A bubble detection method and system for carbon fiber prepreg production

Abstract

The present invention relates to the field of image analysis technology, and in particular to a bubble detection method and system for carbon fiber prepreg production, the method comprising: obtaining a target surface image corresponding to the carbon fiber prepreg to be detected, generating a basically defect-free unit based on the target surface image, and photographing the basically defect-free unit in different postures; dividing suspected bubble areas and non-bubble areas; clustering the suspected bubble areas; determining the degree of morphological deviation corresponding to each suspected bubble area; screening out a set of suspected bubble areas corresponding to each suspected bubble area from the target cluster; performing bubble trend analysis on each suspected bubble area based on the degree of morphological deviation corresponding to each suspected bubble area and the set of suspected bubble areas, and judging whether the suspected bubble area is a real bubble area based on interference compliance. The present invention realizes bubble detection by analyzing the target surface image and improves the accuracy of bubble detection.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a bubble detection method and system for producing carbon fiber prepreg. Background Art

[0002] Detecting bubbles in carbon fiber prepreg is often a critical step in ensuring the performance of composite materials. Bubbles often significantly reduce the mechanical properties of composite materials, such as interlaminar shear strength and fatigue resistance, leading to a decrease in the structural load-bearing capacity or even failure. In high-precision fields such as aerospace and automotive, even very small bubbles can cause stress concentration and endanger overall safety. Bubble detection can often optimize impregnation process parameters, improve resin distribution uniformity, and reduce defect formation. In addition, strict bubble quality inspection often reduces product scrap rates and saves production costs, while meeting the stringent material reliability standards of high-end applications, and has important engineering value in ensuring the quality of end products.

[0003] Currently, when performing defect detection on objects, the commonly used method is to segment the defect area from the object image based on the grayscale value. However, when segmenting the bubble area from the carbon fiber prepreg image based on the grayscale value, the following technical problems often occur:

[0004] Since the grayscale differences between the bubble area, local resin accumulation area and loss area of ​​carbon fiber prepreg are often not large, if only the difference in grayscale values ​​is considered when detecting the bubble area, it may often cause misjudgment of bubble pixels, resulting in poor bubble detection accuracy. Summary of the Invention

[0005] In order to solve the technical problem of poor accuracy in bubble detection, the present invention proposes a bubble detection method and system for carbon fiber prepreg production.

[0006] In a first aspect, the present invention provides a bubble detection method for carbon fiber prepreg production, the method comprising:

[0007] Obtain a target surface image corresponding to the carbon fiber prepreg to be inspected, generate a basic defect-free unit based on the target surface image, and photograph the basic defect-free unit in different postures to obtain a possible state unit;

[0008] Based on the similarity between the preset window corresponding to the pixel point in the target surface image and the possible units in different states, the target surface image is divided into suspected bubble areas and non-bubble areas;

[0009] Based on the non-bubble area, the suspected bubble area is clustered to obtain the target cluster;

[0010] Based on the morphological differences between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs, the degree of morphological deviation corresponding to each suspected bubble region is determined;

[0011] Based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs, a set of suspected bubble regions corresponding to each suspected bubble region is screened out from the target cluster;

[0012] Based on the degree of morphological deviation corresponding to each suspected bubble area and the set of suspected bubble areas, each suspected bubble area is analyzed and processed to obtain the interference conformity corresponding to each suspected bubble area, and based on the interference conformity, it is judged whether the suspected bubble area is a real bubble area.

[0013] In combination with the first aspect above, in a possible implementation, generating a substantially defect-free unit based on the target surface image includes:

[0014] Based on the target surface image, separating the warp yarns and the weft yarns by using a direction-selective filter, and extracting the intersection area in the target surface image as a basic unit;

[0015] After extracting the basic units, a standard defect-free template is generated by merging them through geometric rules;

[0016] Based on a standard defect-free template, a cross region in the standard defect-free template is extracted as a basic defect-free unit through a direction-selective filter.

[0017] In conjunction with the first aspect above, in a possible implementation, photographing substantially defect-free units in different postures to obtain units with possible states includes:

[0018] By adjusting the Euler angle, different postures of the basically defect-free unit can be photographed, and the possible state units corresponding to different Euler angles can be obtained.

[0019] In combination with the first aspect above, in one possible implementation, dividing the target surface image into suspected bubble areas and non-bubble areas based on similarities between a preset window corresponding to a pixel point in the target surface image and possible units in different states includes:

[0020] Determine the NCC coefficient between the preset window corresponding to each pixel point in the target surface image and each state possible unit as the target similarity between each pixel point and each state possible unit;

[0021] Filter out the state possible unit with the greatest target similarity with the pixel from all possible state units and use it as the target matching unit corresponding to the pixel;

[0022] The continuous pixel points with the same corresponding target matching unit in the target surface image constitute a matching analysis area;

[0023] Determine the abnormal distortion index corresponding to each matching analysis area according to the difference between each matching analysis area and its adjacent matching analysis area;

[0024] If the abnormal distortion index corresponding to the matching analysis area is greater than the preset abnormal distortion threshold, the matching analysis area is determined as a suspected bubble area;

[0025] If the abnormal distortion index corresponding to the matching analysis area is less than or equal to the preset abnormal distortion threshold, the matching analysis area is determined to be a non-bubble area.

[0026] In conjunction with the first aspect above, in one possible implementation, determining the abnormal distortion index corresponding to each matching analysis region based on the difference between each matching analysis region and its adjacent matching analysis region includes:

[0027] Determine the Euler angle corresponding to the target matching unit corresponding to the pixel point in each matching analysis area as the target Euler angle corresponding to each matching analysis area;

[0028] Determine any matching analysis region as a marked region, and determine each matching analysis region adjacent to the marked region as a reference region;

[0029] Determine the difference between the target Euler angle corresponding to the marked area and the target Euler angle corresponding to each reference area as the target difference between the marked area and each reference area;

[0030] Determining an angle difference index corresponding to the marked area according to a target difference between the marked area and all reference areas, wherein the target difference is positively correlated with the angle difference index;

[0031] An abnormal distortion index corresponding to the marked area is determined based on the target similarity between all pixels in the marked area and their corresponding target matching units, as well as the angle difference index corresponding to the marked area, wherein the target similarity is negatively correlated with the abnormal distortion index, and the angle difference index is positively correlated with the abnormal distortion index.

[0032] In combination with the first aspect above, in a possible implementation, clustering the suspected bubble areas based on the non-bubble areas to obtain target clusters includes:

[0033] Each edge pixel point of each non-bubble area is determined as a marked pixel point, and template matching is performed using the target matching unit corresponding to the marked pixel point as the matching template to obtain the target matching area corresponding to the marked pixel point;

[0034] Update the target matching units corresponding to all pixels in the target matching area corresponding to the marked pixel to the target matching unit corresponding to the marked pixel;

[0035] Determine an area formed by target matching areas corresponding to all marked pixels as a reference area, and determine each pixel in the reference area as a reference point;

[0036] The area consisting of all pixels except the reference point in each suspected bubble area is determined as a candidate area;

[0037] The reference point and the pixel points in the non-bubble area are recorded as temporary pixel points, and the average of the Euler angles corresponding to the latest target matching units corresponding to all temporary pixels in the preset window corresponding to each edge pixel of each candidate area is determined as the Euler representative angle corresponding to each edge pixel of each candidate area;

[0038] Obtain the state possible unit corresponding to the Euler representative angle corresponding to each edge pixel point of each candidate region as the target possible unit corresponding to each edge pixel point of each candidate region;

[0039] Perform template matching using the target possible unit corresponding to the edge pixel point of the candidate area as the matching template to obtain the target matching area corresponding to the edge pixel point of the candidate area;

[0040] Update the target matching units corresponding to all pixels in the target matching area corresponding to the edge pixel of the candidate area to the target possible unit corresponding to the edge pixel;

[0041] Determine an area formed by target matching areas corresponding to all edge pixels of all candidate areas as a temporary area, and determine each pixel in the temporary area as a temporary point;

[0042] The candidate region is updated to the region consisting of all pixels except the temporary point, and the candidate region updating step is repeated until the final updated candidate region is empty, thereby updating the target matching unit corresponding to each pixel in each suspected bubble region;

[0043] Cluster the updated target matching units corresponding to all pixels in all suspected bubble areas, and determine each obtained cluster as a reference cluster;

[0044] The number of target matching units belonging to each reference cluster in the updated target matching units corresponding to all pixels in each suspected bubble area is determined as the overall number of belongings between each suspected bubble area and each reference cluster;

[0045] The reference cluster with the largest number of overall affiliations with the suspected bubble area is selected from all reference clusters as the best reference cluster corresponding to the suspected bubble area;

[0046] The suspected bubble areas with the same corresponding best reference cluster are divided into the same target cluster.

[0047] In combination with the first aspect above, in one possible implementation, determining the degree of morphological deviation corresponding to each suspected bubble region based on the morphological difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs includes:

[0048] The distance between each two edge pixels of each suspected bubble area is determined as the reference distance, and a reference distance set corresponding to each suspected bubble area is obtained;

[0049] The maximum value in the reference distance set corresponding to each suspected bubble area is determined as the target long side representative distance corresponding to each suspected bubble area;

[0050] The line segment connecting the two edge pixels corresponding to the largest reference distance in the reference distance set corresponding to each suspected bubble area is determined as the target line segment corresponding to each suspected bubble area;

[0051] The angle corresponding to the target line segment corresponding to each suspected bubble area is determined as the target representative angle corresponding to each suspected bubble area;

[0052] Through any point in the suspected bubble region, draw a perpendicular line to the target line segment corresponding to the suspected bubble region, record it as the target perpendicular line corresponding to the suspected bubble region, and slide the target perpendicular line on the suspected bubble region. The line segment obtained by the intersection of the target perpendicular line and the boundary of the suspected bubble region during the sliding process is recorded as a candidate line segment, thereby obtaining a set of candidate line segments corresponding to the suspected bubble region;

[0053] The maximum value in the set of candidate line segments corresponding to each suspected bubble area is determined as the representative distance of the target short side corresponding to each suspected bubble area;

[0054] The ratio of the target long side representative distance to the target short side representative distance corresponding to each suspected bubble area is determined as the shape characteristic factor corresponding to each suspected bubble area;

[0055] The degree of morphological deviation corresponding to each suspected bubble area is determined based on the difference between the shape characteristic factor corresponding to each suspected bubble area and the shape characteristic factors corresponding to other suspected bubble areas in the target cluster to which it belongs, as well as the difference between the target representative angle corresponding to each suspected bubble area and the target representative angle corresponding to other suspected bubble areas in the target cluster to which it belongs.

[0056] In combination with the first aspect above, in one possible implementation, screening out a set of suspected bubble regions corresponding to each suspected bubble region from the target cluster based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs includes:

[0057] Determine any suspected bubble region as a marked suspected region, and determine each suspected bubble region in the target cluster to which the marked suspected region belongs, except the marked suspected region, as a reference suspected region;

[0058] determining a generation consistency factor between the marked suspected region and each reference suspected region based on a distance and area difference between the marked suspected region and each reference suspected region;

[0059] If the generation consistency factor between the marked suspected area and the reference suspected area is greater than a preset consistency threshold, the reference suspected area is determined as the generation consistent area;

[0060] The marked suspected area and all the generated consistent areas are combined to form a set of suspected bubble areas corresponding to the marked suspected area.

[0061] In combination with the first aspect above, in one possible implementation, based on the degree of morphological deviation corresponding to each suspected bubble region and the set of suspected bubble regions, bubble trend analysis is performed on each suspected bubble region to obtain the interference compliance corresponding to each suspected bubble region, including:

[0062] Determine any suspected bubble region as a marked suspected region, and perform straight line fitting on the center points of all suspected bubble regions in the set of suspected bubble regions corresponding to the marked suspected region to obtain a target straight line;

[0063] Draw a perpendicular line to the target straight line through the center point of the suspected marked area, and record it as a reference perpendicular line;

[0064] Using the reference vertical line as a dividing line, the set of suspected bubble regions corresponding to the suspected marked regions is divided into two subsets;

[0065] The mean of the areas of all suspected bubble regions in each subset is determined as the area representative factor corresponding to each subset;

[0066] The subset with the larger area representative factor among the two subsets is determined as the set far from the fiber, and the subset with the smaller area representative factor among the two subsets is determined as the set close to the fiber;

[0067] Determining the proportion of the suspected bubble regions in the distant fiber set whose corresponding areas are larger than the area of ​​the suspected marked region as the first target proportion corresponding to the suspected marked region;

[0068] Determining the proportion of the suspected bubble region whose corresponding area in the near-fiber set is smaller than the area of ​​the suspected marked region as the second target proportion corresponding to the suspected marked region;

[0069] Determine the distance from the center point of the suspected marker area to the target straight line as the marker distance corresponding to the suspected marker area;

[0070] The interference compliance corresponding to the marked suspected area is determined based on the degree of morphological deviation, the proportion of the first target, the proportion of the second target and the marking distance corresponding to the marked suspected area, wherein the degree of morphological deviation and the marking distance are both positively correlated with the interference compliance, and the proportion of the first target and the second target are both negatively correlated with the interference compliance.

[0071] In a second aspect, the present invention provides a bubble detection system for carbon fiber prepreg production, comprising a processor and a memory, wherein the processor is configured to process instructions stored in the memory to implement the method of the first aspect or any possible implementation of the first aspect. Specifically, the system comprises:

[0072] The image and unit acquisition module is used to obtain the target surface image corresponding to the carbon fiber prepreg to be inspected, generate basically defect-free units based on the target surface image, and shoot the basically defect-free units in different postures to obtain units with possible states;

[0073] A region segmentation module is used to segment the target surface image into suspected bubble regions and non-bubble regions based on similarities between a preset window corresponding to a pixel point in the target surface image and possible units of different states;

[0074] The regional clustering module is used to cluster the suspected bubble areas based on the non-bubble areas to obtain the target clusters;

[0075] a morphological deviation degree determination module, configured to determine the morphological deviation degree corresponding to each suspected bubble region based on the morphological differences between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs;

[0076] A region screening module is used to screen out a set of suspected bubble regions corresponding to each suspected bubble region from the target cluster based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs;

[0077] The analysis, processing and judgment module is used to perform bubble trend analysis on each suspected bubble area based on the degree of morphological deviation corresponding to each suspected bubble area and the set of suspected bubble areas, obtain the interference compliance corresponding to each suspected bubble area, and judge whether the suspected bubble area is a real bubble area based on the interference compliance.

[0078] In a third aspect, a server is provided, comprising a memory and a processor. The memory is configured to store executable program code, and the processor is configured to call and execute the executable program code from the memory, so that the device executes the method of the first aspect or any possible implementation of the first aspect.

[0079] In a fourth aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any possible implementation of the first aspect.

[0080] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the above-mentioned first aspect or any possible implementation of the first aspect.

[0081] The present invention has the following beneficial effects:

[0082] The present invention provides a bubble detection method for carbon fiber prepreg production. This method analyzes target surface images to detect bubbles, resolving the technical issue of poor bubble detection accuracy and improving the accuracy of bubble detection. Specifically, the present invention comprehensively considers multiple bubble-related regular characteristics, such as morphological deviation and interference compliance, during bubble detection, thereby achieving relatively objective bubble detection and improving bubble detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0084] Figure 1 This is a flow chart of a bubble detection method for carbon fiber prepreg production according to the present invention;

[0085] Figure 2 This is a schematic diagram of the structure of a bubble detection system for carbon fiber prepreg production according to the present invention;

[0086] Figure 3 The figure is a structural diagram of a computer device of the present invention. DETAILED DESCRIPTION

[0087] To further illustrate the technical means and effects employed by the present invention to achieve its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementations, structures, features, and effects of the technical solutions proposed by the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0088] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0089] Detecting bubbles in carbon fiber prepregs is a key step in ensuring the performance of composite materials. Bubbles can significantly reduce the mechanical properties of composite materials, such as interlaminar shear strength and fatigue resistance, resulting in a decrease in the structural bearing capacity or even failure. In high-precision fields such as aerospace and automobiles, tiny bubbles may cause stress concentration and endanger overall safety. Through detection, the impregnation process parameters can be optimized, the uniformity of resin distribution can be improved, and the formation of defects can be reduced. In addition, strict quality inspection can reduce product scrap rate and save production costs, while meeting the stringent standards of material reliability for high-end applications, and has important engineering value for ensuring the quality of end products. The present invention analyzes the causes of bubbles in carbon fiber prepregs and the changes in morphology and distribution caused by resin flow, judges the suspected bubble area, eliminates interference from other situations, and obtains an accurate bubble area.

[0090] refer to Figure 1 , shows the process of some embodiments of a bubble detection method for carbon fiber prepreg production of the present invention. The bubble detection method for carbon fiber prepreg production includes the following steps:

[0091] Step S1, obtaining a target surface image corresponding to the carbon fiber prepreg to be inspected, generating a basically defect-free unit based on the target surface image, and photographing the basically defect-free unit in different postures to obtain a state-possible unit.

[0092] The carbon fiber prepreg to be inspected can be the same carbon fiber prepreg to be tested for bubbles. Carbon fiber prepreg is made by laminating epoxy resin onto carbon fibers using high-pressure, high-temperature techniques. A composite material, also known as carbon fiber prepreg, is made from carbon fiber yarn, epoxy resin, release paper, and other materials through coating, hot pressing, cooling, laminating, and winding processes. The target surface image can be the surface image of the carbon fiber prepreg to be inspected.

[0093] As an example, this step may include the following steps:

[0094] The first step of obtaining the target surface image corresponding to the carbon fiber prepreg to be inspected may include the following sub-steps:

[0095] In the first sub-step, a surface image of the carbon fiber prepreg to be inspected is collected by a linear array camera to obtain a high-resolution image.

[0096] It should be noted that an industrial-grade line array camera with a resolution of ≥8k can be selected to ensure that the single-line pixel density meets the microscopic detection requirements of carbon fiber texture, such as clear resolution of fiber diameter ≤7μm.

[0097] In the second sub-step, a polarization camera is used to collect images of the carbon fiber prepreg to be inspected in different polarization directions to obtain multi-angle images.

[0098] It should be noted that a polarization camera equipped with a micropolarizer array can be used to synchronously capture images at four polarization directions: 0°, 45°, 90°, and 135°. This provides three-channel information: degree of linear polarization (DOLP), angle of polarization (AOP), and light intensity. Signals are sent via an encoder to synchronize image acquisition between the linear array camera and the polarization camera.

[0099] In the third sub-step, the high-resolution image and the multi-angle image are fused to obtain the target surface image.

[0100] It should be noted that a feature point matching algorithm can be used to spatially align the high-resolution image from the line scan camera with the multi-angle image from the polarization camera to ensure pixel-level correspondence. This feature point matching algorithm can include, but is not limited to, the SIFT (Scale-Invariant Feature Transform) algorithm. Timestamp verification ensures strict consistency between the image acquisition times of the line scan camera and the polarization camera. The acquired image is decomposed into a multi-scale space using the Nonsubsampled Contourlet Transform (NSCT). The texture energy information from the line scan camera is retained in the low-frequency subband, while the edge details of the polarization image are incorporated into the high-frequency subband. The images from the line scan and polarization cameras are fused to obtain an enhanced planar image of the carbon fiber prepreg surface to be inspected, which is recorded as the target surface image.

[0101] In the second step, based on the above target surface image, the warp and weft yarns are separated by a direction-selective filter, and the intersection area in the above target surface image is extracted as a basic unit.

[0102] For example, first, the target surface image can be binarized by setting a grayscale threshold based on the grayscale value. The grayscale threshold can be a pre-set threshold, primarily used for image binarization, and can be 80. Next, the weft yarns can be split horizontally and the warp yarns vertically, extracting the intersection blocks as texture primitives, or basic units.

[0103] In the third step, after extracting the basic units, a standard defect-free template is generated by merging them through geometric rules. This method can retain texture features and simplify data.

[0104] The fourth step is to extract the intersection area in the above standard defect-free template as the basic defect-free unit through a direction-selective filter based on the standard defect-free template.

[0105] For example, a multi-directional Gabor filter is used to extract texture feature vectors, and then a fuzzy C-means algorithm is used for classification to identify basic units of the prepreg pattern and record them as basically defect-free units. The directionally selective filter can be, but is not limited to, a multi-directional Gabor filter.

[0106] It should be noted that the pattern of carbon fiber prepreg itself often exhibits certain texture regularities, and its texture variations exhibit a certain degree of repeatability. During the resin impregnation of the fibers, bubbles can form at various stages of the process. The presence of bubbles often causes distortion of the fiber image in certain areas of the image, disrupting the regularity of the pattern. Therefore, suspected bubble areas can be screened by observing the pattern disruption in the obtained prepreg surface image. A substantially defect-free unit often represents the smallest repeating unit of the pattern without distortion.

[0107] The fifth step is to adjust the Euler angle to achieve different posture shooting of the basically defect-free unit and obtain the possible state units corresponding to different Euler angles.

[0108] The state-possible units may be basic defect-free units at different Euler angles.

[0109] It should be noted that the basic unit of the pattern is photographed in different poses to obtain basic unit images of different poses as possible state units. Specifically, Euler angles are used as a description. Euler angles are one of the most intuitive and commonly used methods to describe the pose of an object. They are adjusted through three rotation angles. Euler angles generally include the yaw angle around the horizontal axis, the pitch angle around the vertical axis, and the roll angle around the vertical axis.

[0110] It should be noted that different state possible units can represent the minimum repeating units in different postures.

[0111] Step S2 , dividing the target surface image into suspected bubble areas and non-bubble areas based on the similarity between the preset window corresponding to the pixel point in the target surface image and the possible units in different states.

[0112] The preset window may be a pre-set window, which may be a 7×7 window.

[0113] As an example, this step may include the following steps:

[0114] In the first step, the NCC (Normalized Cross Correlation) coefficient between the preset window corresponding to each pixel point in the above target surface image and each state possible unit is determined as the target similarity between each pixel point and each state possible unit.

[0115] It should be noted that the matching result of the normalized cross correlation (NCC) is represented by a correlation coefficient matrix, and the coefficient value range is [-1, 1].

[0116] In the second step, the state possible unit with the greatest target similarity with the pixel point is selected from all state possible units as the target matching unit corresponding to the pixel point.

[0117] It should be noted that the target matching unit corresponding to a pixel point can represent the posture of the pixel point to a certain extent.

[0118] In the third step, the continuous pixel points with the same corresponding target matching unit in the target surface image are used to form a matching analysis area.

[0119] The target matching units corresponding to all pixels in the matching analysis area may be the same.

[0120] For example, if the target matching units corresponding to two adjacent pixels in the target surface image are the same, these two adjacent pixels can be divided into the same matching analysis area, thereby realizing the division of all adjacent pixels in the target surface image, and further realizing the division of all pixels in the target surface image.

[0121] It should be noted that the matching analysis area can represent positions with the same posture.

[0122] The fourth step, based on the difference between each matching analysis area and its adjacent matching analysis areas, determines the abnormal distortion index corresponding to each matching analysis area, which may include the following sub-steps:

[0123] In the first sub-step, the Euler angle corresponding to the target matching unit corresponding to the pixel point in each matching analysis area is determined as the target Euler angle corresponding to each matching analysis area.

[0124] In the second sub-step, any matching analysis region is determined as a marked region, and each matching analysis region adjacent to the marked region is determined as a reference region.

[0125] It should be noted that the matching analysis area adjacent to the marked area can be obtained by determining each matching analysis area except the marked area as a calibration area, determining the union of the preset neighborhoods corresponding to all edge pixels of the marked area as the marked surrounding area, and recording the matching analysis area that intersects with the marked surrounding area as the reference area. The preset neighborhood can be a pre-set neighborhood, which can be a 5×5 neighborhood.

[0126] In a third sub-step, a difference between the target Euler angle corresponding to the marked area and the target Euler angle corresponding to each reference area is determined as a target difference between the marked area and each reference area.

[0127] It should be noted that, when the target difference between the marked area and the reference area is greater, it often means that the postures between the marked area and the reference area are more different.

[0128] For example, the formula for determining the target difference between the marked area and the reference area can be:

[0129] ;in, is the target difference between the labeled region and the i-th reference region. i is the ordinal number of the reference region. It is the absolute value function. is the yaw angle included in the target Euler angle corresponding to the marked area. is the yaw angle included in the target Euler angle corresponding to the i-th reference area. is the target Euler angle corresponding to the marked area including the pitch angle. is the pitch angle included in the target Euler angle corresponding to the i-th reference area. is the target Euler angle corresponding to the marked area including the roll angle. is the roll angle included in the target Euler angle corresponding to the i-th reference area.

[0130] The fourth sub-step is to determine an angle difference index corresponding to the marked area according to the target difference between the marked area and all reference areas.

[0131] Among them, the target difference can be positively correlated with the angle difference index.

[0132] It should be noted that when the angle difference index corresponding to the marked area is larger, it often means that the posture between the marked area and most reference areas is more different, the posture between the marked area and the surrounding areas is more different, the more separation there is between the marked area and the surrounding areas, and the more likely the marked area is to be a bubble area that destroys the regular distribution of the pattern.

[0133] For example, the formula for determining the angle difference index corresponding to the marked area can be:

[0134] Where B is the angle difference index corresponding to the marked area. n is the number of reference areas. i is the serial number of the reference area. is the target difference between the labeled region and the i-th reference region.

[0135] The fifth sub-step is to determine the abnormal distortion index corresponding to the marked area according to the target similarity between all pixels in the marked area and their corresponding target matching units, and the angle difference index corresponding to the marked area.

[0136] The target similarity may be negatively correlated with the abnormal distortion index, and the angle difference index may be positively correlated with the abnormal distortion index.

[0137] For example, the formula for determining the abnormal distortion index corresponding to the marked area can be:

[0138] ; Where Q is the abnormal distortion indicator corresponding to the marked area. is the normalization function. B is the angle difference index corresponding to the marked area. is an exponential function with a natural constant as its base. N is the number of pixels in the marked area. j is the pixel index in the marked area. is the target similarity between the jth pixel in the marked area and its corresponding target matching unit.

[0139] It should be noted that when B is larger, it often means that the marked area is more likely to be a bubble area that destroys the regular distribution of the pattern. The larger the value, the more likely the pixels in the marked area are to match their corresponding target matching units, and the more likely the pixels in the marked area are to be normal pixels without bubbles. Therefore, the larger the value of Q, the more likely the marked area is to be a bubble area.

[0140] In the fifth step, if the abnormal distortion index corresponding to the matching analysis area is greater than the preset abnormal distortion threshold, the matching analysis area is determined as a suspected bubble area.

[0141] The preset abnormal distortion threshold may be a preset maximum abnormal distortion index allowed when it is considered that no bubbles occur in the matching analysis area, which may be 0.7.

[0142] In the sixth step, if the abnormal distortion index corresponding to the matching analysis area is less than or equal to the preset abnormal distortion threshold, the matching analysis area is determined to be a non-bubble area.

[0143] Step S3: clustering the suspected bubble areas based on the non-bubble areas to obtain target clusters.

[0144] As an example, this step may include the following steps:

[0145] In the first step, each edge pixel point in each non-bubble area is determined as a marked pixel point, and template matching is performed using the target matching unit corresponding to the marked pixel point as the matching template to obtain the target matching area corresponding to the marked pixel point.

[0146] The target matching area may be an area that matches the target matching unit, that is, an area that is the same as the target matching unit.

[0147] In the second step, the target matching units corresponding to all pixels in the target matching area corresponding to the marked pixel are updated to the target matching units corresponding to the marked pixel.

[0148] It should be noted that in actual situations, since the generation of bubbles often destroys the regularity of the pattern, the generation of bubbles often changes the posture of the pixel. The target matching unit corresponding to the marked pixel can represent the posture of the pixel where no bubble occurs. The target matching area corresponding to the marked pixel can represent the matching area with the same posture as the marked pixel when no bubble occurs. Therefore, when no bubble occurs, the posture corresponding to all pixels in the target matching area is often the same as the posture of the marked pixel. Therefore, the updated target matching unit corresponding to the pixel can, to a certain extent, represent the posture that the pixel should have when no bubble defect occurs.

[0149] In the third step, the area formed by the target matching area corresponding to all the marked pixels is determined as the reference area, and each pixel in the reference area is determined as a reference point.

[0150] It should be noted that the updated target matching units corresponding to all pixels in the reference area can, to a certain extent, represent the postures that these pixels should exhibit when no bubble defects occur.

[0151] In the fourth step, the area consisting of all pixels except the reference point in each suspected bubble area is determined as the candidate area.

[0152] In the fifth step, the reference point and the pixel points in the non-bubble area are recorded as temporary pixel points, and the average of the Euler angles corresponding to the latest target matching units of all temporary pixel points in the preset window corresponding to each edge pixel point of each candidate area is determined as the Euler representative angle corresponding to each edge pixel point of each candidate area.

[0153] It should be noted that, in practice, due to the regularity of carbon fiber prepreg patterns, the pose of pixels within a region without defects is often similar to the pose of surrounding pixels. Therefore, the Euler representative angle corresponding to the edge pixels of the candidate region can, to a certain extent, represent the angle corresponding to the pose of these edge pixels.

[0154] For example, the formula for determining the Euler representative angle corresponding to the edge pixel point can be:

[0155] ;

[0156] in, It is the yaw angle included in the Euler representative angle corresponding to the edge pixel point. It is the pitch angle included in the Euler representative angle corresponding to the edge pixel point. is the roll angle included in the Euler representative angle corresponding to the edge pixel. m is the number of temporary pixels in the preset window corresponding to the edge pixel. a is the sequence number of the temporary pixel in the preset window corresponding to the edge pixel. It is the yaw angle included in the Euler angle corresponding to the latest target matching unit corresponding to the a-th temporary pixel point in the preset window corresponding to the edge pixel point. It is the pitch angle included in the Euler angle corresponding to the latest target matching unit corresponding to the a-th temporary pixel point in the preset window corresponding to the edge pixel point. It is the roll angle included in the Euler angle corresponding to the latest target matching unit corresponding to the a-th temporary pixel in the preset window corresponding to the edge pixel.

[0157] The sixth step is to obtain the state possible unit corresponding to the Euler representative angle corresponding to each edge pixel point of each candidate area as the target possible unit corresponding to each edge pixel point of each candidate area.

[0158] The state possible unit corresponding to the Euler representative angle may be a basic defect-free unit under the Euler representative angle.

[0159] It should be noted that the target possible unit corresponding to the edge pixel point of the candidate area can, to a certain extent, represent the posture that the edge pixel point should exhibit when no bubble defect occurs.

[0160] In the seventh step, template matching is performed using the target possible unit corresponding to the edge pixel points of the candidate area as the matching template to obtain the target matching area corresponding to the edge pixel points of the candidate area.

[0161] The target matching area corresponding to the edge pixel point may be an area that matches the target possible unit corresponding to the edge pixel point, that is, an area that is the same as the target possible unit corresponding to the edge pixel point.

[0162] In the eighth step, the target matching units corresponding to all pixels in the target matching area corresponding to the edge pixels of the candidate area are updated to the target possible units corresponding to the edge pixels.

[0163] In the ninth step, an area formed by target matching areas corresponding to all edge pixels of all candidate areas is determined as a temporary area, and each pixel in the temporary area is determined as a temporary point.

[0164] In the tenth step, the candidate area is updated to the area consisting of all pixels except the temporary point, and the candidate area update step is repeated until the final updated candidate area is empty, thereby realizing the update of the target matching unit corresponding to each pixel in each suspected bubble area.

[0165] It should be noted that based on the target matching unit corresponding to the pixel point in the non-bubble area, the target matching unit corresponding to the pixel point in the suspected bubble area is updated, and the updated target matching unit corresponding to the pixel point in the suspected bubble area is obtained. The updated target matching unit corresponding to the pixel point can, to a certain extent, represent the posture that the pixel point should exhibit when no bubble defect occurs.

[0166] The candidate region updating step may include the following sub-steps:

[0167] In the first sub-step, a state possible unit corresponding to the Euler representative angle corresponding to each edge pixel point of each latest candidate region is obtained as a target possible unit corresponding to each edge pixel point of each latest candidate region.

[0168] Among them, the method for obtaining the Euler representative angle corresponding to the edge pixel point of the latest candidate area can be: each pixel point in the target surface image except all the latest candidate areas is recorded as a reference point; the average of the Euler angles corresponding to the latest target matching units corresponding to all reference points in the preset window corresponding to the edge pixel point of the latest candidate area is determined as the Euler representative angle corresponding to the edge pixel point.

[0169] In the second sub-step, the target possible unit corresponding to the edge pixel point of the latest candidate area is used as the matching template to perform template matching to obtain the target matching area corresponding to the edge pixel point of the latest candidate area.

[0170] In the third sub-step, the target matching units corresponding to all pixels in the target matching area corresponding to the edge pixel of the latest candidate area are updated to the target possible units corresponding to the edge pixel.

[0171] In the fourth sub-step, the temporary area is updated to the area formed by the target matching area corresponding to all edge pixels of all the latest candidate areas, and the temporary point is updated to the pixel point in the latest temporary area.

[0172] The fifth sub-step is to update the candidate region to the region consisting of all pixels except the latest temporary point.

[0173] In the eleventh step, the updated target matching units corresponding to all pixels in all suspected bubble areas are clustered using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, and each resulting cluster is determined as a reference cluster.

[0174] In the twelfth step, the number of target matching units belonging to each reference cluster in the updated target matching units corresponding to all pixels in each suspected bubble area is determined as the overall belonging number between each suspected bubble area and each reference cluster.

[0175] In the thirteenth step, the reference cluster with the largest number of overall affiliations with the suspected bubble area is selected from all reference clusters as the best reference cluster corresponding to the suspected bubble area.

[0176] In the fourteenth step, the suspected bubble areas corresponding to the same best reference cluster are divided into the same target cluster.

[0177] It should be noted that the distribution trends of the suspected bubble areas within the same target cluster are often similar when the influence of bubbles is excluded, which often indicates that the suspected bubble areas within the same target cluster may be located in the same part of the carbon fiber prepreg, and often indicates that the suspected bubble areas within the same target cluster can be analyzed as a whole.

[0178] Step S4: determining the degree of morphological deviation corresponding to each suspected bubble region based on the morphological differences between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs.

[0179] It's important to note that bubbles in the same area often have similar shapes due to the same causes. For example, unvolatilized components in the resin expand during the curing process due to heat, forming bubbles. When the resin flows axially along the fiber, the bubbles are stretched into long strips; when it flows transversely, the bubbles tend to be more round. Unvolatilized components in the resin can be water. Therefore, the differences in the shapes of suspected bubble areas within a single target cluster can be used to identify and eliminate interference from resin gaps or resin accumulation that resemble bubbles.

[0180] As an example, this step may include the following steps:

[0181] In the first step, the distance between every two edge pixels of each suspected bubble area is determined as the reference distance, and a reference distance set corresponding to each suspected bubble area is obtained.

[0182] The reference distance set corresponding to the suspected bubble region may include: the distances between all different edge pixels of the suspected bubble region.

[0183] In the second step, the maximum value in the reference distance set corresponding to each suspected bubble area is determined as the representative distance of the target long side corresponding to each suspected bubble area.

[0184] In the third step, the line segment connecting the two edge pixels corresponding to the largest reference distance in the reference distance set corresponding to each suspected bubble area is determined as the target line segment corresponding to each suspected bubble area.

[0185] The length of the target line segment corresponding to the suspected bubble area may be equal to the maximum reference distance in the reference distance set corresponding to the suspected bubble area.

[0186] In the fourth step, the angle corresponding to the target line segment corresponding to each suspected bubble area is determined as the target representative angle corresponding to each suspected bubble area.

[0187] The angle corresponding to the target segment may be in the range of [0°, 360°], and may be equal to the angle formed when the horizontal right direction is rotated counterclockwise to the target segment.

[0188] The fifth step is to draw a perpendicular line to the target line segment corresponding to the suspected bubble area through any point in the suspected bubble area, which is recorded as the target perpendicular line corresponding to the suspected bubble area, and slide the target perpendicular line on the suspected bubble area. The line segment obtained by the intersection of the target perpendicular line and the boundary of the suspected bubble area during the sliding process is recorded as the candidate line segment, and the candidate line segment set corresponding to the suspected bubble area is obtained.

[0189] It should be noted that the candidate line segment may be: a line segment formed by two intersection points between the target vertical line and the boundary of the suspected bubble area as endpoints.

[0190] In the sixth step, the maximum value in the set of candidate line segments corresponding to each suspected bubble area is determined as the representative distance of the target short side corresponding to each suspected bubble area.

[0191] In the seventh step, the ratio of the target long side representative distance to the target short side representative distance corresponding to each suspected bubble area is determined as the shape characteristic factor corresponding to each suspected bubble area.

[0192] In the eighth step, the degree of morphological deviation corresponding to each suspected bubble area is determined based on the difference between the shape characteristic factors corresponding to each suspected bubble area and the shape characteristic factors corresponding to other suspected bubble areas in the target cluster to which it belongs, as well as the difference between the target representative angle corresponding to each suspected bubble area and the target representative angle corresponding to other suspected bubble areas in the target cluster to which it belongs.

[0193] For example, the formula for determining the degree of morphological deviation corresponding to the suspected bubble area can be:

[0194] ;in, is the degree of morphological deviation corresponding to the fth suspected bubble region in the hth target cluster. h is the target cluster number. f and x are the numbers of different suspected bubble regions in the hth target cluster. is the number of suspected bubble regions in the h-th target cluster. It is the absolute value function. is the shape feature factor corresponding to the fth suspected bubble region in the hth target cluster. is the shape feature factor corresponding to the xth suspected bubble region in the hth target cluster. is the target representative angle corresponding to the fth suspected bubble region in the hth target cluster. is the target representative angle corresponding to the xth suspected bubble region in the hth target cluster.

[0195] It should be noted that when When it is larger, it often means that the shapes and change trends of other suspected bubble areas in the hth target cluster are less similar to those of the fth suspected bubble area, and the fth suspected bubble area is less similar to other suspected bubble areas at the same position. It often means that the fth suspected bubble area is more likely not to be a real bubble area, and the fth suspected bubble area is more likely to be a resin gap or resin accumulation interference area similar to a bubble.

[0196] Step S5 , based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs, a suspected bubble region set corresponding to each suspected bubble region is screened out from the target cluster.

[0197] It should be noted that due to issues with the weaving process, the finished product may have insufficient tension or inconsistent gaps, which can cause bubbles to form during subsequent processing. These bubbles can vary in size due to differences in gap size. Therefore, the present invention uses the difference in bubble size and the distance between them to determine whether two bubbles are generated in the same gap or in close proximity.

[0198] As an example, this step may include the following steps:

[0199] In the first step, any suspected bubble region is determined as a marked suspected region, and each suspected bubble region in the target cluster to which the marked suspected region belongs, except for the marked suspected region, is determined as a reference suspected region.

[0200] In the second step, a consistency factor between the marked suspected region and each reference suspected region is determined based on the distance and area difference between the marked suspected region and each reference suspected region.

[0201] For example, the formula corresponding to the consistency factor between the marked suspected area and the reference suspected area can be:

[0202] ;in, Mark the suspected area and The consistency factor between the reference suspected regions is generated. It is the serial number of the reference suspected area. It is an exponential function with a natural constant as its base. is the absolute value function. S is the area of ​​the marked suspected region. It is The area of ​​the reference suspected area. Mark the suspected area and The distance between the first and second reference suspected regions can be equal to the distance between the center point of the marked suspected region and the The distance between the center points of the reference suspected areas.

[0203] It should be noted that when When the value is larger, it often indicates that the suspected area is different from the first The greater the area difference between the two reference suspected regions, the greater the difference. When the value is larger, it often indicates that the suspected area is different from the first The larger the distance between the reference suspected areas, the larger the distance between the reference suspected areas. When the value is larger, it often indicates that the suspected area is different from the first The smaller the distance between the first and second reference suspected areas, and the smaller the area difference, the smaller the difference between the marked suspected areas and the first The more likely the two reference suspected regions are to be located in the same gap or adjacent gaps.

[0204] In the third step, if the generation consistency factor between the marked suspected area and the reference suspected area is greater than a preset consistency threshold, the reference suspected area is determined as the generation consistent area.

[0205] The preset consistency threshold may be a preset maximum consistency factor allowed when the regions are considered to be located in different gaps or non-adjacent gaps, which may be 0.7.

[0206] The fourth step is to combine the above-mentioned marked suspected area with all the generated consistent areas to form a set of suspected bubble areas corresponding to the above-mentioned marked suspected area.

[0207] It should be noted that the gaps between the suspected bubble regions in the set of suspected bubble regions corresponding to the marked suspected region are often the same as or similar to the gaps between the suspected bubble regions.

[0208] Step S6, based on the morphological deviation degree corresponding to each suspected bubble area and the set of suspected bubble areas, perform bubble trend analysis on each suspected bubble area to obtain the interference conformity corresponding to each suspected bubble area, and based on the interference conformity, determine whether the suspected bubble area is a real bubble area.

[0209] It should be noted that during the resin impregnation process, the resin flow is obstructed in areas close to the fibers due to the dense fiber bundles, making microscopic impregnation difficult and prone to bubble retention. In areas away from the fibers, the macroscopic resin flow is smoother, and bubbles have more space to gather or merge. Therefore, the bubble size tends to gradually increase from close to the fibers to farther away from the fibers. The distribution of bubbles generated in the same fiber gap is affected by the resin flow and often exhibits a linear distribution. Therefore, the bubble conformity of the suspected bubble area can be judged by the size distribution of the suspected bubble area within a single suspected bubble collection and the rationality of the bubble trend.

[0210] As an example, this step may include the following steps:

[0211] In the first step, any suspected bubble area is determined as a marked suspected area, and a straight line is fitted to the center points of all suspected bubble areas in the set of suspected bubble areas corresponding to the marked suspected area to obtain a target straight line.

[0212] The second step is to draw a perpendicular line to the target straight line through the center point of the marked suspected area, which is recorded as the reference perpendicular line.

[0213] In the third step, the set of suspected bubble regions corresponding to the marked suspected regions is divided into two subsets using the reference vertical line as a dividing line.

[0214] Among them, one subset may include: in the suspected bubble area set, all suspected bubble areas on one side of the reference vertical line except the marked suspected area; another subset may include: in the suspected bubble area set, all suspected bubble areas on the other side of the reference vertical line except the marked suspected area.

[0215] The fourth step is to determine the mean of the areas of all suspected bubble regions in each subset as the area representative factor corresponding to each subset.

[0216] In the fifth step, the subset with the larger area representative factor among the two subsets is determined as the set far from the fiber, and the subset with the smaller area representative factor among the two subsets is determined as the set close to the fiber.

[0217] In the sixth step, the proportion of the suspected bubble region in the above-mentioned distant fiber set whose corresponding area is larger than the area of ​​the above-mentioned marked suspected region is determined as the first target proportion corresponding to the above-mentioned marked suspected region.

[0218] For example, the formula for determining the first target ratio corresponding to the marked suspected area may be:

[0219] ;in, It is the proportion of the first target corresponding to the marked suspected area. It is the number of suspected bubble regions in the set far away from the fibers, whose corresponding area is larger than the area of ​​the marked suspected region. is the total number of all suspected bubble regions in the set away from the fibers.

[0220] It should be noted that when The larger the The closer , which often indicates that the larger the area of ​​the suspected bubble region far away from the fiber set is than the marked suspected region, the more likely the distribution of the suspected bubble region far away from the fiber set is to conform to the distribution of bubble size changes.

[0221] In the seventh step, the proportion of the suspected bubble area in the above-mentioned close fiber set whose corresponding area is smaller than the area of ​​the above-mentioned marked suspected area is determined as the second target proportion corresponding to the above-mentioned marked suspected area.

[0222] For example, the formula for determining the proportion of the second target corresponding to the marked suspected area can be:

[0223] ;in, It is the proportion of the second target corresponding to the marked suspected area. It is the number of suspected bubble regions in the fiber set whose corresponding area is smaller than the area of ​​the marked suspected region. is the total number of all suspected bubble regions in the fiber collection.

[0224] It should be noted that when The larger the The closer , which often indicates that the smaller the area of ​​the suspected bubble region close to the fiber set is than the marked suspected region, the more likely the distribution of the suspected bubble region close to the fiber set is to conform to the distribution of bubble size changes.

[0225] In the eighth step, the distance from the center point of the suspected mark area to the target straight line is determined as the mark distance corresponding to the suspected mark area.

[0226] In the ninth step, the interference compliance corresponding to the above-mentioned suspected marked area is determined according to the morphological deviation degree, the first target ratio, the second target ratio and the marking distance corresponding to the above-mentioned suspected marked area.

[0227] The degree of morphological deviation and the marking distance may both be positively correlated with the interference conformity, while the proportion of the first target and the proportion of the second target may both be negatively correlated with the interference conformity.

[0228] For example, the formula for determining the interference compliance corresponding to the marked suspected area can be:

[0229] ; Where F is the interference compliance corresponding to the marked suspected area. is a normalization function. D is the distance between the markers corresponding to the suspected region. H is the degree of morphological deviation corresponding to the suspected region. It is the proportion of the first target corresponding to the marked suspected area. It is the proportion of the second target corresponding to the marked suspected area. It is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0, which can be 0.001.

[0230] It should be noted that when H is larger, it is more likely that the suspected marked area is not a real bubble area, and it is more likely that the suspected marked area is a resin gap or resin accumulation interference area similar to a bubble. When the value is larger, it often indicates that the distribution of suspected bubble areas far away from the fiber collection is more likely to conform to the distribution of bubble size changes. A larger value indicates that the distribution of suspected bubbles near the fiber assembly is more likely to conform to the distribution of bubble size variations. A smaller value for D indicates that the marked suspected area is more likely to conform to the linear distribution of bubble variation. Therefore, a larger value for F indicates that the marked suspected area is less likely to be a true bubble area and is more likely to be a resin gap or resin accumulation interference area similar to a bubble.

[0231] The tenth step is to determine whether the suspected bubble area is a real bubble area based on the interference compliance.

[0232] For example, if the interference conformity corresponding to the suspected bubble region is less than or equal to a preset interference threshold, the suspected bubble region can be determined to be a real bubble region. The preset interference threshold can be a preset maximum interference conformity allowed when a bubble is considered to exist in the suspected bubble region, which can be 0.6.

[0233] refer to Figure 2 Based on the same inventive concept as the above method embodiment, the present invention provides a bubble detection system for carbon fiber prepreg production. The system includes a processor and a memory. The processor is configured to process instructions stored in the memory to implement a bubble detection method for carbon fiber prepreg production. The steps may include:

[0234] The image and unit acquisition module 201 is used to acquire a target surface image corresponding to the carbon fiber prepreg to be inspected, generate a substantially defect-free unit based on the target surface image, and photograph the substantially defect-free unit in different postures to obtain a unit with possible states;

[0235] A region segmentation module 202 is configured to segment the target surface image into suspected bubble regions and non-bubble regions based on similarities between a preset window corresponding to a pixel point in the target surface image and possible units of different states;

[0236] A region clustering module 203 is used to cluster the suspected bubble regions based on the non-bubble regions to obtain target clusters;

[0237] A morphological deviation degree determining module 204 is configured to determine a morphological deviation degree corresponding to each suspected bubble region based on morphological differences between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs;

[0238] The region screening module 205 is configured to screen out a set of suspected bubble regions corresponding to each suspected bubble region from the target cluster based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs;

[0239] The analysis, processing and judgment module 206 is used to perform bubble trend analysis on each suspected bubble area based on the degree of morphological deviation corresponding to each suspected bubble area and the set of suspected bubble areas, obtain the interference compliance corresponding to each suspected bubble area, and judge whether the suspected bubble area is a real bubble area based on the interference compliance.

[0240] Figure 3 FIG. 1 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. For example, Figure 3 As shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302, wherein when the processor 302 executes the computer program 303, the computer device can execute any one of the bubble detection methods for carbon fiber prepreg production introduced above.

[0241] Based on the same inventive concept as the above-described method embodiments, the present invention provides a server comprising a memory and a processor. The memory is configured to store executable program code, and the processor is configured to retrieve and execute the executable program code from the memory, thereby enabling the server to perform any of the above-described bubble detection methods for carbon fiber prepreg production.

[0242] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer program product, which includes: computer program code, which, when executed on a computer, enables the computer to execute any one of the above-mentioned bubble detection methods for carbon fiber prepreg production.

[0243] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer-readable storage medium, which stores computer program code. When the computer program code is run on a computer, the computer executes any one of the above-mentioned bubble detection methods for carbon fiber prepreg production.

[0244] In summary, the present invention comprehensively considers multiple regular characteristics related to bubbles when performing bubble detection, such as the degree of morphological deviation and interference compliance, thereby achieving bubble detection relatively objectively and improving the accuracy of bubble detection.

[0245] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A bubble detection method for carbon fiber prepreg production, characterized in that: The following steps are involved: Acquiring a target surface image corresponding to the carbon fiber prepreg to be inspected, generating a basic defect-free unit based on the target surface image, and photographing the basic defect-free unit in different postures to obtain state-possible units, including: photographing the basic defect-free unit in different postures by adjusting the Euler angle to obtain state-possible units corresponding to different Euler angles; Based on the similarity between the preset window corresponding to the pixel point in the target surface image and the possible units in different states, the target surface image is divided into suspected bubble areas and non-bubble areas; Based on the non-bubble area, the suspected bubble area is clustered to obtain the target cluster; Based on the morphological differences between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs, the degree of morphological deviation corresponding to each suspected bubble region is determined; Based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs, a set of suspected bubble regions corresponding to each suspected bubble region is screened out from the target cluster; Based on the degree of morphological deviation corresponding to each suspected bubble area and the set of suspected bubble areas, each suspected bubble area is analyzed and processed to obtain the interference conformity corresponding to each suspected bubble area, and based on the interference conformity, it is determined whether the suspected bubble area is a real bubble area; The method of dividing the target surface image into suspected bubble areas and non-bubble areas based on similarities between the preset window corresponding to the pixel point in the target surface image and the possible units in different states includes: Determine the NCC coefficient between the preset window corresponding to each pixel point in the target surface image and each state possible unit as the target similarity between each pixel point and each state possible unit; Filter out the state possible unit with the greatest target similarity with the pixel from all possible state units and use it as the target matching unit corresponding to the pixel; The continuous pixel points with the same corresponding target matching unit in the target surface image constitute a matching analysis area; Determine the abnormal distortion index corresponding to each matching analysis area according to the difference between each matching analysis area and its adjacent matching analysis area; If the abnormal distortion index corresponding to the matching analysis area is greater than the preset abnormal distortion threshold, the matching analysis area is determined as a suspected bubble area; If the abnormal distortion index corresponding to the matching analysis area is less than or equal to the preset abnormal distortion threshold, the matching analysis area is determined to be a non-bubble area.

2. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: The method of generating a substantially defect-free unit based on the target surface image includes: Based on the target surface image, separating the warp yarns and the weft yarns by using a direction-selective filter, and extracting the intersection area in the target surface image as a basic unit; After extracting the basic units, a standard defect-free template is generated by merging them through geometric rules; Based on a standard defect-free template, a cross region in the standard defect-free template is extracted as a basic defect-free unit through a direction-selective filter.

3. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: Determining the abnormal distortion index corresponding to each matching analysis region based on the difference between each matching analysis region and its adjacent matching analysis region includes: Determine the Euler angle corresponding to the target matching unit corresponding to the pixel point in each matching analysis area as the target Euler angle corresponding to each matching analysis area; Determine any matching analysis region as a marked region, and determine each matching analysis region adjacent to the marked region as a reference region; Determine the difference between the target Euler angle corresponding to the marked area and the target Euler angle corresponding to each reference area as the target difference between the marked area and each reference area; Determining an angle difference index corresponding to the marked area according to a target difference between the marked area and all reference areas, wherein the target difference is positively correlated with the angle difference index; An abnormal distortion index corresponding to the marked area is determined based on the target similarity between all pixels in the marked area and their corresponding target matching units, as well as the angle difference index corresponding to the marked area, wherein the target similarity is negatively correlated with the abnormal distortion index, and the angle difference index is positively correlated with the abnormal distortion index.

4. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: The method clusters the suspected bubble areas based on the non-bubble areas to obtain target clusters, including: Each edge pixel point of each non-bubble area is determined as a marked pixel point, and template matching is performed using the target matching unit corresponding to the marked pixel point as the matching template to obtain the target matching area corresponding to the marked pixel point; Update the target matching units corresponding to all pixels in the target matching area corresponding to the marked pixel to the target matching unit corresponding to the marked pixel; Determine an area formed by target matching areas corresponding to all marked pixels as a reference area, and determine each pixel in the reference area as a reference point; The area consisting of all pixels except the reference point in each suspected bubble area is determined as a candidate area; The reference point and the pixel points in the non-bubble area are recorded as temporary pixel points, and the average of the Euler angles corresponding to the latest target matching units corresponding to all temporary pixels in the preset window corresponding to each edge pixel of each candidate area is determined as the Euler representative angle corresponding to each edge pixel of each candidate area; Obtain the state possible unit corresponding to the Euler representative angle corresponding to each edge pixel point of each candidate region as the target possible unit corresponding to each edge pixel point of each candidate region; Perform template matching using the target possible unit corresponding to the edge pixel point of the candidate area as the matching template to obtain the target matching area corresponding to the edge pixel point of the candidate area; Update the target matching units corresponding to all pixels in the target matching area corresponding to the edge pixel of the candidate area to the target possible unit corresponding to the edge pixel; Determine an area formed by target matching areas corresponding to all edge pixels of all candidate areas as a temporary area, and determine each pixel in the temporary area as a temporary point; The candidate region is updated to the region consisting of all pixels except the temporary point, and the candidate region updating step is repeated until the final updated candidate region is empty, thereby updating the target matching unit corresponding to each pixel in each suspected bubble region; Cluster the updated target matching units corresponding to all pixels in all suspected bubble areas, and determine each obtained cluster as a reference cluster; The number of target matching units belonging to each reference cluster in the updated target matching units corresponding to all pixels in each suspected bubble area is determined as the overall number of belongings between each suspected bubble area and each reference cluster; The reference cluster with the largest number of overall affiliations with the suspected bubble area is selected from all reference clusters as the best reference cluster corresponding to the suspected bubble area; The suspected bubble areas with the same corresponding best reference cluster are divided into the same target cluster.

5. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: The determining of the degree of morphological deviation corresponding to each suspected bubble region based on the morphological difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs includes: The distance between each two edge pixels of each suspected bubble area is determined as the reference distance, and a reference distance set corresponding to each suspected bubble area is obtained; The maximum value in the reference distance set corresponding to each suspected bubble area is determined as the target long side representative distance corresponding to each suspected bubble area; The line segment connecting the two edge pixels corresponding to the largest reference distance in the reference distance set corresponding to each suspected bubble area is determined as the target line segment corresponding to each suspected bubble area; The angle corresponding to the target line segment corresponding to each suspected bubble area is determined as the target representative angle corresponding to each suspected bubble area; Through any point in the suspected bubble region, draw a perpendicular line to the target line segment corresponding to the suspected bubble region, record it as the target perpendicular line corresponding to the suspected bubble region, and slide the target perpendicular line on the suspected bubble region. The line segment obtained by the intersection of the target perpendicular line and the boundary of the suspected bubble region during the sliding process is recorded as a candidate line segment, thereby obtaining a set of candidate line segments corresponding to the suspected bubble region; The maximum value in the set of candidate line segments corresponding to each suspected bubble area is determined as the representative distance of the target short side corresponding to each suspected bubble area; The ratio of the target long side representative distance to the target short side representative distance corresponding to each suspected bubble area is determined as the shape characteristic factor corresponding to each suspected bubble area; The degree of morphological deviation corresponding to each suspected bubble area is determined based on the difference between the shape characteristic factor corresponding to each suspected bubble area and the shape characteristic factors corresponding to other suspected bubble areas in the target cluster to which it belongs, as well as the difference between the target representative angle corresponding to each suspected bubble area and the target representative angle corresponding to other suspected bubble areas in the target cluster to which it belongs.

6. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: The method of screening out a set of suspected bubble regions corresponding to each suspected bubble region from the target cluster based on the area difference between each suspected bubble region and other suspected bubble regions in the target cluster to which it belongs includes: Determine any suspected bubble region as a marked suspected region, and determine each suspected bubble region in the target cluster to which the marked suspected region belongs, except the marked suspected region, as a reference suspected region; determining a generation consistency factor between the marked suspected region and each reference suspected region based on a distance and area difference between the marked suspected region and each reference suspected region; If the generation consistency factor between the marked suspected area and the reference suspected area is greater than a preset consistency threshold, the reference suspected area is determined as the generation consistent area; The marked suspected area and all the generated consistent areas are combined to form a set of suspected bubble areas corresponding to the marked suspected area.

7. The bubble detection method for carbon fiber prepreg production according to claim 1, characterized in that: The method of performing bubble trend analysis on each suspected bubble region based on the degree of morphological deviation corresponding to each suspected bubble region and the set of suspected bubble regions to obtain the interference compliance corresponding to each suspected bubble region includes: Determine any suspected bubble region as a marked suspected region, and perform straight line fitting on the center points of all suspected bubble regions in the set of suspected bubble regions corresponding to the marked suspected region to obtain a target straight line; Draw a perpendicular line to the target straight line through the center point of the suspected marked area, and record it as a reference perpendicular line; Using the reference vertical line as a dividing line, the set of suspected bubble regions corresponding to the suspected marked regions is divided into two subsets; The mean of the areas of all suspected bubble regions in each subset is determined as the area representative factor corresponding to each subset; The subset with the larger area representative factor among the two subsets is determined as the set far from the fiber, and the subset with the smaller area representative factor among the two subsets is determined as the set close to the fiber; Determining the proportion of the suspected bubble regions in the distant fiber set whose corresponding areas are larger than the area of ​​the suspected marked region as the first target proportion corresponding to the suspected marked region; Determining the proportion of the suspected bubble region whose corresponding area in the near-fiber set is smaller than the area of ​​the suspected marked region as the second target proportion corresponding to the suspected marked region; Determine the distance from the center point of the suspected marker area to the target straight line as the marker distance corresponding to the suspected marker area; The interference compliance corresponding to the marked suspected area is determined based on the degree of morphological deviation, the proportion of the first target, the proportion of the second target and the marking distance corresponding to the marked suspected area, wherein the degree of morphological deviation and the marking distance are both positively correlated with the interference compliance, and the proportion of the first target and the second target are both negatively correlated with the interference compliance.

8. A bubble detection system for carbon fiber prepreg production, characterized in that: The method comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement a bubble detection method for carbon fiber prepreg production according to any one of claims 1 to 7.

Images