Method and device for identifying microstructure of three-dimensional braided composite materials using machine vision

Through machine vision technology, independent images are constructed and recognition results are corrected, and the identification errors and misclassification problems caused by the complex movement direction of yarn in the mesoporological structure recognition of three-dimensional braided composite materials are solved, and the identification errors and misclassification problems are achieved, which is efficient and accurate identification results are achieved, reducing manual intervention. , simplifies the material design process.

CN118658161BActive Publication Date: 2024-11-08JINCHENG NANJING ELECTROMECHANICAL HYDRAULIC PRESSURE ENG RES CENT AVIATION IND OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411149591.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-08
Estimated Expiration
2044-08-21

AI Technical Summary

Technical Problem

The existing mesoscopic structure recognition methods of three-dimensional braided composite materials have problems such as slow recognition speed, low accuracy and large artificial dependence, especially due to the complex direction of yarn movement, there are many identification errors and misclassification areas.

Method used

Through machine vision technology, independent images are constructed based on the yarn motion direction of the original image, the recognition results are corrected, the interference of the yarn motion direction is eliminated, the recognition accuracy is improved, and through automated identification and marking, manual intervention is reduced to achieve three-dimensional weaving composite Efficient identification of material mesoscopic structure.

Benefits of technology

It improves the identification accuracy and efficiency of the meticulous structure of three-dimensional braided composite materials, reduces the dependence of manual counting, simplifies the iteration cycle of material design, and reduces the identification error.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118658161B_ABST
    Figure CN118658161B_ABST
Patent Text Reader

Abstract

The present application provides a method and device for identifying the microstructure of a three-dimensional woven composite material under machine vision. The method provided by the present application obtains a first recognition result based on the original image of the three-dimensional woven composite material; constructs an independent image corresponding to each original image according to the movement direction of the yarn in the original image, and one original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, and the yarn movement direction in each independent image is the same, and the yarn movement directions of multiple independent images corresponding to one original image are different from each other, and the first recognition result is corrected based on the independent image to obtain a second recognition result; and the number of the first object in the original image is identified according to the second recognition result. The method and device provided by the present application are used to efficiently and accurately identify the microstructure of a three-dimensional woven composite material.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of material modeling, and in particular to a method and device for identifying the microstructure of three-dimensional woven composite materials under machine vision. Background Art

[0002] Since the internal microstructure of three-dimensional woven composites is relatively complex, rapid prediction of their material properties to construct a three-dimensional model can help shorten the iteration cycle of material design.

[0003] At present, commonly used three-dimensional models include the ideal Represent Volume Element (RVE) model and the model based on X-ray Computed Tomography (XCT). The RVE model assumes that the material can be represented by an infinitely repeated structure. It observes the image containing the microstructure of the material, measures the geometric structure parameters of the yarn, and establishes a parameterized three-dimensional RVE model based on the mathematical correlation between parameters such as fiber volume fraction and cross-sectional size. At the same time, some parameters that affect the final mechanical behavior of the material are added to the model to make the model closer to the real structure. The XCT model identifies the microstructure of the image and establishes the model through manual labeling or computer vision methods.

[0004] Although the RVE model can conveniently perform parameterized modeling of idealized RVE with arbitrary braiding angles and fiber volume fractions based on mathematical relationships, it cannot describe the internal defects of the material caused by the preparation process due to a large number of simplifications and assumptions in the structure. The modeling results of the XCT model are relatively accurate, but the establishment process involves multiple disciplines, and the steps are relatively cumbersome and the workload is large, resulting in this technology not being widely used. Summary of the invention

[0005] In view of this, the present application provides a method and device for identifying the mesostructure of a three-dimensional woven composite material under machine vision, so as to efficiently and accurately identify the mesostructure of a three-dimensional woven composite material.

[0006] Specifically, the present application is implemented through the following technical solutions:

[0007] The first aspect of the present application provides a method for identifying a microstructure of a three-dimensional woven composite material under machine vision, the method comprising:

[0008] Obtaining a first recognition result based on an original image of the three-dimensional woven composite material, wherein the original image is a plurality of images, each original image corresponds to a first recognition result, each original image corresponds to a different thickness position of the three-dimensional woven composite material, and each original image includes yarns in a plurality of movement directions;

[0009] constructing an independent image corresponding to each original image according to the movement direction of the yarn in the original image, wherein one original image corresponds to a plurality of independent images, each independent image includes a portion of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarns in the plurality of independent images corresponding to one original image are different from each other, and correcting the first recognition result based on the independent images to obtain a second recognition result;

[0010] The number of the first objects in the original image is identified according to the second identification result.

[0011] The second aspect of the present application provides a device for identifying the microstructure of a three-dimensional woven composite material under machine vision, the device comprising: an identification module and a correction module; wherein the identification module is used to obtain a first identification result based on an original image of the three-dimensional woven composite material, the original images are multiple, each original image corresponds to a first identification result, each original image corresponds to a different thickness position of the three-dimensional woven composite material, and each original image includes yarns in multiple movement directions;

[0012] The correction module is used to construct an independent image corresponding to each original image according to the movement direction of the yarn in the original image, one original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarns between the multiple independent images corresponding to one original image are different from each other, and the first recognition result is corrected based on the independent images to obtain a second recognition result;

[0013] The recognition module is further configured to recognize the quantity of the first objects in the original image according to the second recognition result.

[0014] The present application provides a method and device for identifying the microstructure of a three-dimensional woven composite material under machine vision. In the first aspect, after obtaining the initial first recognition result, in the face of the characteristics of two-to-two recognition errors of yarns in different movement directions, based on the multi-dimensional and multi-directional characteristics of the three-dimensional woven composite material, the yarns in the same movement direction are reorganized, so that the information interference of yarns in other directions is eliminated in the reorganized image recognition result, and the accuracy of recognition is improved. The first recognition result is automatically corrected based on the second recognition result after eliminating the interference, which effectively improves the situation where there are many and small misclassified areas, and improves the accuracy of recognition. In the second aspect, the internal structure of the material is accurately and automatically identified through the correction of the first recognition result by the second recognition result, thereby providing data support for the automatic construction of the model, reducing the dependence on manual counting, and through the number statistics of the three-dimensional woven composite material, the efficiency of machine counting is also improved. On this basis, the original image provided by the present invention is a batch recognition method for a large number of original images under different thicknesses of composite materials. Based on automatic recognition and marking, for a large number of original images, the workload of manual annotation is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of Embodiment 1 of the method for identifying the microstructure of a three-dimensional braided composite material under machine vision provided in the present application;

[0016] Figure 2 An original image including four yarn movement directions is shown for an exemplary embodiment of the present application;

[0017] Figure 3 This is a schematic diagram of a second recognition result shown in an exemplary embodiment of the present application;

[0018] Figure 4 This is a schematic diagram of a third recognition result shown in an exemplary embodiment of the present application;

[0019] Figure 5 This is a schematic diagram of a cropped third recognition result shown in an exemplary embodiment of the present application;

[0020] Figure 6 This is a schematic diagram showing the number of first objects in a motion direction in an original image according to an exemplary embodiment of the present application;

[0021] Figure 7 This is a schematic diagram showing the number of first objects in three other motion directions in an original image according to an exemplary embodiment of the present application;

[0022] Figure 8 A schematic diagram of a three-dimensional model of a three-dimensional woven composite material shown as an exemplary embodiment of the present application;

[0023] Fig. 9This is a schematic structural diagram of Example 1 of a device for identifying the microstructure of a three-dimensional woven composite material under machine vision provided in the present application. DETAILED DESCRIPTION

[0024] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0025] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0026] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0027] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0028] Figure 1 This is a flow chart of the first embodiment of the method for identifying the microstructure of a three-dimensional woven composite material under machine vision provided by this application. Please refer to Figure 1 , the method provided in this embodiment may include:

[0029] S101. Obtain a first recognition result based on an original image of the three-dimensional woven composite material, wherein there are multiple original images, each of which corresponds to a first recognition result, each of which corresponds to a different thickness position of the three-dimensional woven composite material, and each of which includes yarns in multiple movement directions.

[0030] Specifically, a three-dimensional woven composite material is a complex three-dimensional structure woven from a specific type of yarn through a three-dimensional braiding machine. The original image of the three-dimensional woven composite material includes images of different movement directions of the yarn of the three-dimensional woven composite material at different time points, which are obtained by scanning the image of the three-dimensional woven composite material. The first recognition result is used to identify the first object of the three-dimensional woven composite material, where the first object refers to the yarn, wherein the yarn is obtained by depositing chemical substances on a fiber bundle, and the fiber bundle is obtained by aggregating a number of fiber filaments. The first recognition result includes the movement direction, shape and position of the yarn. Taking a three-dimensional four-way braided composite material as an example, Figure 2 The original image including four yarn movement directions shown for an exemplary embodiment of the present application is not limited thereto, and the composite material may include any number of directions.

[0031] It should be noted that, in a preferred embodiment, the first object may also be a fiber bundle, a woven structural unit, a pore, an interface between layers, a crack, impurities or foreign matter, a resin or matrix material, a boundary area, etc.

[0032] In the specific implementation, the original image is first preprocessed (including denoising, adjusting contrast and brightness, etc.) to improve the quality of the original image and the accuracy of subsequent recognition. Furthermore, the original image is recognized by the dense optical flow method, which calculates the movement of each pixel in the original image, and infers the movement direction and position change of the yarn based on this information, that is, the first recognition result is obtained. Specifically, the first recognition result is the yarn recognition result of each movement direction.

[0033] S102. Construct an independent image corresponding to each original image according to the movement direction of the yarn in the original image, wherein one original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarn between the multiple independent images corresponding to one original image are different from each other, and correct the first recognition result based on the independent images to obtain a second recognition result.

[0034] Specifically, each original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, and the movement direction of the yarn in each independent image is the same. For different independent images, the movement directions of the yarns in each independent image are different, that is, each independent image contains yarns in one movement direction. It should be noted that the number of independent images is consistent with the number of categories of the movement direction of the yarn, and the size of the independent image is the same as the size of the original image. The original image is obtained from a specific thickness scan of the composite material. Since the morphology of the fibers at different thicknesses of the composite material is anisotropic, the thickness values ​​are different, the original image content is different, and there are multiple original images. From an independent point of view, each original image needs to identify the yarn. At this time, for a large number of original images, the workload of yarn identification and indexing is large. The method provided by the present invention can automatically realize the accurate identification and indexing of the yarn, and provides support for the identification of a large number of composite material structures. From the overall point of view, the distribution of a yarn runs through different thicknesses. At this time, when the yarn is counted and marked, it is necessary to combine the yarn identification of the original images before and after for processing.

[0035] An independent image is an image created from the yarn in one direction of motion in the original image, and has the same size as the original image.

[0036] In a specific implementation, the step of constructing an independent image corresponding to each original image according to the movement direction of the yarn in the original image includes:

[0037] (1) Identify and classify the movement direction of the yarn in the original image.

[0038] In specific implementation, combined with the above description, the original image is first preprocessed (including denoising, adjusting contrast and brightness, etc.), and the original image is recognized by the dense optical flow method, which calculates the movement of each pixel in the original image and infers the movement direction of the yarn based on this information. Furthermore, the original images with the same movement direction are classified into the same category, and the original images with different movement directions are classified into different categories.

[0039] (2) Creating an independent image corresponding to the original image according to the number of categories of the motion direction, wherein the independent image has the same size as the original image.

[0040] Specifically, the number of independent images is consistent with the number of categories of the movement directions of the yarn, and the size of the independent images is the same as the size of the original images.

[0041] In a specific implementation, according to the classified original image, the number of categories of different categories is counted, and an independent image with the same number of categories and the same size as the original image is created.

[0042] For example, in one embodiment, the original image contains yarns in four moving directions (direction A, direction B, direction C and direction D), and accordingly, contains four types of independent images, each type of independent image corresponds to yarns in one moving direction (independent image 1 corresponds to direction A, independent image 2 corresponds to direction B, independent image 3 corresponds to direction C, and independent image 4 corresponds to direction D). For an original image, it contains four independent images 1, 2, 3, and 4. For N original materials of a three-dimensional four-directional composite material, the number of independent images is 4*4*N.

[0043] (3) The yarns with the same movement direction in the original image are placed into independent images with corresponding directions.

[0044] In a specific implementation, the yarns in the original image (which may be one original image or multiple original images) belonging to the same type of movement direction are placed into independent images of corresponding directions.

[0045] (4) Determining an identification color according to the movement direction of the yarn corresponding to the independent image, and identifying the independent image based on the identification color.

[0046] Specifically, the identification color is set according to the moving direction of the yarn, and this is not limited in this embodiment. It should be noted that the number of categories of identification colors is the same as the number of categories of the moving directions of the yarn.

[0047] In a specific implementation, based on the movement direction of the yarn corresponding to each independent image and the corresponding relationship between the movement direction and the identification color, the identification color corresponding to the independent image is determined, and the independent image is represented by the identification color.

[0048] For example, in one embodiment, the yarn in the original image contains four movement directions, and accordingly, four independent images (independent image 1, independent image 2, independent image 3, and independent image 4) are created. The four independent images contain four identification colors (red, gray, green, and pink), and each independent image corresponds to an identification color (independent image 1 corresponds to red, independent image 2 corresponds to gray, independent image 3 corresponds to green, and independent image 4 corresponds to pink). For independent image 1, based on the corresponding relationship, it is determined that the identification color of independent image 1 is red, and independent image 1 is identified with red.

[0049] Furthermore, for each independent image, all regions in the independent image are numbered by the connected region marking method, and the number of pixels contained in each region is counted. Based on the statistical results, the region with a pixel number less than a preset value (the preset value is set according to actual needs, and is not limited in this embodiment. For example, in one embodiment, the preset value is 500) is determined as a misclassified region, and the misclassified region in the independent image is defined as a pore and assigned a value of 0, and other regions in the independent image are filled with the identification color. After the independent images are processed, all independent images are merged into one image to complete the correction of the yarns in the same moving direction in the first recognition result, so as to eliminate the information interference of the yarns in other moving directions. Figure 3 This is a schematic diagram of a second recognition result shown in an exemplary embodiment of the present application.

[0050] For example, in combination with the above example, four independent images are marked as red, gray, green and pink respectively. Three category correction operations are performed on these four independent images. Taking the red image as an example, all regions in the red image are numbered by the connected region labeling method, and the number of pixels contained in each region is counted. Regions with less than 500 pixels are determined as misclassified regions. These regions in such independent images are defined as pores and assigned a value of 0, and the corresponding positions in the gray image are assigned pixel values. This operation is defined as [red, gray], which becomes a correction of the red image. The first operation includes four correction steps: [red, gray], [gray, red], [green, pink], and [pink, green]. The second operation includes four correction steps: [red, pink], [pink, red], [gray, green], and [green, gray]. The third operation includes two correction steps: [gray, red] and [red, gray]. After the above operations, the local misclassification phenomenon in the independent image will be completely eliminated, and the distribution of yarns in four motion directions in the independent image will be obtained, completing the category correction of the microstructure.

[0051] The method provided in this embodiment processes and analyzes yarns with different movement directions separately. On the one hand, it can organize and manage data more clearly, reduce the complexity of image structure caused by mixed directions, and more accurately evaluate and understand the characteristics and behavior of the yarn. On the other hand, by processing yarns with different movement directions separately, it is easier to correct the misclassification of independent images, and to identify, check quality and correct independently in one direction, so as to perform single-direction object recognition in a clearly structured image, and optimize the data storage structure and improve the data processing speed. When correcting the first recognition result, multiple corrections were performed between independent images in different directions, which fully eliminated the interference of yarns with different movement directions and improved the accuracy of original image recognition.

[0052] Optionally, after obtaining the second recognition result, the method further includes:

[0053] (1) Smoothing and filtering the original image.

[0054] In the specific implementation, the original image is first smoothed by a Gaussian low-pass filter to eliminate the sharp edges of the yarn in the original image. Furthermore, based on the pixel value of each pixel in the original image and its size relationship with the preset threshold, when the pixel value is less than the preset threshold, the pixel is determined to be a pore, that is, its pixel value is 0. When the pixel value is not less than the preset threshold, the pixel value of the pixel is determined to be 255.

[0055] (2) Determine a structural element based on the shape and movement direction of the first object, and perform a dilation operation on the original image based on the determined structural element to eliminate gaps between the first objects.

[0056] In the specific implementation, based on the shape and movement direction of the yarn, a linear structural element consistent with the movement direction of the yarn is selected. If the movement direction of the yarn is uneven or has multiple directions, a multi-directional structural element needs to be designed. Further, the determined structural element is used to open the original image to remove isolated points and yarn edge burrs in the original image. In addition, since the shape of the yarn is roughly hexagonal, the "hexagonal" structural element is further used to operate the original image to make the yarns in close contact. Further, the corresponding position of the original image at this time is assigned to 0 according to the pore position index.

[0057] (3) Identify the original image based on an edge detection algorithm to obtain position information of the second object.

[0058] Specifically, in this embodiment, the second object refers to the pores in the original image, and the pores refer to the space filled by fibers or other matrix materials in the three-dimensional woven composite material, which is usually manifested as black spots or blank areas in the area. By detecting the second object, i.e., the pores, the first object, i.e., the yarn, can be better located and identified. Therefore, in this step, the position information of the second object needs to be identified.

[0059] In a specific implementation, a Canny edge detector or a Sobel operator is used to identify the boundary of the pore. Furthermore, through morphological operations such as dilation and erosion, the boundary of the pore is further clarified and improved to obtain the position information of the second object.

[0060] (4) Detecting position information of the first object in the second recognition result based on an edge detection algorithm.

[0061] In specific implementation, combined with the above description, the Canny edge detector or the Sobel operator is continued to be used to detect the edge of the yarn, and further, the texture analysis method (such as gray level co-occurrence matrix, local binary pattern) is used to distinguish the yarn from the background to obtain the position information of the first object.

[0062] (5) Obtaining a third recognition result as a basis for identifying the number of first objects based on the position information of the second object, the position information of the first object, and a pre-constructed spatial relationship model between the first object and the second object.

[0063] Specifically, the third recognition result refers to the result after optimizing the second recognition result, which includes the movement direction, shape and position of the yarn.

[0064] The pre-built spatial relationship model between the first object and the second object is used to identify which first objects surround or cross the second object. By substituting the position information of the first object and the position information of the second object into the spatial relationship model, a preliminary recognition result can be obtained. Further, the preliminary recognition result can be processed using an image segmentation algorithm (such as Watershed or Graph Cut) to further distinguish the first object from the second object and obtain a third recognition result. Figure 4 This is a schematic diagram of the third recognition result of an exemplary embodiment of the present application. It should be noted that, since the yarn recognition effect around the original image is poor at this time, the third recognition result needs to be cropped and the second object is assigned to the recognition result again. Figure 5 This is a schematic diagram of a cropped third recognition result shown in an exemplary embodiment of the present application.

[0065] The method provided in this embodiment optimizes the recognition result of the first object according to the position of the second object, can more accurately distinguish the first object from the second object, avoid recognizing the second object as the first object, and improve the accuracy and reliability of the recognition result. Furthermore, through the position of the second object, more targeted operations can be performed in the image preprocessing stage, thereby reducing misjudgments and improving the overall performance and efficiency of the automation system. When the third recognition result is identified, the smoothness of the yarn boundary is first improved by noise smoothing and burr filtering, thereby reducing the difficulty of boundary recognition and improving the accuracy of yarn recognition.

[0066] S103: Identify the number of the first objects in the original image according to the second recognition result.

[0067] In a specific implementation, the step of identifying the number of the first objects in the original image according to the second recognition result includes:

[0068] (1) The original image includes a plurality of images, one original image is identified in each cycle, and the number of first objects in the original images identified in adjacent cycles is compared according to the third recognition results of the current cycle and the previous cycle to obtain a comparison result, and the comparison result is used to characterize the relationship between the number of first objects in the original images of adjacent cycles.

[0069] Specifically, the original image includes multiple images, and the third recognition result in each period can be obtained by periodically recognizing the original image, that is, the third recognition result is the recognition result of one period for one original image. The comparison result represents the quantitative relationship of the first object in the original images of adjacent periods, which includes three comparison results: equal to, less than, and greater than.

[0070] It should be noted that, since the multiple original images are arranged in sequence, the multiple original images are identified in sequence. The method of identifying each original image is the same, that is, the process of identifying the original image is periodic, and an original image is identified in each cycle, and the original image identified in each cycle is different. The method of identifying the original image in the current cycle is the same as the method of identifying the original image in the previous cycle.

[0071] For example, in one embodiment, the previous cycle recognizes the first original image, labels the regions in the first original image, and recognizes the number of first objects in the first original image. The current cycle recognizes the second original image, labels the regions in the second original image, and recognizes the number of first objects in the second original image.

[0072] Optionally, after obtaining the comparison result, the method further includes:

[0073] For the first object in the original image recognized in the current cycle, based on the position information of the center of gravity of the first object and the position information of the center of gravity of all the first objects in the original image recognized in the previous cycle, the distance between the center of gravity of the first object and the center of gravity of all the first objects is calculated.

[0074] It should be noted that, in order to ensure the consistency of the calculated distance and the rationality of the comparison, in this embodiment, the centroid of the first object is used as a reference point to calculate the distance, thereby avoiding the influence of the irregular shape of the first object on the calculation result.

[0075] In a specific implementation, for the first object in the original image identified in the current period and all the first objects in the original image identified in the previous period, the center of gravity of these first objects is first determined, and further, the distance between the center of gravity of the first object in the current period and the center of gravity of all the first objects in the previous period is calculated through a distance calculation formula.

[0076] (2) Determining a quantitative statistical method based on the comparison results.

[0077] Specifically, the quantitative statistical method corresponds to the comparison result. In combination with the above description, when the comparison result includes three comparison results of equal, less than and greater than, correspondingly, the quantitative statistical method includes three quantitative statistical methods.

[0078] (3) Determine the number of the first objects in the original image recognized in the current cycle according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle.

[0079] Specifically, in a possible implementation, determining the number of the first objects in the original image recognized in the current cycle according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes:

[0080] (1) When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being equal to the number of first objects in the third recognition result of the previous cycle, and the distance is less than a first preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle.

[0081] Specifically, when the comparison result is that the number of first objects in the third recognition result of the current cycle is equal to the number of first objects in the third recognition result of the previous cycle, the corresponding quantitative statistical method is a method corresponding to the number of first objects in the third recognition result of the current cycle being equal to the number of first objects in the third recognition result of the previous cycle. Under this quantitative statistical method, the number of first objects in the original image recognized in the current cycle is determined based on the relationship between the distance and the first preset threshold. Among them, the first preset threshold is set according to actual needs, and in this embodiment, it is not limited to this. For example, in one embodiment, the first preset threshold is 20.

[0082] In a specific implementation, if the distance is less than a first preset threshold, the number of first objects in the original image identified in the current cycle is determined based on the number of first objects in the original image identified in the current cycle in the original image of the previous cycle.

[0083] (2) When the distance is not less than a first preset threshold, determining that the number of the first objects in the original image recognized in the current cycle is a value that is a successor of the number of the first objects in the original image recognized in the previous cycle.

[0084] In a specific implementation, if the distance is not less than a first preset threshold, the number of first objects in the original image identified in the current cycle is determined based on the number of first objects in the original image identified in the previous cycle, that is, the value obtained by adding 1 to the number of first objects in the original image identified in the previous cycle is determined as the number of first objects in the original image identified in the current cycle.

[0085] It should be noted that for the first object located at the edge of the original image, the center of gravity of the first object in the original image identified in the current period may be 0 in the original image of the previous period. At this time, the horizontal coordinate or the vertical coordinate of the center of gravity of the first object automatically moves nearby separately or simultaneously, thereby determining the number of the first object.

[0086] Optionally, in another possible implementation, determining the number of the first objects in the original image recognized in the current cycle according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes:

[0087] When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being less than the number of first objects in the third recognition result of the previous cycle, and the distance is less than a second preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle.

[0088] Specifically, when the comparison result is that the number of first objects in the third recognition result of the current cycle is less than the number of first objects in the third recognition result of the previous cycle, the corresponding quantitative statistical method is a method corresponding to the number of first objects in the third recognition result of the current cycle being less than the number of first objects in the third recognition result of the previous cycle. Under this quantitative statistical method, the number of first objects in the original image recognized in the current cycle is determined based on the relationship between the distance and the second preset threshold. The second preset threshold is set according to actual needs and is not limited in this embodiment. For example, in one embodiment, the second preset threshold is 10.

[0089] In a specific implementation, if the distance is less than the second preset threshold, the number of first objects in the original image identified in the current cycle is determined based on the number of first objects in the original image identified in the current cycle in the original image of the previous cycle.

[0090] Optionally, in another possible implementation, determining the number of the first objects in the original image recognized in the current cycle according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes:

[0091] (1) When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being greater than the number of first objects in the third recognition result of the previous cycle, and the distance is less than a third preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle.

[0092] Specifically, when the comparison result is that the number of first objects in the third recognition result of the current cycle is greater than the number of first objects in the third recognition result of the previous cycle, the corresponding quantitative statistical method is a method corresponding to the number of first objects in the third recognition result of the current cycle being greater than the number of first objects in the third recognition result of the previous cycle. Under this quantitative statistical method, the number of first objects in the original image recognized in the current cycle is determined based on the relationship between the distance and the third preset threshold. Among them, the third preset threshold is set according to actual needs, and in this embodiment, it is not limited to this. For example, in one embodiment, the third preset threshold is 30.

[0093] In a specific implementation, if the distance is less than a third preset threshold, the number of first objects in the original image identified in the current cycle is determined based on the number of first objects in the original image identified in the current cycle in the original image of the previous cycle.

[0094] (2) When the distance is not less than a third preset threshold, determining that the number of the first objects in the original image recognized in the current cycle is a value that is a successor of the number of the first objects in the original image recognized in the previous cycle.

[0095] In specific implementation, if the distance is not less than the third preset threshold, the number of first objects in the original image identified in the current cycle is determined based on the number of first objects in the original image identified in the previous cycle, that is, the value obtained by adding 1 to the number of first objects in the original image identified in the previous cycle is determined as the number of first objects in the original image identified in the current cycle. Figure 6 is a schematic diagram showing the number of first objects in a motion direction in an original image according to an exemplary embodiment of the present application, Figure 7 It is a schematic diagram showing the number of first objects in three other motion directions in an original image according to an exemplary embodiment of the present application.

[0096] The method provided in this embodiment tracks the spatial position change of the first object of the current cycle in the subsequent cycle in real time according to the Euclidean distance between the center of gravity of the first objects in the original images identified in adjacent cycles, calculates whether it is necessary to add or subtract the number of first objects in the current cycle according to the Euclidean distance between the center of gravity of the first objects in the original images identified in adjacent cycles, and uses the number of the center of gravity of the first objects in the current cycle in the previous cycle to determine the number of first objects in the original image identified in the current cycle, thereby realizing the labeling statistics problem of three-dimensional woven composite materials and providing data support for subsequent modeling.

[0097] Optionally, after identifying the number of the first objects in the original image according to the second recognition result, the method further includes:

[0098] (1) Based on the number of the first objects, the three-dimensional woven composite material is reconstructed using a marching cubes algorithm to generate a three-dimensional mesh.

[0099] Specifically, the original image is processed using a marching cube algorithm to reconstruct the microstructure of a three-dimensional woven composite material from a three-dimensional scalar field. The marching cube algorithm will divide the original image into multiple small cubes according to a set threshold value, thereby extracting an isosurface. It should be noted that when using the marching cube algorithm to extract a three-dimensional grid, the number and distribution of the first objects will affect the selection and processing of the isosurface. For example, different numbers of first objects may require different threshold settings to optimize the extraction of isosurfaces to ensure that the grid can accurately represent the position and direction of movement of each first object.

[0100] (2) Based on the number of the first objects, the three-dimensional mesh, and the material properties of the three-dimensional woven composite material, construct a three-dimensional model using simulation software.

[0101] In the specific implementation, the generated three-dimensional mesh is assigned attributes according to the material properties of the three-dimensional woven composite material (such as elastic modulus, tensile strength, etc.). It should be noted that the number of first objects needs to be considered when assigning attributes, because different first objects may have different material properties, especially in three-dimensional woven composite materials using multiple types of first objects. Furthermore, the three-dimensional model with material attributes is imported into engineering simulation software (such as ANSYS, ABAQUS, etc.) to perform simulation analysis of mechanical, thermodynamic and other properties to obtain a three-dimensional model of the three-dimensional woven composite material. Figure 8 A schematic diagram of a three-dimensional model of a three-dimensional woven composite material according to an exemplary embodiment of the present application.

[0102] The method provided in this embodiment, on the one hand, establishes an automated correction method based on the characteristics of the existence of pairwise recognition errors in yarns with different movement directions, and effectively improves the situation where there are many and small misclassified areas. On the other hand, the recognition result of the first object is optimized according to the position of the second object, which can more accurately distinguish the first object and the second object, avoid identifying the second object as the first object, and improve the accuracy and reliability of the recognition result. Further, through the position of the second object, more targeted operations can be performed in the image preprocessing stage, thereby reducing misjudgments and improving the overall performance and efficiency of the automation system. On the third hand, by real-time tracking the spatial position change of the first object of the current cycle in the subsequent cycle according to the Euclidean distance between the centers of gravity of the first object in the original image identified in adjacent cycles, the Euclidean distance between the centers of gravity of the first object in the original image identified in adjacent cycles is used to calculate whether it is necessary to add or subtract the number of the first object in the current cycle, and the number of the centers of gravity of the first object in the current cycle in the previous cycle is used to determine the number of the first object in the original image identified in the current cycle, so as to realize the labeling statistics problem of three-dimensional woven composite materials and provide data support for subsequent modeling. The above processes are all completed automatically by computer programs without human participation.

[0103] Corresponding to the aforementioned embodiment of the method for identifying the mesostructure of a three-dimensional woven composite material under machine vision, the present application also provides an embodiment of a device for identifying the mesostructure of a three-dimensional woven composite material under machine vision.

[0104] Fig. 9 This is a schematic diagram of the structure of the first embodiment of the device for identifying the microstructure of a three-dimensional woven composite material under machine vision provided by this application. Fig. 9 , the device provided in this embodiment includes: an identification module 910 and a correction module 920;

[0105] The recognition module 910 is used to obtain a first recognition result based on an original image of the three-dimensional woven composite material, wherein the original image is a plurality of images, each of which corresponds to a first recognition result, each of which corresponds to a different thickness position of the three-dimensional woven composite material, and each of which includes yarns in multiple movement directions;

[0106] The correction module 920 is used to construct an independent image corresponding to each original image according to the movement direction of the yarn in the original image, one original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarns between the multiple independent images corresponding to one original image are different from each other, and the first recognition result is corrected based on the independent images to obtain a second recognition result;

[0107] The recognition module 910 is further configured to recognize the number of the first objects in the original image according to the second recognition result.

[0108] The device provided in this embodiment, in the first aspect, after obtaining the initial first recognition result, in the face of the characteristics of the yarns with different movement directions having two-to-two recognition errors, the yarns with the same movement direction are reorganized, so that the information interference of the yarns in other directions is eliminated in the reorganized image recognition result, the recognition accuracy is improved, and the first recognition result is automatically corrected based on the second recognition result after eliminating the interference, which effectively improves the situation where there are many and small misclassified areas, and improves the recognition accuracy. In the second aspect, the recognition result of the first object is optimized according to the position of the second object, and the first object and the second object can be distinguished more accurately, avoiding the second object from being identified as the first object, and improving the accuracy and reliability of the recognition result, that is, the recognition result is optimized by different attributes of the same object, and the recognition result is optimized by different objects. Further, through the position of the second object, more targeted operations can be performed in the image preprocessing stage, that is, it is conducive to improving the pertinence in the post-image processing link, thereby reducing misjudgment and improving the overall performance and efficiency of the automation system. In the third aspect, the counting statistics of objects are performed through the distance between the centers of adjacent image target objects, reducing the dependence on manual counting, and through the label statistics of three-dimensional woven composite materials, the efficiency of machine counting is also improved, providing data support for subsequent modeling.

[0109] The device of this embodiment can be used to perform Figure 1 The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0110] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0111] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0112] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for identifying the microstructure of three-dimensional woven composite materials using machine vision, characterized in that: The method comprises: Obtaining a first recognition result based on an original image of the three-dimensional woven composite material, wherein the original image is a plurality of images, each original image corresponds to a first recognition result, each original image corresponds to a different thickness position of the three-dimensional woven composite material, and each original image includes yarns in a plurality of movement directions; constructing an independent image corresponding to each original image according to the movement direction of the yarn in the original image, wherein one original image corresponds to a plurality of independent images, each independent image includes a portion of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarns in the plurality of independent images corresponding to one original image are different from each other, and correcting the first recognition result based on the independent images to obtain a second recognition result; identifying the number of first objects in the original image according to the second recognition result; After obtaining the second recognition result, the method further includes: Performing smoothing and filtering processing on the original image; determining a structural element based on the shape and movement direction of the first object, and performing a dilation operation on the original image based on the determined structural element to eliminate gaps between the first objects; Identify the original image based on an edge detection algorithm to obtain position information of a second object; Based on an edge detection algorithm, detecting position information of the first object in the second recognition result; Obtaining a third recognition result as a basis for identifying the number of the first objects according to the position information of the second object, the position information of the first object, and a pre-constructed spatial relationship model between the first object and the second object; The step of identifying the number of the first objects in the original image according to the second recognition result includes: The original image includes a plurality of images, one original image is identified in each cycle, and the number of first objects in the original images identified in adjacent cycles is compared according to the third identification results of the current cycle and the previous cycle to obtain a comparison result, and the comparison result is used to characterize the relationship between the number of first objects in the original images of adjacent cycles; Determining a quantitative statistical method based on the comparison results; The number of the first objects in the original image identified in the current cycle is determined according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle.

2. The method according to claim 1, characterized in that The step of constructing an independent image corresponding to each original image according to the movement direction of the yarn in the original image comprises: Identifying and classifying the movement direction of the yarn in the original image; Creating an independent image corresponding to the original image according to the number of categories of the motion direction, wherein the independent image has the same size as the corresponding original image; The yarns with the same movement direction in the original image are placed into independent images with corresponding directions; The identification color is determined according to the moving direction of the yarn corresponding to the independent image, and the independent image is identified based on the identification color.

3. The method according to claim 1, characterized in that After obtaining the comparison result, the method further includes: For the first object in the original image recognized in the current cycle, based on the position information of the center of gravity of the first object and the position information of the center of gravity of all the first objects in the original image recognized in the previous cycle, the distance between the center of gravity of the first object and the center of gravity of all the first objects is calculated.

4. The method according to claim 3, characterized in that Determining the number of the first objects in the original image identified in the current cycle according to the number counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes: When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being equal to the number of first objects in the third recognition result of the previous cycle, and the distance is less than a first preset threshold, determining that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle; When the distance is not less than a first preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is a value subsequent to the number of the first objects in the original image recognized in the previous cycle.

5. The method according to claim 3, characterized in that: Determining the number of the first objects in the original image identified in the current cycle according to the number counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes: When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being less than the number of first objects in the third recognition result of the previous cycle, and the distance is less than a second preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle.

6. The method according to claim 3, characterized in that Determining the number of the first objects in the original image identified in the current cycle according to the number counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle includes: When the quantity counting method is a method corresponding to the number of first objects in the third recognition result of the current cycle being greater than the number of first objects in the third recognition result of the previous cycle, and the distance is less than a third preset threshold, determining that the number of the first objects in the original image recognized in the current cycle is the number of the center of gravity of the first object in the original image of the previous cycle; When the distance is not less than a third preset threshold, it is determined that the number of the first objects in the original image recognized in the current cycle is a value subsequent to the number of the first objects in the original image recognized in the previous cycle.

7. The method according to claim 1, characterized in that After identifying the number of the first objects in the original image according to the second recognition result, the method further includes: Based on the number of the first objects, reconstructing the three-dimensional woven composite material using a marching cubes algorithm to generate a three-dimensional mesh; A three-dimensional model is constructed using simulation software based on the number of the first objects, the three-dimensional mesh, and material properties of the three-dimensional woven composite material.

8. A device for identifying the microstructure of three-dimensional woven composite materials under machine vision, characterized in that: The device comprises: an identification module and a correction module; wherein, The recognition module is used to obtain a first recognition result based on an original image of the three-dimensional woven composite material, wherein the original image is a plurality of images, each of which corresponds to a first recognition result, each of which corresponds to a different thickness position of the three-dimensional woven composite material, and each of which includes yarns in a plurality of movement directions; The correction module is used to construct an independent image corresponding to each original image according to the movement direction of the yarn in the original image, one original image corresponds to multiple independent images, each independent image includes part of the yarn in the original image, the movement direction of the yarn in each independent image is the same, and the movement directions of the yarns between the multiple independent images corresponding to one original image are different from each other, and the first recognition result is corrected based on the independent images to obtain a second recognition result; The recognition module is further used to recognize the number of first objects in the original image according to the second recognition result; After obtaining the second recognition result, the recognition module is further used to: Performing smoothing and filtering processing on the original image; determining a structural element based on the shape and movement direction of the first object, and performing a dilation operation on the original image based on the determined structural element to eliminate gaps between the first objects; Identify the original image based on an edge detection algorithm to obtain position information of a second object; Based on an edge detection algorithm, detecting position information of the first object in the second recognition result; Obtaining a third recognition result as a basis for identifying the number of the first objects according to the position information of the second object, the position information of the first object, and a pre-constructed spatial relationship model between the first object and the second object; The step of identifying the number of the first objects in the original image according to the second recognition result includes: The original image includes a plurality of images, one original image is identified in each cycle, and the number of first objects in the original images identified in adjacent cycles is compared according to the third identification results of the current cycle and the previous cycle to obtain a comparison result, and the comparison result is used to characterize the relationship between the number of first objects in the original images of adjacent cycles; Determining a quantitative statistical method based on the comparison results; The number of the first objects in the original image identified in the current cycle is determined according to the quantity counting method and the position information of the first objects in the third recognition result of the current cycle and the position information of the first objects in the third recognition result of the previous cycle.