An intelligent segmentation method for enhancing images of basalt fiber composites with defects

Abstract

The present invention relates to an intelligent segmentation method for enhancing images of defective basalt fiber composites, comprising the following steps: 1), collecting XCT digital images, randomly selecting a part of the images as the real image set, annotating the yarns, cracks, and bubbles in the images to generate a label set corresponding to the real image set; 2), selecting multiple vertices of the real yarn contour for polygon smooth fitting to generate a random virtual yarn contour, placing it in the pseudo-label image according to the real yarn arrangement, and then adding bubble and crack features to obtain a pseudo-label set; 3), using the DP-GAN style conversion network to convert the pseudo-labels into pseudo-images, and then using the Palette network model for yarn edge reconstruction; 4), mixing the real training set and the pseudo-training set, inputting them into the Mask DINO segmentation network for training, and after the training is completed, using the training weights to perform inference on the original XCT image to obtain the segmentation result. The present invention can enhance image data, effectively improve the network segmentation ability, and save the manual annotation cost.

Description

Technical Field

[0001] The present invention relates to an intelligent segmentation method for enhancing images of basalt fiber composites with defects, which can complete automatic data enhancement of features such as yarns, bubbles, and fissures in three-dimensional braided basalt fiber composite images, and belongs to the field of image processing technology. Background Art

[0002] Three-dimensional braided basalt fiber composite is a new type of high-performance composite material composed of basalt fibers as reinforcement and resin as matrix through a three-dimensional braiding process.

[0003] However, in the actual application process, the microstructure of three-dimensional braided basalt fiber composites is complex. In order to more comprehensively understand the relationship between the mesoscopic structural damage evolution and macroscopic mechanical properties inside the composite material, it is necessary to use deep learning segmentation algorithms to spatially divide each phase inside the composite material. However, it is costly and difficult to obtain a large number of labeled data. Therefore, it has important theoretical and practical significance to perform data enhancement based on real datasets and improve the segmentation ability of each internal phase.

[0004] Therefore, to solve the above technical problems, it is indeed necessary to provide an innovative intelligent segmentation method for enhancing images of basalt fiber composites with defects to overcome the defects in the prior art. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent segmentation method for enhancing images of basalt fiber composites with defects. It generates a pseudo-label set from a partial real image set obtained by XCT technology, and then combines a DP-GAN style transfer network and a Palette repair network to generate a pseudo-image set with smooth transition features, thereby realizing data enhancement of the XCT image segmentation dataset of three-dimensional braided basalt fiber composites and achieving precise segmentation at the single yarn level of fiber composites.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is: an intelligent segmentation method for enhancing images of basalt fiber composites with defects, which includes the following steps:

[0007] 1), Collect digital images of basalt fiber composites through XCT scanning technology, construct an original dataset, screen based on these original data, select a part of them as a real image set, and apply different colors to the yarn area, fissure area, and bubble area in the real images for manual annotation to generate a real label set corresponding to the real image set;

[0008] 2), Based on the above-mentioned real label set, select six vertices in each contour of the yarn to form a polygon, and smoothly fit to generate a pseudo-yarn contour; then randomly place it on a blank label to form a pseudo-label image; then randomly select points to generate ellipses of different shapes and angles to simulate initial bubbles; finally, use Perlin noise to generate pseudo-cracks on the yarn cross-section, and then export the pseudo-label set;

[0009] 3), Use the DP-GAN style conversion network to convert the pseudo-label image into a pseudo-image, and then use the Palette network model to reconstruct the yarn edge to obtain a set of pseudo-images after yarn edge reconstruction;

[0010] 4), Combine the real image set and the real label set to form a real training set, combine the pseudo-image set and the pseudo-label set to form a pseudo-training set, and put the two training sets into the Mask DINO segmentation network for training; after the training is completed, use the.pth weight file of the trained network model to perform inference on the XCT original data set, divide the yarn area, crack area, and bubble area in the image, and attach a color mask to obtain the segmentation result.

[0011] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: The specific steps of step 1) are as follows:

[0012] 1-1), Collect basalt fiber composite samples through XCT scanning, obtain XCT original data, and select some tomography data to construct a real image set;

[0013] 1-2), In the real image set, manually annotate the yarn area, crack area, and bubble area with different colors. The colors can be randomly selected, and the same color can be used for the same type of area, so as to generate a real marked image set;

[0014] 1-3), The smeared areas in the above real marked image set become connected areas with the same gray value. The connected areas with different features express different gray values. Use Boolean operations in the OpenCV library to extract the marks, and reassign the different feature gray values, convert the 24-bit to 8-bit gray image, and obtain a real label set containing matrix, yarn, crack, and bubble features. The image is an 8-bit.png file, and the gray scale standard is: background: 0, matrix: 1, yarn: 2, crack: 3, bubble: 4. Different gray areas in the image represent different feature divisions.

[0015] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: In step 1-1), the basalt fiber composite is specifically a composite material fabricated by three-dimensional four-way four-step braiding and then encapsulated by the VARTM process; the yarn used for braiding is composed of 1200Tex basalt fiber bundles.

[0016] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: in the step 1-1), the XCT raw data is stored in the.raw format with a size of 16 bits; then, in the Avizo software, the 16-bit data is reduced to 8 bits, and after rotation correction, cropping, and contrast enhancement, 1350 8-bit.tiff-format tomography raw data are obtained, and 111 data are randomly selected to construct a.tiff-format real image set.

[0017] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: the step 2) is specifically as follows:

[0018] 2-1), taking the real yarn contour as a template, selecting the upper and lower two vertices and the left and right four vertices to make a polygon, and smoothly fitting to generate a random virtual yarn contour;

[0019] 2-2), putting the randomly generated virtual yarn contour into a blank label to form a pseudo-label image;

[0020] 2-3), using the OpenCV library to generate ellipses of different sizes, angles, and shapes to simulate the bubbles in the real sample, and randomly putting them into the virtual yarn contour avoided in the pseudo-label image in step 2-2);

[0021] 2-4), using the Perlin noise library to generate virtual cracks of different lengths, curvatures, and branch numbers, and putting them into the virtual yarn contour in the pseudo-label image; finally, a pseudo-label set containing yarns, bubbles, and cracks is obtained, where the gray value standard is the same as that of the real label set.

[0022] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: in the step 2-1), according to the six vertex coordinates, they are concatenated to obtain a closed hexagon, noise points are added around the hexagon connection line, and the noise point perturbation amplitude is controlled, that is, the amplitude of the noise points around the hexagon, and the Bezier curve is used to fit and smooth the noise point connection line to obtain a smooth virtual yarn contour without sharp corners.

[0023] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention is further as follows: the step 3) is specifically as follows:

[0024] 3-1), using the DP-GAN style conversion network to convert the pseudo-label image into a pseudo-image;

[0025] 3-2), using the images after dilation and erosion of the pseudo-label set to perform contour Boolean subtraction operation to obtain a pseudo-edge mask image;

[0026] 3-3), use the Palette network model to reconstruct the yarn edges of the pseudo-images; the Palette network detects and optimizes the edges in the pseudo-images through multi-scale feature extraction and fusion, and finally generates a set of pseudo-images after yarn edge reconstruction.

[0027] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention further includes: the specific process of the yarn edge reconstruction is as follows: cover the pseudo-images with the pseudo-edge masks described in 3-2), and the Palette network model regenerates the covered yarn edge regions according to the features of the pseudo-images.

[0028] The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention may also be: the specific process of step 4) is as follows:

[0029] 4-1), form a real training set with the real image set and the real label set, form a pseudo-training set with the pseudo-image set and the pseudo-label set, and put the two training sets into the Mask DINO segmentation network for training. The network divides the feature regions in the corresponding images according to the guidance of the label markings, and then calculates the loss function by comparing the segmentation results with the label images in terms of regions; through continuous periodic iteration, finally obtain a segmentation model weight file with higher accuracy;

[0030] 4-2), after the training is completed, use the trained network model weight.pth file to perform inference on the entire.png format XCT original image, divide different regions in the original image, and mark them as color masks to obtain the segmentation results.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention uses different deep learning networks such as the DP-GAN style conversion network and the Palette network, which can automatically generate a pseudo-training set for data augmentation, realize the accurate segmentation at the single yarn level of fiber composites, provide data support for subsequent simulation and improvement of various properties of composites by methods such as finite element analysis, optimize the composite formula and process, greatly improve the three-dimensional model accuracy of composites, and has good generalization ability and application prospects.

[0033] 2. The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention introduces the DP-GAN style conversion network and the Palette repair network, which can effectively alleviate the problems of difficult annotation and high production cost of the XCT image segmentation data set of three-dimensional braided basalt fiber composites.

[0034] 3. The intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention uses the MaskDINO segmentation network for training, which can effectively improve the network segmentation ability. Description of the Drawings

[0035] Figure 1 is the flowchart of the intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention.

[0036] Figure 2 is the schematic diagram of the virtual contour generated by polygon smoothing fitting according to the vertexes of the yarn contour in the real image in step 2) of the present invention.

[0037] Figure 3 is the working principle diagram of the DP-GAN style conversion network in step 3) of the present invention.

[0038] Figure 4 is the working principle diagram of the Palette network in step 3) of the present invention.

[0039] Figure 5 is the gray-scale comparison diagram of the pseudo-image and the real image generated in the present invention.

[0040] Figure 6 is the XCT original image obtained by the XCT scanning technology of the present invention.

[0041] Figure 7 is the image after data augmentation and re-segmentation by the present invention. Detailed Embodiments

[0042] Please refer to the attached Figure 1 shown in the specification, which is an intelligent segmentation method for enhancing the images of defective basalt fiber composites of the present invention, and includes the following steps:

[0043] 1), Collect digital images of basalt fiber composites through XCT scanning technology, as shown in the attached Figure 6 shown in the specification, and construct an original data set. Based on these original data, screen and select a part of them as the real image set, and apply different colors to the yarn area, crack area, and bubble area in the real image for manual annotation to generate a real label set corresponding to the real image set.

[0044] The basalt fiber composite is specifically a composite material fabricated by three-dimensional four-directional four-step weaving and encapsulated by the VARTM process. The yarn used for weaving is composed of 1200Tex basalt fiber bundles.

[0045] Specifically, the step 1) is implemented in the following manner:

[0046] 1-1), The basalt fiber composite material samples are collected by XCT scanning to obtain the original XCT data, and a part of the tomographic data is selected to construct a real image set.

[0047] XCT (X-ray Computed Tomography), namely X-ray computed tomography technology, is a three-dimensional imaging technology that uses X-rays to penetrate an object and reconstruct its internal structure by a computer. Through XCT scanning, the microstructure of the three-dimensional braided basalt fiber composite material can be obtained, including fiber distribution, fiber orientation, and existing defects (fissures, bubbles), etc.

[0048] In this embodiment, the original XCT data is stored in the.raw format with a size of 16 bits; then the 16-bit data is reduced to 8 bits in Avizo software, and after rotation correction, cropping, and contrast enhancement, 1350 pieces of 8-bit tomographic original data in the.png format are obtained, and 111 pieces of data are randomly selected to construct a real image set in the.png format.

[0049] 1-2), In the real image set, different colors are used for manual annotation of the yarn area, fissure area, and bubble area. The colors can be randomly selected, and the same color is used for the same type of area, so as to generate a real marked image set.

[0050] 1-3), The smeared areas in the above real marked image set become connected areas with the same gray value, and the connected areas with different features express different gray values. The marker extraction is carried out on it using Boolean operations in the OpenCV library, and different feature gray values are re-assigned. Specifically, the 24-bit is converted into an 8-bit grayscale image to obtain a real label set containing matrix, yarn, fissure, and bubble features. The image is an 8-bit.png file, and the gray scale standard is: background: 0, matrix: 1, yarn: 2, fissure: 3, bubble: 4. Different gray areas in the image represent different feature divisions.

[0051] 2), Based on the above real label set, six vertices are selected from each contour of the yarn to make a polygon, and the pseudo-yarn contour is generated by smooth fitting. Then the pseudo-yarn contour is randomly placed on a blank label to form a pseudo-label image. Then randomly selected points are used to generate ellipses with different shapes and angles to simulate initial bubbles. Finally, pseudo-fissures are generated on the yarn cross-section using Perlin noise, and then the pseudo-label set is exported.

[0052] The specific steps of step 2) are as follows:

[0053] 2-1), As shown in the attached instructions Figure 2As shown, taking the real yarn contour as a template, select the upper and lower two vertices and the left and right four vertices to form a polygon, and smoothly fit to generate a random virtual yarn contour. Specifically, according to the coordinates of the six vertices, a closed hexagon is obtained after concatenation. After adding noise points around the hexagon connection lines, controlling the noise perturbation amplitude, that is, the amplitude of the noise around the hexagon, can make the contour shape more diverse. However, too large a perturbation amplitude will cause the virtual yarn contour to be distorted, and too small will result in a higher similarity between the generated virtual yarn contours. Then, use Bezier curves to fit and smooth the noise connection lines to obtain a smooth virtual yarn contour without sharp corners.

[0054] (2-2), Put the randomly generated virtual yarn contour into a blank label to form a pseudo-label image.

[0055] (2-3), Use the OpenCV library to generate ellipses of different sizes, angles, and shapes to simulate the bubbles in the real sample, and randomly place them avoiding the virtual yarn contours in the pseudo-label image obtained in step (2-2).

[0056] (2-4), Use the Perlin noise library to generate virtual cracks of different lengths, curvatures, and branch numbers, and place them in the virtual yarn contours of the pseudo-label image; finally, obtain a pseudo-label set containing yarns, bubbles, and cracks, where the gray value standard is the same as that of the real label set.

[0057] (3), Use the DP-GAN style transfer network to convert the pseudo-label image into a pseudo-image, and then use the Palette network model to reconstruct the yarn edge to obtain a set of pseudo-images after yarn edge reconstruction.

[0058] Among them, as shown in the attached instructions Figure 3 The DP-GAN (Blended Diffusion Model) style transfer network is an image style transfer model based on the generative adversarial network (GAN). This model can convert the pseudo-label image in the three-dimensional braided basalt fiber composite material into a pseudo-image with the style of a real XCT image.

[0059] This step is specifically as follows:

[0060] (3-1), Use the DP-GAN style transfer network to convert the pseudo-label image into a pseudo-image. The DP-GAN network retains the yarn, bubble, and crack structures in the pseudo-label during the conversion process, and at the same time integrates the style features of the real XCT image to generate a pseudo-image similar to the real XCT image.

[0061] (3-2), Use the dilated and eroded images of the pseudo-label set to perform contour Boolean subtraction operations to obtain a pseudo-edge mask image.

[0062] (3-3), As shown in the attached instructions Figure 4As shown, the Palette network model is used to reconstruct the yarn edges of the pseudo-images. The Palette inpainting network is a deep learning-based image inpainting technique designed to address the problem of uneven transitions at the yarn interfaces in the pseudo-images generated by DP-GAN, making the transitions from yarns to other features in the pseudo-images more natural and closer to real images. The Palette network detects and optimizes the edges in the pseudo-images through multi-scale feature extraction and fusion, and finally generates a set of pseudo-images with reconstructed yarn edges, as shown in the appended Figure 5 As shown, it is the grayscale comparison between the pseudo-image in the pseudo-image set and the real image. It can be seen that the grayscale distributions of the real and pseudo-images highly overlap, proving the high verisimilitude of the pseudo-images.

[0063] Among them, the yarn edge reconstruction is specifically as follows: The pseudo-image is covered with the pseudo-edge mask described in 3-2), and the Palette network model regenerates the covered yarn edge area according to the pseudo-image features. Through the above edge reconstruction, the problem of uneven transitions in the interface areas between different features in the pseudo-images is solved.

[0064] 4), The real image set and the real label set are combined to form a real training set, and the pseudo-image set and the pseudo-label set are combined to form a pseudo-training set. The two training sets are combined to form a mixed training set and input into the Mask DINO segmentation network for training; after the training is completed, the.pth weight file of the trained network model is used to perform inference on the XCT original data set, divide the yarn areas, crack areas, and bubble areas in the images, and attach color masks to obtain the segmentation results, that is, the images after data augmentation and re-segmentation, as shown in the appended Figure 7 As shown.

[0065] The specific steps of step 4) are as follows:

[0066] 4-1), The real image set and the real label set are combined to form a real training set, and the pseudo-image set and the pseudo-label set are combined to form a pseudo-training set. The two training sets are combined to form a mixed training set and input into the Mask DINO segmentation network for training. The network divides the feature areas in the corresponding images according to the guidance of the labels, and then calculates the loss function by comparing the segmentation results with the label images in terms of regions; through continuous periodic iteration, a segmentation model weight file with relatively high accuracy is finally obtained.

[0067] 4-2), After the training is completed, the entire.png-format XCT original image is inferred using the trained network model weight.pth file, the different regions in the original image are divided, and marked with color masks to obtain the segmentation results.

[0068] In summary, the intelligent segmentation method for enhancing the images of defective basalt fiber composites in the present invention obtains the images of relevant fiber composites through XCT scanning, obtains the real image set, and extracts the real label images by using algorithms after manual marking; subsequently, the yarn features of the real label images are used to fit the virtual yarn contour by using algorithms, and the Berlin noise is added to the cracks and the simulated bubble features are added by randomly placing ellipses to obtain the pseudo-label set; then, the DP-GAN style conversion network is used to convert the pseudo-label set into a pseudo-image set, and the yarn edge reconstruction is carried out on the yarn cross-section area in the pseudo-image set through the Palette repair network to obtain the pseudo-image set after yarn edge reconstruction; finally, the real image set and the real label set form the real training set, the pseudo-image set and the pseudo-label set form the pseudo-training, and the real training set and the pseudo-training set are combined into a mixed training set and input into the MaskDINO segmentation network for training. After the training is completed, the model weights are obtained, and the three-dimensional reconstruction model after segmentation is obtained by using the weights to perform inference segmentation on the XCT original data.

[0069] The intelligent segmentation method for enhancing the images of defective basalt fiber composites in the present invention can automatically generate a pseudo-training set for data augmentation by combining different deep learning networks, and introduces the DP-GAN style conversion network and the Palette repair network. These networks can effectively alleviate the problems of difficult annotation and high production cost of the XCT image segmentation dataset of three-dimensional braided basalt fiber composites, and achieve accurate segmentation at the single yarn level of fiber composites, providing data support for subsequent simulation and improvement of various properties of composites by combining methods such as finite element analysis, and optimizing the composite formula and process. It greatly improves the three-dimensional model accuracy of composites and has good generalization ability and application prospects.

[0070] The above specific embodiments are only the preferred embodiments of this creation, and are not used to limit this creation. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this creation shall be included within the protection scope of this creation.

Claims

1. An intelligent segmentation method for enhancing defective basalt fiber composite images, characterized by: The steps include: 1) Using XCT scanning technology to collect digital images of basalt fiber composites, constructing an original data set, screening based on these original data, selecting a part of them as a real image set, painting different colors on the yarn area, crack area, and bubble area in the real image for manual annotation, and generating a real label set corresponding to the real image set; 2), based on the above-mentioned real label set, six vertices are selected from each contour of the yarn to form a polygon, and a pseudo yarn contour is generated by smooth fitting; Then, random placement is performed on a blank label to form a pseudo-label image; then, ellipses of different shapes and angles are randomly selected to simulate initial bubbles; finally, pseudo cracks are generated on the yarn cross section using Perlin noise, and then a pseudo-label set is derived; Specifically: 2-1), using the real yarn profile as a template, selecting the upper and lower two vertices and the left and right four vertices as polygons, and smoothly fitting to generate a random virtual yarn profile; that is, according to the coordinates of the six vertices, a closed hexagon is obtained after being connected in series, and after adding noise points around the hexagonal lines, the noise disturbance amplitude is controlled, that is, the amplitude of the noise points around the hexagon, and the noise point lines are fitted and smoothed using the Bezier curve to obtain a smooth virtual yarn profile without edges and corners; 2-2), placing the randomly generated virtual yarn profile into a blank label to form a pseudo label image; 2-3), using the OpenCV library to generate ellipses of different sizes, angles, and shapes to simulate bubbles in real samples, and randomly placing them into the pseudo-label image of step 2-2) to avoid the virtual yarn contour; 2-4), using the Perlin noise library to generate virtual cracks of different lengths, curvatures, and branch numbers, and placing them in the virtual yarn contour of the pseudo-label image; finally, a pseudo-label set containing yarn, bubbles, and cracks is obtained, in which the gray value standard is the same as the real label set; 3) The pseudo-label image is converted into a pseudo-image using the DP-GAN style transfer network, and then the yarn edge is reconstructed using the Palette network model to obtain a pseudo-image set after the yarn edge is reconstructed; 4) The real image set and the real label set are combined into a real training set, the pseudo image set and the pseudo label set are combined into a pseudo training set, and the two training sets are combined into a mixed training set and put into the Mask DINO segmentation network for training; After the training, the .pth weight file of the trained network model is used to infer the XCT original data set, divide the yarn area, crack area, and bubble area in the image, and add a color mask to obtain the segmentation result, which is as follows: 4-1), the real image set and the real label set are combined into a real training set, the pseudo image set and the pseudo label set are combined into a pseudo training set, and the two training sets are combined into a mixed training set and put into the Mask DINO segmentation network for training. The network divides the feature area in the corresponding image according to the label guidance, and then compares the segmentation result with the label image to calculate the loss function; through continuous cycle iteration, a segmentation model weight file with high accuracy is finally obtained; 4-2), after the training is completed, the trained network model weight .pth file is used to infer the entire .png format XCT original image, divide the different areas in the original image, and mark them into color masks to obtain the segmentation result.

2. The intelligent segmentation method of reinforced defective basalt fiber composite material image according to claim 1, characterized in that: The step 1) is specifically as follows: 1-1), collect basalt fiber composite material samples through XCT scanning, obtain XCT raw data, and select part of the tomography data to construct a real image set; 1-2), in the real image set, the yarn area, crack area, and bubble area are manually labeled with different colors, and the colors can be randomly selected, and the same type of area can use the same color, thereby generating a real labeled image set; 1-3), the smeared areas in the above-mentioned real labeled image set become connected areas with the same grayscale value. Connected areas with different features express different grayscale values. The Boolean operation in the OpenCV library is used to extract the labels, and the grayscale values ​​of different features are reassigned. The 24-bit image is converted into an 8-bit grayscale image to obtain a real label set containing matrix, yarn, crack, and bubble features. The image is an 8-bit .png file, and the grayscale standard is: background: 0, matrix: 1, yarn: 2, crack: 3, bubble:

4. Different grayscale areas in the image represent different feature divisions.

3. The intelligent segmentation method of reinforced defective basalt fiber composite material image according to claim 2, characterized in that: In the step 1-1), the basalt fiber composite material is specifically a composite material produced by three-dimensional four-way four-step weaving and then encapsulating by VARTM process; the yarn used for weaving is composed of 1200Tex basalt fiber bundles.

4. The intelligent segmentation method of reinforced defective basalt fiber composite material image according to claim 2, characterized in that: In the step 1-1), the XCT raw data is stored in .raw format and 16 bits in size; then the 16-bit data is reduced to 8 bits in Avizo software, and after rotation correction, cropping, and contrast enhancement, 1350 8-bit .png format tomography raw data are obtained, and 111 data are randomly selected to construct a .png format real image set.

5. The intelligent segmentation method of reinforced defective basalt fiber composite material image according to claim 1, characterized in that: The step 3) is specifically as follows: 3-1), using the DP-GAN style transfer network to convert pseudo-label images into pseudo-images; 3-2), using the pseudo label set after expansion and erosion of the image to perform contour Boolean subtraction operation to obtain a pseudo edge mask image; 3-3), use the Palette network model to reconstruct the yarn edges of the pseudo image; the Palette network detects and optimizes the edges in the pseudo image through multi-scale feature extraction and fusion, and finally generates a set of pseudo images after yarn edge reconstruction.

6. The intelligent segmentation method of reinforced defective basalt fiber composite material image according to claim 5, characterized in that: The yarn edge reconstruction is specifically as follows: the pseudo image is covered with the pseudo edge mask described in 3-2), and the Palette network model regenerates the masked yarn edge area according to the pseudo image features.

Images