Wafer defect classification method and device and computer equipment

Abstract

The invention provides a wafer defect classification method and device and computer equipment. The wafer defect classification method comprises the following steps: acquiring an image set composed of wafer defect images obtained by the same wafer manufacturer in a wafer manufacturing process; and inputting the image set into a weight generation model, and generating a weight for controlling the attention degree of the classification model to different positions of the image. And finally, applying the generated weight to a classification model, and inputting the image set into the classification model to obtain the defect type of each image. And clustering the image subset corresponding to each defect type, and taking a clustering result as a wafer defect classification result of the image. Self-adaptive weights are generated for wafer defect images of different manufacturers, so that the same classification model can be applied to different manufacturers. Through the method of clustering after classification, different subdivision granularities can be selected according to different manufacturers, so that the scheme can be applied to different manufacturers, and the classification accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a wafer defect classification method, device and computer equipment. Background Art

[0002] As one of the core driving forces of modern science and technology, the complexity and precision of the semiconductor industry's production process places extremely high demands on product quality. Wafers are the basic raw materials for manufacturing semiconductor products, and the quality of wafers will affect the performance and reliability of semiconductor products. During the wafer manufacturing process, various defects often appear on the wafer surface due to the influence of various factors, such as material defects, process parameter fluctuations, equipment failures, etc. These defects will affect the yield rate of the wafer and the quality of the semiconductor materials manufactured based on the wafer.

[0003] In the semiconductor manufacturing process, defect detection is a key link to ensure product quality. Different types of wafer defects are often caused by different reasons. Classifying wafer defects helps to quickly discover problems in the wafer defect manufacturing process, so as to take measures to reduce wafer defects as soon as possible and ensure the quality of finished products. However, with the continuous advancement of semiconductor manufacturing technology, the types and number of defects are also increasing, which brings huge challenges to defect detection. Traditional defect detection methods mainly rely on manual experience and simple image processing technology. These methods often find it difficult to guarantee the accuracy and efficiency of detection when faced with large-scale and diverse defect data.

[0004] It can be seen that the current wafer defect classification is inaccurate.

[0005] In recent years, with the rapid development of artificial intelligence technology, especially the advancement of deep learning and computer vision technology, new solutions have been provided for semiconductor defect detection. However, the existing defect detection methods still have some problems in practical applications.

[0006] First, the classification and identification process of defect data lacks a unified standard, which makes it difficult to compare and migrate the inspection results between different manufacturers. Second, because different companies have different requirements for the granularity of defect segmentation, existing methods are difficult to flexibly adapt to such diverse needs. Summary of the invention

[0007] To overcome the problems existing in the related art, this specification provides a wafer defect classification method, device and computer equipment.

[0008] According to a first aspect of an embodiment of this specification, a wafer defect classification method is provided, the method comprising: Acquire an image set including a plurality of images; the plurality of images in the image set are wafer defect images obtained by the same wafer manufacturing manufacturer during a wafer manufacturing process within a preset time period; Inputting the image set into a weight generation model to obtain weight data corresponding to the image set; the weight data indicates the degree of attention paid by the classification model to different positions of the image; Applying the weight to the classification model, and inputting the image set into the classification model configured with the weight data, wherein the classification model performs weighted processing on the extracted image features based on the weight data, and performs classification of the first-level defect type based on the weighted image features to obtain the first-level defect type of each image in the image set; For each first-level defect type, the images corresponding to the first-level defect type in the image set are clustered according to the target cluster number, and the clustering result is used as the second-level wafer defect classification result of the image set.

[0009] According to a second aspect of an embodiment of this specification, a wafer defect classification device is provided, the device comprising: An image set acquisition module, used to acquire an image set including a plurality of images; the plurality of images in the image set are wafer defect images obtained by the same wafer manufacturing manufacturer during the wafer manufacturing process within a preset time period; A weight data acquisition module, used to input the image set into a weight generation model to obtain weight data corresponding to the image set; the weight data indicates the degree of attention paid to different positions of the image; a classification module, configured to input the image set into a classification model configured with the weight data, wherein the classification model performs weighted processing on the extracted image features based on the weight data, and performs classification of the first-level defect type based on the weighted image features to obtain the first-level defect type of each image in the image set; The clustering module is used to cluster the images corresponding to the first-level defect type in the image set according to the target cluster number for each first-level defect type, and use the clustering result as the second-level defect classification result of the image set.

[0010] According to a third aspect of the embodiments of this specification, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the wafer defect classification method as described in the first aspect of the embodiments of this specification is implemented.

[0011] According to a fourth aspect of an embodiment of this specification, a computer device is provided, the computer device comprising: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the wafer defect classification method described in the first aspect.

[0012] This specification provides a wafer defect classification method. An image set consisting of wafer defect images obtained by the same wafer manufacturer during the wafer manufacturing process is obtained. The image set is input into a warrant generation model to generate weights for controlling the degree of attention paid by the classification model to different positions of the image. Finally, the generated weights are applied to the classification model, and the image set is input into the classification model to obtain the defect type of each image. Clustering is performed for the image subset corresponding to each defect type, and the clustering result is used as the wafer defect classification result of the image.

[0013] Since the defects in wafers produced by the same wafer manufacturer within a period of time are generally caused by the same reasons, the positions of the defects in the images of the wafers of the same wafer manufacturer are close. However, the positions of defects in different wafer manufacturers may differ due to different causes of the defects. The above method takes into account the different positions and manifestations of defects in wafers manufactured by different wafer manufacturers in the image, and generates adaptive weights for wafer defect images of different manufacturers, so that the same classification model can be applied to different manufacturers. And using the weight data for classification can make the classification model pay more attention to the location of the defects, thereby making the large categories obtained by classification more accurate. Moreover, by using the method of clustering after classification, different subdivision granularity can be selected according to different manufacturers, so that the solution can be applied to different manufacturers and improve classification accuracy.

[0014] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0016] Figure 1 The flowchart of a wafer defect classification method is shown in this specification according to an exemplary embodiment.

[0017] Figure 2 This is a wafer defect image shown in this specification.

[0018] Figure 3 It is a block diagram of a wafer defect classification device according to an exemplary embodiment of the present specification.

[0019] Figure 4It is a hardware structure diagram of a computer device shown in this specification according to an exemplary embodiment. DETAILED DESCRIPTION

[0020] Exemplary embodiments will be described in detail herein, examples of which 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 this specification. Instead, they are merely examples of devices and methods consistent with some aspects of this specification as detailed in the appended claims.

[0021] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. The singular forms "a", "the" and "the" used in this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0022] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0023] Based on the existing problems, this specification provides a wafer defect classification method. Obtain an image set consisting of wafer defect images obtained by the same wafer manufacturer during the wafer manufacturing process. Input the image set into a warrant generation model to generate weights for controlling the degree of attention paid by the classification model to different positions of the image. Finally, apply the generated weights to the classification model, and input the image set into the classification model to obtain the defect type of each image. And cluster the image subsets corresponding to each defect type, and use the clustering results as the wafer defect classification results of the image.

[0024] Since the defects in wafers produced by the same wafer manufacturer within a period of time are generally caused by the same reasons, the positions of the defects in the images of the wafers of the same wafer manufacturer are close. However, the positions of defects in different wafer manufacturers may be different due to different causes of the defects. The above method takes into account the different positions and manifestations of defects in wafers manufactured by different wafer manufacturers in the image. Adaptive weights are generated for wafer defect images of different manufacturers, and the weights are used for classification, which can make the classification model pay more attention to the location of the defects, thereby making the large categories obtained by classification more accurate. In addition, the accuracy of the final classification results can be improved by clustering after classification.

[0025] Next, the embodiments of this specification are described in detail.

[0026] like Figure 1 As shown, Figure 1 is a flowchart of a method according to an exemplary embodiment of the present specification, comprising the following steps: Step 101: Acquire an image set including multiple images.

[0027] The multiple images in the image set are wafer defect images obtained by the same wafer manufacturer during the wafer manufacturing process within a preset time period.

[0028] Specifically, since defect locations in images from different manufacturers are different, this specification requires generating weights for images from the same wafer manufacturer, so defect images of wafers produced by the same wafer manufacturer, i.e., wafer defect images, are first obtained.

[0029] Among them, the wafer manufacturer is the manufacturer that produces and manufactures wafers. The wafer defect image can be: for the wafer manufactured by the wafer manufacturer, the wafer image is obtained through a scanning electron microscope (SEM), and the wafer defect image with defects in the wafer is screened out. The scanning electron microscope can take pictures and scan the wafer and identify the defects therein. However, it can only identify the existence of defects and the location of the defects, and cannot specifically determine the type of defects.

[0030] Images of wafer defects can be found at Figure 2 . Figure 2 It can be seen that there is a defect in the image where part of the long strip shape is not etched enough. The coarse classification defect category corresponding to this defect can be called pattern fail.

[0031] For different wafer manufacturers, there are differences in the manufacturing equipment, SEM, production process and problems in the production process used by different wafer manufacturers, which may cause defects to be located in different positions in the wafer defect image. For example, due to problems with the wafer manufacturing equipment settings of a certain manufacturer, defects are generally located in the center of the wafer; for example, the SEMs of different manufacturers have different crosshairs, which causes the captured images to be offset in different directions, such as some manufacturers' crosshairs are biased to the lower right corner, and some manufacturers' crosshairs are biased to the lower left corner. All these make the defect positions in the wafer defect images of different manufacturers different, while the defect positions of the wafer defect images of the same manufacturer are similar.

[0032] In addition, different wafer manufacturers have different production requirements, which results in different graphics on the produced wafers. This results in some wafer defect images from different wafer manufacturers having complex backgrounds and others having simple backgrounds. This means that during the wafer defect classification process, some manufacturers’ wafer defect images may have complex backgrounds, causing the complex background to affect the classification results of wafer defects. Therefore, this method can make the classification model pay more attention to the location of the wafer defect rather than the background, thereby avoiding the influence of the background on the wafer defect classification results.

[0033] After the image set is acquired in step 101, the image may be preprocessed. In an optional implementation, image preprocessing methods corresponding to various wafer manufacturers may be preset; the method further includes: determining the preprocessing method corresponding to the image set according to the preset image preprocessing methods corresponding to various wafer manufacturers, and preprocessing the images of the image set by using the determined preprocessing method.

[0034] Since the images corresponding to different wafer manufacturers are different, they generally need to be processed using different preprocessing methods. For example, if the image noise of a certain manufacturer is high, filtering can be performed, and if the image clarity of a certain manufacturer is low, data enhancement methods can be used for enhancement. Therefore, by presetting the preprocessing methods corresponding to different manufacturers, you can get preprocessed images that better meet your needs.

[0035] Step 103: input the image set into a weight generation model to obtain weight data corresponding to the image set.

[0036] The weight data indicates the degree of attention paid to different positions of the image.

[0037] Specifically, in order to adapt to the different locations of wafer defects in wafer defect images from different manufacturers, the image set needs to be input into the weight generation model. The weight generation model generates adaptive weights for the classification model based on the feature distribution characteristics of the images in the image set. This allows the classification model to pay more attention to the wafer defects themselves in the image, rather than other locations, to avoid other locations affecting the wafer defect classification results.

[0038] Regarding the specific form of the weight generation model, in an optional implementation, the weight generation model includes a first module, a second module and a third module; step 103 specifically includes: the first module extracts the features of each image in the image set; the second module calculates the mean and variance of the features of each image based on the features of each image; the third module generates weight data corresponding to the image set based on the mean and variance.

[0039] The extracted features may be in the form of feature vectors.

[0040] Specifically, the first module is an encoder, which is used to encode the wafer defect image into the form of a feature vector. In an optional embodiment, the first module can reuse the feature extraction module of the classification model. The second module is used to calculate the mean and variance of the feature vectors of each image. Through the mean of the feature vector, the average representation of multiple images can be obtained, and the typical representation of the objects corresponding to a group of images can also be obtained. The variance of the feature vector can be used to obtain the degree of change of each feature in the image. The mean and variance of a group of image feature vectors can be used to describe the characteristics of a group of images.

[0041] Then, the third module can generate the weight of the image set corresponding to the group of images according to the mean and variance. The third module can be expressed as a function of the mean and variance, W=f(μ,σ), where W represents the weight of the image set, and μ and σ represent the mean and variance of the feature vector of the image set, respectively. In one embodiment, the third module can implement the above function f through a neural network.

[0042] The generated weight matrix can be applied to the classification model, so that the classification model can pay more attention to the defective parts in the image, making the classification result of the classification model more accurate.

[0043] Step 105: input the image set into a classification model configured with the weight data; the classification model performs weighted processing on the extracted image features based on the weight data, and performs classification of the first-level defect type based on the weighted image features to obtain the first-level defect type of each image in the image set.

[0044] Specifically, the weight obtained in step 103 can be applied to the classification model. Thus, the images in the above image set can be more accurately classified for wafer defects through the classification model. By introducing an adaptive weight mechanism, the classification model can automatically adjust the weight according to the data characteristics of different manufacturers, thereby improving the portability and adaptability of the classification model.

[0045] The weight is equivalent to applying a mask to the image. For the mask, the value can be 1 at the location where some defects are more likely to exist, and 0 or a decimal between 0 and 1 at other locations. In other words, the image features and weights can be multiplied to obtain the image data after masking. In this way, the model can pay different attention to different locations in different images, and pay more attention to the location of the defects.

[0046] The classification model may be a model for classifying wafer defects obtained by training based on a training data set, wherein the training data set includes a plurality of wafer defect images, each of which has a corresponding label.

[0047] In an optional implementation, the classification model may be a vision transformer (ViT). ViT performs well in the classification of large quantities of data. In the scenario used in this specification, the amount of wafer defect image data is large, and ViT can achieve good results. This solves the problems of processing efficiency and accuracy in large-scale defect data in related technologies.

[0048] As for the application of weights, weights are applied to each layer of the classification model, and the weights of different positions of the image when they are input into each layer can be controlled in a weighted manner.

[0049] When the classification model is ViT, the generated adaptive weights can be applied to each transformer block of the ViT model. Specifically, in each transformer block of the ViT model, adaptive weights are introduced to adjust the input image features in a weighted manner, thereby adjusting the output of each transformer block. In this way, the model can automatically adjust the weights according to the data features of different manufacturers, thereby improving the adaptability and transferability of the model.

[0050] In other words, the classification model includes a plurality of transformer blocks for performing self-attention processing, and each transformer block is configured with a weight matrix corresponding to the weight data. Step 103 may include: each transformer block of the classification model performs self-attention processing on the image features input therein, obtains an attention matrix or an image feature matrix as an intermediate matrix, and performs positional multiplication of the intermediate matrix with the weight matrix to obtain the image features output by the transformer block.

[0051] The first level defect type in step 105 refers to a rough classification. After the rough classification is completed, in step 107, each defect type can be finely classified within the defect type to obtain a second level defect type, so that the classification result can be more accurate.

[0052] In the related art, if the classification model of machine learning is directly used for fine classification, it may be impossible to use one model to classify the wafer defect images of all manufacturers due to the different classification standards of different manufacturers. Moreover, if a new type of wafer defect appears, the classification model needs to be adjusted. These may lead to inaccurate classification results. The method of this specification first uses the classification model to perform coarse classification. For coarse classification, the types of coarse classification of different manufacturers are generally the same, but there are differences in fine classification. For example, there will be a defect category of pattern fail, but under this type, different manufacturers have different fine classification categories. Coarse classification is generally determined based on fixed standards in the semiconductor industry, and coarse classification generally does not add new categories. This method can improve the classification accuracy and can be applied to different manufacturers.

[0053] In addition, in an optional embodiment, the weight generation model and the classification model are jointly trained based on a training data set, and the training data set includes wafer defect images of multiple different wafer manufacturers and classification labels of each wafer defect image. That is, the training process is similar to the above-mentioned application process. The image set is first input into the weight generation model to obtain weight data, and then the weight data is applied to the classification model, and the image set is classified using the classification model. After the classification is completed, the parameters of the weight generation model and the classification model are adjusted according to the label and classification results using a preset loss function.

[0054] In other words, the weight generation model and the classification model are trained in the following manner: a wafer defect image of any wafer manufacturer in a training data set is input into the weight generation model to obtain first weight data generated by the weight generation model; the training data set includes wafer defect images of multiple different wafer manufacturers and a first-level classification label of each wafer defect image; the wafer defect image of the wafer manufacturer is input into a classification model configured with the first weight data, the classification model performs weighted processing on the extracted image features based on the first weight data, and performs first-level defect type classification based on the weighted image features; the parameters of the weight generation model and the classification model are adjusted according to the first-level classification labels of each wafer image and the first-level defect type of each image output by the classification model.

[0055] In the joint training process, the model parameters can be optimized by minimizing the classification error and weight generation error. During the training process, the cross-validation method is used to ensure the generalization ability of the model on data from different manufacturers.

[0056] This allows the generated weights to be better applied to the classification model, making the classification results more accurate.

[0057] In addition, the classification model can be evaluated after training. Specifically, the classification accuracy and transferability of the model are evaluated by performing classification tasks on data from multiple manufacturers. The evaluation results show that the improved ViT algorithm performs well on data from different manufacturers and has high classification accuracy and transferability.

[0058] Step 107 , for each first-level defect type, cluster the images corresponding to the first-level defect type in the image set according to the target cluster number, and use the clustering result as the second-level wafer defect classification result of the image set.

[0059] As mentioned above, a rough classification may be performed first, and then for each defect type of the rough classification, clustering may be performed within the image corresponding to the defect type, thereby obtaining a wafer defect classification result for each image.

[0060] For a specific clustering algorithm, it may be a K-Means clustering algorithm. Of course, this example does not limit the present specification.

[0061] With respect to the number of target clusters for each defect type, in an optional implementation, the number of target clusters is preset for the wafer manufacturing company and each first-level defect type corresponding to the image set.

[0062] Since different manufacturers have different requirements for defect segmentation granularity and different standards for defect subdivision, it is difficult to migrate the detection results of different manufacturers. In the above method, the number of clusters is preset according to the needs of each manufacturer, so that the classification method can be applied to the data of each manufacturer.

[0063] In another optional implementation, some manufacturers do not have a complete classification standard. Specifically, the classification standard is generally determined based on past production experience, and the specific sub-classification is determined based on the defect types that existed in the past production process. For some new wafer manufacturers, they have no experience in wafer classification and may not be sure how many specific categories each defect type needs to be subdivided into.

[0064] In the above case, the number of clusters corresponding to each defect type can be determined according to the characteristics of the image set. Specifically, for each defect type and each first-level defect type, the images corresponding to the first-level defect type in the image set are input into the cluster number generation model to obtain the target cluster number; the target cluster number is applied to the clustering process of the images corresponding to the first-level defect type.

[0065] In an optional embodiment, the cluster number generation model can be similar to the weight generation model and also include three modules. For an image corresponding to a certain defect type, the first module is first used to obtain the feature vector of the image corresponding to the defect type. Specifically, the feature vector extracted in the previous step is reused here. The second module is used to calculate the mean and variance between the feature vectors obtained. The third module is used to generate the number of clusters based on the mean and variance. The third module can be specifically expressed as K=h(μ,σ), where μ and σ represent the mean and variance of the feature distribution, respectively, and K represents the number of adaptive clusters generated.

[0066] This can help new manufacturers who do not know how to classify themselves, and by generating a model based on the number of clusters, the clustering method can be applied to various manufacturers, thus improving flexibility.

[0067] This specification proposes a semiconductor defect classification method that combines an improved ViT algorithm and an optimized clustering algorithm. This method uses a two-step process of classification followed by clustering, which can not only improve the accuracy and efficiency of defect detection, but also flexibly adapt to the needs of different manufacturers for defect segmentation granularity. By introducing the improved ViT for preliminary classification and taking advantage of its advantages in large-scale data processing, the accuracy of the classification results is ensured, while being able to be well migrated between different manufacturers; then a clustering algorithm is used within each large category to adapt to the different segmentation granularities of different companies. This method not only improves the versatility and flexibility of defect detection, but also provides an efficient and portable solution for defect classification in the semiconductor industry.

[0068] Corresponding to the embodiments of the aforementioned method, this specification also provides embodiments of an apparatus and a computer device to which it is applied.

[0069] like Figure 3 As shown, Figure 3 1 is a block diagram of a wafer defect classification device according to an exemplary embodiment of the present specification, the device comprising: An image set acquisition module 310 is used to acquire an image set including a plurality of images; the plurality of images in the image set are wafer defect images obtained by the same wafer manufacturing manufacturer during the wafer manufacturing process within a preset time period; The weight data acquisition module 320 is used to input the image set into the weight generation model to obtain weight data corresponding to the image set; the weight data indicates the degree of attention paid to different positions of the image; A classification module 330, configured to input the image set into a classification model configured with the weight data, wherein the classification model performs weighted processing on the extracted image features based on the weight data, and performs classification of the first-level defect type based on the weighted image features to obtain the first-level defect type of each image in the image set; The clustering module 340 is used to cluster the images corresponding to the first-level defect type in the image set according to the target cluster number for each first-level defect type, and use the clustering result as the second-level defect classification result of the image set.

[0070] In an optional embodiment, the device also includes a cluster number generation module 350 (not shown in the figure), which is used to input the images corresponding to the first-level defect type in the image set into a cluster number generation model for each first-level defect type to obtain a target cluster number.

[0071] In an optional implementation, the number of target clusters is preset for the wafer manufacturing company and each first-level defect type corresponding to the image set.

[0072] In an optional implementation, the classification model is a visual transformer ViT.

[0073] In an optional embodiment, the classification model includes multiple transformer blocks for self-attention processing, and each transformer block is configured with a weight matrix corresponding to the weight data; the classification module 330 is specifically used for each transformer block of the classification model to perform self-attention processing on the image features input therein, and obtain an attention matrix or an image feature matrix as an intermediate matrix, and perform bitwise multiplication of the intermediate matrix with the weight matrix to obtain the image features output by the transformer block.

[0074] In an optional embodiment, the weight data acquisition module 320 includes a first module, a second module and a third module; the first module extracts the features of each image in the image set; the second module calculates the mean and variance of the features of each image based on the features of each image; the third module generates weight data corresponding to the image set based on the mean and variance.

[0075] In an optional embodiment, the device also includes a preprocessing module 300 (not shown in the figure), which is used to determine the preprocessing method corresponding to the image set according to the preset image preprocessing method corresponding to each wafer manufacturer, and preprocess the images of the image set using the determined preprocessing method.

[0076] In an optional embodiment, the device includes a training module 360 ​​(not shown in the figure), which is used to input a wafer defect image of any wafer manufacturer in a training data set into a weight generation model to obtain first weight data generated by the weight generation model; the training data set includes wafer defect images of multiple different wafer manufacturers and a first-level classification label of each wafer defect image; the wafer defect image of the wafer manufacturer is input into a classification model configured with a first weight data, the classification model performs weighted processing on the extracted image features based on the first weight data, and performs first-level defect type classification based on the weighted image features; the parameters of the weight generation model and the classification model are adjusted according to the first-level classification labels of each wafer image and the first-level defect type of each image output by the classification model.

[0077] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, which will not be repeated here.

[0078] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this specification. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0079] like Figure 4 As shown, Figure 4 A hardware structure diagram of a computer device in which a wafer defect classification device of an embodiment is located is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other in communication within the device through the bus 1050.

[0080] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0081] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0082] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0083] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0084] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010 , the memory 1020 , the input / output interface 1030 , and the communication interface 1040 ).

[0085] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0086] The embodiments of this specification also provide a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the wafer defect classification method as described in the first aspect of the embodiments of this specification is implemented.

[0087] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0088] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0089] Those skilled in the art will readily appreciate other embodiments of the specification after considering the specification and practicing the invention claimed herein. The specification is intended to cover any variations, uses or adaptations of the specification that follow the general principles of the specification and include common knowledge or customary techniques in the art that are not claimed in the specification. The specification and examples are to be considered exemplary only, and the true scope and spirit of the specification are indicated by the following claims.

[0090] It should be understood that the present description is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present description is limited only by the appended claims.

[0091] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

Claims

1. A wafer defect classification method, characterized in that: The method comprises: Acquire an image set including a plurality of images; the plurality of images in the image set are wafer defect images obtained by the same wafer manufacturing manufacturer during a wafer manufacturing process within a preset time period; Inputting the image set into a weight generation model to obtain weight data corresponding to the image set; the weight data indicates the degree of attention paid to different positions of the image; Inputting the image set into a classification model configured with the weight data, the classification model performs weighted processing on the extracted image features based on the weight data, and classifies the first-level defect types based on the weighted image features to obtain the first-level defect types of each image in the image set; For each first-level defect type, the images corresponding to the first-level defect type in the image set are clustered according to the target cluster number, and the clustering result is used as the second-level defect classification result of the image set.

2. The method according to claim 1, characterized in that The method further comprises: For each first-level defect type, the images corresponding to the first-level defect type in the image set are input into the cluster number generation model to obtain the target cluster number.

3. The method according to claim 1, characterized in that The number of target clusters is preset according to the wafer manufacturing company and each first-level defect type corresponding to the image set.

4. The method according to claim 1, characterized in that: The classification model is a visual transformer ViT.

5. The method according to claim 1, characterized in that The classification model includes a plurality of transformer blocks for performing self-attention processing, each transformer block being configured with a weight matrix corresponding to the weight data; The classification model performs weighted processing on the extracted image features based on the weight data, including: Each transformer block of the classification model performs self-attention processing on the image features input therein to obtain an attention matrix or an image feature matrix as an intermediate matrix, and performs bitwise multiplication of the intermediate matrix with the weight matrix to obtain the image features output by the transformer block.

6. The method according to claim 1, characterized in that The weight generation model includes a first module, a second module and a third module; The step of inputting the image set into a weight generation model to obtain weight data corresponding to the image set includes: The first module extracts features of each image in the image set; The second module calculates the mean and variance of the features of each image according to the features of each image; The third module generates weight data corresponding to the image set according to the mean and variance.

7. The method according to claim 1, characterized in that Also includes: According to the preset image preprocessing methods corresponding to each wafer manufacturer, the preprocessing method corresponding to the image set is determined, and the images of the image set are preprocessed by the determined preprocessing method.

8. The method according to claim 1, characterized in that The weight generation model and the classification model are trained in the following manner: Inputting a wafer defect image of any wafer manufacturer in a training data set into a weight generation model to obtain first weight data generated by the weight generation model; the training data set includes wafer defect images of multiple different wafer manufacturers and a first-level classification label of each wafer defect image; Inputting the wafer defect image of the wafer manufacturer into a classification model configured with first weight data, the classification model weighting the extracted image features based on the first weight data, and performing first-level defect type classification based on the weighted image features; According to the first-level classification labels of each wafer image and the first-level defect types of each image output by the classification model, the parameters of the weight generation model and the classification model are adjusted.

9. A wafer defect classification device, characterized in that: The device comprises: An image set acquisition module, used to acquire an image set including a plurality of images; the plurality of images in the image set are wafer defect images obtained by the same wafer manufacturing manufacturer during the wafer manufacturing process within a preset time period; A weight data acquisition module, used to input the image set into a weight generation model to obtain weight data corresponding to the image set; the weight data indicates the degree of attention paid to different positions of the image; a classification module, configured to input the image set into a classification model configured with the weight data, wherein the classification model performs weighted processing on the extracted image features based on the weight data, and performs classification of the first-level defect type based on the weighted image features to obtain the first-level defect type of each image in the image set; The clustering module is used to cluster the images corresponding to the first-level defect type in the image set according to the target cluster number for each first-level defect type, and use the clustering result as the second-level defect classification result of the image set.

10. A computer device, characterized in that: The computer device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

Images