A method for segmenting a cross-sectional image of a fiber preform

Abstract

The present application relates to a kind of fiber preform cross section image segmentation method, first to the woven composite material physical object CT scanning is carried out, to obtain warp and weft cross section information;Then a certain amount of CT image is selected to make neural network training data set, train deep convolutional neural network to be used for the yarn cross section in CT image segmentation;Again, the yarn cross section after neural network segmentation is identified, and the cross section that occurs overlap is found and cross section classification is carried out;Finally, the overlapping region of overlapping cross section is segmented.The present application is high in segmentation efficiency, and segmentation speed is fast, need not image to have higher resolution and contrast, and the present application uses a variety of yarn separation algorithm to be combined, can be well to overlapping yarn cross section separation, effectively reduces the artificial and time cost before data pre-processing before modeling, and provides convenience for subsequent parameterized modeling.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for segmenting cross-sectional images of fiber preforms. Background Technology

[0002] Textile composite materials are frequently used in high-tech industries due to their high strength-to-weight ratio, excellent mechanical properties, customizability, and versatility. By altering the weaving pattern and fabric structure of woven composite materials, a range of materials with different mechanical properties can be obtained for various applications. However, the different microstructures of textile composite materials significantly influence their mechanical properties, making it difficult to predict their mechanical properties and service life in many applications due to their complex structures. Modeling and simulation of composite textile materials is an effective means of studying their mechanical properties and damage evolution. Model reconstruction is particularly important, especially composite textile material models based on CT image reconstruction, which are frequently used for model reconstruction and analysis due to their high accuracy and adaptability. One of the most crucial steps in the process of material model reconstruction based on CT images is the segmentation of yarn cross-sections in the image.

[0003] Traditional methods for segmenting yarns in CT images primarily rely on manual separation of yarn cross-sections, which is extremely labor-intensive. Other segmentation methods are based on structure tensors and gray-level co-occurrence matrices (GLCMs) of CT image grayscale, but these require high CT image resolution and significant contrast between different phases. Therefore, manual intervention is often necessary in practice, and existing methods cannot separate overlapping cross-sections. Summary of the Invention

[0004] This invention discloses a method for segmenting cross-sectional images of fiber preforms, which can separate overlapping yarn cross-sections and overlay them onto the original CT image, enabling subsequent model reconstruction to be used for reconstruction using more diverse modeling methods.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for segmenting cross-sectional images of fiber preforms, comprising:

[0007] Step 1: Obtain CT images of the warp and weft cross sections of the woven composite material;

[0008] Step 2: Build and train a neural network, use the trained neural network to segment the warp cross-section CT image and the weft cross-section CT image, and display the segmented warp cross-section CT image and weft cross-section CT image separately;

[0009] Step 3: Calculate the width and height of each yarn cross section in all segmented CT images, determine whether there is cross section overlap, and classify the overlapping cross sections. The overlapping cross sections are classified into transverse overlapping cross sections, longitudinal overlapping cross sections, and mixed overlapping cross sections.

[0010] Step 4: Use the erosion operation in image morphology to erode the horizontally overlapping sections and the mixed overlapping sections, record the position information of the overlapping sections, obtain the contour points of the sections based on the projection of the position information onto the original section image, calculate the ratio of the area of ​​the section to the background in the rectangular judgment region between two adjacent contour points to determine the dividing line, and then separate the overlapping sections.

[0011] Step 5: Use column lookup to traverse the vertically overlapping sections to obtain the coordinates of the contour points on each column of the image. Determine the coordinates of the inner concave points in the overlapping area of ​​the sections by judging the position information of the midpoint coordinates between each pair of contour points. Use the ratio of the area of ​​the cross-section background in the circular judgment area distributed on the cross-section contour to determine the coordinates of the outer concave points of the overlapping sections. Draw the dividing line with the obtained inner and outer concave point coordinates, and then separate the overlapping sections.

[0012] Step 6: Integrate the yarn cross-sections after steps 4 and 5 and display them on the original CT image.

[0013] Furthermore, the neural network employs a U-net deep convolutional neural network.

[0014] Furthermore, in step 3, the transverse overlapping section refers to a yarn section that overlaps only in the transverse direction, the longitudinal overlapping section refers to a yarn section that overlaps only in the longitudinal direction, and the mixed overlapping section refers to a yarn section that overlaps in both the transverse and longitudinal directions.

[0015] Furthermore, in step 3, the width and height of the yarn cross-section are calculated according to the following formulas:

[0016] |W|=W max -W min (Formula 1)

[0017] |H|=H max -H min (Formula 2)

[0018] W max =Max{C i x}, W min =Min{C i x}, i = 1 ，2，3 ...n (Formula 3)

[0019] H max =Max{Ci y}, H min =Min{C i y}, i = 1 ，2，3 ...n (Formula 4)

[0020] In the above formula, W is the width, H is the height, and C is the height. i x Let C be the x-coordinate of the cross-sectional profile point C. i y Let y be the coordinate of point C on the cross-section profile, and n be the number of coordinates of the profile points.

[0021] Furthermore, in step 4, erosion operation is first performed according to formula 5 and convolution operation is performed on the image using a convolution kernel. The minimum value of the pixel points in the area covered by the convolution kernel is calculated and assigned to the reference point in the convolution kernel. The position information of the separation line at the cross-section overlap is obtained through multiple iterative operations.

[0022]

[0023] In the above formula, the symbol The erosion operation is represented by A, which is the original image after binarization, B is the convolution kernel, and z is the coordinate information of the z point in A contained in B after B erodes A.

[0024] The obtained separation line is then projected onto the original overlapping section and intersects the section profile at multiple profile points;

[0025] Set a rectangle between the contour points, and use Formula 6 to calculate the proportion of the cross-sectional area in the rectangle. If the calculated proportion is greater than the set first threshold, record the two contour points as the endpoint coordinates of the dividing line. Perform the above process for each dividing line. Finally, connect the endpoints of the dividing lines of the corresponding overlapping areas to separate the overlapping sections.

[0026]

[0027] In the above formula, R sqe A is the ratio of the cross-sectional area to the area of ​​the rectangular frame region. sqe Let p' be the area of ​​the rectangle, and p' be the number of pixels on the cross-section.

[0028] Furthermore, the first threshold is set to 80%.

[0029] Furthermore, in step 5, the contour points of the longitudinally overlapping cross section are found by traversing the image column pixels, and the coordinates of the midpoint located outside the cross section with the two contour points as line segments are calculated by formula 7.

[0030]

[0031] In the above formula, k represents the outline point number of each column's cross-sectional profile, and P j The coordinates of the midpoint between two contour points in column j;

[0032] Then, set the pixel values ​​for the outside and inside of the cross section, and then use Formula 8 to calculate the left and right pixel values ​​of the midpoint coordinates to determine whether the midpoint coordinates are the most concave point at the overlap of the two cross sections. The most concave point is the determined indentation point.

[0033]

[0034] In the above formula, dst(X, Y) are the coordinates of the concave point;

[0035] Furthermore, a circular judgment area is set on the contour of the overlapping section. The area of ​​the circular judgment area and the area of ​​the cross section located within the circular judgment area are calculated. Then, the ratio of the area of ​​the cross section located within the circular judgment area to the area of ​​the circular judgment area is calculated according to Formula 9. If the ratio is greater than the set second threshold, the contour point is recorded. Among the recorded contour points, the contour point with the maximum area ratio is selected as the outer concave point. The concave points determined by the column search and the circular judgment area are connected as the cutting line to separate the overlapping area.

[0036]

[0037] In the above formula, R int A is the ratio of the cross-sectional area to the area of ​​the circular judgment region. int Let p' be the area of ​​the circular region, and C be the number of pixels on the cross-section. i x Let C be the x-coordinate of the cross-sectional profile point C. i y Let y be the coordinate of point C on the cross-section profile.

[0038] Furthermore, the second threshold is set to 70%.

[0039] The segmentation method disclosed in this invention is highly efficient and fast. It first uses a deep convolutional neural network for preliminary cross-sectional segmentation, which does not require the image to have high resolution and contrast, while traditional segmentation methods have high requirements for the resolution and contrast of CT images. Traditional segmentation methods cannot separate overlapping yarn cross-sections, which makes it difficult to extract the boundary contour data of individual yarn cross-sections. However, the segmentation method disclosed in this invention combines multiple yarn separation algorithms to deal with different situations, and can ultimately separate overlapping yarn cross-sections very well. It can be widely applied to the segmentation of cross-sections of CT images of various textile composite materials with similar structures. Attached Figure Description

[0040] Figure 1 This is a flowchart of the segmentation method used in this invention;

[0041] Figure 2 A schematic diagram of a textile composite material to which the segmentation method of the present invention is applied;

[0042] Figure 3 A cross-sectional view of the weft yarn after segmentation using the neural network of this invention;

[0043] Figure 4 This is a cross-sectional view of the warp yarn after segmentation using the neural network of this invention;

[0044] Figure 5 Schematic diagrams of different types of overlapping sections;

[0045] Figure 6 This is a schematic diagram illustrating the separation of laterally overlapping sections;

[0046] Figure 7 This is a schematic diagram illustrating the use of column search to locate indentation points when separating longitudinally overlapping sections;

[0047] Figure 8 This is a schematic diagram illustrating the use of a circular judgment area to locate indentation points when separating longitudinally overlapping sections;

[0048] Figure 9 This is a comparison diagram showing the effect of separating the overlapping cross-section of the weft yarns before and after using the segmentation method of this invention;

[0049] Figure 10 This is a comparison diagram showing the effect of separating the overlapping sections of warp yarns using the segmentation method of the present invention before and after. Detailed Implementation

[0050] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0051] This embodiment discloses a method for segmenting cross-sectional images of fiber preforms, which is targeted at... Figure 2 The design of the cross-sectional image segmentation method for the textile composite material shown is based on the cross-sectional image segmentation of the textile composite material, which is woven from warp and weft yarns. Figure 1 As shown, the main steps include:

[0052] Step 1: Obtain CT images of the warp and weft sections of the woven composite material.

[0053] Specifically, X-ray microtomography was used to... Figure 2 The woven composite material was scanned in both the warp and weft directions to obtain the original warp cross-section CT images and weft cross-section CT images.

[0054] Step 2: Build and train a neural network, use the trained neural network to segment the warp cross-section CT image and the weft cross-section CT image, and display the segmented warp cross-section CT image and weft cross-section CT image separately.

[0055] Specifically, a neural network is first constructed to perform preliminary segmentation on the CT images obtained in step 1, reducing the high requirements for resolution and contrast of the CT images. In this embodiment, a U-net deep convolutional neural network is used. This network consists of a backbone feature extraction network and an enhanced feature extraction network. The backbone feature extraction network uses VGG16 as its backbone, with each layer consisting of a convolutional layer and a max-pooling layer, for a total of 5 layers. The enhanced feature extraction network has a similar structure to the backbone feature extraction network. A portion of the CT images obtained in step 1 is selected as the training dataset for the neural network, and the neural network is trained using this dataset. The trained neural network performs preliminary segmentation on the warp and weft cross-section CT images, respectively. The preliminarily segmented weft cross-section image is shown below. Figure 3 As shown, the cross-sectional diagram of the warp yarns after preliminary segmentation is as follows: Figure 4 As shown.

[0056] Step 3: Calculate the width and height of each yarn cross section in all CT images after preliminary segmentation, determine whether there is cross section overlap, and classify the overlapping cross sections. The overlapping cross sections are classified into transverse overlapping cross sections, longitudinal overlapping cross sections, and mixed overlapping cross sections.

[0057] Specifically, the method for identifying overlapping sections is as follows: On the segmented image obtained after step 2, the coordinates of the contour points of each section are acquired one by one to calculate the width and height information of the section. The width and height of the yarn section are calculated according to Formula 1 and Formula 2, respectively. Based on the comparison between the calculated width and height values ​​and the set threshold, it is determined whether there are overlapping sections.

[0058] |W|=W max -W min (Formula 1)

[0059] |H|=H max -H min (Formula 2)

[0060] W max =Max{C i x}, W min =Min{C i x}, i = 1 ，2，3 ...n (Formula 3)

[0061] H max=Max{C i y}, H min =Min{C i y}, i = 1 ，2，3 ...n (Formula 4)

[0062] In the above formula, W is the width, H is the height, and c i x Let c be the x-coordinate of point C on the cross-sectional profile. i y Let y be the coordinate of point C on the cross-section profile, and n be the number of coordinates of the profile points.

[0063] After identifying the overlapping sections, the overlapping sections are classified and labeled. The overlapping section types in this invention are divided into three categories, namely: Figure 5 The transverse overlapping section shown in Figure A above, such as Figure 5 The longitudinal overlapping section shown in Figure B in the middle, as... Figure 5 The mixed overlapping section is shown in Figure C below. A transverse overlapping section refers to a yarn section that overlaps only in the transverse direction (occurring between warp yarn sections), a longitudinal overlapping section refers to a yarn section that overlaps only in the longitudinal direction (occurring between warp yarn sections and weft yarn sections), and a mixed overlapping section refers to a yarn section that overlaps in both the transverse and longitudinal directions (there is overlap between warp yarn sections and overlap between warp yarn sections and weft yarn sections).

[0064] Step 4: Divide the transversely overlapping sections. For the sections that overlap transversely in the mixed overlapping sections, the division should also be carried out in accordance with Step 4.

[0065] Specifically, the erosion operation in image morphology is used to erode the horizontally overlapping sections and the mixed overlapping sections, and the position information of the overlapping sections is recorded. The contour points of the sections are obtained by projecting the position information onto the original section image. The ratio of the area of ​​the section to the background in the rectangular judgment region between two adjacent contour points is calculated to determine the dividing line, and then the overlapping sections are separated.

[0066] Combination Figure 6 As shown, the division of the transversely overlapping section is explained in more detail below: Figure 6 Image A in the diagram shows the horizontal overlapping cross-section before segmentation. First, erosion is performed according to Formula 5, and then convolution is applied to the image using a convolution kernel. The minimum pixel value in the area covered by the convolution kernel is calculated and assigned to a reference point in the kernel. Through multiple iterative calculations, the location information of the separation line at the overlapping section is obtained, such as... Figure 5 The location information of the three separation lines C-1, C-2, and C-3 can be determined as shown in Figure B.

[0067]

[0068] In the above formula, the symbol The erosion operation is represented by A, which is the original image after binarization, B is the convolution kernel, and z is the coordinate information of the z point in A contained in B after B erodes A.

[0069] Then, the three separation lines obtained above are projected onto the original overlapping section (e.g., Figure 5 (As shown in Figure C) and intersects the cross-sectional profile at multiple profile points, so as Figure 5 Taking the separation line C-1 in Figure C as an example, the separation line intersects the cross-sectional profile at four profile points: a, b, c, and d (e.g., ...). Figure 5 (As shown in Figure D). A rectangular frame is set between the contour points. Formula 6 is used to calculate the proportion of the cross-sectional area within the rectangular frame. If the calculated proportion is greater than the set first threshold (the first threshold is set to 80% in this embodiment), the two contour points are recorded as the endpoint coordinates of the dividing line. The above process is performed for each dividing line. Finally, the endpoints of the dividing lines of the corresponding overlapping areas are connected to separate the overlapping cross-sections.

[0070]

[0071] In the above formula, R sqe A is the ratio of the cross-sectional area to the area of ​​the rectangular frame region. sqe p is the area of ​​the rectangle, and p' is the number of pixels on the cross section (one pixel occupies a certain area).

[0072] Step 5: Divide the longitudinally overlapping sections. For the longitudinally overlapping sections in the mixed overlapping sections, the division should also be carried out in accordance with Step 5.

[0073] Specifically, a column search method is used to traverse the longitudinally overlapping sections to obtain the coordinates of the contour points on each column of the image. The position information of the midpoint coordinates between each pair of contour points is used to determine the coordinates of the inner concave points in the overlapping area of ​​the section. The ratio of the area of ​​the cross-section background in the circular judgment area distributed on the cross-section contour is used to determine the coordinates of the outer concave points of the overlapping section. The obtained inner and outer concave point coordinates are used to draw the dividing line, and then the longitudinally overlapping sections are separated.

[0074] Combination Figure 7 A detailed explanation of the column search method for finding concave points is provided: For the longitudinally overlapping cross section, the cross section contour points are found by traversing the image column pixels. The coordinates of the midpoint located outside the cross section with the two contour points as line segments are calculated using Formula 7.

[0075]

[0076] In the above formula, k represents the outline point number of each column's cross-sectional profile, and P jLet be the coordinates of the midpoint between two contour points in column j.

[0077] Then, the pixel values ​​for the outside and inside of the cross-section are set. In this embodiment, the pixel value for the outside of the cross-section is 0, and the pixel value for the inside of the cross-section is 255. Then, the left and right pixel values ​​of the midpoint coordinates are calculated using Formula 8 to determine whether the midpoint coordinates are the most concave point at the overlap of the two cross-sections. This most concave point is the determined indentation point. For example... Figure 7 As shown in the figure, the dashed line is the midpoint line formed by the discrete midpoint coordinates, and a, b, c, and d are the concave points of the longitudinally overlapping cross section.

[0078]

[0079] In the above formula, dst(X, Y) are the coordinates of the concave point.

[0080] Combination Figure 8 A detailed explanation of using a circular judgment area method to find depression points: When the coordinates of the depression point cannot be determined using the above method, a circular judgment area (e.g., ...) is set on the contour of the longitudinally overlapping section. Figure 8 The circle in the diagram is used to calculate the area of ​​the circular judgment region and the cross-sectional area within the circular judgment region. Then, the ratio of the cross-sectional area within the circular judgment region to the area of ​​the circular judgment region is calculated according to Formula 9. If the ratio is greater than the set second threshold (the second threshold in this embodiment is 70%), the contour point is recorded. Among the recorded contour points, the contour point with the largest area ratio is selected as the outer concave point. The concave points determined by the column search and the circular judgment region are connected as the cutting line to separate the longitudinal overlap.

[0081]

[0082] In the above formula, R int A is the ratio of the cross-sectional area to the area of ​​the circular judgment region. int Let p' be the area of ​​the circular region, p' be the number of pixels on the cross-section (one pixel occupies a certain area), and C be the area of ​​the circular region. i x Let C be the x-coordinate of the cross-sectional profile point C. i y Let y be the coordinate of point C on the cross-section profile.

[0083] Step 6: Finally, the yarn cross-sections after steps 4 and 5 are integrated and displayed on the original CT image, providing favorable conditions for subsequent yarn modeling of textile composite materials.

[0084] The effect diagram after segmenting the overlapping section of the weft yarns according to the segmentation method of the present invention is shown below. Figure 9 The effect diagram after segmenting the warp yarn overlap section according to the segmentation method of the present invention is shown in the figure. Figure 10The upper image in both figures is the CT image before segmentation, and the lower image is the result of segmentation using the method of the present invention. From the comparison of the two images before and after segmentation, it can be seen that there is cross-section overlap before segmentation, while the overlapping cross-sections have been separated after segmentation, which verifies that the segmentation method of the present invention is very effective.

[0085] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for segmenting cross-sectional images of a fiber preform, characterized in that, include: Step 1: Obtain CT images of the warp and weft cross sections of the woven composite material; Step 2: Build and train a neural network, use the trained neural network to segment the warp cross-section CT image and the weft cross-section CT image, and display the segmented warp cross-section CT image and weft cross-section CT image separately; Step 3: Calculate the width and height of each yarn cross section in all segmented CT images, determine whether there is cross section overlap, and classify the overlapping cross sections. The overlapping cross sections are classified into transverse overlapping cross sections, longitudinal overlapping cross sections, and mixed overlapping cross sections. Step 4: Use the erosion operation in image morphology to erode the horizontally overlapping sections and the mixed overlapping sections, record the position information of the overlapping sections, obtain the contour points of the sections based on the projection of the position information onto the original section image, calculate the ratio of the area of ​​the section to the background in the rectangular judgment region between two adjacent contour points to determine the dividing line, and then separate the overlapping sections. Step 5: Use column lookup to traverse the vertically overlapping sections to obtain the coordinates of the contour points on each column of the image. Determine the coordinates of the inner concave points in the overlapping area of ​​the sections by judging the position information of the midpoint coordinates between each pair of contour points. Use the ratio of the area of ​​the cross-section background in the circular judgment area distributed on the cross-section contour to determine the coordinates of the outer concave points of the overlapping sections. Draw the dividing line with the obtained inner and outer concave point coordinates, and then separate the overlapping sections. Step 6: Integrate the yarn cross-sections after steps 4 and 5 and display them on the original CT image.

2. The method for segmenting a cross-sectional image of a fiber preform according to claim 1, characterized in that: The neural network used is a U-net deep convolutional neural network.

3. The method for segmenting a cross-sectional image of a fiber preform according to claim 1, characterized in that: In step 3, the transverse overlapping section refers to a yarn section that overlaps only in the transverse direction, the longitudinal overlapping section refers to a yarn section that overlaps only in the longitudinal direction, and the mixed overlapping section refers to a yarn section that overlaps in both the transverse and longitudinal directions.

4. The method for segmenting a cross-sectional image of a fiber preform according to claim 1, characterized in that: In step 3, the width and height of the yarn cross-section are calculated using the following formulas: |W| = W max -W min (Official 1) |H|=H max -H min (Formula 2) W max = Max{Ci x}, W min = Min{Ci x}, i = 1, 2, 3...n (Formula 3) H max = Max{Ci y}, H min = Min{Ci y}, i = 1, 2, 3...n (Formula 4) In the above formula, W is the width, H is the height, and c i x Let c be the x-coordinate of point C on the cross-sectional profile. i y Let y be the coordinate of point C on the cross-section profile, and n be the number of coordinates of the profile points.

5. The method for segmenting a cross-sectional image of a fiber preform according to claim 1, characterized in that: In step 4, erosion operation is first performed according to formula 5 and convolution operation is performed on the image using convolution kernel. The minimum value of the pixel points in the area covered by the convolution kernel is calculated and assigned to the reference point in the convolution kernel. The position information of the separation line at the cross-section overlap is obtained through multiple iterative operations. In the above formula, the symbol The erosion operation is represented by A, which is the original image after binarization, B is the convolution kernel, and z is the coordinate information of the z point in A contained in B after B erodes A. The obtained separation line is then projected onto the original overlapping section and intersects the section profile at multiple profile points; Set a rectangle between the contour points, and use Formula 6 to calculate the proportion of the cross-sectional area in the rectangle. If the calculated proportion is greater than the set first threshold, record the two contour points as the endpoint coordinates of the dividing line. Perform the above process for each dividing line. Finally, connect the endpoints of the dividing lines of the corresponding overlapping areas to separate the overlapping sections. In the above formula, R sqe A is the ratio of the cross-sectional area to the area of ​​the rectangular frame region. sqe Let p' be the area of ​​the rectangle, and p' be the number of pixels on the cross-section.

6. The method for segmenting a cross-sectional image of a fiber preform according to claim 5, characterized in that: The first threshold is set to 80%.

7. The method for segmenting a cross-sectional image of a fiber preform according to claim 1, characterized in that: In step 5, the contour points of the longitudinally overlapping cross section are found by traversing the image column pixels, and the coordinates of the midpoint located outside the cross section with the two contour points as line segments are calculated by formula 7. In the above formula, k represents the outline point number of each column's cross-sectional profile, and P j The coordinates of the midpoint between two contour points in column j; Then, set the pixel values ​​for the outside and inside of the cross section, and then use Formula 8 to calculate the left and right pixel values ​​of the midpoint coordinates to determine whether the midpoint coordinates are the most concave point at the overlap of the two cross sections. The most concave point is the determined indentation point. In the above formula, dst(X, Y) are the coordinates of the concave point; Furthermore, a circular judgment area is set on the contour of the overlapping section. The area of ​​the circular judgment area and the area of ​​the cross section located within the circular judgment area are calculated. Then, the ratio of the area of ​​the cross section located within the circular judgment area to the area of ​​the circular judgment area is calculated according to Formula 9. If the ratio is greater than the set second threshold, the contour point is recorded. Among the recorded contour points, the contour point with the maximum area ratio is selected as the outer concave point. The concave points determined by the column search and the circular judgment area are connected as the cutting line to separate the overlapping area. In the above formula, R int A is the ratio of the cross-sectional area to the area of ​​the circular judgment region. int Let p' be the area of ​​the circular region, and C be the number of pixels on the cross-section. i x Let C be the x-coordinate of the cross-sectional profile point C. i y Let y be the coordinate of point C on the cross-section profile.

8. The method for segmenting a cross-sectional image of a fiber preform according to claim 7, characterized in that: The second threshold is set to 70%.

Images