-dimensional reconstruction, the physiological characteristics of bones can be reproduced more "realistically", thereby improving the quality of visualization. By cutting and other operations on the reconstructed bone model, its internal structure and the specific situation of the injured part can be observed more intuitively, which is convenient for doctors to make better judgments.

[0003] At present, the main methods for reconstructing three-dimensional bone models using tomographic medical images can be divided into two categories: three-dimensional reconstruction methods based on volume data and three-dimensional reconstruction methods based on surface meshes. The former mainly includes ray casting algorithms and voxel reconstruction methods. The disadvantages of both are that the clarity of the reconstructed three-dimensional model is poor and the subsequent mesh editing is subject to greater restrictions, which limits the scope of application of the reconstructed model. The latter includes MC algorithm pavement modeling. Due to the ambiguity of the topological structure of the MC algorithm, the establishment of a three-dimensional model will produce a step effect, which affects the surface accuracy. Although the step effect can be weakened by interpolating new sections, this will reduce the data processing efficiency. In view of the above shortcomings, the present invention proposes a bone model reconstruction method based on tomographic medical images. Summary of the invention

[0004] In view of the above situation, the present invention proposes a bone model reconstruction method based on tomographic medical images, which is a method for converting two-dimensional tomographic medical images into three-dimensional models. It can be used in related scientific research or medical work to improve the quality of visualization and facilitate doctors' diagnosis.

[0005] The technical scheme of the present invention is: firstly, a DICOM format tomographic medical image is obtained by CT scanning, and then the voxels between the faults are increased by unidirectional linear interpolation to establish a three-dimensional data field. Then, the target tissue is segmented by using the gray threshold method and the regional growing method to separate it from other tissues, and then edge detection and polygon approximation are performed on each slice one by one to obtain a discrete point set of the target bone contour. Then, the moving least square solution is used, and the cubic spline function is selected as the compact support weighting function, and the discrete point set of the bone contour obtained by polygon approximation is fitted to a curve, and finally the shortest span method is used as the connection criterion to perform triangular paving connection between two adjacent contour lines to obtain a three-dimensional bone model.

[0006] The present invention is a bone model reconstruction method based on tomographic medical images, which is characterized by comprising the following steps:

[0007] 1. Obtain DICOM-format tomographic medical images of bones through CT scanning, and set the scanning slice distance as small as possible to reduce the workload of subsequent unidirectional linear interpolation between slices.

[0008] 2. According to the spatial position parameters, each 2D tomographic image is placed in the same 2D coordinate system and arranged in space in the direction of CT scanning. Then, the unidirectional linear interpolation method is used to interpolate between each tomographic layer to generate new voxels, and finally form a 3D data field.

[0009] 3. Use the grayscale threshold method to separate bones, muscles, epidermis and soft tissues with different grayscale values, and then use the region growing segmentation method to separate single bones by removing other bone pixels connected to the target bone to obtain an independent target bone.

[0010] 4. Obtain the pixel point set on the outermost edge through edge detection, thereby obtaining the closed contour line of the target bone. Then perform polygonal approximation to reduce the number of pixels on the closed contour line, remove edge burrs caused by noise and other reasons, and finally obtain the discrete point set of the target bone contour.

[0011] 5. Using the moving least squares solution, the cubic spline function is selected as the compact support weighting function, and a curve is fitted to the discrete point set of the bone contour obtained by polygonal approximation. Finally, the shortest span method is used as the connection criterion to connect the two adjacent contour lines with triangles to obtain a three-dimensional bone model.

[0012] For step 2, the scanned tomographic medical image is two-dimensional data. The subsequent segmentation and modeling are all based on volume data, so it is necessary to interpolate between each fault to generate new voxels, and then arrange them in space to form a three-dimensional data field. One-way linear interpolation is used between two faults to add voxels to form a new fault.

[0013] Let H k (i, j) is the gray value of the pixel at point (i, j) on the k-th layer image, H k+1 (i, j) is the gray value of the pixel at point (i, j) on the k+1 layer image. The gray value of the pixel at point (i, j) on the new fault generated by interpolation between the k layer and the k+1 layer is:

[0014] H d (i, j) = (1-d)H k (i, j)+d×H k+1 (i, j) (1)

[0015] Among them, let the interval between the k layer and the interpolation layer be l k , the interval between the k+1 layer and the interpolation layer is l k+1, d = l k / (l k +l k+1 ).

[0016] For step 3, the gray threshold method uses the different gray values ​​of human tissues to select a suitable threshold range to separate bones from muscles, epidermis, and soft tissues. The reference CT value of human bones is generally greater than 400. The threshold range can be adjusted according to the segmentation result to achieve the best segmentation result. Then observe the tomographic image, and then use the regional growing segmentation method to remove other bone pixels connected to the target bone layer by layer to separate the target bone independently. Check the segmentation result. If you are not satisfied with the result, perform the above operation multiple times.

[0017] For step 4, edge detection is to set all pixel points of the target bone in the tomographic image to 1, and all other background pixel points to 0, select an initial pixel, and then search for the target bone pixels adjacent to the background pixels in the clockwise direction to obtain the pixel point set on the outermost edge, thereby obtaining the closed contour line of the target bone. Next, polygon approximation is performed.

[0018] First, perform a collinearity check. Given three points a1(i1, j1), a2(i2, j2), and a3(i3, j3) on the contour line, the length calculation formula of the line segment a1a3 is:

[0019]

[0020] The area calculation formula of Δa1a2a3 is:

[0021]

[0022] The vertical distance from point a2 to line segment a1a3 is:

[0023]

[0024] Through the maximum vertical distance d max To detect the collinearity of each point: given a string of points N on the contour line a , N a+1 ,…,N b-1 , N b , calculate the distance from each point in the point sequence to the line segment N a N b The vertical distance d, find the largest vertical distance d max As an approximation error, the allowable error ω is specified according to the actual situation. If the maximum vertical distance d max >ω, then the point is not collinear and is defined as N m ; If the maximum vertical distance d max<ω, then the point is considered to be collinear.

[0025] Assume that the set of pixels on the contour line is {N1, N2, ..., N n}, start from N1 and perform polygonal approximation in segments. During the approximation, perform a collinearity check. If it is determined to be collinear, a line segment passing through the beginning and end of the line segment can be used as the edge of the approximating polygon; if it is determined to be non-collinear, start from point N m The segment is split and rechecked as the end of the segment, and the segment is split and checked again until it is determined to be collinear. In this way, the amount of data is reduced and a discrete point set of the target bone contour is obtained.

[0026] For step 5, assume that the fitting function is in the form of formula (5):

[0027]

[0028] Among them, Q(i)=(Q1(i), Q2(i),…,Q m (i)) T is the coefficient vector; P(i)=(P1(i), P2(i),…, P m (i)) T is the basis function and has the following form:

[0029] First-order polynomial: P(i) = (1, i, j) T , m = 3

[0030] Quadratic polynomial: P(i) = (1, i, j, i 2 ,ij,j 2 ) T , m = 6

[0031] Define the weighted discrete norm:

[0032]

[0033] Where r is the radius of the support domain; α(|ii z | / r) is the weight function.

[0034] Derivative Q(i) in equation (6) yields:

[0035]

[0036] Q(i)=M -1 (i)N(i)j (7)

[0037] In the formula, N(i)=[α(|i-i1| / r)P(i1), α(|i-i2| / r)P(i2), ..., α(|i-i3| / r)P(i3)].

[0038] Substituting formula (7) into formula (5), we can obtain the expression of the moving least squares method:

[0039]

[0040] In the formula, k is the order of the basis function, is the shape function,

[0041] The node values ​​are calculated using formula (8), and the cubic spline function is used as the compact support weight function to fit a curve to the discrete points of the skeleton contour obtained by polygonal approximation. Finally, the shortest span method is used as the connection criterion to connect the two adjacent contour lines with triangle paving.

[0042] According to the fitting curve, the points on the adjacent contour lines are determined to correspond to each other. A triangular face is formed by two points on one contour line and one point on the adjacent contour line. The shortest span method is used to determine whether the two vertices of the same layer on the triangular mesh surface exist on the upper contour line or the lower contour line. That is, the triangular face with the shortest span is selected for paving. Assuming that the point set {D1, D2, ..., D n}, the point set on the upper contour line {U1, U2, ..., U n}, if the pavement has reached point D i , then there are two cases for the next pavement triangle: ΔD i U i+1 U i or ΔD i D i+1 U i , at this time judge and Length, if Then connect ΔD i D i+1 U i ; On the contrary, connect ΔD i U i+1 U i According to this idea, triangle mesh pavement is connected between all contour lines to finally obtain the 3D model of the skeleton.

[0043] The method of the present invention has the following advantages:

[0044] The method of the present invention can conveniently obtain a three-dimensional model by approximating and fitting the tomographic image, and is also applicable to other models with complex surfaces. The moving least squares method is used for fitting, and a compact support weighting function is introduced. The smoothness and accuracy of the fitting curve and surface can be controlled by changing the weight function. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Flow chart of a bone model reconstruction method based on tomographic medical images described in the present invention

[0046] Figure 2 The contour of the human tibia obtained by edge detection

[0047] Figure 3 Collinearity check diagram

[0048] Figure 4 The contour line set after moving least squares fitting

[0049] Figure 5 3D reconstruction model of human tibia DETAILED DESCRIPTION

[0050] In order to facilitate those skilled in the art to better understand the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] The overall flow chart of the bone model reconstruction method based on tomographic medical images of the present invention is as follows: Figure 1 As shown, the details are as follows:

[0052] 1. Scan data in the hospital CT room to obtain cross-sectional medical images from the upper end of the knee joint to the lower end of the ankle joint. The volunteer information is: male, 34 years old, 176 cm tall, 70 kg in weight, healthy, and no bone disease in the lower limbs. The slice thickness is 0.625 mm, the slice increment is 1 mm, and the number of slices is 415. The save format is DICOM.

[0053] 2. According to the spatial position parameters, each two-dimensional tomographic image is placed in the same two-dimensional coordinate system, and then arranged in space to form a three-dimensional data field. The vertical direction parameter k∈(1, 2, ..., 415) uses unidirectional linear interpolation to perform weighted averaging on each adjacent tomographic layer. After interpolation, the number of tomographic medical image layers is 829.

[0054] Let H k (i, j) is the gray value of the pixel at point (i, j) on the k-th layer image, H k+1 (i, j) is the grayscale value of the pixel at point (i, j) on the k+1 layer image. The grayscale value of the pixel at point (i, j) on the new fault generated by interpolation between the k layer and the k+1 layer is:

[0055] H d (i, j) = (1-d)H k (i, j)+d×H k+1 (i, j) (1)

[0056] Among them, let the distance between the k layer and the interpolation layer be l k , the distance between the k+1 layer and the interpolation layer is lk+1, d = l k / (l k +l k+1 ).

[0057] 3. Select the layer where the human tibia is located, a total of 746 layers. Use the grayscale threshold method, select the CT value range of 226 to 2603, separate the bones from muscles, epidermis, and soft tissues, observe the image, and when the segmentation result is ideal, use the region growing segmentation method to remove other pixels connected to the target bone to separate the target bone. Check the segmentation result. If you are not satisfied with the result, perform the above operation multiple times.

[0058] 4. Set all the target bone pixels in the tomographic image to 1, and all the other background pixels to 0. Use edge detection to select an initial pixel, and then search for the target bone pixels adjacent to the background pixels in the clockwise direction to obtain the pixel set on the outermost edge, thereby obtaining the closed contour line of the target bone, as shown in the attached manual. Figure 2 shown.

[0059] When performing polygonal approximation, a collinearity check is required. Given three points a1(i1, j1), a2(i2, j2), and a3(i3, j3) on the contour line, the length calculation formula of the line segment a1a3 is:

[0060]

[0061] The area calculation formula of Δa1a2a3 is:

[0062]

[0063] The vertical distance from point a2 to line segment a1a3 is:

[0064]

[0065] Through the maximum vertical distance d max To detect the collinearity of each point, d max The judgment of Figure 3 Shown: a series of points N on a given contour line a , N a+1 ,…,N b-1 , N b , calculate the distance from each point in the point sequence to the line segment N a N b The vertical distance d, find the largest vertical distance d max As the approximation error, set the allowable error ω = 0.1. If the maximum vertical distance d max >ω, then the point is not collinear and is defined as N m ; If the maximum vertical distance dmax <ω, then the point is considered to be collinear.

[0066] Assume that the set of pixels on the contour line is {N1, N2, ..., N n}, start from N1 and perform polygon approximation in segments. During the approximation, perform a collinearity check. If it is determined to be collinear, a line segment passing through the beginning and end of the line segment can be used as the edge of the approximation polygon; if it is determined to be non-collinear, start from point N m The segment is split and rechecked as the end of the segment, and the segment is split and checked again until it is determined to be collinear. In this way, the amount of data is reduced and a discrete point set of the target bone contour is obtained.

[0067] 5. Assume that the fitting function is as shown in formula (5):

[0068]

[0069] Among them, Q(i)=(Q1(i), Q2(i),…,Q m (i)) T is the coefficient vector; P(i)=(P1(i), P2(i),…, P m (i)) T is the basis function and has the following form:

[0070] First-order polynomial: P(i) = (1, i, j) T , m = 3

[0071] Quadratic polynomial: P(i) = (1, i, j, i 2 ,ij,j 2 ) T , m = 6

[0072] Define the weighted discrete norm:

[0073]

[0074] Where r is the radius of the support domain; α(|ii z | / r) is the weight function.

[0075] Derivative Q(i) in equation (6) yields:

[0076]

[0077] Q(i)=M -1 (i)N(i)j (7)

[0078] In the formula, N(i)=[α(|i-i1| / r)P(i1), α(|i-i2| / r)P(i2), ..., α(|i-i3| / r)P(i3)].

[0079] Substituting formula (7) into formula (5), we can obtain the expression of the moving least squares method:

[0080]

[0081] In the formula, k is the order of the basis function, is the shape function,

[0082] Formula (8) is used to calculate the node value, and the cubic spline function is used as the compact support weight function. The discrete points of the skeleton contour obtained by polygonal approximation are fitted into a curve, and the following is obtained: Figure 4 The tibia contour line set is shown. Finally, the shortest span method is used as the connection criterion to perform triangular paving connection between two adjacent contour lines.

[0083] According to the fitting curve, the points on the adjacent contour lines are determined to correspond to each other. A triangular face is formed by two points on one contour line and one point on the adjacent contour line. The shortest span method is used to determine whether the two vertices of the same layer on the triangular mesh surface exist on the upper contour line or the lower contour line. That is, the triangular face with the shortest span is selected for paving. Assuming that the point set {D1, D2, ..., D n}, the point set on the upper contour line {U1, U2, ..., U n}, if the pavement has reached point D i , then there are two cases for the next pavement triangle: ΔD i U i+1 U i or ΔD i D i+1 U i , at this time judge and Length, if Then connect ΔD i D i+1 U i ; On the contrary, connect ΔD i U i+1 U i According to this idea, triangle mesh paving is performed between all contour lines to finally obtain the three-dimensional model of the tibia, as shown in Figure 5 shown.

[0084] The above only describes the basic principles and preferred implementations of the method of the present invention. Those skilled in the art may make various changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A bone model reconstruction method based on tomographic medical images, Features The steps include: 1) Obtain DICOM-format tomographic medical images of bones through CT scanning, and set a small scanning slice distance to reduce the workload of subsequent unidirectional linear interpolation between slices; 2) According to the spatial position parameters, each 2D tomographic image is placed in the same 2D coordinate system and arranged in space in sequence according to the CT scanning direction; then, a one-way linear interpolation method is used to interpolate between each tomographic layer to generate new voxels, and finally a 3D data field is formed; 3) Use the grayscale threshold method to segment bones, muscles, epidermis and soft tissues with different grayscale values, and then use the region growing segmentation method to separate the single bone by removing other bone pixels connected to the target bone to obtain an independent target bone; 4) Obtain the pixel point set on the outermost edge through edge detection, so as to obtain the closed contour line of the target bone; then perform polygonal approximation to reduce the number of pixels on the closed contour line, and remove the edge burrs caused by noise, and finally obtain the discrete point set of the target bone contour; 5) Using the moving least squares solution, the cubic spline function is selected as the compact support weighting function, and a curve is fitted to the discrete point set of the bone contour obtained by polygonal approximation. Finally, the shortest span method is used as the connection criterion to connect the two adjacent contour lines with triangles to obtain a three-dimensional bone model.

2. A method for reconstructing a bone model based on tomographic medical images as claimed in claim 1, It is characterized in that Step 2) The unidirectional linear interpolation is used to interpolate between the tomographic medical images to generate a new tomographic layer; the tomographic medical image obtained by scanning is two-dimensional data, and the subsequent segmentation and modeling are all performed on the basis of volume data, so it is necessary to interpolate between each tomographic layer to generate a new voxel, and then arrange them in sequence in space to form a three-dimensional data field; unidirectional linear interpolation is used between two tomographic layers to add voxels to form a new tomographic layer; Let H k (i, j) is the gray value of the pixel at point (i, j) on the k-th layer image, H k+1 (i, j) is the gray value of the pixel at point (i, j) on the k+1 layer image. The gray value of the pixel at point (i, j) on the new fault generated by interpolation between the k layer and the k+1 layer is: H d (i,j)=(1-d)H k (i,j)+d×H k+1 (i,j) (1) Among them, let the interval between the k layer and the interpolation layer be l k , the interval between the k+1 layer and the interpolation layer is l k+1 , d = l k / (l k +l k+1 ).

3. A bone model reconstruction method based on tomographic medical images as claimed in claim 1, It is characterized in that Step 3) Apply the gray threshold method and the region growing segmentation method when performing image segmentation; the gray threshold method uses the different characteristics of gray values ​​of various tissues of the human body to select a suitable threshold range to separate bones from muscles, epidermis and soft tissues. The reference CT value of human bones is greater than 400. The threshold range can be adjusted according to the segmentation result to achieve the best segmentation result; then observe the tomographic image, and then use the region growing segmentation method to remove other bone pixels connected to the target bone layer by layer to independently segment the target bone; check the segmentation result, and if you are not satisfied with the result, perform the above operation multiple times.

4. A bone model reconstruction method based on tomographic medical images as claimed in claim 1, It is characterized in that Step 4) applying edge detection and polygonal approximation to obtain a discrete point set of the target bone contour; edge detection is to set all pixel points of the target bone in the tomographic image to 1, and all other background pixel points to 0, select an initial pixel, and then search for target bone pixels adjacent to the background pixels in a clockwise direction to obtain a pixel point set on the outermost edge, thereby obtaining a closed contour line of the target bone; then polygonal approximation is performed; First, perform a collinearity check; given three points a on the contour line 1 (i 1 , j 1 ), a 2 (i 2 , j 2 ), a 3 (i 3 , j 3 ), line segment a 1 a 3 The length calculation formula is: Δa 1 a 2 a 3 The area calculation formula is: Point a 2 To line segment a 1 a 3 The vertical distance is: Through the maximum vertical distance d max To detect the collinearity of each point: given a string of points N on the contour line a , N a+1 ,…,N b-1 , N b , calculate the distance from each point in the point sequence to the line segment N a N b The vertical distance d, find the largest vertical distance d max As an approximation error, the allowable error ω is specified according to the actual situation. If the maximum vertical distance d max >ω, then the point is not collinear and is defined as N m ; If the maximum vertical distance d max <ω, then the point is considered collinear; Suppose the set of pixels on the contour line is {N 1 , N 2 ,…,N n }, from N 1 Start to approximate the polygon in segments, and perform collinearity check during approximation. If it is determined to be collinear, a line segment passing through the beginning and end of the line segment can be used as the edge of the approximation polygon; if it is determined to be non-collinear, start from point N m Disconnect it and use it as the end of the segment to re-check the collinearity. Continue to split and check until it is determined to be collinear; this reduces the amount of data and obtains a discrete point set of the target bone contour.

5. A method for reconstructing a bone model based on tomographic medical images as claimed in claim 4, It is characterized in that Step 5) Apply the moving least squares method and the shortest span method to fit the curve and surface to obtain a three-dimensional bone model; assume that the fitting function is as shown in formula (5): Where Q(i)=(Q 1 (i), Q 2 (i), …, Q m (i)) T is the coefficient vector; P(i)=(P 1 (i), P 2 (i), …, P m (i)) T is the basis function and has the following form: First-order polynomial: P(i) = (1, i, j) T , m = 3 Quadratic polynomial: P(i) = (1, i, j, i 2 ,ij,j 2 ) T , m = 6 Define the weighted discrete norm: Where r is the radius of the support domain; α(|ii z | / r) is the weight function; Derivative Q(i) in equation (6) yields: Q(i)=M -1 (i)N(i)j (7) wherein, N(i) = [α(|i - i 1 | / r)P(i 1 ) , α(|i - i 2 | / r)P(i 2 ) , … , α(|i - i 3 | / r)P(i 3 )]; Substituting formula (7) into formula (5), we can obtain the expression of the moving least squares method: In the formula, k is the order of the basis function, is the shape function, The node values ​​are calculated using formula (8), and the cubic spline function is used as the tight support weight function to fit a curve to the discrete points of the skeleton contour obtained by polygonal approximation. Finally, the shortest span method is used as the connection criterion to perform triangular paving connection between two adjacent contour lines. According to the fitting curve, the points on the adjacent contour lines are determined to correspond to each other. A triangular face is formed by two points on one contour line and one point on the adjacent contour line. The shortest span method is used to determine whether the two vertices of the same layer on the triangular mesh surface exist on the upper contour line or the lower contour line. That is, the triangular face with the shortest span is selected for paving. Assuming that the point set {D 1 , D 2 , …, D n }, the point set on the upper contour line {U 1 , U 2 , …, U n }, if the pavement has reached point D i , then there are two cases for the next pavement triangle: ΔD i U i+1 U i or ΔD i D i+1 U i , at this time judge and Length, if Then connect ΔD i D i+1 U i ; On the contrary, connect ΔD i U i+1 U i ; According to this idea, triangle mesh pavement is connected between all contour lines to finally obtain the three-dimensional skeleton model.