Metallized film for buffer capacitor and preparation method thereof

Abstract

Disclosed are a metallized film for a buffer capacitor and a method for preparing the same. The method comprises: providing an organic plastic film as a dielectric film; depositing a metallized layer on the dielectric film to obtain a metallized film; and cutting the metallized film into predetermined dimensions. In this manner, the metallized film for the buffer capacitor is produced.

Description

Technical Field

[0001] The present application relates to the field of intelligent production, and more specifically, to a metallized film for buffer capacitors and a preparation method thereof. Background Art

[0002] The metallized film used in snubber capacitors is a composite material consisting of a metal layer deposited on a dielectric film, which improves the performance and reliability of the capacitor. Metallized film capacitors are characterized by their self-healing ability. When the capacitor experiences a local breakdown due to overvoltage or defects, the metal layer at the breakdown point melts and evaporates due to arcing or short-circuit current, forming a metal-free zone. This restores the insulation state and allows the capacitor to continue operating. This improves the lifespan and stability of the capacitor.

[0003] The metallized film manufacturing process involves heating a high-purity metal in a vacuum, vaporizing it and depositing it onto a dielectric film. The film then undergoes slitting, inspection, and packaging. During the metallized film manufacturing process, the surface morphology and quality of the metallized layer affect its electrical performance and stability. Therefore, during the production process, it is desirable to inspect the surface characteristics of the metallized layer to determine whether its molding quality meets predetermined standards. Summary of the Invention

[0004] To address the above-mentioned technical problems, the present application is proposed. Embodiments of the present application provide a metallized film for a buffer capacitor and a method for preparing the same. The method comprises: providing an organic plastic film as a dielectric film; depositing a metallized layer on the dielectric film to obtain a metallized film; and cutting the metallized film into predetermined dimensions. In this manner, a metallized film for a buffer capacitor is produced.

[0005] According to one aspect of the present application, a method for preparing a metallized film for a buffer capacitor is provided, comprising:

[0006] Providing an organic plastic film as a dielectric film;

[0007] depositing a metallization layer on the dielectric film to obtain a metallized film; and

[0008] The metallized film is cut into predetermined sizes.

[0009] In the above-mentioned method for preparing the metallized film for the buffer capacitor, depositing a metallized layer on the dielectric film to obtain the metallized film comprises:

[0010] Acquiring a surface image of the metallization layer captured by a camera;

[0011] Calculating a directional gradient histogram of a surface image of the metallized layer;

[0012] aggregating the surface image of the metallized layer and the histogram of directional gradients to obtain a multi-channel image;

[0013] Performing image block processing on the multi-channel image to obtain a sequence of image blocks;

[0014] Passing the sequence of image blocks through a ViT model including an embedding layer to obtain a plurality of context image block semantic feature vectors;

[0015] Calculating the cosine similarity between every two context image block semantic feature vectors in the plurality of context image block semantic feature vectors to obtain a classification feature vector composed of a plurality of cosine similarities;

[0016] Performing probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector; and

[0017] The optimized classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the surface quality of the metallization layer meets a predetermined standard.

[0018] In the above-mentioned method for preparing the metallized film for buffer capacitors, performing image block processing on the multi-channel image to obtain a sequence of image blocks includes:

[0019] The multi-channel image is uniformly divided into blocks to obtain a sequence of image blocks, wherein each image block in the sequence of image blocks has the same size.

[0020] In the above-mentioned method for preparing the metallized film for buffer capacitors, the sequence of image blocks is passed through a ViT model including an embedding layer to obtain a plurality of context image block semantic feature vectors, including:

[0021] Using the embedding layer to perform embedding coding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors; and

[0022] The ViT model is used to perform contextual semantic encoding based on a self-attention mechanism on the sequence of the image block embedding vectors to obtain the multiple contextual image block semantic feature vectors.

[0023] In the above-mentioned method for preparing the metallized film for the buffer capacitor, using the embedding layer to embed the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors includes:

[0024] Expanding the two-dimensional pixel matrix of each image block into a one-dimensional pixel expansion feature vector; and

[0025] The one-dimensional pixel expansion feature vector is fully connected encoded using the embedding layer to obtain the image block embedding vector.

[0026] In the above-mentioned method for preparing the metallized film for buffer capacitors, using the ViT model to perform contextual semantic encoding based on the self-attention mechanism on the sequence of image block embedding vectors to obtain the multiple contextual image block semantic feature vectors, including:

[0027] Arranging the sequence of image block embedding vectors in one dimension to obtain a global image block vector;

[0028] Calculating the product of the global image block vector and the transposed vector of each image block embedding vector in the sequence of image block embedding vectors to obtain a plurality of self-attention association matrices;

[0029] Normalizing each of the plurality of self-attention association matrices to obtain a plurality of normalized self-attention association matrices;

[0030] Passing each of the plurality of normalized self-attention association matrices through a Softmax classification function to obtain a plurality of probability values; and

[0031] Each image block embedding vector in the sequence of image block embedding vectors is weighted using each probability value in the multiple probability values as a weight to obtain the multiple context image block semantic feature vectors.

[0032] In the above-mentioned method for preparing the metallized film for the buffer capacitor, calculating the cosine similarity between every two context image block semantic feature vectors in the plurality of context image block semantic feature vectors to obtain a classification feature vector composed of the plurality of cosine similarities includes:

[0033] Calculating the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors using the following cosine formula to obtain multiple cosine similarities;

[0034] Wherein, the cosine formula is:

[0035]

[0036] Among them, V i and V j represents every two context image block semantic feature vectors in the multiple context image block semantic feature vectors, and Represents the semantic feature vector V of the context image block i and V j The characteristic value of each position in d(Vi ,V j ) represents the cosine distance between any two context image block semantic feature vectors among the multiple context image block semantic feature vectors; and

[0037] The multiple cosine similarities are arranged to obtain the classification feature vector.

[0038] In the above-mentioned method for preparing the metallized film for the buffer capacitor, performing probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector includes:

[0039] The probability density discreteness of the classification feature vector is optimized using the following optimization formula to obtain an optimized classification feature vector;

[0040] Wherein, the optimization formula is:

[0041]

[0042] Wherein, μ and σ are the mean and standard deviation of the eigenvalue set at each position in the classification feature vector, V is the classification feature vector, and v i is the eigenvalue of the ith position of the classification feature vector, and v i ′ is the eigenvalue of the i-th position of the optimized classification feature vector.

[0043] In the above-mentioned method for preparing a metallized film for a buffer capacitor, the optimized classification feature vector is passed through a classifier to obtain a classification result. The classification result is used to indicate whether the surface quality of the metallized layer meets a predetermined standard, including:

[0044] Performing full-connection encoding on the optimized classification feature vector using multiple fully-connected layers of the classifier to obtain an encoded classification feature vector; and

[0045] The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

[0046] According to another aspect of the present application, a metallized film for a buffer capacitor is provided. The metallized film for a buffer capacitor is produced by the above-mentioned method for producing a metallized film for a buffer capacitor.

[0047] Compared to the prior art, the present application provides a metallized film for a buffer capacitor and a method for preparing the same. The method comprises: providing an organic plastic film as a dielectric film; depositing a metallized layer on the dielectric film to obtain a metallized film; and cutting the metallized film into predetermined sizes. In this manner, a metallized film for a buffer capacitor is produced. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the technical solutions of the embodiments of this application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of this application. A person skilled in the art can derive other drawings based on these drawings without inventive effort. The following drawings are not intentionally scaled to actual size; their focus is on illustrating the main purpose of this application.

[0049] Figure 1 Flowchart of a method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0050] Figure 2 This is an application scenario diagram of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0051] Figure 3 Flowchart of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0052] Figure 4 FIG. 1 is a schematic diagram of the structure of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0053] Figure 5 Flowchart of sub-step S125 of the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0054] Figure 6 This is a flow chart of sub-step S1251 of the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0055] Figure 7 Flowchart of sub-step S1252 of the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0056] Figure 8 Flowchart of sub-step S128 of the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application.

[0057] Figure 9 4 is a block diagram of a system for preparing a metallized film for a buffer capacitor according to an embodiment of the present application. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts also fall within the scope of protection of this application.

[0059] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0060] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0061] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0062] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0063] like Figure 1 As shown, in the technical solution of the present application, the method for preparing the metallized film for the buffer capacitor includes: S110, providing an organic plastic film as a dielectric film; S120, depositing a metallized layer on the dielectric film to obtain a metallized film; and, S130, cutting the metallized film into a predetermined size.

[0064] In response to the above technical requirements, the technical concept of the present application is: to collect the surface image of the metallized layer through a camera, and use the converter-based ViT model to process and analyze the surface image of the metallized layer to obtain an image analysis result indicating whether the surface quality of the metallized layer meets the predetermined standards. In this way, during the preparation process of the metallized layer, the surface features of the metallized layer are detected to ensure product quality.

[0065] Specifically, in the technical solution of this application, a surface image of the metallized layer is first acquired by a camera. Acquiring the surface image of the metallized layer is a prerequisite for determining the surface quality of the metallized layer. During the preparation of the metallized layer, parameters such as the thickness and composition of the metal film significantly impact the electrical properties and stability of the film. Therefore, during the preparation process, it is necessary to promptly acquire the surface image of the metallized layer in order to further determine whether the quality of the metallized layer meets predetermined standards.

[0066] Next, a histogram of oriented gradients of the surface image of the metallized layer is calculated. Considering that, in the technical solution of this application, the texture features of the metallized layer are of great significance for determining whether its molding quality meets predetermined standards, the purpose of calculating the histogram of oriented gradients of the surface image of the metallized layer in the technical solution of this application is to extract feature information from aspects such as angles, edges, and texture of the metallized layer surface.

[0067] Specifically, the Histogram of Oriented Gradients (HOG) is a commonly used feature description method that can perform statistics on local gradients in an image and represent the gradient distribution through a histogram. By calculating the histogram of directional gradients for the surface image of the metallization layer, the local features of the image can be effectively extracted, including feature information such as edges, angles, and textures. At the same time, the histogram of directional gradients is also invariant, that is, images with the same gradient distribution pattern have similar histograms of directional gradients. Therefore, during the preparation process of the metallization layer, different surface images of the metallization layer may have some common or similar features. By calculating the histogram of directional gradients, these common or similar features can be extracted.

[0068] Next, the surface image of the metallization layer and the histogram of oriented gradients are aggregated to produce a multi-channel image. Here, the surface image of the metallization layer and the histogram of oriented gradients are aggregated along the channel dimension, combining two different types of information to form a single image containing multiple channels, which better describes the characteristic information of the metallization layer surface. Specifically, the surface image of the metallization layer provides more intuitive and comprehensive surface information of the metallization layer, while the histogram of oriented gradients provides more localized and detailed surface information of the metallization layer.

[0069] Next, the multi-channel image is subjected to image block processing to obtain a sequence of image blocks, and the sequence of image blocks is passed through a ViT model including an embedding layer to obtain multiple contextual image block semantic feature vectors. In particular, in the technical solution of the present application, the surface quality of the metallized layer can be represented by the difference and consistency information between the local regions of the metallized layer. It should be understood that if the molding quality of the metallized layer meets predetermined requirements, the consistency between the local regions of the metallized layer is better, that is, the difference is smaller.

[0070] Specifically, in the technical solution of the present application, the multi-channel image is first subjected to uniform image block processing to obtain a sequence of image blocks, wherein the image blocks in the sequence of image blocks have the same size. Then, the ViT model containing the embedding layer is used to perform contextual semantic association encoding based on the transformer mechanism on the sequence of image blocks to obtain the multiple contextual image block semantic feature vectors. It is worth mentioning here that ViT is a deep learning model based on the Transformer architecture, which is mainly used to process visual tasks. It adopts a structure similar to the Transformer encoder in the field of natural language processing, takes the image block sequence as input, and converts it into a set of high-level semantic feature vector sequences. Unlike traditional convolutional neural networks, ViT can capture more global and long-range image information, and has stronger generalization and adaptability.

[0071] Then, the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors is calculated to obtain a classification feature vector composed of multiple cosine similarities. That is, in the technical solution of the present application, the surface quality of the metallization layer is represented by the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors, which can be expressed by the difference and consistency information between the local areas of the metallization layer. Cosine similarity is a common similarity measurement method that can be used to measure the degree of similarity between two vectors. In the present technical solution, for the semantic feature vector of each context image block, the cosine similarity with other context image blocks can be calculated to form a classification feature vector composed of multiple cosine similarities. Such a classification feature vector can reflect the similarities and differences between different context image blocks, further extract important feature information of the surface quality, and thus more accurately distinguish the quality of the surface of the metallization layer.

[0072] The classification feature vector is then passed through a classifier to obtain a classification result, which indicates whether the surface quality of the metallization layer meets a predetermined standard. In other words, the classifier is used to determine a class probability label to which the classification feature vector belongs, wherein the class probability label indicates whether the surface quality of the metallization layer meets the predetermined standard.

[0073] More specifically, the classifier model is a machine learning model that automatically learns the mapping relationship between features and labels. In this technical solution, a large amount of metallization layer surface data and corresponding label information can be used to train a classifier model using modern machine learning algorithms. This classifier model can then determine whether the quality of the metallization layer surface meets predetermined standards based on the classification feature vector.

[0074] In particular, in the technical solution of the present application, for performing image block processing on the multi-channel image to obtain a sequence of image blocks, since the image block processing cannot guarantee the uniformity of the source image semantic similarity between each image block, after the sequence of image blocks passes through the ViT model including the embedding layer, although the image blocks are subjected to image semantic context association encoding of the source image semantics, the image feature semantic similarities of the obtained multiple context image block semantic feature vectors still have large non-uniformity, resulting in the overall feature distribution of the classification feature vector composed of multiple cosine similarities obtained by calculating the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors having probability density discreteness in the class probability expression of the classifier, that is, the convergence in the probability density space is poor, which affects the classification effect of the classification feature vector passing through the classifier.

[0075] Therefore, preferably, the applicant of the present application performs a baseline meshing of the manifold surface of Gaussian probability density on the classification feature vector, which is specifically expressed as follows:

[0076]

[0077] where μ and σ are the eigenvalue sets v i ∈V, where V is the classification feature vector, and v i ′ is the eigenvalue of the i-th position of the optimized classification feature vector.

[0078] Here, the baseline meshing of the manifold surface of the Gaussian probability density uses the statistical characteristics of the high-dimensional feature set of the classification feature vector, namely the mean and standard deviation, as the benchmark anchor points for the probability density measurement, and performs meshing along the local linear embedding direction of the manifold surface to obtain a low-dimensional constrained expression of the neighborhood network of the local probability density extreme values, thereby reconstructing the probability density expression of the manifold surface to constrain the local distribution of the high-dimensional features based on the benchmark-based relative spatial position relationship based on the neighborhood distribution. In this way, the spatial convergence of the probability density of the high-dimensional features of the classification feature vector is improved, thereby improving the classification effect of the classification feature vector through the classifier, namely the training speed of the classifier and the accuracy of the classification results.

[0079] Figure 2 FIG is an application scenario diagram of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application. Figure 2 As shown, in this application scenario, first, obtain the image data from the camera (for example, Figure 2 C) collected metallization layer (e.g., Figure 2 N) shown in FIG (e.g., Figure 2 D), then, the surface image of the metallized layer is input into a server that is equipped with a preparation algorithm for the metallized film for buffer capacitors (e.g., Figure 2 S) as shown in , wherein the server is capable of processing the surface image of the metallization layer using the preparation algorithm of the metallization film for the buffer capacitor to obtain a classification result indicating whether the surface quality of the metallization layer meets a predetermined standard.

[0080] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0081] Figure 3 FIG. 1 is a flow chart of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application. Figure 3As shown, in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application, the step of depositing a metallized layer on the dielectric film to obtain a metallized film includes: S121, obtaining a surface image of the metallized layer captured by a camera; S122, calculating a directional gradient histogram of the surface image of the metallized layer; S123, aggregating the surface image of the metallized layer and the directional gradient histogram to obtain a multi-channel image; S124, performing image block processing on the multi-channel image to obtain a sequence of image blocks; S125, passing the sequence of image blocks through a ViT model including an embedding layer to obtain multiple context image block semantic feature vectors; S126, calculating the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors to obtain a classification feature vector composed of multiple cosine similarities; S127, performing probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector; and S128, passing the optimized classification feature vector through a classifier to obtain a classification result, which is used to indicate whether the surface quality of the metallized layer meets a predetermined standard.

[0082] Figure 4 FIG. 1 is a schematic diagram of the structure of step S120 in the method for preparing a metallized film for a buffer capacitor according to an embodiment of the present application. Figure 4 As shown, in this network architecture, first, a surface image of the metallization layer captured by a camera is obtained; then, a histogram of directional gradients of the surface image of the metallization layer is calculated; then, the surface image of the metallization layer and the histogram of directional gradients are aggregated to obtain a multi-channel image; then, the multi-channel image is subjected to image block processing to obtain a sequence of image blocks; then, the sequence of image blocks is passed through a ViT model containing an embedding layer to obtain a plurality of context image block semantic feature vectors; then, the cosine similarity between each two context image block semantic feature vectors in the plurality of context image block semantic feature vectors is calculated to obtain a classification feature vector composed of a plurality of cosine similarities; then, the classification feature vector is subjected to probability density discreteness optimization to obtain an optimized classification feature vector; finally, the optimized classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the surface quality of the metallization layer meets a predetermined standard.

[0083] More specifically, in step S121, a surface image of the metallized layer is captured by a camera. Acquiring the surface image of the metallized layer is a prerequisite for determining the surface quality of the metallized layer. During the metallized layer preparation process, parameters such as the thickness and composition of the metal film significantly impact the film's electrical properties and stability. Therefore, it is necessary to promptly acquire the surface image of the metallized layer during the preparation process to further determine whether the metallized layer's quality meets predetermined standards.

[0084] More specifically, in step S122, a histogram of oriented gradients (HGs) of the surface image of the metallized layer is calculated. The texture features of the metallized layer are of great significance for determining whether its molding quality meets predetermined standards. Therefore, in the technical solution of the present application, the purpose of calculating the HGs of the surface image of the metallized layer is to extract feature information from aspects such as the angle, edge, and texture of the metallized layer surface. Specifically, the Histogram of Oriented Gradients (HOG) is a commonly used feature description method that can statistically analyze local gradients in an image and represent the gradient distribution using a histogram. Calculating the HGs of the surface image of the metallized layer can effectively extract local features of the image, including feature information such as edges, angles, and texture. Furthermore, the HGs of HOGs are invariant, meaning that images with the same gradient distribution pattern have similar HGs. Therefore, during the metallized layer preparation process, different surface images of the metallized layer may share common or similar features. By calculating the HGs of HOGs, these common or similar features can be extracted.

[0085] More specifically, in step S123, the surface image of the metallization layer and the histogram of oriented gradients are aggregated to obtain a multi-channel image. By aggregating the surface image of the metallization layer and the histogram of oriented gradients along the channel dimension, the two different types of information are combined to form an image containing multiple channels to better describe the characteristic information of the metallization layer surface. Specifically, the surface image of the metallization layer provides more intuitive and comprehensive surface information of the metallization layer, while the histogram of oriented gradients provides more localized and detailed surface information of the metallization layer.

[0086] More specifically, in step S124, the multi-channel image is subjected to image block processing to obtain a sequence of image blocks. The surface quality of the metallized layer can be represented by the difference and consistency information between each local area of the metallized layer. It should be understood that if the forming quality of the metallized layer meets the predetermined requirements, the consistency between each local area of the metallized layer is better, that is, the difference is smaller.

[0087] Accordingly, in a specific example, performing image block processing on the multi-channel image to obtain a sequence of image blocks includes: performing uniform block processing on the multi-channel image to obtain a sequence of image blocks, wherein each image block in the sequence of image blocks has the same size.

[0088] More specifically, in step S125, the sequence of image blocks is passed through a ViT model including an embedding layer to obtain a plurality of contextual image block semantic feature vectors. The sequence of image blocks is subjected to contextual semantic association encoding based on a transformer mechanism using the ViT model including an embedding layer to obtain the plurality of contextual image block semantic feature vectors. It is worth mentioning here that ViT is a deep learning model based on the Transformer architecture, which is mainly used to process visual tasks. It adopts a structure similar to the Transformer encoder in the field of natural language processing, takes an image block sequence as input, and converts it into a set of high-level semantic feature vector sequences. Unlike traditional convolutional neural networks, ViT can capture more global and long-range image information, and has stronger generalization and adaptability.

[0089] Accordingly, in a specific example, Figure 5 As shown, the sequence of image blocks is passed through a ViT model including an embedding layer to obtain a plurality of contextual image block semantic feature vectors, including: S1251, using the embedding layer to perform embedding encoding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors; and, S1252, using the ViT model to perform contextual semantic encoding based on a self-attention mechanism on the sequence of image block embedding vectors to obtain the plurality of contextual image block semantic feature vectors.

[0090] Accordingly, in a specific example, Figure 6 As shown, the embedding layer is used to perform embedding encoding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors, including: S12511, expanding the two-dimensional pixel matrix of each image block into a one-dimensional pixel expansion feature vector; and, S12512, using the embedding layer to perform fully connected encoding on the one-dimensional pixel expansion feature vector to obtain the image block embedding vector.

[0091] Accordingly, in a specific example, Figure 7As shown, the ViT model is used to perform contextual semantic encoding based on the self-attention mechanism on the sequence of image block embedding vectors to obtain the multiple context image block semantic feature vectors, including: S12521, arranging the sequence of image block embedding vectors in one dimension to obtain a global image block vector; S12522, calculating the product between the global image block vector and the transposed vector of each image block embedding vector in the sequence of image block embedding vectors to obtain multiple self-attention association matrices; S12523, respectively normalizing each self-attention association matrix in the multiple self-attention association matrices to obtain multiple standardized self-attention association matrices; S12524, passing each standardized self-attention association matrix in the multiple standardized self-attention association matrices through a Softmax classification function to obtain multiple probability values; and, S12525, respectively weighting each image block embedding vector in the sequence of image block embedding vectors using each probability value in the multiple probability values as a weight to obtain the multiple context image block semantic feature vectors.

[0092] More specifically, in step S126, the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors is calculated to obtain a classification feature vector composed of multiple cosine similarities. That is, in the technical solution of the present application, the surface quality of the metallization layer is represented by the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors, which can be expressed by the difference and consistency information between the local areas of the metallization layer. Cosine similarity is a common similarity measurement method that can be used to measure the degree of similarity between two vectors. In the present technical solution, for the semantic feature vector of each context image block, the cosine similarity with other context image blocks can be calculated to form a classification feature vector composed of multiple cosine similarities.

[0093] Accordingly, in a specific example, calculating the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors to obtain a classification feature vector composed of multiple cosine similarities includes: calculating the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors using the following cosine formula to obtain multiple cosine similarities; wherein the cosine formula is:

[0094]

[0095] Among them, V i and V j represents every two context image block semantic feature vectors in the multiple context image block semantic feature vectors, and Represents the semantic feature vector V of the context image block i and V j The characteristic value of each position in d(V i ,V j ) represents the cosine distance between any two context image block semantic feature vectors among the multiple context image block semantic feature vectors; and, the multiple cosine similarities are arranged to obtain the classification feature vector.

[0096] More specifically, in step S127 , the probability density discreteness of the classification feature vector is optimized to obtain an optimized classification feature vector.

[0097] In particular, in the technical solution of the present application, for performing image block processing on the multi-channel image to obtain a sequence of image blocks, since the image block processing cannot guarantee the uniformity of the source image semantic similarity between each image block, after the sequence of image blocks passes through the ViT model containing the embedding layer, although the image blocks are subjected to image semantic context association encoding of the source image semantics, the image feature semantic similarities of the obtained multiple context image block semantic feature vectors still have large non-uniformity, resulting in the overall feature distribution of the classification feature vector composed of multiple cosine similarities obtained by calculating the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors having probability density discreteness in the class probability expression of the classifier, that is, poor convergence in the probability density space, affecting the classification effect of the classification feature vector passing through the classifier. Therefore, preferably, the applicant of the present application performs baseline meshing of the manifold surface of the Gaussian probability density on the classification feature vector.

[0098] Accordingly, in a specific example, performing probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector includes: performing probability density discreteness optimization on the classification feature vector using the following optimization formula to obtain an optimized classification feature vector; wherein the optimization formula is:

[0099]

[0100] Wherein, μ and σ are the mean and standard deviation of the eigenvalue set at each position in the classification feature vector, V is the classification feature vector, and v i is the eigenvalue of the ith position of the classification feature vector, and v i ′ is the eigenvalue of the i-th position of the optimized classification feature vector.

[0101] Here, the baseline meshing of the manifold surface of the Gaussian probability density uses the statistical characteristics of the high-dimensional feature set of the classification feature vector, namely the mean and standard deviation, as the benchmark anchor points for the probability density measurement, and performs meshing along the local linear embedding direction of the manifold surface to obtain a low-dimensional constrained expression of the neighborhood network of the local probability density extreme values, thereby reconstructing the probability density expression of the manifold surface to constrain the local distribution of the high-dimensional features based on the benchmark-based relative spatial position relationship based on the neighborhood distribution. In this way, the spatial convergence of the probability density of the high-dimensional features of the classification feature vector is improved, thereby improving the classification effect of the classification feature vector through the classifier, namely the training speed of the classifier and the accuracy of the classification results.

[0102] More specifically, in step S128, the optimized classification feature vector is passed through a classifier to obtain a classification result, which is used to indicate whether the surface quality of the metallization layer meets the predetermined standard. In other words, the classifier is used to determine the class probability label to which the classification feature vector belongs, wherein the class probability label is used to indicate whether the surface quality of the metallization layer meets the predetermined standard.

[0103] That is, in the technical solution of the present application, the labels of the classifier include whether the surface quality of the metallization layer meets the predetermined standard (first label), and whether the surface quality of the metallization layer does not meet the predetermined standard (second label), wherein the classifier determines to which classification label the classification feature vector belongs by a soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether the surface quality of the metallization layer meets the predetermined standard". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the surface quality of the metallization layer meets the predetermined standard is actually converted from the classification label to a binary classification class probability distribution that conforms to natural laws. In essence, what is used is the physical meaning of the natural probability distribution of the label, rather than the linguistic text meaning of "whether the surface quality of the metallization layer meets the predetermined standard".

[0104] More specifically, the classifier model is a machine learning model that automatically learns the mapping relationship between features and labels. In this technical solution, a large amount of metallization layer surface data and corresponding label information can be used to train a classifier model using modern machine learning algorithms. This classifier model can then determine whether the quality of the metallization layer surface meets predetermined standards based on the classification feature vector.

[0105] Accordingly, in a specific example, Figure 8As shown, the optimized classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the surface quality of the metallization layer meets a predetermined standard, including: S1281, using multiple fully connected layers of the classifier to perform fully connected encoding on the optimized classification feature vector to obtain an encoded classification feature vector; and, S1282, passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0106] In summary, based on the preparation method of the metallized film for the buffer capacitor according to the embodiment of the present application, in the step of depositing the metallized layer on the dielectric film to obtain the metallized film, it first obtains the surface image of the metallized layer captured by the camera, then calculates the directional gradient histogram of the surface image of the metallized layer, then aggregates the surface image of the metallized layer and the directional gradient histogram to obtain a multi-channel image, then performs image block processing on the multi-channel image to obtain a sequence of image blocks, then passes the sequence of image blocks through a ViT model including an embedding layer to obtain multiple context image block semantic feature vectors, then calculates the cosine similarity between each two context image block semantic feature vectors in the multiple context image block semantic feature vectors to obtain a classification feature vector composed of multiple cosine similarities, then performs probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector, and finally, passes the optimized classification feature vector through a classifier to obtain a classification result indicating whether the surface quality of the metallized layer meets a predetermined standard.

[0107] Correspondingly, in the technical solution of the present application, the metallized film for the buffer capacitor is prepared by any of the aforementioned methods for preparing the metallized film for the buffer capacitor.

[0108] Figure 9 FIG. 1 is a block diagram of a system 100 for preparing a metallized film for a buffer capacitor according to an embodiment of the present application. Figure 9 As shown, a system 100 for preparing a metallized film for a buffer capacitor according to an embodiment of the present application includes: a film providing module 110 for providing an organic plastic film as a dielectric film; a deposition metallization module 120 for depositing a metallized layer on the dielectric film to obtain a metallized film; and a cutting module 130 for cutting the metallized film into a predetermined size.

[0109] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the deposition metallization module 120 includes: a surface image acquisition unit for acquiring a surface image of the metallized layer captured by a camera; a gradient histogram calculation unit for calculating a directional gradient histogram of the surface image of the metallized layer; an aggregation unit for aggregating the surface image of the metallized layer and the directional gradient histogram to obtain a multi-channel image; an image block processing unit for performing image block processing on the multi-channel image to obtain a sequence of image blocks; an embedded coding unit for encoding the sequence of image blocks by including A ViT model of the embedding layer is used to obtain multiple context image block semantic feature vectors; a cosine similarity calculation unit is used to calculate the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors to obtain a classification feature vector composed of multiple cosine similarities; an optimization unit is used to perform probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector; and a surface quality judgment unit of the metallization layer is used to pass the optimized classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the surface quality of the metallization layer meets the predetermined standard.

[0110] In one example, in the above-mentioned system 100 for preparing the metallized film for buffer capacitors, the image block processing unit is used to: uniformly block the multi-channel image to obtain a sequence of image blocks, wherein each image block in the sequence of image blocks has the same size.

[0111] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the embedding coding unit is used to: use the embedding layer to perform embedding coding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors; and use the ViT model to perform contextual semantic coding based on the self-attention mechanism on the sequence of image block embedding vectors to obtain the multiple contextual image block semantic feature vectors.

[0112] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the embedding layer is used to embed the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors, including: expanding the two-dimensional pixel matrix of each image block into a one-dimensional pixel expansion feature vector; and using the embedding layer to fully connect encode the one-dimensional pixel expansion feature vector to obtain the image block embedding vector.

[0113] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the ViT model is used to perform contextual semantic encoding based on the self-attention mechanism on the sequence of the image block embedding vectors to obtain the multiple contextual image block semantic feature vectors, including: arranging the sequence of the image block embedding vectors in one dimension to obtain a global image block vector; calculating the product between the global image block vector and the transposed vector of each image block embedding vector in the sequence of the image block embedding vectors to obtain multiple self-attention association matrices; normalizing each of the multiple self-attention association matrices to obtain multiple standardized self-attention association matrices; passing each of the multiple standardized self-attention association matrices through a Softmax classification function to obtain multiple probability values; and weighting each image block embedding vector in the sequence of the image block embedding vectors using each of the multiple probability values as a weight to obtain the multiple contextual image block semantic feature vectors.

[0114] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the cosine similarity calculation unit is configured to calculate the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors using the following cosine formula to obtain multiple cosine similarities; wherein the cosine formula is:

[0115]

[0116] Among them, V i and V j represents every two context image block semantic feature vectors in the multiple context image block semantic feature vectors, and Represents the semantic feature vector V of the context image block i and V j The characteristic value of each position in d(V i ,V j ) represents the cosine distance between any two context image block semantic feature vectors among the multiple context image block semantic feature vectors; and, the multiple cosine similarities are arranged to obtain the classification feature vector.

[0117] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the optimization unit is configured to perform probability density discreteness optimization on the classification feature vector using the following optimization formula to obtain an optimized classification feature vector; wherein the optimization formula is:

[0118]

[0119] Wherein, μ and σ are the mean and standard deviation of the eigenvalue set at each position in the classification feature vector, V is the classification feature vector, and v i is the eigenvalue of the ith position of the classification feature vector, and v i ′ is the eigenvalue of the i-th position of the optimized classification feature vector.

[0120] In one example, in the above-mentioned system 100 for preparing the metallized film for the buffer capacitor, the surface quality judgment unit of the metallized layer is used to: use multiple fully connected layers of the classifier to fully connect encode the optimized classification feature vector to obtain an encoded classification feature vector; and pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0121] Here, those skilled in the art will appreciate that the specific functions and operations of the various modules in the above-mentioned system for preparing the metallized film for buffer capacitors 100 have been described in detail above. Figures 2 to 8 The method for preparing the metallized film for the buffer capacitor has been described in detail, and therefore, repeated description thereof will be omitted.

[0122] As described above, the system 100 for preparing a metallized film for a buffer capacitor according to an embodiment of the present application can be implemented in various wireless terminals, such as a server equipped with an algorithm for preparing a metallized film for a buffer capacitor. In one example, the system 100 for preparing a metallized film for a buffer capacitor according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the system 100 for preparing a metallized film for a buffer capacitor can be a software module within the operating system of the wireless terminal, or an application developed specifically for the wireless terminal. Of course, the system 100 for preparing a metallized film for a buffer capacitor can also be one of the many hardware modules of the wireless terminal.

[0123] Alternatively, in another example, the system 100 for preparing the metallized film for the buffer capacitor and the wireless terminal may be separate devices, and the system 100 for preparing the metallized film for the buffer capacitor may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0124] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the method described above can be executed.

[0125] The program portion of the technology can be considered a "product" or "article of manufacture" in the form of executable code and / or related data, implemented or implemented through computer-readable media. Tangible, permanent storage media can include any memory or storage used by a computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives, or any similar device that can provide storage for software.

[0126] All or part of the software may sometimes be communicated over a network, such as the Internet or other communication network. Such communication can load the software from one computer device or processor to another. For example: loading from a server or host computer of a video target detection device to a hardware platform of a computer environment, or other computer environment that implements the system, or a system with similar functions related to providing information required for target detection. Therefore, another medium capable of transmitting software elements can also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., which are transmitted through cables, optical cables or air. Physical media used to carry carriers, such as cables, wireless connections or optical cables and the like, can also be considered as media that carry software. As used herein, unless limited to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media that participate in the process of executing any instructions by the processor.

[0127] This application uses specific terms to describe the embodiments of this application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different places in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0128] In addition, it will be understood by those skilled in the art that various aspects of the present application can be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application can be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0129] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or highly formal sense, unless expressly defined as such herein.

[0130] The above is an illustration of the present invention and should not be considered as limiting thereof. Although several exemplary embodiments of the present invention have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined by the claims. It should be understood that the above is an illustration of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A method for preparing a metallized film for a buffer capacitor, characterized in that: include: Providing an organic plastic film as a dielectric film; depositing a metallized layer on the dielectric film to obtain a metallized film; as well as cutting the metallized film into predetermined sizes; Depositing a metallization layer on the dielectric film to obtain a metallized film, comprising: Acquiring a surface image of the metallization layer captured by a camera; Calculating a directional gradient histogram of a surface image of the metallized layer; aggregating the surface image of the metallized layer and the histogram of directional gradients to obtain a multi-channel image; Performing image block processing on the multi-channel image to obtain a sequence of image blocks; Passing the sequence of image blocks through a ViT model including an embedding layer to obtain a plurality of context image block semantic feature vectors; Calculating the cosine similarity between every two context image block semantic feature vectors in the plurality of context image block semantic feature vectors to obtain a classification feature vector composed of a plurality of cosine similarities; Performing probability density discreteness optimization on the classification feature vector to obtain an optimized classification feature vector; and The optimized classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the surface quality of the metallization layer meets a predetermined standard.

2. The method for preparing a metallized film for a buffer capacitor according to claim 1, wherein: Performing image block processing on the multi-channel image to obtain a sequence of image blocks, comprising: The multi-channel image is uniformly divided into blocks to obtain a sequence of image blocks, wherein each image block in the sequence of image blocks has the same size.

3. The method for preparing a metallized film for a buffer capacitor according to claim 2, wherein: Passing the sequence of image blocks through a ViT model including an embedding layer to obtain multiple context image block semantic feature vectors, including: Using the embedding layer to perform embedding coding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors; and The ViT model is used to perform contextual semantic encoding based on a self-attention mechanism on the sequence of the image block embedding vectors to obtain the multiple contextual image block semantic feature vectors.

4. The method for preparing a metallized film for a buffer capacitor according to claim 3, wherein: Using the embedding layer to perform embedding encoding on the two-dimensional pixel matrix of each image block in the sequence of image blocks to obtain a sequence of image block embedding vectors, including: Expanding the two-dimensional pixel matrix of each image block into a one-dimensional pixel expansion feature vector; and The one-dimensional pixel expansion feature vector is fully connected encoded using the embedding layer to obtain the image block embedding vector.

5. The method for preparing a metallized film for a buffer capacitor according to claim 4, wherein: Using the ViT model to perform contextual semantic encoding based on a self-attention mechanism on the sequence of the image block embedding vectors to obtain the multiple contextual image block semantic feature vectors, including: Arranging the sequence of image block embedding vectors in one dimension to obtain a global image block vector; Calculating the product of the global image block vector and the transposed vector of each image block embedding vector in the sequence of image block embedding vectors to obtain a plurality of self-attention association matrices; Normalizing each of the plurality of self-attention association matrices to obtain a plurality of normalized self-attention association matrices; Passing each of the plurality of normalized self-attention association matrices through a Softmax classification function to obtain a plurality of probability values; and Each image block embedding vector in the sequence of image block embedding vectors is weighted using each probability value in the multiple probability values as a weight to obtain the multiple context image block semantic feature vectors.

6. The method for preparing a metallized film for a buffer capacitor according to claim 5, wherein: Calculating the cosine similarity between every two context image block semantic feature vectors in the plurality of context image block semantic feature vectors to obtain a classification feature vector composed of the plurality of cosine similarities, including: Calculating the cosine similarity between every two context image block semantic feature vectors in the multiple context image block semantic feature vectors using the following cosine formula to obtain multiple cosine similarities; Wherein, the cosine formula is: in, and represents every two context image block semantic feature vectors in the multiple context image block semantic feature vectors, and Represents the semantic feature vector of the context image block and The eigenvalues at each position in , represents a cosine distance between any two context image block semantic feature vectors among the plurality of context image block semantic feature vectors; and The multiple cosine similarities are arranged to obtain the classification feature vector.

7. The method for preparing a metallized film for a buffer capacitor according to claim 6, wherein: Optimizing the probability density discreteness of the classification feature vector to obtain an optimized classification feature vector includes: The probability density discreteness of the classification feature vector is optimized using the following optimization formula to obtain an optimized classification feature vector; Wherein, the optimization formula is: in, and is the mean and standard deviation of the eigenvalue set at each position in the classification feature vector, is the first The eigenvalues at the positions, and is the first The eigenvalues at each position.

8. The method for preparing a metallized film for a buffer capacitor according to claim 7, wherein: Passing the optimized classification feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the surface quality of the metallization layer meets a predetermined standard, includes: Performing full-connection encoding on the optimized classification feature vector using multiple fully-connected layers of the classifier to obtain an encoded classification feature vector; and The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

9. A metallized film for a buffer capacitor, characterized in that: The metallized film for a buffer capacitor is prepared by the method for preparing a metallized film for a buffer capacitor according to any one of claims 1 to 8.

Images