Photoresist defect detection method and system based on image recognition and cloud platform

Abstract

The invention relates to the technical field of image processing, and particularly discloses a photoresist defect detection method and system based on image recognition and a cloud platform, and the method comprises the steps: S1, obtaining target image data of a plurality of target images based on photoetching in a specified time period, and obtaining photoresist image data of all target images; s2, determining a plurality of types of photoetching data and a type label of each type of photoetching data; s3, determining a first defect fitting matrix of each target image in the category photoetching data, and determining a second defect fitting matrix of each category label; s4, constructing a defect detection model of each category label; and S5, acquiring a real-time photoetching image to be detected, and determining defect data. By combining artificial intelligence and computer vision technologies and utilizing an automatic defect fitting and classifying method, the accuracy and efficiency of defect identification in the photoetching process are improved, the automation degree of defect detection is improved, and the production efficiency and the product quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a method, a system and a cloud platform for detecting photoresist defects based on image recognition. Background Art

[0002] Lithography technology is widely used in high-precision manufacturing fields such as semiconductors and microelectronics, especially in the processing of integrated circuits (ICs) and photoresist materials. With the rapid development of microelectronics technology, defect detection in the lithography process has become an important link in the production process. Traditional lithography defect detection relies on manual observation and empirical judgment, which has large errors and efficiency problems. With technological progress, automated detection methods based on image recognition have emerged, especially with the help of artificial intelligence and computer vision technologies, which have become important tools for improving detection accuracy and efficiency. However, there are still many deficiencies in the existing technologies, mainly reflected in the low recognition accuracy of complex defects and the insufficient degree of detection automation, etc., and the processing ability of real-time lithography images is limited.

[0003] Therefore, the present invention proposes a method, a system and a cloud platform for detecting photoresist defects based on image recognition. Summary of the Invention

[0004] The present invention provides a method, a system and a cloud platform for detecting photoresist defects based on image recognition. By analyzing the determined image data and photoresist image data, multiple categories of lithography data and the category labels of each category of lithography data are determined. The category lithography data is analyzed to determine the second defect fitting matrix of the category lithography data of each category label, and a defect detection model for each category label is constructed to realize the defect detection of real-time lithography images. By combining artificial intelligence and computer vision technologies and using an automated defect fitting and classification method, the accuracy and efficiency of defect recognition in the lithography process are improved, quality problems in the production process are avoided, manual intervention is reduced, the degree of automation of defect detection is improved, and production efficiency and product quality are increased.

[0005] The present invention provides a method for detecting photoresist defects based on image recognition, including:

[0006] S1: Obtain the target image data of multiple target images based on lithography within a specified time period, and obtain the photoresist image data of all target images;

[0007] S2: Analyze the target image data and the photoresist image data to determine multiple categories of lithography data and the category labels of each category of lithography data;

[0008] S3: Analyze the category lithography data to determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label;

[0009] S4: Based on the category lithography data of each category label and the second defect fitting matrix, construct a defect detection model for each category label;

[0010] S5: Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

[0011] Preferably, for the detection method of photoresist defects based on image recognition, obtain the target image data of multiple target images based on lithography within a specified time period, including:

[0012] Obtain the target image sub-data of each target image based on lithography within a specified time period, where the target image sub-data includes the target image type, target image lithography requirements, and target image lithography attributes of the target image;

[0013] Based on the target image sub-data of all target images obtained within the specified time period, determine the target image data.

[0014] Preferably, for the detection method of photoresist defects based on image recognition, obtain the photoresist image data of all target images, including:

[0015] Obtain all the photoresist images after development of each target image within the specified time period, and determine the photoresist image sub-data of each target image;

[0016] Based on the photoresist image sub-data of all target images, determine the photoresist image data.

[0017] Preferably, for the detection method of photoresist defects based on image recognition, analyze the target image data and the photoresist image data, and determine multiple category lithography data and the category labels of each category lithography data, including:

[0018] Preprocess the target image data and the photoresist image data respectively;

[0019] Extract the target image lithography requirements in the target image sub-data of each target image in the preprocessed target image data, determine multiple requirement features of each target image, and determine the requirement quantization value of each requirement feature of each target image;

[0020] Extract the target image lithography attributes in the target image sub-data of each target image in the preprocessed target image data, determine multiple attribute features of each target image, and determine the attribute quantization value of each attribute feature of each target image;

[0021] Based on the demand quantization values of all demand features and the attribute quantization values of all attribute features of each target image in the preprocessed target image data, determine the demand attribute level of each target image;

[0022] Extract the target image type from the target image sub-data of each target image in the preprocessed target image data, and based on the extracted target image type of the target image and the determined demand attribute level, determine the type level label of each target image;

[0023] Based on the preprocessed photoresist image sub-data of all target images with the same type level label, determine the category photolithography data, and determine the category label for the category photolithography data with the type level label;

[0024] Preferably, for the detection method of photoresist defects based on image recognition, analyze the category photolithography data to determine the first defect fitting matrix of each target image in the category photolithography data, including:

[0025] Extract the defect features from all the photoresist images in the preprocessed photoresist image sub-data of each target image in the category photolithography data, and determine the defect feature matrix of all the photoresist images of each target image in the category photolithography data;

[0026]

[0027] Among them, represents the defect feature matrix of the i-th target image of the category photolithography data with the category label of type a and level b, respectively represent the first defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category photolithography data with the category label of type a and level b, where iN2 represents the number of defects of the i-th target image, respectively represent the k-th defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category photolithography data with the category label of type a and level b, where N1 represents the number of defect features, respectively represent the N1-th defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category photolithography data with the category label of type a and level b, that is, the defect type;

[0028] Extract the defect types from the defect feature matrix of all the photoresist images of each target image to determine the defect type set of each target image;

[0029] Based on the defect feature matrix and the defect type set of all the photoresist images of each target image, determine the first defect fitting matrix of each target image in the category photolithography data;

[0030]

[0031] Among them, represents the first defect fitting matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the first fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the k-th fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the N1 - 1-th fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, where iN4 represents the number of defects in the defect type set of the i-th target image, represents the p-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, represents the first exponential function.

[0032] Preferably, for the method for detecting photoresist defects based on image recognition, determining the second defect fitting matrix of the category lithography data for each category label includes:

[0033] Based on the defect type sets of all target images in each category lithography data, determine the category defect sets of each category lithography data;

[0034] Based on the category defect sets of each category lithography data and the first defect fitting matrices of all target images of each category lithography data, determine the second defect fitting matrix of the category lithography data for each category label;

[0035]

[0036] Among them, CM ab represents the second defect fitting matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the first fitting features of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the k-th fitting features of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the (N1 - 1)-th fitting feature of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of category lithography data with category label of type a and grade b, where abN5 represents the number of category defects in the category defect set of category lithography data with category label of type a and grade b. represents the q-th defect type in the category defect set of category lithography data with category label of type a and grade b. represents the second exponential function, Nu represents the number of target images in the category lithography data, W a represents the type weight with category label of type a, W ab represents the grade weight with category label of type a and grade b.

[0037] Preferably, for the method for detecting photoresist defects based on image recognition, obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models, including:

[0038] Determine the real-time target image based on the real-time lithography image to be detected, and determine the real-time label of the real-time lithography image based on the real-time image type, real-time image lithography requirement, and real-time image lithography attribute of the real-time target image;

[0039] Select the defect detection model with the same category label as the real-time label of the real-time lithography image as the defect detection model of the real-time lithography image;

[0040] Input the real-time lithography image into the determined defect detection model, and determine the defect data of the real-time lithography image based on the output result of the defect detection model.

[0041] The present invention provides a system for detecting photoresist defects based on image recognition, which is used to execute any one of the methods for detecting photoresist defects based on image recognition in Embodiments 1 to 7, including:

[0042] Acquisition module: Acquire the target image data of multiple target images based on lithography within a specified time period, and acquire the photoresist image data of all target images;

[0043] Analysis module: Analyze the target image data and the photoresist image data to determine multiple category lithography data and the category label of each category lithography data;

[0044] Determination module: Analyze the category lithography data, determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label;

[0045] Building module: Based on the category lithography data for each category label and the second defect fitting matrix, build a defect detection model for each category label;

[0046] Detection module: Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

[0047] The present invention provides a detection cloud platform for photoresist defects based on image recognition, which is used to execute any one of the detection methods for photoresist defects based on image recognition in Embodiments 1 to 7.

[0048] The beneficial effects of the present invention compared with the prior art are as follows: By analyzing the determined image data and photoresist image data, determine multiple categories of lithography data and the category labels for each category of lithography data, analyze the category lithography data, determine the second defect fitting matrix for the category lithography data of each category label, and build a defect detection model for each category label to achieve defect detection of the real-time lithography image. By combining artificial intelligence and computer vision technologies, using an automated defect fitting and classification method, the accuracy and efficiency of defect recognition in the lithography process are improved, quality problems in the production process are avoided, manual intervention is reduced, the degree of automation of defect detection is increased, and production efficiency and product quality are improved.

[0049] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0050] The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0051] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the accompanying drawings:

[0052] Figure 1 is a flowchart of the detection method for photoresist defects based on image recognition in an embodiment of the present invention;

[0053] Figure 2 is a schematic diagram of the detection system for photoresist defects based on image recognition in an embodiment of the present invention. Detailed Embodiments

[0054] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0055] Example 1:

[0056] The present invention provides a method for detecting photoresist defects based on image recognition, referring to Figure 1 , including:

[0057] S1: Obtain the target image data of multiple target images based on lithography within a specified time period, and obtain the photoresist image data of all target images;

[0058] S2: Analyze the target image data and the photoresist image data to determine multiple categories of lithography data and the category labels of each category of lithography data;

[0059] S3: Analyze the category lithography data, determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label;

[0060] S4: Based on the category lithography data of each category label and the second defect fitting matrix, construct a defect detection model for each category label;

[0061] S5: Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

[0062] In this embodiment, the photoresist image data refers to the image of the photoresist after development, which shows the pattern and the defect information therein.

[0063] In this embodiment, the specified time period refers to collecting image data within a specific time window to ensure the timeliness and accuracy of the data.

[0064] In this embodiment, by analyzing the target image data and the photoresist image data, they are classified into different categories of lithography data. Each category of data will be assigned a category label according to the characteristics of the image, such as lithography requirements, attributes, etc.

[0065] In this embodiment, the defect fitting matrix is a data structure used to represent the characteristics and patterns of defects in the image. Based on the category lithography data, first fit the defects of all photoresist images of each target image to construct the first defect fitting matrix. Then, based on the data of all target images, generate the second defect fitting matrix of each category label.

[0066] In this embodiment, using the aforementioned second defect fitting matrix, construct a defect detection model associated with each category label. This model can accurately identify and classify the defect characteristics in different categories, thereby improving the detection accuracy.

[0067] In this embodiment, a real-time lithography image to be detected is obtained and processed through a pre-constructed defect detection model to output defect data of the image, which helps to monitor defects in the lithography process in real time.

[0068] In this embodiment, the defect detection model is a machine learning or deep learning model constructed based on existing category lithography data and a defect fitting matrix. Through training, the model can learn and identify various defect patterns that appear in new input images.

[0069] In this embodiment, each category label corresponds to a dedicated detection model, and the model will perform defect identification based on the input real-time lithography image and the feature information of this category. These models, based on a large amount of training data (i.e., category lithography data) and the second defect fitting matrix, can very accurately capture the defect features in this category.

[0070] The beneficial effects of the above technology are as follows: By analyzing the determined image data and photoresist image data, multiple category lithography data and the category labels of each category lithography data are determined. By analyzing the category lithography data, the second defect fitting matrix of the category lithography data of each category label is determined, and a defect detection model for each category label is constructed to achieve defect detection of real-time lithography images. By combining artificial intelligence and computer vision technologies and using an automated defect fitting and classification method, the accuracy and efficiency of defect identification in the lithography process are improved, quality problems in the production process are avoided, manual intervention is reduced, the automation degree of defect detection is increased, and production efficiency and product quality are improved.

[0071] Embodiment 2:

[0072] Based on the method for detecting photoresist defects based on image recognition in Embodiment 1, target image data of multiple target images based on lithography within a specified time period is obtained, including:

[0073] Target image sub-data of each target image based on lithography within the specified time period is obtained, where the target image sub-data includes the target image type, target image lithography requirements, and target image lithography attributes of the target image;

[0074] Based on the target image sub-data of all target images obtained within the specified time period, the target image data is determined.

[0075] In this embodiment, the specified time period is a period of time before and including the current time. For example, the specified time period can be 10 days. The time difference of 10 days apart may lead to significant differences in the defect type or incidence rate, and such differences may stem from dynamic changes in process conditions, equipment status, environmental factors, etc.

[0076] In this embodiment, the target image sub-data of each image collected within a specified time period contains multiple important pieces of information related to the lithography process. Specifically: The target image type can be an image of a transistor active region, a gate structure image, a metal interconnect image, a via image, an alignment mark image, etc.; The lithography requirements of the target image refer to the specific requirements that the image needs to meet during the lithography process, such as resolution, accuracy, the thickness of the photoresist layer, etc.; The lithography attributes of the target image describe the physical attributes during the lithography process, such as exposure time, light intensity, light source type, etc.

[0077] In this embodiment, the target image sub-data of all the collected target images are comprehensively determined to obtain the target image data. This data will integrate all the relevant information of all the target images.

[0078] The beneficial effects of the above technology are as follows: Obtaining the target image data of multiple target images based on lithography within a specified time period can provide data support for determining category lithography data and category labels.

[0079] Embodiment 3:

[0080] Based on the method for detecting photoresist defects based on image recognition in Embodiment 1, the photoresist image data of all the target images is obtained, including:

[0081] All the photoresist images after development of each target image within a specified time period are obtained, and the photoresist image sub-data of each target image is determined;

[0082] The photoresist image data is determined based on the photoresist image sub-data of all the target images.

[0083] In this embodiment, multiple photoresist images of each target image after development are obtained within a specified time period. After the development process, the photoresist images will reflect the patterns formed and possible defects during the lithography process.

[0084] In this embodiment, the photoresist image sub-data of all the target images are integrated to generate a comprehensive photoresist image data set. This data set will contain all the information of the photoresist images after development of all the target images.

[0085] The beneficial effects of the above technology are as follows: Obtaining the photoresist image data of all the target images can provide data support for determining category lithography data and category labels.

[0086] Embodiment 4:

[0087] Based on the method for detecting photoresist defects based on image recognition in Embodiment 1, the target image data and the photoresist image data are analyzed to determine multiple category lithography data and the category labels of each category lithography data, including:

[0088] Preprocess the target image data and the photoresist image data respectively;

[0089] Extract the target image lithography requirements in the target image sub - data of each target image in the preprocessed target image data, determine multiple requirement features of each target image, and determine the requirement quantization value of each requirement feature of each target image;

[0090] Extract the target image lithography attributes in the target image sub - data of each target image in the preprocessed target image data, determine multiple attribute features of each target image, and determine the attribute quantization value of each attribute feature of each target image;

[0091] Based on the requirement quantization values of all requirement features and the attribute quantization values of all attribute features of each target image in the preprocessed target image data, determine the requirement - attribute level of each target image;

[0092] Extract the target image type in the target image sub - data of each target image in the preprocessed target image data. Based on the extracted target image type and the determined requirement - attribute level, determine the type - level label of each target image;

[0093] Based on the preprocessed photoresist image sub - data of all target images with the same type - level label, determine the category lithography data, and determine the category label for the category lithography data with the type - level label.

[0094] In this embodiment, the preliminary processing of the target image data and the photoresist image data usually includes denoising, normalization, data augmentation, etc., to improve the data quality and prepare for subsequent analysis.

[0095] In this embodiment, the requirement features of the target image can be the minimum line - width requirement, alignment accuracy requirement, number of lithography layers requirement, mask complexity requirement, etc.

[0096] In this embodiment, the attribute features of the target image can be line density, pattern curvature, edge sharpness, expansion coefficient, photoresist thickness uniformity, etc.

[0097] In this embodiment, the requirement - attribute level can include level one, level two, level three, etc.

[0098] In this embodiment, to determine the requirement - attribute level of the target image, the requirement - attribute evaluation value of the target image can be calculated first, and the requirement - attribute level of the target image is determined according to the requirement - attribute evaluation value. The requirement - attribute evaluation value can be calculated by calculating the requirement quantization values of all requirement features of the corresponding target image, the requirement - feature weights, the attribute quantization values of all attribute features, and the attribute - quantization weights.

[0099] In this embodiment, according to the target image type and the requirement attribute level, the type-level label of each image is determined, which helps to classify the images.

[0100] In this embodiment, based on the type-level label of the target image, all photoresist image sub-data are classified to obtain the corresponding category photolithography data, and the label of each category is determined.

[0101] The beneficial effects of the above technology are as follows: By analyzing the target image data and the photoresist image data, determining multiple category photolithography data and the category labels of each category of photolithography data, refined classification of photolithography images can be achieved, the accuracy of defect detection in the photolithography process can be improved, and the product quality and production efficiency can be enhanced.

[0102] Embodiment 5:

[0103] Based on the method for detecting photoresist defects based on image recognition in Embodiment 1, analyzing the category photolithography data to determine the first defect fitting matrix of each target image in the category photolithography data, including:

[0104] Extract the defect features of all photoresist images in the preprocessed photoresist image sub-data of each target image in the category photolithography data to determine the defect feature matrix of all photoresist images of each target image in the category photolithography data;

[0105]

[0106] Among them, represents the defect feature matrix of the i-th target image in the category photolithography data with the category label of type a and level b, respectively represent the first defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image in the category photolithography data with the category label of type a and level b. iN2 represents the number of defects of the i-th target image, respectively represent the k-th defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image in the category photolithography data with the category label of type a and level b. N1 represents the number of defect features, respectively represent the N1-th defect feature of the first defect, the j-th defect, and the iN2-th defect of the i-th target image in the category photolithography data with the category label of type a and level b, that is, the defect type;

[0107] Extract the defect types of the defect feature matrices of all photoresist images of each target image to determine the defect type set of each target image;

[0108] Based on the defect feature matrix and the defect type set of all photoresist images of each target image, determine the first defect fitting matrix of each target image in the category lithography data;

[0109]

[0110] Among them, represents the first defect fitting matrix of the i-th target image in the category lithography data with category label of type a and grade b, respectively represent the first fitting features of the 1st defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image in the category lithography data with category label of type a and grade b, respectively represent the k-th fitting features of the 1st defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image in the category lithography data with category label of type a and grade b, respectively represent the N1-1-th fitting features of the 1st defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image in the category lithography data with category label of type a and grade b. iN4 represents the number of defects in the defect type set of the i-th target image, represents the p-th defect type in the defect type set of the i-th target image in the category lithography data with category label of type a and grade b, represents the first exponential function.

[0111] In this embodiment, the first exponential function represents the indicator function of the j-th defect of the i-th target image in the category lithography data with category label of type a and grade b based on the p-th defect type.

[0112] In this embodiment, defect features are extracted from each photoresist image in the preprocessed photoresist image sub-data of each target image. These features generally include the shape, size, position, edge features, irregularities in the photoresist pattern, brightness changes, defect types, etc. of all defects. The defect features of all extracted defects are organized into a defect feature matrix, which reflects the defect information of each target image and facilitates subsequent analysis and detection.

[0113] In this embodiment, the defect types in the defect feature matrix of each target image are extracted. For example, possible defect types include scratches, offsets, exposures, etc., and the defect type set of each target image is determined.

[0114] In this embodiment, combining the defect feature matrix and the defect type set of each target image, determine the first defect fitting matrix.

[0115] The beneficial effects of the above technology are as follows: By analyzing category lithography data and determining the first defect fitting matrix of each target image in the category lithography data, it can provide data support for determining the second defect fitting matrix of the category lithography data of each category label, perform efficient and accurate defect classification for each target image, and improve the accuracy of defect detection in the lithography process.

[0116] Example 6:

[0117] Based on the method for detecting photoresist defects based on image recognition in Example 1, determining the second defect fitting matrix of the category lithography data of each category label includes:

[0118] Based on the set of defect types of all target images in each category lithography data, determine the category defect set of each category lithography data;

[0119] Based on the category defect set of each category lithography data and the first defect fitting matrix of all target images of each category lithography data, determine the second defect fitting matrix of the category lithography data of each category label;

[0120]

[0121] Among them, CM ab represents the second defect fitting matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the first fitting feature of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the k-th fitting feature of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the N1 - 1-th fitting feature of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, where abN5 represents the number of category defects in the category defect set of the category lithography data with the category label of type a and grade b, represents the q-th defect type in the category defect set of the category lithography data with the category label of type a and grade b, represents the second exponential function, Nu represents the number of target images in the category lithography data, W a represents the type weight of the category label of type a, W ab represents the grade weight of the category label of type a and grade b.

[0122] In this example, The number of defects of the p-th defect type in the i-th target image representing the category lithography data with the category label of type a and grade b.

[0123] In this embodiment, the second exponential function Represents the q-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, based on the indicator function of the p-th defect type in the defect type set of the i-th target image.

[0124] In this embodiment, by analyzing the defect types of all target images in each category of lithography data, the category defect set of each category of lithography data is identified and determined.

[0125] In this embodiment, by combining the category defect set of each category of lithography data with the first defect fitting matrix of each target image in this category of lithography data, the second defect fitting matrix of each category label is generated.

[0126] The beneficial effects of the above technology are as follows: Determining the second defect fitting matrix of the category lithography data of each category label can significantly improve the accuracy of defect recognition and classification, optimize the lithography quality control in the production process, and improve production efficiency and product quality.

[0127] Embodiment 7:

[0128] Based on the method for detecting photoresist defects based on image recognition in Embodiment 1, a real-time lithography image to be detected is obtained, and defect data of the real-time lithography image is determined based on the real-time lithography image and all defect detection models, including:

[0129] Based on the real-time lithography image to be detected, a real-time target image is determined, and based on the real-time image type, real-time image lithography requirements, and real-time image lithography attributes of the real-time target image, the real-time label of the real-time lithography image is determined;

[0130] Select the defect detection model with the same category label as the real-time label of the real-time lithography image as the defect detection model of the real-time lithography image;

[0131] The real-time lithography image is input into the determined defect detection model, and the defect data of the real-time lithography image is determined based on the output result of the defect detection model.

[0132] In this embodiment, a real-time lithography image to be detected is obtained, a real-time target image is determined, feature extraction is performed on the real-time image lithography requirements and real-time image lithography attributes of the real-time target image, and the real-time label is determined.

[0133] In this embodiment, according to the real-time label, a defect detection model with the same category label is selected.

[0134] In this embodiment, the real-time lithography image is input into the selected defect detection model, and the model will process the image and output defect data, that is, information such as the type, location, and size of the defects in the image.

[0135] The beneficial effects of the above technology are as follows: By obtaining the real-time lithography image to be detected and determining the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models, the accuracy and timeliness of defect detection can be improved, the defect recognition efficiency of lithography images can be enhanced, human intervention and errors can be reduced, and the production stability and product quality can be improved.

[0136] Embodiment 8:

[0137] The present invention provides a detection system for photoresist defects based on image recognition, which is used to execute any one of the detection methods for photoresist defects based on image recognition in Embodiments 1 to 7. Refer to Figure 2 , and includes:

[0138] Acquisition module: Acquire the target image data of multiple target images based on lithography within a specified time period, and acquire the photoresist image data of all target images;

[0139] Analysis module: Analyze the target image data and the photoresist image data to determine multiple categories of lithography data and the category labels of each category of lithography data;

[0140] Determination module: Analyze the category lithography data, determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label;

[0141] Construction module: Based on the category lithography data of each category label and the second defect fitting matrix, construct a defect detection model for each category label;

[0142] Detection module: Acquire the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

[0143] The beneficial effects of the above technology are as follows: By analyzing and determining the image data and the photoresist image data, determining multiple categories of lithography data and the category labels of each category of lithography data, analyzing the category lithography data, determining the second defect fitting matrix of the category lithography data of each category label, and constructing a defect detection model for each category label, the defect detection of real-time lithography images is realized. By combining artificial intelligence and computer vision technologies and using automated defect fitting and classification methods, the accuracy and efficiency of defect recognition in the lithography process are improved, quality problems in the production process are avoided, human intervention is reduced, the automation degree of defect detection is enhanced, and production efficiency and product quality are improved.

[0144] Embodiment 9:

[0145] The present invention provides a cloud platform for detecting photoresist defects based on image recognition, which is used to execute any one of the methods for detecting photoresist defects based on image recognition in Embodiments 1 to 7.

[0146] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A method for detecting photoresist defects based on image recognition, characterized in that, Including: S1: Obtain the target image data of multiple target images based on lithography within a specified time period, and obtain the photoresist image data of all target images; S2: Analyze the target image data and the photoresist image data to determine multiple categories of lithography data and the category labels of each category of lithography data; S3: Analyze the category lithography data, determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label; S4: Based on the category lithography data of each category label and the second defect fitting matrix, construct a defect detection model for each category label; S5: Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

2. The detection method of photoresist defects based on image recognition according to claim 1, characterized in that, Obtaining the target image data of multiple target images based on lithography within a specified time period includes: Obtain the target image sub-data of each target image based on lithography within a specified time period, where the target image sub-data includes the target image type, target image lithography requirements, and target image lithography attributes of the target image; Based on the target image sub-data of all target images obtained within the specified time period, determine the target image data.

3. The detection method of photoresist defects based on image recognition according to claim 2, wherein, Obtaining the photoresist image data of all target images includes: Obtain all photoresist images after development of each target image within a specified time period, and determine the photoresist image sub-data of each target image; Based on the photoresist image sub-data of all target images, determine the photoresist image data.

4. The method for detecting photoresist defects based on image recognition according to claim 3, wherein, Analyzing the target image data and the photoresist image data to determine multiple categories of lithography data and the category labels of each category of lithography data includes: Preprocess the target image data and the photoresist image data respectively; Extract the target image lithography requirements in the target image sub-data of each target image in the preprocessed target image data, determine multiple demand features of each target image, and determine the demand quantization value of each demand feature of each target image; Extract the target image lithography attributes in the target image sub-data of each target image in the preprocessed target image data, determine multiple attribute features of each target image, and determine the attribute quantization value of each attribute feature of each target image; Based on the demand quantization values of all demand features and the attribute quantization values of all attribute features of each target image in the preprocessed target image data, determine the demand attribute level of each target image; Extract the target image type in the target image sub-data of each target image in the preprocessed target image data, and based on the extracted target image type and the determined demand attribute level, determine the type level label of each target image; Based on the preprocessed photoresist image sub-data of all target images with the same type level label, determine the category lithography data, and determine the type level label as the category label of the category lithography data.

5. The detection method of photoresist defects based on image recognition according to claim 4, wherein, Analyzing the category lithography data to determine the first defect fitting matrix of each target image in the category lithography data includes: Extract defect features from all photoresist images in the preprocessed photoresist image sub-data of each target image in the category lithography data, and determine the defect feature matrix of all photoresist images of each target image in the category lithography data; Among them, denotes the defect feature matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively denote the first defect features of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category lithography data with the category label of type a and grade b. iN2 represents the number of defects in the i-th target image, respectively denote the k-th defect features of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category lithography data with the category label of type a and grade b. N1 represents the number of defect features, respectively denote the N1-th defect features of the first defect, the j-th defect, and the iN2-th defect of the i-th target image of the category lithography data with the category label of type a and grade b, that is, the defect type; Extract the defect types of the defect feature matrices of all photoresist images of each target image, and determine the defect type set of each target image; Based on the defect feature matrix and the defect type set of all photoresist images of each target image, determine the first defect fitting matrix of each target image in the category lithography data; Among them, represents the first defect fitting matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the first fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the k-th fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the N1 - 1-th fitting features of the first defect type, the p-th defect type, and the iN4-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, where iN4 represents the number of defects in the defect type set of the i-th target image, represents the p-th defect type in the defect type set of the i-th target image of the category lithography data with the category label of type a and grade b, represents the first exponential function.

6. The detection method of photoresist defects based on image recognition according to claim 5, wherein, Determine the second defect fitting matrix of the category lithography data of each category label, including: Based on the defect type sets of all target images in each category lithography data, determine the category defect set of each category lithography data; Based on the category defect set of each category lithography data and the first defect fitting matrices of all target images of each category lithography data, determine the second defect fitting matrix of the category lithography data of each category label; Among them, CM ab represents the second defect fitting matrix of the i-th target image of the category lithography data with the category label of type a and grade b, respectively represent the first fitting features of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the k-th fitting features of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, respectively represent the N1 - 1-th fitting features of the first category defect, the q-th category defect, and the abN5-th category defect in the category defect set of the category lithography data with the category label of type a and grade b, where abN5 represents the number of category defects in the category defect set of the category lithography data with the category label of type a and grade b, represents the q-th defect type in the category defect set of the category lithography data with the category label of type a and grade b, represents the second exponential function, Nu represents the number of target images in the category lithography data, W a represents the type weight with the category label of type a, W ab represents the grade weight with the category label of type a and grade b.

7. The detection method of photoresist defects based on image recognition according to claim 1, wherein Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models, including: Determine the real-time target image based on the real-time lithography image to be detected, and determine the real-time label of the real-time lithography image based on the real-time image type, the real-time lithography requirement, and the real-time lithography attribute of the real-time target image; Select the defect detection model of the category label that is the same as the real-time label of the real-time lithography image as the defect detection model of the real-time lithography image; Input the real-time lithography image into the determined defect detection model, and determine the defect data of the real-time lithography image based on the output result of the defect detection model.

8. A detection system for photoresist defects based on image recognition, characterized in that, For implementing any one of the methods for detecting photoresist defects based on image recognition in Embodiments 1 to 7, including: Acquisition module: Acquire the target image data of multiple target images based on lithography within a specified time period, and acquire the photoresist image data of all target images; Analysis module: Analyze the target image data and the photoresist image data to determine multiple category lithography data and the category label of each category lithography data; Determination module: Analyze the category lithography data, determine the first defect fitting matrix of each target image in the category lithography data, and determine the second defect fitting matrix of the category lithography data of each category label; Construction module: Based on the category lithography data of each category label and the second defect fitting matrix, construct the defect detection model of each category label; Detection module: Obtain the real-time lithography image to be detected, and determine the defect data of the real-time lithography image based on the real-time lithography image and all defect detection models.

9. A cloud platform for detecting photoresist defects based on image recognition, characterized in that For implementing any one of the methods for detecting photoresist defects based on image recognition in Claims 1 to 7.

Images