A composite material nondestructive testing method and system based on machine vision

Abstract

The application relates to the field of woven material image detection, and discloses a composite nondestructive testing method based on machine vision, which comprises the following steps: obtaining an image of a target carbon fiber woven material after weaving is completed and carrying out pretreatment to obtain a woven surface image; carrying out overall detection on the woven surface image, determining a weaving area in the woven surface image, and determining a fiber coverage rate of the woven surface image according to the weaving area; wherein the weaving area corresponds to a region with carbon fibers in the woven surface image; carrying out corner point detection on the weaving area, determining the position of a corner point in the weaving area, and determining a weaving angle of the weaving area according to the position of the corner point of the weaving area; and determining the fiber line width and the fiber spacing of the weaving area according to the position of the corner point in the weaving area. According to the application, various defects of the carbon fiber woven material are detected and calculated in a reasonable order, and the detection efficiency of the carbon fiber woven material is improved.

Description

Technical Field

[0001] This application relates to the field of image detection technology for woven materials, and more specifically, to a machine vision-based non-destructive testing method and system for composite materials. Background Technology

[0002] Carbon fiber is an advanced material with high strength, lightweight, and corrosion resistance, attracting significant attention due to its wide application in aerospace, automotive, sporting goods, and construction. Carbon fiber itself possesses very high tensile strength and modulus; stacking multiple layers of carbon fiber fabric together creates composite structures with even higher strength and stiffness, and multilayer carbon fiber composites can resist fatigue failure. Simultaneously, multilayer carbon fiber composites exhibit excellent corrosion resistance, showing superior performance against chemical corrosion from water, acids, and alkalis, and demonstrating good stability in harsh environments such as seawater and chemical plants. The overall lightweight nature of multilayer carbon fiber composites makes them ideal for product designs and applications requiring weight reduction. By controlling the orientation and stacking order of each carbon fiber layer, directional design of strength and stiffness characteristics in different directions can be achieved, thereby meeting the specific needs of composite materials in practical engineering applications.

[0003] Quality inspection of multi-layer carbon fiber woven materials can be performed using image recognition. For example, patent CN114565589B (application number: CN202210203660.4) provides a method and apparatus for detecting missing or entangled carbon fiber warp yarns. This method obtains the column containing the carbon fiber warp yarns from an image of the carbon fiber warp yarns, then determines whether the spacing between adjacent columns is greater than a preset threshold for the width of a single warp yarn, thus identifying whether there are missing yarns. Similarly, it determines whether the pixel width of the entangled column is greater than a preset threshold for the width of the entangled yarn, effectively detecting either missing or entangled yarns. However, the method in patent CN114565589B can only detect entanglement in single-layer carbon fiber woven materials and cannot detect various defects in the material. For example, it cannot efficiently detect defects such as weaving angle, coverage, fiber line width, fiber spacing, and fiber fuzz, nor can it efficiently inspect the quality of multi-layer carbon fiber woven materials. Summary of the Invention

[0004] The purpose of this application is to provide a machine vision-based non-destructive testing method and system for composite materials, which solves the technical problem of not being able to efficiently detect various defects in carbon fiber woven materials, such as weaving angle, coverage, fiber line width, fiber spacing and fiber fuzz, and achieves the technical effect of being able to efficiently detect various defects in carbon fiber woven materials.

[0005] This application provides a machine vision-based non-destructive testing method and system for composite materials. The method includes: acquiring an image of the target carbon fiber woven material after weaving is completed and preprocessing it to obtain a fabric image; performing overall detection on the fabric image to determine the woven area in the fabric image, and determining the fiber coverage of the fabric image based on the woven area; wherein the woven area corresponds to the area in the fabric image where carbon fibers exist; performing corner detection on the woven area to determine the corner position in the woven area, and determining the woven angle of the woven area based on the corner position; and determining the fiber line width and fiber spacing of the woven area based on the corner position.

[0006] In one possible implementation, the method further includes: performing hairy region of interest detection on the woven area to determine the hairy region of interest in the woven area; wherein the hairy region of interest includes regions of interest in the woven area where hairy defects may exist; performing corner point detection on the hairy region of interest to determine the corner point position in the hairy region of interest, and determining the woven angle of the hairy region of interest based on the corner point position; and determining the fiber line width and fiber spacing of the hairy region of interest based on the corner point position.

[0007] In another possible implementation, the method further includes: after sequentially performing image enhancement and erosion operations on the feathering region of interest, determining an enhanced feathering region of interest within the feathering region of interest; wherein the erosion operation is used to separate connected regions and noise-removing regions within the feathering region of interest; determining the feathering region within the enhanced feathering region of interest, and determining the feathering area occupied by the feathering region within the enhanced feathering region of interest.

[0008] In another possible implementation, a region of interest detection is performed on the woven area to determine the region of interest for the hair, including: determining the fiber backbone region in the woven area after performing an erosion operation on the woven area; determining multiple connection regions related to the fiber backbone region in the woven area according to a region growing algorithm or an edge connection algorithm, and determining the region of interest for the hair based on the multiple connection regions; wherein the connection region corresponds to the region where the hair is located in the woven area.

[0009] In another possible implementation, determining the feather interest region based on multiple connected regions includes: clustering the multiple connected regions according to shape and category to obtain multiple feather interest regions, each feather interest region including one or more connected regions.

[0010] In another possible implementation, the method further includes: correcting the fiber coverage of the fabric image based on the area of ​​hair occupied by the hair region in the enhanced hair region of interest.

[0011] In another possible implementation, corner detection is performed on the region of interest (ROI) to determine the corner positions within the ROI, including: identifying fabric primitives that do not include ROI regions within the ROI; wherein the fabric primitives are complete units of carbon fiber bundles exposed within the ROI; and corner detection is performed on the fabric primitives to determine the corner positions within the fabric primitives, which are then used as the corner positions within the ROI.

[0012] In another possible implementation, the method further includes: performing hairy region of interest detection on the woven area to determine multiple hairy regions of interest in the woven area; performing corner point detection on each of the multiple hairy regions of interest to determine the corner point position in each hairy region of interest, and determining the woven angle of the hairy region of interest based on the corner point position of each hairy region of interest; when the difference between the woven angle of the first hairy region of interest and the woven angle of the second hairy region of interest is greater than a preset woven angle difference, stopping the detection of fiber line width and fiber spacing of the hairy region of interest, and marking the target carbon fiber woven material as unqualified.

[0013] In another possible implementation, the method further includes: when the difference between the weaving angle of the first feathering region of interest and the weaving angle of the second feathering region of interest is less than or equal to a preset weaving angle difference, determining the fiber linewidth and fiber spacing of each feathering region of interest based on the corner point position in each feathering region of interest; when the difference between the fiber linewidth of the first feathering region of interest and the fiber linewidth of the second feathering region of interest is greater than a preset fiber linewidth difference, or when the difference between the fiber spacing of the first feathering region of interest and the fiber spacing of the second feathering region of interest is greater than a preset fiber spacing difference, marking the target carbon fiber braided material as unqualified.

[0014] In another possible implementation, the method further includes: determining the fiber coverage of the region of interest based on the fiber linewidth and fiber spacing of the region of interest; and correcting the fiber coverage of the fabric image based on the fiber coverage of the region of interest.

[0015] In another possible implementation, the fiber coverage of the fabric image is corrected based on the fiber coverage of the region of interest in the feathers, including: calculating the fiber coverage of the region of interest in the feathers and the fiber coverage of the fabric image outside the region of interest in the knitted area; and weighting and summing the fiber coverage of the region of interest in the feathers and the area weight of the region of interest in the feathers, as well as the fiber coverage of the fabric image outside the region of interest in the feathers and the area weight of the fabric image outside the region of interest in the feathers, to obtain the fiber coverage of the fabric image.

[0016] This application also provides a machine vision-based nondestructive testing system for composite materials, including units for implementing the machine vision-based nondestructive testing method for composite materials as described above.

[0017] The beneficial effects of the embodiments in this application compared with the prior art are:

[0018] This application provides a machine vision-based non-destructive testing method for composite materials. The method includes: acquiring an image of the target carbon fiber woven material after weaving is completed and preprocessing it to obtain a fabric image; performing overall detection on the fabric image to determine the woven area in the fabric image, and determining the fiber coverage of the fabric image based on the woven area; wherein the woven area corresponds to the area in the fabric image where carbon fibers exist; performing corner detection on the woven area to determine the corner position in the woven area, and determining the weaving angle of the woven area based on the corner position; and determining the fiber line width and fiber spacing of the woven area based on the corner position. This application can detect carbon fiber woven composite material images and sequentially detect coverage, weaving angle, fiber line width, and fiber spacing, reducing the computational load in image detection and improving the detection efficiency of carbon fiber woven composite materials. It is suitable for efficient detection of multilayer carbon fiber composite materials. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic flowchart of a machine vision-based nondestructive testing method for composite materials in an embodiment of this application.

[0021] Figure 2 This is a schematic diagram of the fabric image obtained in the embodiments of this application;

[0022] Figure 3 This is a schematic diagram showing the locations of multiple corner points in the weaving area in an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of the fiber backbone region in an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the image detection process for feather regions of interest in an embodiment of this application;

[0025] In the figure, 1 is the fabric image; 11 is the weaving area; 11a is the fiber backbone area; 11b is the connecting area; 111 is the area of ​​interest in hairiness; 111a is the fabric primitive; 112 is the area of ​​interest in enhancing hairiness; and 113 is the hairiness area. Detailed Implementation

[0026] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0027] It should be noted that when a component or structure is referred to as being "fixed to" or "set on" another component or structure, it can be directly on or indirectly on the other component or structure. When a component or structure is referred to as being "connected to" another component or structure, it can be directly connected to or indirectly connected to the other component or structure.

[0028] It should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device, component, or structure referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0029] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0030] Quality inspection of multi-layer carbon fiber woven materials can be performed using image recognition. However, existing quality inspection methods cannot efficiently detect various defects in carbon fiber woven materials, such as weave angle, coverage, fiber line width, fiber spacing, and fiber fuzz, nor can they efficiently inspect the quality of multi-layer carbon fiber woven materials.

[0031] Based on the above reasons, this application provides a machine vision-based non-destructive testing method for composite materials. The method includes: acquiring an image of the target carbon fiber woven material after weaving is completed and preprocessing it to obtain a fabric image; performing overall detection on the fabric image to determine the woven area in the fabric image, and determining the fiber coverage of the fabric image based on the woven area; wherein the woven area corresponds to the area in the fabric image where carbon fibers exist; performing corner detection on the woven area to determine the corner position in the woven area, and determining the weaving angle of the woven area based on the corner position; and determining the fiber line width and fiber spacing of the woven area based on the corner position. This application can detect carbon fiber woven composite material images and sequentially detect coverage, weaving angle, fiber line width, and fiber spacing, reducing the computational load in image detection and improving the detection efficiency of carbon fiber woven composite materials. It is suitable for efficient detection of multilayer carbon fiber composite materials.

[0032] In some scenarios, a machine vision-based non-destructive testing method for composite materials according to an embodiment of this application can be applied to the quality inspection of carbon fiber woven composite materials. Due to the reduced computational load, it can be applied to image quality inspection of each layer of a multi-layered composite structure, thereby improving the inspection efficiency of carbon fiber woven composite materials.

[0033] The following describes in detail, with specific examples, a machine vision-based nondestructive testing method and system for composite materials provided in the embodiments of this application.

[0034] Figure 1 This is a flowchart illustrating a machine vision-based nondestructive testing method for composite materials, as shown in an embodiment of this application. Figure 1 As shown, the machine vision-based non-destructive testing method for composite materials in this application includes steps S110 to S140, which will be described in detail below.

[0035] S110. Acquire an image of the target carbon fiber woven material after weaving is completed and preprocess it to obtain the fabric image 1.

[0036] In use, images of the target carbon fiber woven material can be captured to perform image recognition and detection of the weaving quality. After acquiring images of the target carbon fiber woven material after weaving is completed, preprocessing such as highlight removal and image enhancement can be performed to obtain the fabric image 1. For example, Figure 2 This is a schematic diagram of the fabric image obtained in an embodiment of this application. Figure 2 It contains multiple interwoven carbon fiber woven materials.

[0037] S120. Perform overall detection on the fabric image 1 to determine the woven area 11 in the fabric image 1, and determine the fiber coverage of the fabric image 1 based on the woven area 11. The woven area 11 corresponds to the area in the fabric image 1 where carbon fibers exist.

[0038] After obtaining the fabric image 1, an overall detection can be performed on the fabric image 1. The overall detection is a preliminary detection of the fabric image 1. Through the overall detection, the woven area 11 in the fabric image 1 can be determined.

[0039] For example, a threshold-based segmentation algorithm can be used to perform overall detection on the fabric image 1. Specifically, the image can be divided into different regions based on the grayscale values ​​of pixels to determine the woven region 11 in the fabric image 1 based on overall detection. Then, the coverage rate corresponding to the woven region 11 can be obtained by calculating the ratio between the area of ​​the woven region 11 and the area of ​​the non-woven region in the fabric image 1. Here, the woven region 11 corresponds to the region in the fabric image 1 where carbon fibers exist.

[0040] For example, the fiber coverage of the fabric image 1 is the ratio of the area of ​​the woven region 11 to the area of ​​the fabric image 1.

[0041] S130. Perform corner detection on the knitting area 11, determine the corner positions in the knitting area 11, and determine the knitting angle of the knitting area 11 based on the corner positions.

[0042] For example, the automatic surface parameter measurement system for three-dimensional braided composite material preforms in patent CN105387814B (application number: CN201510888073.3) can detect the corner points of the braided area 11 and calculate the braiding angle of the braided area 11 according to the process of steps 1-9 in patent CN105387814B.

[0043] like Figure 2 As shown, the corner positions A1-A4 corresponding to the carbon fiber bundles can be detected in the braided area 11.

[0044] For example, the Harris corner detection algorithm can also be used to detect corners in the woven region 11. This algorithm detects corners by using grayscale changes in local areas of the image, calculating the corner response function at each pixel location in the image, and then using non-maximum suppression to filter corners.

[0045] S140. Determine the fiber width and fiber spacing of the weaving area 11 based on the corner positions in the weaving area 11.

[0046] For example, such as Figure 2 and Figure 3 As shown, Figure 3 This is a schematic diagram showing the corner positions in the knitting area in an embodiment of this application, such as... Figure 3 As shown, after obtaining the corner positions A1-A4 in the weaving area 11, Figure 3 The corner points A1-A4 correspond to Figure 2 By identifying the corner points A1-A4 in the braided area 11, we can obtain the parallelogram A corresponding to a carbon fiber bundle. The boundary of parallelogram A can be determined using an edge detection algorithm. Then, the height of parallelogram A can be determined based on the position and length information of the boundary of parallelogram A, and the height of parallelogram A can be used as the fiber linewidth.

[0047] For example, when determining the fiber line width and fiber spacing corresponding to parallelogram A based on the position and length information of the boundary of parallelogram A, the area of ​​parallelogram A can first be calculated by an image detection algorithm, and a side of parallelogram A can be selected as a baseline. The area of ​​parallelogram A can be divided by the length of the baseline to obtain the height D of parallelogram A, and the height D of parallelogram A can be used as the fiber line width in the weaving area 11.

[0048] For example, after detecting parallelogram A in the braided area 11, after detecting corner positions B1-B4 in the braided area 11, corner positions B1-B4 correspond to parallelogram B corresponding to another carbon fiber bundle in the braided area 11, and the distance between parallelogram A and parallelogram B can be calculated as the fiber spacing d.

[0049] The beneficial effect of the above implementation method is that, when performing image detection on woven materials, the coverage of the woven area of ​​the woven material is first determined by overall detection, and then the corner point detection of the woven area is performed to obtain the woven angle of the woven material. Then, the fiber width and fiber spacing of the woven area can be detected based on the corner points of the woven area, realizing the reuse of the corner points of the woven area and improving the efficiency of image detection on woven materials.

[0050] In some implementations, the above method also includes S210 to S230, which will be described in detail below.

[0051] S210. Perform a hair and feather region of interest detection on the knitted area 11 to determine the hair and feather region of interest 111 in the knitted area 11. The hair and feather region of interest 111 includes the regions of interest in the knitted area 11 where hair and feather defects may exist.

[0052] In the weaving of carbon fiber materials, fuzz defects refer to defects caused by the breakage of carbon fibers or the adhesion of impurities in the carbon fiber bundles. Fuzz defects can affect the quality of carbon fiber materials. During use, after obtaining the weaving area 11 following S120, the fuzz region of interest in the weaving area 11 can be detected, thereby identifying fuzz defects within the weaving area 11.

[0053] In use, the detection of the feather interest region in the knitted area 11 can be directly obtained by performing overall detection in S120. The feather interest region detection can be performed by using a convolutional neural network (CNN) for target detection and training a model to identify the feather interest region in the knitted area. Through data annotation and model training, accurate detection of feathers can be achieved.

[0054] S220. Perform corner detection on the feather interest region 111, determine the corner position in the feather interest region 111, and determine the weaving angle of the feather interest region 111 based on the corner position.

[0055] When in use, after obtaining the feather interest region 111, corner point detection can be performed directly in the feather interest region 111 to determine the corner point position in the feather interest region 111, and the weaving angle of the feather interest region 111 can be determined based on the corner point position of the feather interest region 111.

[0056] In use, the corner detection method for the feather interest region 111 can be implemented in a similar way to the corner detection method for the knitting region 11, thereby determining the corner position in the feather interest region 111 and determining the knitting angle of the feather interest region 111 based on the corner position.

[0057] S230. Determine the fiber linewidth and fiber spacing of the feather interest region 111 based on the corner positions in the feather interest region 111.

[0058] When using it, a method similar to that in S140 can be adopted to determine the fiber linewidth and fiber spacing of the feather interest region 111 based on the corner position of the feather interest region 111. This will not be elaborated here.

[0059] The beneficial effect of the above implementation method is that the region of interest (ROI) for hairy fibers contains the hairy defect features in the weaving process of carbon fiber materials. After corner point detection of the ROI to obtain the weaving angle, and after detecting the fiber line width and fiber spacing of the ROI, the image detection of the hairy defect features can be directly performed using the ROI. The ROI can be reused without repeated calculations, thus improving the computational efficiency of image detection of carbon fiber materials.

[0060] In some implementations, the above method also includes S240 to S250, which are described below.

[0061] S240. After performing image enhancement and erosion operations sequentially on the feathering region of interest 111, an enhanced feathering region of interest 112 is determined within the feathering region of interest 111. The erosion operation is used to separate connected regions and noise-removing regions within the feathering region of interest 111.

[0062] In use, after obtaining the feather interest region 111, image enhancement and erosion operations can be performed on the feather interest region 111 in sequence, thereby determining the enhanced feather interest region 112 in the feather interest region 111, which can improve the accuracy of feather defect identification.

[0063] For example, the feather interest region 111 is coarsely obtained through an image recognition neural network, while the enhanced feather interest region 112 can be identified in the feather interest region 111 through an image recognition algorithm based on fine segmentation.

[0064] S250. Determine the feather region 113 in the enhanced feather interest region 112, and determine the feather area occupied by the feather region 113 in the enhanced feather interest region 112.

[0065] In use, a neural network model can be used to determine the feather region 113 in the enhanced feather region of interest 112. The feather region 113 is an image region containing feather defects. Thus, the feather region 113 corresponding to the feather defects in the enhanced feather region of interest 112 can be identified, thereby achieving the effect of identifying feather defects.

[0066] When in use, the area of ​​feathers occupied by feather region 113 in the enhanced feather interest region 112 is determined, thereby enabling quantitative identification of feather defects in feather region 113 and improving the identification effect of feather defects.

[0067] It should be noted that S240 to S250 can be executed after S210 to S230 to enable the reuse of the feather interest region without repeated calculation, thus reducing the computational load for feather region identification.

[0068] The beneficial effect of the above implementation method is that by directly using the region of interest of the hair to perform image detection of the hair defect features, the region of interest of the hair can be reused without repeated calculation, thus improving the computational efficiency of image detection of carbon fiber materials.

[0069] In some implementations, the above-mentioned S210, which involves detecting the feather interest region in the knitting area 11 and determining the feather interest region 111 in the knitting area 11, specifically includes S211 to S212. S211 to S212 will be explained in detail below.

[0070] S211. After performing an etching operation on the braided area 11, the fiber backbone area 11a in the braided area 11 is determined.

[0071] In this embodiment of the application, the fiber backbone region 11a in the braided region 11 can be determined by performing an erosion operation on the braided region 11. When identifying the fiber backbone region 11a, texture feature analysis methods (such as gray-level co-occurrence matrix (GLCM)) can be used to extract the texture features of the fiber backbone region 11a to distinguish the fiber from other regions, thereby achieving the identification of the fiber backbone region 11a.

[0072] S212. Based on the region growth algorithm or the edge connection algorithm, determine multiple connection regions 11b related to the fiber backbone region 11a in the weaving region 11, and determine the hair interest region 111 based on the multiple connection regions 11b. Among them, the connection region 11b corresponds to the region where the hair is located in the weaving region 11.

[0073] In the fabric image 1, the fuzz defects often coincide with the carbon fiber bundles. That is, the fiber backbone region 11a in the weaving region 11 often coincides with the region corresponding to the fuzz. Therefore, multiple connection regions 11b related to the fiber backbone region 11a can be determined in the weaving region 11 by the region growing algorithm or the edge connection algorithm. The connection regions 11b correspond to the regions where the fuzz is located in the weaving region 11.

[0074] When using, Figure 4 This is a schematic diagram of the fiber backbone region in an embodiment of this application, as shown below. Figure 4 As shown, after identifying the fiber backbone region 11a in the weaving region 11, multiple connecting regions 11b connected to the fiber backbone region 11a can be identified by a region growth algorithm or an edge connection algorithm. The multiple connecting regions 11b correspond to the regions where the hairs are located in the weaving region 11.

[0075] For example, such as Figure 4 As shown, after obtaining multiple connected regions 11b, the smallest rectangular region surrounding the multiple connected regions 11b can be determined, and the smallest rectangular region surrounding the multiple connected regions 11b is taken as the feather interest region 111.

[0076] The beneficial effect of the above implementation method is that by identifying the fiber backbone region, multiple connecting regions connected to the fiber backbone region can then be identified, and the region including multiple connecting regions can be regarded as the feather interest region, thus realizing efficient identification of the feather interest region and improving the identification efficiency of the feather interest region.

[0077] In some implementations, determining the feather interest region 111 based on multiple connected regions 11b includes: clustering the multiple connected regions 11b according to shape and category to obtain multiple feather interest regions 111, each feather interest region 111 including one or more connected regions 11b.

[0078] In use, feather defects often exhibit clustering characteristics. Therefore, after obtaining multiple connected regions 11b, in order to scientifically and rationally determine the feather interest region 111, the multiple connected regions 11b can be clustered according to shape and category to obtain multiple feather interest regions 111. Each feather interest region 111 includes one or more connected regions 11b. Each feather interest region 111 can include one or more connected regions 11b. Subsequently, the feather interest regions 111 can be processed to further process the feather defects in the feather interest regions 111.

[0079] For example, the feather interest region 111 can be a region enclosed by a polygon or a curve.

[0080] The beneficial effect of the above implementation method is that by performing clustering processing on multiple connected regions, a feather interest region with clustering characteristics of multiple connected regions is obtained, which can reduce the area of ​​the feather interest region and reduce the computational load of processing the feather interest region in subsequent processing, thereby improving the computational load of processing carbon fiber material images.

[0081] In some implementations, the above method further includes: correcting the fiber coverage of the fabric image 1 based on the area of ​​hair occupied by the hair region 113 in the enhanced hair interest region 112.

[0082] In use, after obtaining the enhanced hairiness region of interest 112 and the hairiness region 113, the area of ​​the hairiness region 113 can be calculated. By calculating the ratio of the area of ​​the hairiness region 113 to the area of ​​the enhanced hairiness region of interest 112, the fiber coverage of the enhanced hairiness region of interest 112 can be calculated. Subsequently, the fiber coverage of the fabric image 1 can be corrected based on the hairiness area occupied by the hairiness region 113 in the enhanced hairiness region of interest 112.

[0083] In use, the area occupied by the hair region 113 in the enhanced hair interest region 112 is used to correct the fiber coverage of the fabric image 1, thereby improving the accuracy of the fiber coverage of the fabric image 1.

[0084] For example, the area ratio of the enhanced feathering interest region 112 in the fabric image 1 can be calculated first, and the coverage rate of the feathering region 113 in the enhanced feathering interest region 112 can be calculated. Then, the fiber coverage rate of the fabric image 1 can be weighted and corrected according to the coverage rate of the feathering region 113 in the enhanced feathering interest region 112 and the area ratio of the enhanced feathering interest region 112 in the fabric image 1, so as to obtain a more accurate value of the fiber coverage rate of the fabric image 1.

[0085] The beneficial effect of the above implementation method is that it corrects the fiber coverage of the fabric image to obtain a more accurate value of the fiber coverage of the fabric image, which can further improve the detection accuracy of the fiber coverage of carbon fiber materials and improve the detection accuracy of weaving quality.

[0086] In some implementations, Figure 5 This is a schematic diagram illustrating the image detection process for feather regions of interest in an embodiment of this application, as shown below. Figure 5 As shown, in the above-mentioned S220, corner detection is performed on the feather interest region 111 to determine the corner position in the feather interest region 111. S221 and S222 are also included. S221 and S222 will be explained in detail below.

[0087] S221. In the feather interest region 111, determine the fabric texture element 111a that does not include the feather region 113.

[0088] In order to further improve the accuracy of corner detection during use, interference from hair defects can be eliminated during corner detection. Specifically, a fabric texture element 111a that does not include the hair region 113 can be determined in the hair region of interest 111. The fabric texture element 111a is an image region that does not contain hair defects, which can provide a clearer image for corner detection and improve the accuracy of corner detection.

[0089] S222. Perform corner detection on the fabric texture element 111a to determine the corner position in the fabric texture element 111a, which is then used as the corner position in the feather interest region 111.

[0090] In use, corner detection is performed on the fabric texture element 111a to eliminate the influence of hair defects on corners, and then the corner positions in the fabric texture element 111a are determined as the corner positions in the hair interest region 111, which improves the accuracy of corner detection and subsequent determination of fiber line width and fiber spacing.

[0091] The beneficial effect of the above implementation method is that it eliminates the influence of fuzz defects on corner points, and improves the accuracy of corner point detection and subsequent determination of fiber linewidth and fiber spacing.

[0092] In some implementations, the above method also includes S310 to S320, which will be described in detail below.

[0093] S310. Perform feather interest region detection on the knitting area 11 to determine multiple feather interest regions 111 in the knitting area 11.

[0094] In actual testing, multiple feather interest areas 111 may be identified in the weaving area 11, and the weaving quality of the weaving material can be tested by comparing the weaving quality of the multiple feather interest areas 111.

[0095] S320. Perform corner detection on multiple feather interest regions 111 respectively, determine the corner position in each feather interest region 111, and determine the weaving angle of the feather interest region 111 based on the corner position of each feather interest region 111.

[0096] In actual testing, multiple feather interest regions 111 can be tested, and then the corner position of each feather interest region 111 can be determined to determine the weaving angle of the feather interest region 111.

[0097] S330 When the difference between the weaving angle of the first hair interest region and the weaving angle of the second hair interest region is greater than the preset weaving angle difference, stop detecting the fiber line width and fiber spacing of the hair interest region 111, and mark the target carbon fiber weaving material as unqualified.

[0098] During the weaving process, if the difference between the weaving angle of the first hair interest area and the weaving angle of the second hair interest area is greater than the preset weaving angle difference, it indicates that the weaving angle deviation of the weaving material in different areas is too large, which may lead to the low weaving quality of the weaving material. The detection of the fiber line width and fiber spacing of the hair interest area 111 can be stopped, and the target carbon fiber weaving material can be marked as unqualified.

[0099] For example, the preset weave angle difference can be 3° to 5°.

[0100] The beneficial effect of the above implementation method is that, by detecting the weaving angle of the area with fuzz defects, when the difference between the weaving angle of the first fuzz interest area and the weaving angle of the second fuzz interest area is greater than the preset weaving angle difference, the detection of the fiber line width and fiber spacing of the fuzz interest area is stopped, and the target carbon fiber braided material is marked as unqualified.

[0101] The beneficial effect of the above implementation method is that, in the process of detecting the difference in the knitting angle of different regions, the difference in the knitting angle of the region with fuzz defects is used for detection. This allows for targeted comparison of the knitting angle of the region with fuzz defects, that is, the knitting angle is detected from the region most likely to have poor quality, thereby improving the accuracy of detecting the knitting quality of the knitting material.

[0102] In some implementations, the method also includes S340 to S350, which will be explained in detail below.

[0103] S340 When the difference between the weaving angle of the first feather interest region and the weaving angle of the second feather interest region is less than or equal to the preset weaving angle difference, the fiber line width and fiber spacing of each feather interest region 111 are determined according to the corner position in each feather interest region 111.

[0104] In actual testing, when the difference between the weaving angle of the first area of ​​interest in hairiness and the weaving angle of the second area of ​​interest in hairiness is less than or equal to the preset weaving angle difference, it indicates that the area with hairiness defects in terms of weaving angle meets the quality requirements. Next, the limit line width and fiber spacing of the area with hairiness can be tested, and then the quality of the woven material can be tested from the perspective of the limit line width and fiber spacing.

[0105] S350. When the difference between the fiber linewidth in the first area of ​​interest and the fiber linewidth in the second area of ​​interest is greater than a preset fiber linewidth difference, or when the difference between the fiber spacing in the first area of ​​interest and the fiber spacing in the second area of ​​interest is greater than a preset fiber spacing difference, the target carbon fiber woven material is marked as unqualified.

[0106] During the inspection, when the fiber line width is measured from the perspective of the limit line width, if the difference between the fiber line width in the first area of ​​interest and the fiber line width in the second area of ​​interest is greater than the preset fiber line width difference, the target carbon fiber woven material is marked as unqualified.

[0107] For example, the preset fiber line width difference can be 0.2mm to 0.5mm.

[0108] During testing, the target carbon fiber woven material can be marked as unqualified if the difference between the fiber spacing in the first area of ​​interest and the fiber spacing in the second area of ​​interest is greater than a preset fiber spacing difference.

[0109] For example, the preset fiber spacing difference can be from 0.1 mm to 0.6 mm.

[0110] The beneficial effect of the above implementation method is that it can detect the fiber width and fiber spacing of the woven material in areas with fuzz defects. It can detect the fiber width and fiber spacing in defective areas, that is, detect the fiber width and fiber spacing from the areas where the quality is most likely to be poor, thus improving the reliability of the woven quality detection.

[0111] In some implementations, the above method also includes S410 to S420, which are described in detail below.

[0112] S410. Determine the fiber coverage of the feather interest region 111 based on the fiber linewidth and fiber spacing of the feather interest region 111.

[0113] During the detection, the fiber coverage of the hairy region of interest 111 can be detected from the perspectives of fiber line width and fiber spacing, thereby achieving a more accurate detection of fiber coverage.

[0114] For example, the fiber linewidth can correspond to the area covered by the fiber, and the fiber spacing can correspond to the area not covered by the fiber. The fiber coverage rate can be obtained by dividing the fiber linewidth by the sum of the fiber linewidth and the fiber spacing.

[0115] S420. Correct the fiber coverage of the fabric image 1 based on the fiber coverage of the hair interest region 111.

[0116] In use, the fiber coverage of the fabric image 1 can be further corrected based on the fiber coverage of the hair interest region 111 to improve the detection accuracy of the fiber coverage of the fabric image 1.

[0117] In some implementations, the correction of the fiber coverage of the fabric image 1 based on the fiber coverage of the hair interest region 111 in the above-mentioned S420 also includes S421 to S422, which will be specifically explained below.

[0118] S421. Calculate the fiber coverage of the hair interest region 111 and the fiber coverage of the fabric image 1 in the weaving region 11 excluding the hair interest region 111.

[0119] During detection, the fiber coverage of the feather interest region 111 and the fiber coverage of the fabric image 1 excluding the feather interest region 111 in the weaving region 11 can be calculated and determined separately, so that the fiber coverage of the weaving region 11 can be corrected according to the fiber coverage of the feather interest region 111.

[0120] S422. The fiber coverage of the feather interest region 111 and the area weight of the feather interest region 111, the fiber coverage of the fabric image 1 other than the feather interest region 111 and the area weight of the fabric image 1 other than the feather interest region 111 are weighted and summed to obtain the fiber coverage of the fabric image 1.

[0121] In the detection, the area of ​​the feather interest region 111 and the area of ​​the fabric image 1 excluding the feather interest region 111 can be calculated. The fiber coverage of the fabric image 1 is obtained by weighted summation based on the fiber coverage of the feather interest region 111 and the area weight of the feather interest region 111, as well as the fiber coverage of the fabric image 1 excluding the feather interest region 111 and the area weight of the fabric image 1 excluding the feather interest region 111.

[0122] The beneficial effect of the above implementation method is that after obtaining multiple feather interest regions, the fiber coverage of the fabric image can be corrected according to the fiber coverage of the multiple feather interest regions, thereby improving the detection quality of the fiber coverage of the fabric image and improving the detection effect of the fabric image.

[0123] Some implementations also include,

[0124] Acquire and preprocess the surface image of the carbon fiber woven material; detect the woven area and calculate the original fiber coverage.

[0125] Perform corner detection and calculate the weaving angle, and calculate the fiber line width and fiber spacing based on the corner location;

[0126] The corner detection is calculated based on an adaptive meshing corner extraction algorithm, including:

[0127] The fiber principal direction is detected by Hough transform; a virtual mesh is generated based on the principal direction, and the coordinates of the mesh intersection points are used as a candidate corner point set; sub-pixel level corner point localization is performed on the candidate corner point set.

[0128] The calculation of fiber linewidth includes: extracting the fiber center skeleton line; dynamically measuring the edge distance along the normal direction of the skeleton line;

[0129] Dynamic edge distance measurement includes: setting N sampling points along the skeleton line, calculating the local normal direction of each sampling point i; detecting the edge points on both sides along the local normal direction, and calculating the line width at each point to obtain the final line width W; the formula is expressed as follows:

[0130]

[0131] Where W is the fiber linewidth; N is the number of sampling points for the skeleton line; and is the edge point on both sides of the normal direction; i is the spatial coordinate of the i-th sampling point on the skeleton line.

[0132] By independently calculating the normal direction at different points i, the fiber linewidth calculation is adapted to fiber bending, further improving the accuracy of the calculation.

[0133] A local coordinate system is established at each sampling point of the skeleton line to determine the direction of the X-axis and the direction of the normal. The direction of the normal is obtained by rotating the X-axis by 90 degrees, and the edge points are determined by the extreme values ​​of the gray-level gradient.

[0134] To further optimize the process, a feathering impact factor model is established. Based on this model, the fiber coverage rate is corrected to quantify the actual impact of feathering on strength. The fiber coverage rate correction specifically includes: constructing a feathering impact factor model; using this model to quantify the actual impact of feathering on strength, thereby quantifying the impact of feathering on the actual effective coverage rate (or strength) of the fiber material; and calculating the corrected fiber coverage rate accordingly. This fiber coverage rate more accurately reflects the effective coverage area of ​​the material after considering feathering defects.

[0135] More specifically, it includes:

[0136] Step 1, Calculate the area ratio of feather defects: Obtain the total area of ​​the feather defect area and the total area of ​​the detected area, and calculate the proportion of the total area of ​​the feather defect area in the total area of ​​the detected area; thus obtaining the area ratio of feather defects.

[0137] Step 2: Calculate the coverage loss factor caused by feathers using the material empirical coefficient α;

[0138] Step 3: Obtain the effective coverage retention factor based on the coverage loss factor, and obtain the corrected fiber coverage based on the effective coverage retention factor and the original fiber coverage.

[0139] Step 2 also includes: combining the material empirical coefficient 'a' and the feather area ratio obtained in Step 1 to calculate the coverage loss factor caused by feathers.

[0140] Among them, the material empirical coefficient 'a' is an empirical constant determined through experiments or historical data. It is used to characterize the sensitivity of the material to feather defects and quantifies the impact of the proportion of unit feather area on the actual effective coverage loss rate.

[0141] The intensity of material property degradation caused by fuzzy defects per unit area for the empirical coefficient 'a' is:

[0142] When the material empirical coefficient a = 1, the proportion of feather area is completely equivalent to the performance loss.

[0143] When the material empirical coefficient a>1, fuzzing causes stress concentration, and small-area defects lead to a significant performance degradation.

[0144] When the empirical coefficient a < 1, the material is not sensitive to hairiness.

[0145] The fiber coverage impact factor model, based on the original coverage rate, modifies the original coverage rate by incorporating a correction factor determined by the total area ratio of fiber defects and the material's sensitivity to fiber defects (material empirical coefficient 'a'). This results in a more realistic coverage rate value, thus achieving the goal of correcting fiber coverage. Through a complete data processing chain—image preprocessing → region segmentation → corner detection → parameter calculation → fiber-specific correction—the results of non-destructive testing are greatly improved, making them more accurate.

[0146] This application also provides a machine vision-based nondestructive testing system for composite materials, including units for implementing the machine vision-based nondestructive testing method for composite materials described above. The beneficial effects of this nondestructive testing system for composite materials have already been explained in the above method description and will not be repeated here.

[0147] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A machine vision-based nondestructive testing method for composite materials, characterized in that, The method includes: The target carbon fiber braided material was captured after the braiding was completed and preprocessed to obtain the fabric image (1). The fabric image (1) is detected as a whole to determine the woven area (11) in the fabric image (1) and the fiber coverage of the fabric image (1) is determined according to the woven area (11); wherein the woven area (11) corresponds to the area in the fabric image (1) where carbon fibers exist. Corner detection is performed on the knitting area (11) to determine the position of the corner in the knitting area (11), and the knitting angle of the knitting area (11) is determined based on the position of the corner in the knitting area (11); Based on the corner positions in the weaving area (11), determine the fiber width and fiber spacing of the weaving area (11); The knitted area (11) is subjected to a feather interest region detection to determine the feather interest region (111) in the knitted area (11); wherein the feather interest region (111) includes the interest region in the knitted area (11) where there may be feather defects. Corner detection is performed on the feather interest region (111) to determine the corner positions in the feather interest region (111); After performing image enhancement and erosion operations on the feather interest region (111) in sequence, the enhanced feather interest region (112) is determined in the feather interest region (111); wherein, the erosion operation is used to separate the connected regions and the noise removal regions in the feather interest region (111). Determine the feather region (113) in the enhanced feather interest region (112), and determine the feather area occupied by the feather region (113) in the enhanced feather interest region (112). Corner detection is performed on the hairy region of interest (111) to determine the corner position in the hairy region of interest (111), including: determining the fabric element (111a) in the hairy region of interest (111) that does not include the hairy region (113); wherein, the fabric element (111a) is a complete unit of carbon fiber bundle exposed in the hairy region of interest (111).

2. The machine vision-based nondestructive testing method for composite materials as described in claim 1, characterized in that, The method further includes: The weaving angle of the feather interest area (111) is determined based on the corner position of the feather interest area (111); Based on the corner positions in the feather interest region (111), the fiber linewidth and fiber spacing of the feather interest region (111) are determined.

3. The machine vision-based nondestructive testing method for composite materials as described in claim 1, characterized in that, Feather interest region detection is performed on the woven area (11) to determine the feather interest region (111) in the woven area (11), including: After performing an etching operation on the braided area (11), the fiber backbone area (11a) in the braided area (11) is determined. Based on the region growth algorithm or the edge connection algorithm, multiple connection regions (11b) related to the fiber backbone region (11a) are determined in the weaving region (11), and the hair interest region (111) is determined based on the multiple connection regions (11b); wherein, the connection region (11b) corresponds to the region where the hair is located in the weaving region (11).

4. The machine vision-based nondestructive testing method for composite materials as described in claim 3, characterized in that, The feather interest region (111) is determined based on multiple connected regions (11b), including: Multiple connected regions (11b) are clustered according to shape and category to obtain multiple feather interest regions (111), each feather interest region (111) including one or more connected regions (11b).

5. The machine vision-based nondestructive testing method for composite materials as described in claim 4, characterized in that, The method further includes: The fiber coverage of the fabric image (1) is corrected based on the area of ​​hair occupied by the hair region (113) in the enhanced hair interest region (112).

6. The machine vision-based nondestructive testing method for composite materials as described in claim 5, characterized in that, Corner detection is performed on the feather interest region (111) to determine the corner positions in the feather interest region (111), and the method also includes: Corner detection is performed on the fabric texture element (111a) to determine the corner positions in the fabric texture element (111a), which are then used as the corner positions in the feather interest region (111).

7. The machine vision-based nondestructive testing method for composite materials as described in claim 6, characterized in that, The method further includes: Feather interest region detection is performed on the woven area (11) to determine multiple feather interest regions (111) in the woven area (11). Corner detection is performed on multiple feather interest regions (111) respectively to determine the corner position in each feather interest region (111), and the weaving angle of the feather interest region (111) is determined according to the corner position of each feather interest region (111). When the difference between the weaving angle of the first hair interest region and the weaving angle of the second hair interest region is greater than the preset weaving angle difference, the detection of the fiber line width and fiber spacing of the hair interest region (111) is stopped, and the target carbon fiber weaving material is marked as unqualified.

8. The machine vision-based nondestructive testing method for composite materials as described in claim 7, characterized in that, The method further includes: When the difference between the weaving angle of the first feather interest region and the weaving angle of the second feather interest region is less than or equal to the preset weaving angle difference, the fiber line width and fiber spacing of each feather interest region (111) are determined according to the corner position in each feather interest region (111). If the difference between the fiber linewidth in the first area of ​​interest and the fiber linewidth in the second area of ​​interest is greater than a preset fiber linewidth difference, or if the difference between the fiber spacing in the first area of ​​interest and the fiber spacing in the second area of ​​interest is greater than a preset fiber spacing difference, the target carbon fiber woven material is marked as unqualified.

9. A machine vision-based nondestructive testing system for composite materials, characterized in that, It includes a unit for implementing the machine vision-based nondestructive testing method for composite materials as described in any one of claims 1 to 8.

Images