A defect detection method for copper clad laminate process

Abstract

The present application relates to the field of image data processing technology, and in particular to a defect detection method for a copper clad laminate process. The method comprises: obtaining a surface grayscale image of a copper clad laminate to be detected, and performing block processing on the surface grayscale image to obtain a plurality of image blocks, so as to determine the grayscale difference value of a target image block among the plurality of image blocks; determining the gradient direction difference value of the target image block; taking the product of the grayscale difference value of the target image block and the gradient direction difference value as the first parameter value of a pixel point in the target image block, so as to determine the second parameter value of a target pixel point in the surface grayscale image; obtaining an image of a rust region where the rust is located in the surface grayscale image according to the second parameter value of the pixel point in the surface grayscale image. Through the above technical scheme, the difficulty of rust detection on unetched copper clad laminates can be reduced.

Description

Technical Field

[0001] The present application relates to the technical field of image data processing, and in particular to a defect detection method for a copper clad laminate process. Background Art

[0002] Copper clad laminates are mainly used in the manufacture of printed circuit boards. The copper layer formed by the copper foil on the copper clad laminate can be used as a conductor for the circuit. By etching the copper clad laminate, conductive paths can be formed on the surface of the copper clad laminate for connecting various electronic components to form circuits. These conductive paths can be signal lines, power lines or ground lines.

[0003] If the copper clad laminate is exposed to a humid environment for a long time during storage, the air in the humid environment will react chemically with copper to form rust on the surface of the copper clad laminate. The rusted part of the copper clad laminate will affect the conductive performance of the conductive path formed by the subsequent etching process. Therefore, it is necessary to detect surface rust defects on the unetched copper clad laminate.

[0004] In order to facilitate the detection of surface rust defects of products with copper surfaces, a Chinese patent application document with publication number CN118135328A provides an intelligent segmentation system for copper plate and strip defect images, including: a data acquisition module: obtaining a data set containing normal copper plate and strip images and strip defect images; constructing a normal image data set; classifying defect sample data into sample data sets of pore defects, crack defects, deformation defects and oxidation corrosion defects; a data supplement module: used to generate sample data of the sample data set through an adversarial network; an image reconstruction module: used to reconstruct the sample data and the normal image, and to fuse the images in the normal image data set; a defect judgment module: used to judge the defects of the real-time copper plate and strip image data through a trained deep learning network model; a defect segmentation module: outputs image defects according to the model construction module.

[0005] In the related art, a large number of surface image samples of the product to be tested are mainly collected, and the defect area is segmented by a defect segmentation model trained with diverse surface image samples. However, it is difficult to collect diverse rust images on the surface of the copper clad laminate, which makes it difficult to detect surface rust on the unetched copper clad laminate. Summary of the invention

[0006] In order to overcome the difficulty in detecting rust on the surface of an unetched copper clad laminate in the related art, the present application provides a defect detection method for a copper clad laminate process, comprising:

[0007] A surface grayscale image of a copper clad laminate to be inspected is obtained, and the surface grayscale image is processed into blocks to obtain a plurality of image blocks, so as to determine a grayscale difference value of a target image block among the plurality of image blocks; the grayscale difference value is used to characterize the degree of difference in grayscale value between pixels in the target image block; a gradient direction difference value of the target image block is determined, and the gradient direction difference value is used to characterize the degree of difference between a gradient vector of the target image block and a gradient vector of a pixel in the target image block; the product of the grayscale difference value of the target image block and the gradient direction difference value is used as a first parameter value of a pixel in the target image block; and a second parameter value of the target pixel in the surface grayscale image is determined. , norm is the normalization function, m is the number of other pixels in the neighborhood of the target pixel, T is the first parameter value of the target pixel, H is the average value of the first parameter value of all pixels in the surface grayscale image, is the first parameter value of the qth other pixel in the neighborhood of the target pixel; according to the second parameter value of the pixel in the surface grayscale image, the rust area image where the rust is located in the surface grayscale image is obtained.

[0008] In this way, the pixel value of the pixel point in the target image is equal to the product of the grayscale value of the pixel point in the surface grayscale image of the copper clad laminate to be detected and the second parameter value. The surface rust of the copper clad laminate has a higher grayscale value in the rust surface grayscale image and a higher second parameter value. Therefore, the contrast between the pixel points of the rust and other pixel points other than the rust can be improved in the target image. Threshold segmentation of the target image can obtain a more accurate rust area image, and the rust area image is extracted according to the characteristics of the rust in the copper clad laminate. Compared with collecting a large number of diverse rust sample images for the copper clad laminate, the difficulty of rust detection on the unetched copper clad laminate is reduced.

[0009] The second parameter value of the target pixel is obtained by comparing the target pixel with the first parameter values ​​of other pixels. The first parameter value can combine the gradient direction change and grayscale value characteristics of the pixel. The first parameter value can enlarge the difference between the pixel of the rust in the surface grayscale image and the pixel of the copper clad laminate itself. Compared with the comparison between pixels only by grayscale value, there is a greater difference between the first parameter value of the pixel corresponding to the rust in the surface grayscale image and the first parameter value of the pixel of the copper clad laminate itself. The rust in the copper clad laminate can be determined more accurately, thereby reducing the impact of the rust that may exist in the copper clad laminate.

[0010] Optionally, the gradient direction difference value of the target image block is determined by: , where M is the gradient direction difference value of the target image block, n is the number of pixels in the image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

[0011] In this way, by comparing the gradient vector of the target image block and the gradient vector of the pixel point in the target image block, the obtained gradient direction difference value can better characterize the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel point in the target image block, so as to determine the rust area in the surface grayscale image according to the gradient direction difference value.

[0012] Optionally, the gradient vector of the target image block is determined in the following manner: determine the gradient vector of the pixel points in the target image block, perform principal component analysis on the gradient vectors of all the pixel points in the target image block, and use the vector obtained after the principal component analysis as the gradient vector of the target image block.

[0013] In this way, by performing principal component analysis on the gradient vectors of all pixels in the target image block, the gradient vector of the target image block obtained can better represent the gradient direction of the pixels in the target image block, so that the gradient vector of the target image block obtained can be easily compared with the gradient vector of the pixels in the target image block.

[0014] Optionally, the gradient vector of the target image block is determined in the following manner: determining the gradient direction of the pixel points in the target image block, and performing statistics on the gradient directions of all the pixel points in the target image block to obtain a gradient direction histogram; taking the gradient direction with the highest frequency in the gradient direction histogram as the target gradient direction, and taking the vector with a length equal to a preset length in the target gradient direction as the gradient vector of the target image block.

[0015] Optionally, the grayscale difference value of the target image block is determined by: , P is the grayscale difference value of the target image block, E is the average grayscale value of the target image block, n is the number of pixels in the target image block, is the gray value of the jth pixel in the target image block, and D is the median of the gray values ​​of the pixels in the target image block.

[0016] In this way, since the grayscale difference and the average grayscale value of the pixels in the image block containing rust and the image block not containing rust are different, the grayscale difference values ​​of the image block containing rust and the image block not containing rust are different. Therefore, the obtained grayscale difference value can facilitate the distinction between the image block with rust and the image block without rust.

[0017] Optionally, according to the second parameter value of the pixel point in the surface grayscale image, a rust area image where the rust is located in the surface grayscale image is obtained, including: taking the ratio of the grayscale value of the pixel point in the surface grayscale image to the corresponding second parameter value as the pixel value of the pixel point in the target image to map the surface grayscale image to the target image; and performing threshold segmentation on the target image to obtain the rust area image.

[0018] In this way, compared with the surface grayscale image, the difference between the pixel value of the pixel point with a larger second parameter value and the pixel value of the pixel point with a smaller second parameter value in the target image is amplified; in the obtained target image, the contrast between the pixel points with higher rust and the pixel points with lower rust is improved, and the rust that may exist on the surface of the copper clad laminate to be inspected can be more easily determined through the target image.

[0019] Optionally, threshold segmentation is performed on the target image to obtain a rust area image, including: using Otsu's method to determine a target segmentation threshold corresponding to the target image; using the target segmentation threshold to threshold segment the target image to obtain a rust area image, wherein the grayscale value of a pixel in the rust area image is less than or equal to the target segmentation threshold.

[0020] Optionally, according to the second parameter value of the pixel point in the surface grayscale image, a rust area image where the rust is located in the surface grayscale image is obtained, including: taking the pixel points in the surface grayscale image whose second parameter value is greater than or equal to a preset parameter threshold as the rust pixel points in the surface grayscale image, so as to obtain a rust area image composed of rust pixel points.

[0021] Optionally, the gradient direction difference value of the target image block is determined by: , where M is the gradient direction difference value of the target image block, n is the number of pixels in the image block, is the frequency ratio of the gradient direction of the j-th pixel in the target image block in the target image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

[0022] Optionally, the surface grayscale image of the copper clad laminate to be inspected is acquired in the following manner: acquiring a surface image of the copper clad laminate acquired by an image acquisition device for the copper clad laminate to be inspected; inputting the surface image of the copper clad laminate into a pre-trained copper clad laminate region extraction model to obtain a copper clad laminate region image, wherein the copper clad laminate region extraction model is used to set the pixel values ​​of other pixel points in the copper clad laminate surface image except for the region where the copper clad laminate is located to 0; and graying the copper clad laminate region image to obtain a surface grayscale image.

[0023] The technical solution provided by the embodiments of the present application may include the following beneficial effects: the pixel value of a pixel point in the target image is equal to the product of the grayscale value of the pixel point in the surface grayscale image of the copper clad laminate to be detected and the second parameter value, and the surface rust of the copper clad laminate has a higher grayscale value in the rust surface grayscale image and a higher second parameter value. Therefore, the contrast between the pixel points of the rust and other pixel points other than the rust can be improved in the target image, and a more accurate rust area image can be obtained by threshold segmentation of the target image. The rust area image is extracted based on the characteristics of the rust in the copper clad laminate, which reduces the difficulty of rust detection on unetched copper clad laminates compared to collecting a large number of diverse rust sample images for the copper clad laminates.

[0024] The second parameter value of the target pixel is obtained by comparing the target pixel with the first parameter values ​​of other pixels. The first parameter value can combine the gradient direction change and grayscale value characteristics of the pixel. The first parameter value can enlarge the difference between the pixel of the rust in the surface grayscale image and the pixel of the copper clad laminate itself. Compared with the comparison between pixels only by grayscale value, there is a greater difference between the first parameter value of the pixel corresponding to the rust in the surface grayscale image and the first parameter value of the pixel of the copper clad laminate itself. The rust in the copper clad laminate can be determined more accurately, thereby reducing the impact of the rust that may exist in the copper clad laminate.

[0025] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0027] Figure 1 The present invention is a flow chart showing a defect detection method for a copper clad laminate process according to an exemplary embodiment. DETAILED DESCRIPTION

[0028] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0029] First, the application scenario of the embodiment of the present application is briefly introduced. In the application scenario of the present application, the copper clad laminate may produce rust of the same or different degrees on the surface of the copper clad laminate due to poor storage conditions, such as high humidity in the storage environment. The conductivity of the area where the rust exists in the copper clad laminate will decrease, which may affect the conductivity of the conductive path formed by the copper clad laminate after etching, thereby affecting the conduction between the electronic components on the path where the conductive path is located.

[0030] In order to detect possible rust in the copper clad laminate, the detection of rust areas can be achieved through threshold segmentation in the relevant technology. However, the segmentation ability of threshold segmentation on rust areas is limited, and it is difficult to accurately detect possible rust areas in the copper clad laminate.

[0031] Alternatively, a large number of diverse rust sample images can be collected from the surface of the copper clad laminate. When the richness of the rust sample images meets certain requirements, the network model can be trained to obtain a recognition model with higher accuracy. However, in actual work, it is difficult to collect a large number of diverse rust sample images from the surface of the copper clad laminate. Therefore, it is difficult to detect the rust that may exist in the copper clad laminate.

[0032] In view of the above technical problems, the present application provides a defect detection method for a copper clad laminate process. Figure 1 is a flow chart of a defect detection method for a copper clad laminate process according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps.

[0033] In step S101, a surface grayscale image of a copper clad laminate to be inspected is obtained, and the surface grayscale image is divided into blocks to obtain a plurality of image blocks, so as to determine a grayscale difference value of a target image block among the plurality of image blocks.

[0034] In order to detect possible rust on the surface of the copper clad laminate to be inspected, the copper clad laminate to be inspected can be an unetched copper clad laminate, and the surface image of the copper clad laminate to be inspected can be acquired by an image acquisition device. After conversion processing of the surface image, a surface grayscale image of the copper clad laminate can be obtained, so that the surface grayscale image can be used to detect rust on the surface of the copper clad laminate.

[0035] By processing the surface grayscale image in blocks, for example, dividing the surface grayscale image into multiple image blocks of preset sizes, different image blocks together constitute a complete surface grayscale image. The preset size can be a pre-set or real-time set size such as 7×7, 8×8 and 9×9.

[0036] Since the rust in the copper clad laminate appears as a rust area, the surface grayscale image is divided into blocks to obtain multiple image blocks. This can facilitate obtaining more refined features of the pixel points in the image blocks obtained after the block processing, so as to more accurately detect the rust in the copper clad laminate.

[0037] The target image block may be any image block among the multiple image blocks, or the target image block may be any image block whose grayscale difference value has not been determined; the grayscale difference value is used to characterize the degree of difference in grayscale value of pixels in the target image block.

[0038] The larger the grayscale difference value of the target image block, the greater the difference in grayscale values ​​of the pixels in the target image block, and the higher the probability that the target image block contains rust in the copper clad laminate; conversely, the smaller the grayscale difference value of the target image block, the smaller the difference in grayscale values ​​of the pixels in the target image block, and the lower the probability that the target image block contains rust in the copper clad laminate.

[0039] In one embodiment, the grayscale difference value of the target image block is determined by: , P is the grayscale difference value of the target image block, E is the average grayscale value of the target image block, n is the number of pixels in the target image block, is the gray value of the jth pixel in the target image block, and D is the median of the gray values ​​of the pixels in the target image block.

[0040] When there is rust on the surface of the copper clad laminate, the rust has a weaker ability to reflect light than the pixels of the copper clad laminate itself, which makes the grayscale value of the rust pixels smaller in the grayscale image of the surface of the copper clad laminate, and makes the grayscale mean of the image block where the rust pixels are located smaller. Therefore, the smaller the grayscale mean of the target image block, the higher the probability that there are rust pixels in the target image block, or the greater the probability that there are deeper rust in the target image block.

[0041] The pixel values ​​of the pixels of the copper clad laminate without rust in the surface grayscale image of the copper clad laminate have high consistency, and since the degree of rust on the surface of the copper clad laminate may be more diverse, for example, there may be two or more areas with different degrees of rust on the surface of the copper clad laminate. Therefore, compared with the target image block that only includes the unrusted area, the difference between the grayscale values ​​of the pixels in the target image block that includes both the rusted area and the unrusted area is greater.

[0042] Here, the median can represent the general level of the grayscale values ​​of the pixels in the target image block. The grayscale values ​​of the pixels in the target image block are compared with the median of the grayscale values ​​of the target image block. The greater the difference between the grayscale values ​​of the pixels in the target image block and the median of the grayscale values ​​of the target image block, the more uneven the grayscale values ​​of the pixels in the target image block are, and the higher the probability of rust existing in the target image block.

[0043] The norm function is used to normalize the obtained grayscale difference value to the range of 0 to 1, so as to compare the grayscale difference values ​​of different image blocks in the surface grayscale image.

[0044] In this way, since the grayscale difference and the average grayscale value of the pixels in the image block containing rust and the image block not containing rust are different, the grayscale difference values ​​of the image block containing rust and the image block not containing rust are different. Therefore, the obtained grayscale difference value can facilitate the distinction between the image block with rust and the image block without rust.

[0045] In step S102, the gradient direction difference value of the target image block is determined.

[0046] The gradient direction difference value is used to characterize the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel points in the target image block; the larger the gradient direction difference value of the target image block, the greater the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel points in the target image block; conversely, the smaller the gradient direction difference value of the target image block, the smaller the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel points in the target image block.

[0047] When there is no rust in the target image block, since there is a high consistency between the pixel values ​​of the copper clad laminate itself, the gradient value of the pixel in the target image block is small, and the gradient direction of each pixel also has a high consistency.

[0048] When there is partial rust in the target image block, that is, when the target image block includes both pixels of the rust and the copper clad laminate itself, there is a certain difference in pixel value between the pixels of the copper clad laminate itself and the pixels of the rust, so that the target image block has a large grayscale change at least at the edge of the rust, or there is a certain grayscale difference between the pixels at the edge of the rust and the pixels inside the rust, so that the gradient directions of the pixels in the target image block are more diverse, and the gradient direction difference value becomes larger. Therefore, compared with the image block without rust, the gradient direction difference value of the target image block with partial rust is larger.

[0049] When the target image block is entirely rust, there are certain differences between the pixel values ​​of pixels at different positions in the rust. Compared with the image block without rust, the change direction of the pixel values ​​of the pixels in the target image block is more diverse, making the gradient direction of the pixels in the target image block more diverse. Therefore, compared with the image block without rust, the gradient direction difference value of the target image block entirely rust is greater.

[0050] By determining the gradient direction difference value of the target image block, the gradient direction difference value is used to characterize the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel point in the target image block; it helps to combine the gradient direction difference value of the target image block with the grayscale difference value to more accurately determine whether there is rust in the target image block, so as to more accurately detect rust in the grayscale image of the surface of the copper clad laminate.

[0051] In step S103, the product of the grayscale difference value and the gradient direction difference value of the target image block is used as the first parameter value of the pixel in the target image block to determine the second parameter value of the target pixel in the surface grayscale image.

[0052] The second parameter value of the target pixel in the surface grayscale image is determined in the following way: , Z is the second parameter value of the target pixel, norm is the normalization function, m is the number of other pixels in the neighborhood of the target pixel, T is the first parameter value of the target pixel, and H is the average value of the first parameter values ​​of all pixels in the surface grayscale image. is the first parameter value of the qth pixel in the neighborhood of the target pixel.

[0053] The size of the neighborhood range can be an area of ​​3×3, 5×5, etc. It should be understood that the size of the neighborhood range is smaller than the size of the image block obtained after the surface grayscale image is segmented. Those skilled in the art can adaptively set the size of the neighborhood range and the size of the image block according to actual needs, which will not be elaborated here.

[0054] In the denominator of the calculation formula for the second parameter value of the target pixel point, the first parameter value of the target pixel point is compared with the average value of the first parameter values ​​of all pixels in the surface grayscale image, so that the denominator of the calculation formula for the second parameter value can reflect the relative difference between the target pixel point and the overall surface grayscale image in the first parameter value.

[0055] In the numerator of the calculation formula of the second parameter value of the target pixel, the first parameter value of the target pixel is compared with the first parameter values ​​of other pixels in the neighborhood of the target pixel, so that the numerator of the calculation formula of the second parameter value can reflect the absolute difference between the target pixel and other pixels in the neighborhood of the target pixel in the first parameter value, so that the second parameter value can better reflect the relative difference between the pixel and the pixels in the neighborhood. Compared with the absolute difference obtained by only comparing the first parameter value of the pixel with other pixels in the neighborhood, the second parameter value can better reflect the probability that the target pixel belongs to rust.

[0056] For example, when the contrast of the surface grayscale image increases or decreases as a whole due to reasons such as lighting or shooting angle, the difference between the target pixel and other pixels in the neighborhood will increase or decrease. Since the denominator of the calculation formula of the second parameter value reflects the comparison result between the target pixel and the image as a whole, the second parameter value is still sufficient to better characterize the probability that the target pixel belongs to rust, so that the second parameter value can better characterize the probability that the target pixel belongs to rust in different usage scenarios, thereby improving the robustness of detecting rust that may exist on the surface of the copper clad laminate in different scenarios.

[0057] Furthermore, in the calculation formula for the second parameter value of the target pixel in the surface grayscale image, the denominator of the calculation formula for the second parameter value of the target pixel includes the first parameter values ​​of different pixels, and the numerator of the calculation formula for the second parameter value of the target pixel also includes the first parameter values ​​of different pixels. This can eliminate the dimension corresponding to the first parameter value in the obtained second parameter value, thereby facilitating better characterization of the probability that the target pixel belongs to rust through the second parameter value of the target pixel.

[0058] In this way, the first parameter value of the target pixel is compared with the average value of the first parameter values ​​of all pixels in the surface grayscale image, reflecting the overall relative difference between the target pixel and the surface grayscale image in the first parameter value; the first parameter value of the target pixel is compared with the first parameter values ​​of other pixels in the neighborhood, reflecting the overall absolute difference between the target pixel and other pixels in the neighborhood in the first parameter value, so that the obtained second parameter value can more accurately reflect the relative difference between the target pixel and other pixels in the neighborhood, and the second parameter value can more accurately characterize the probability that the target pixel belongs to rust on the surface of the copper clad laminate.

[0059] In step S104, a rust region image where the rust is located in the surface grayscale image is obtained according to the second parameter value of the pixel point in the surface grayscale image.

[0060] Since the second parameter value can more accurately characterize the probability that the target pixel point belongs to the rust on the surface of the copper clad laminate, the rust area image where the rust is located in the surface grayscale image can be obtained according to the second parameter value of the pixel point in the surface grayscale image, thereby realizing the detection of the rust area on the surface of the copper clad laminate.

[0061] For example, when the area of ​​the rust region image is greater than or equal to a preset area threshold, a first prompt information can be output, and the first prompt information is used to prompt that rust exists in the copper clad laminate to be detected; when the area of ​​the rust region image is less than the preset area threshold, a second prompt information can be output, and the second prompt information is used to prompt that there is no rust in the copper clad laminate to be detected.

[0062] Among them, the first prompt information or the second prompt information can be at least one of text prompt information, audio prompt information, light prompt information and vibration prompt information. The preset area threshold can be set according to the area of ​​the rust area image, and can also be adaptively configured according to actual needs. For example, the preset area threshold can be 0, which will not be repeated here.

[0063] When it is determined according to the rust region image that rust exists in the copper clad laminate to be detected, a control instruction may be sent to the rust removal mechanism so that the rust removal mechanism processes the rust in the copper clad laminate.

[0064] In one embodiment, according to the second parameter value of the pixel point in the surface grayscale image, a rust area image where the rust is located in the surface grayscale image is obtained, including: taking the ratio of the grayscale value of the pixel point in the surface grayscale image to the corresponding second parameter value as the pixel value of the pixel point in the target image to map the surface grayscale image to the target image; performing threshold segmentation on the target image to obtain the rust area image.

[0065] In the surface grayscale image of the copper clad laminate, the pixel values ​​of the pixels in the rusted area of ​​the copper clad laminate are larger than the pixel values ​​of the pixels of the copper clad laminate itself that has not rusted, and the pixel points with larger second parameter values ​​are more likely to be rusty. The ratio of the grayscale value of the pixel point in the surface grayscale image to the corresponding second parameter value is used as the pixel value of the pixel point in the target image.

[0066] In one embodiment, obtaining a rust region image where the rust is located in the surface grayscale image according to the second parameter value of the pixel point in the surface grayscale image includes:

[0067] The second parameter value of the pixel point in the surface grayscale image is normalized to the range of 0 to 1, and the difference between 1 and the normalized second parameter value of the pixel point is used as the weight value of the pixel point; the product of the grayscale value of the pixel point in the surface grayscale image and the corresponding weight is used as the pixel value of the pixel point in the target image to map the surface grayscale image to the target image; the target image is threshold segmented to obtain the rust area image.

[0068] In this way, compared with the surface grayscale image, the difference between the pixel value of the pixel point with a larger second parameter value and the pixel value of the pixel point with a smaller second parameter value in the target image is amplified; in the obtained target image, the contrast between the pixel points with higher rust and the pixel points with lower rust is improved, and the rust that may exist on the surface of the copper clad laminate to be inspected can be more easily determined through the target image.

[0069] In one embodiment, threshold segmentation is performed on a target image to obtain a rust area image, including: using Otsu's method to determine a target segmentation threshold corresponding to the target image; using the target segmentation threshold to threshold segment the target image to obtain a rust area image, wherein the grayscale value of a pixel in the rust area image is less than or equal to the target segmentation threshold.

[0070] Otsu's method is a method of image segmentation threshold that can adapt to the actual situation of the image. The purpose of Otsu's method is to minimize the intra-class variance and maximize the inter-class variance. In the target image obtained from the surface grayscale image of the copper clad laminate, the target image is segmented using the target segmentation threshold obtained by Otsu's method, which can better determine the rust area that may exist in the copper clad laminate to be detected.

[0071] In one embodiment, according to the second parameter value of the pixel point in the surface grayscale image, obtaining the rust area image where the rust is located in the surface grayscale image includes: taking the pixel points in the surface grayscale image whose second parameter value is greater than or equal to the preset parameter threshold as the rust pixel points in the surface grayscale image, so as to obtain the rust area image composed of the rust pixel points.

[0072] In this way, since the second parameter value can be used to characterize the probability that the target pixel point belongs to rust, the pixel points whose second parameter value in the surface grayscale image is greater than or equal to the preset parameter threshold are taken as the rust pixel points in the surface grayscale image, so that the image of the rust area that may exist in the copper clad laminate to be detected can be obtained more easily.

[0073] The preset parameter threshold can be set according to actual needs. For example, when the value range of the second parameter is between 0 and 1, the preset parameter threshold can be between 0.8 and 0.9.

[0074] Through the defect detection method of the copper clad laminate process provided by the embodiment of the present application, the pixel value of the pixel point in the target image is equal to the product of the grayscale value of the pixel point in the surface grayscale image of the copper clad laminate to be detected and the second parameter value, and the surface rust of the copper clad laminate has a higher grayscale value and a higher second parameter value in the rust surface grayscale image. Therefore, the contrast between the pixel points of the rust and other pixel points other than the rust can be improved in the target image, and a more accurate rust area image can be obtained by threshold segmentation of the target image. The rust area image is extracted according to the characteristics of the rust in the copper clad laminate, which reduces the difficulty of rust detection on unetched copper clad laminates compared to collecting a large number of diverse rust sample images for copper clad laminates.

[0075] The second parameter value of the target pixel is obtained by comparing the target pixel with the first parameter values ​​of other pixels. The first parameter value can combine the gradient direction change and grayscale value characteristics of the pixel. The first parameter value can enlarge the difference between the pixel of the rust in the surface grayscale image and the pixel of the copper clad laminate itself. Compared with the comparison between pixels only by grayscale value, there is a greater difference between the first parameter value of the pixel corresponding to the rust in the surface grayscale image and the first parameter value of the pixel of the copper clad laminate itself. The rust in the copper clad laminate can be determined more accurately, thereby reducing the impact of the rust that may exist in the copper clad laminate.

[0076] In one embodiment, the gradient direction difference value of the target image block is determined by: , where M is the gradient direction difference value of the target image block, n is the number of pixels in the image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

[0077] The gradient vector of the target image block can reflect the overall gradient direction of the pixels in the target image block. The gradient vector of the target image block and the gradient vector of the pixels in the target image block can be vectors whose lengths are normalized to the same length, so as to determine the diversity of the gradient directions in the target image.

[0078] Compared with the target image block without rust, the directions of the gradient vectors of the pixels in the target image block with rust are more diverse; after the gradient vector of the target image block and the gradient vector of the pixels in the target image block are subtracted, the larger the norm of the vector obtained by the vector subtraction operation, the greater the difference between the gradient vector of the target image block and the gradient vector of the pixels in the target image block, and the greater the degree of difference in direction between the gradient vector of the target image block and the gradient vector of the pixels in the target image block.

[0079] In this way, by comparing the gradient vector of the target image block and the gradient vector of the pixel point in the target image block, the obtained gradient direction difference value can better characterize the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel point in the target image block, so as to determine the rust area in the surface grayscale image according to the gradient direction difference value.

[0080] In one embodiment, the gradient direction difference value of the target image block is determined by: , where M is the gradient direction difference value of the target image block, n is the number of pixels in the image block, is the frequency ratio of the gradient direction of the j-th pixel in the target image block in the target image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

[0081] In the numerator of the calculation formula of the gradient direction difference value of the target image block, since the frequency ratio of the gradient direction of the pixel points in the target image block in the target image block is introduced, and the larger the frequency ratio of the gradient direction of the pixel points in the target image block in the target image block, the more dominant the gradient direction of the pixel points in the target image block is in the target image block, and the contribution of the gradient direction with a larger frequency ratio in the target image block to the gradient direction difference value of the target image block can be strengthened.

[0082] When the frequency of the gradient direction of the pixel points in the target image block accounts for a small proportion in the target image block, it means that the gradient vector of this part of the pixel points has little influence on the entire target image block. In addition, when there is rust in the target image block, the target image block usually includes multiple pixel points with grayscale changes, and the pixel points with grayscale changes do not exist alone in the target image block. If the frequency of the gradient direction of the pixel points in the target image block accounts for a small proportion in the target image block, the contribution of the gradient direction of this part of the pixel points with a small frequency proportion to the gradient direction difference value can be weakened.

[0083] because is the frequency ratio of the gradient direction of the j-th pixel in the target image block in the target image block. The cumulative sum of the frequency ratios of the gradient directions of all pixels in the target image block in the target image block may not be equal to 1. By using the denominator of the calculation formula for the gradient direction difference value of the target image block, the frequency ratio of the gradient direction of the pixel in the target image block in the target image block can be normalized to facilitate comparison of the gradient direction difference values ​​of different target image blocks.

[0084] In this way, since the frequency ratio of the gradient direction of the pixel points in the target image block is taken into account, a more accurate gradient direction difference value of the target image block can be obtained, so that the obtained gradient direction difference value can better characterize the degree of difference between the gradient vector of the target image block and the gradient vector of the pixel points in the target image block.

[0085] In one embodiment, the gradient vector of the target image block is determined in the following manner: the gradient vector of the pixel points in the target image block is determined, and the principal component analysis is performed on the gradient vectors of all the pixel points in the target image block, and the vector obtained after the principal component analysis is used as the gradient vector of the target image block.

[0086] In this way, by performing principal component analysis on the gradient vectors of all pixels in the target image block, the gradient vector of the target image block obtained can better represent the gradient direction of the pixels in the target image block, so that the gradient vector of the target image block obtained can be easily compared with the gradient vector of the pixels in the target image block.

[0087] In one embodiment, the gradient vector of the target image block is determined in the following manner: determining the gradient direction of the pixel points in the target image block, and obtaining a gradient direction histogram by statistically analyzing the gradient directions of all the pixel points in the target image block; taking the gradient direction with the highest frequency in the gradient direction histogram as the target gradient direction, and taking the vector with a length equal to a preset length in the target gradient direction as the gradient vector of the target image block.

[0088] In this way, by determining the gradient direction histogram corresponding to the gradient direction of the pixel points in the target image block, and taking the gradient direction with the highest frequency in the gradient direction histogram as the target gradient direction, the target gradient direction can better represent the gradient direction of the pixel points in the target image block, so that the gradient vector of the target image block obtained can be easily compared with the gradient vector of the pixel points in the target image block.

[0089] The preset length can be set according to actual needs, for example, the preset length can be set to 1.

[0090] In one embodiment, a surface grayscale image of a copper clad laminate to be inspected is acquired in the following manner: an image of a copper clad laminate surface acquired by an image acquisition device for the copper clad laminate to be inspected is acquired; the surface image of the copper clad laminate is input into a pre-trained copper clad laminate region extraction model to acquire a copper clad laminate region image, wherein the copper clad laminate region extraction model is used to set pixel values ​​of other pixel points in the copper clad laminate surface image except for the region where the copper clad laminate is located to 0; and the copper clad laminate region image is grayed to acquire a surface grayscale image.

[0091] In this way, by inputting the copper clad laminate surface image into the pre-trained copper clad laminate area extraction model to obtain the copper clad laminate area image, the adverse effects of the pixel values ​​of other pixel points in the copper clad laminate surface image except for the area where the copper clad laminate is located can be eliminated, so as to more accurately determine the rust area in the copper clad laminate.

[0092] The copper clad laminate surface image captured by the image acquisition device for the copper clad laminate to be inspected can be marked, and the pixel values ​​of other pixels in the copper clad laminate surface image except the area where the copper clad laminate is located can be marked as 0 to obtain a mask image corresponding to the copper clad laminate surface image.

[0093] By taking the copper clad laminate surface image as the input of a pre-built network model and taking the mask image corresponding to the copper clad laminate surface image as the output of the pre-built network model, the network model can be trained. The trained network model can be used as a copper clad laminate region extraction model, so that the copper clad laminate region extraction model can be used to process more copper clad laminate surface images.

[0094] It should be understood that the features of some embodiments of the various applications described herein may be combined with each other, unless specifically stated otherwise.

[0095] Although terms such as "first", "second", and "third" may be used herein to describe various components, parts, regions, layers, or sections, these components, parts, regions, layers, or sections are not limited to these terms. Instead, these terms are only used to distinguish one component, part, region, layer, or section from another component, part, region, layer, or section. Therefore, without departing from the teachings of the examples described herein, the first component, part, region, layer, or section mentioned in the examples may also be referred to as the second component, part, region, layer, or section.

[0096] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description herein, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0097] In addition, the word "exemplary" is used herein to mean serving as an example, instance, or diagram. Any aspect or design described herein as "exemplary" is not necessarily to be understood as being advantageous over other aspects or designs. Instead, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or".

[0098] Likewise, although the present application has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. With particular regard to the various functions performed by the components (e.g., elements, resources, etc.) described above, unless otherwise noted, the terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific functions of the described components, even if not structurally equivalent to the disclosed structures.

[0099] Additionally, while particular features of the present application may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and advantageous for any given or particular application.

[0100] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary technical means in the art that are not disclosed in the present application, and the specification and embodiments are only considered as exemplary.

[0101] It should be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A defect detection method for a copper clad laminate process, used to detect surface rust defects on an unetched copper clad laminate, characterized in that: include: Acquire a surface grayscale image of the copper clad laminate to be inspected, and perform block processing on the surface grayscale image to obtain multiple image blocks, so as to determine a grayscale difference value of a target image block among the multiple image blocks; the grayscale difference value is used to characterize the degree of difference in grayscale value of pixels in the target image block; Determine a gradient direction difference value of a target image block, where the gradient direction difference value is used to characterize the degree of difference between a gradient vector of the target image block and a gradient vector of a pixel point in the target image block; The product of the grayscale difference value of the target image block and the gradient direction difference value is used as the first parameter value of the pixel in the target image block to determine the second parameter value of the target pixel in the surface grayscale image. , norm is the normalization function, m is the number of other pixels in the neighborhood of the target pixel, T is the first parameter value of the target pixel, H is the average value of the first parameter value of all pixels in the surface grayscale image, is the first parameter value of the qth pixel in the neighborhood of the target pixel; According to the second parameter value of the pixel point in the surface grayscale image, an image of a rust region where the rust is located in the surface grayscale image is obtained; Among them, the gradient direction difference value of the target image block , n is the number of pixels in the image block, is the frequency ratio of the gradient direction of the j-th pixel in the target image block in the target image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

2. The defect detection method of the copper clad laminate process according to claim 1, characterized in that: The gradient direction difference value of the target image block is determined in the following way: , where M is the gradient direction difference value of the target image block, n is the number of pixels in the image block, is the gradient vector of the j-th pixel in the target image block, is the gradient vector of the target image block, is the norm of the vector.

3. The defect detection method for copper clad laminate process according to claim 1, characterized in that: The gradient vector of the target image block is determined as follows: The gradient vector of the pixel points in the target image block is determined, and the principal component analysis is performed on the gradient vectors of all the pixel points in the target image block, and the vector obtained after the principal component analysis is used as the gradient vector of the target image block.

4. The defect detection method for copper clad laminate process according to claim 1, characterized in that: The gradient vector of the target image block is determined as follows: Determine the gradient direction of the pixel points in the target image block, and obtain a gradient direction histogram by statistically analyzing the gradient directions of all the pixel points in the target image block; The gradient direction with the highest frequency in the gradient direction histogram is taken as the target gradient direction, and the vector with a length equal to a preset length in the target gradient direction is taken as the gradient vector of the target image block.

5. The defect detection method of the copper clad laminate process according to claim 1, characterized in that: The grayscale difference value of the target image block is determined in the following way: , P is the grayscale difference value of the target image block, E is the average grayscale value of the target image block, n is the number of pixels in the target image block, is the gray value of the jth pixel in the target image block, and D is the median of the gray values ​​of the pixels in the target image block.

6. The defect detection method for copper clad laminate process according to claim 1, characterized in that: According to the second parameter value of the pixel point in the surface grayscale image, an image of a rust region where the rust is located in the surface grayscale image is obtained, including: The ratio of the gray value of a pixel point in the surface gray image to the corresponding second parameter value is used as the pixel value of the pixel point in the target image to map the surface gray image into the target image; Perform threshold segmentation on the target image to obtain the rust area image.

7. The defect detection method for copper clad laminate process according to claim 6, characterized in that: Perform threshold segmentation on the target image to obtain the rust area image, including: The target segmentation threshold corresponding to the target image is determined using the Otsu method; The target image is subjected to threshold segmentation using the target segmentation threshold to obtain a rust region image; the grayscale value of the pixel point in the rust region image is less than or equal to the target segmentation threshold.

8. The defect detection method for copper clad laminate process according to claim 1, characterized in that: According to the second parameter value of the pixel point in the surface grayscale image, an image of a rust region where the rust is located in the surface grayscale image is obtained, including: The pixel points in the surface grayscale image whose second parameter value is greater than or equal to the preset parameter threshold are taken as the rust pixel points in the surface grayscale image to obtain a rust area image composed of rust pixel points.

9. The defect detection method for copper clad laminate process according to claim 1, characterized in that: The surface grayscale image of the copper clad laminate to be inspected is obtained in the following way: Acquire a surface image of the copper clad laminate to be inspected, which is captured by an image acquisition device; Inputting the copper clad laminate surface image into a pre-trained copper clad laminate region extraction model to obtain a copper clad laminate region image, wherein the copper clad laminate region extraction model is used to set the pixel values ​​of other pixel points in the copper clad laminate surface image except for the region where the copper clad laminate is located to 0; The copper clad laminate area image is grayed to obtain a surface grayscale image.

Images