Silicon wafer quality evaluation method based on image features

Abstract

The invention discloses a silicon wafer quality evaluation method based on image features, relates to the technical field of image processing, and solves the problem that feature extraction and feature quantification are difficult to perform according to a surface image. Secondly, feature selection is difficult to generate feature vectors according to feature parameters; next, silicon wafer defects are difficult to analyze and identify according to the feature vectors, and silicon wafer deformation detection is difficult to carry out; secondly, the distribution of dopants in the ROI region is difficult to analyze, and the metal pollution of the silicon wafer is difficult to analyze; and finally, the quality evaluation index of the silicon wafer is difficult to comprehensively analyze according to various data. According to the invention, defects on the surface of the silicon wafer can be identified through image processing and feature extraction technologies, and the deformation of the silicon wafer is measured through a deformation detection technology; the performance change rate of the silicon wafer is obtained by analyzing dopant distribution of the ROI, and the metal pollution rate is obtained through a metal pollution detection technology; and integrating various data to evaluate the quality of the silicon wafer.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and in particular is a silicon wafer quality assessment method based on image features. Background Art

[0002] With the rapid development of image processing technology, silicon wafer quality assessment methods based on image features have gradually become a research hotspot. Silicon wafers are key materials for solar cells, semiconductor devices, etc., and their quality directly affects the performance and reliability of products. Therefore, accurate and rapid quality assessment of silicon wafers is an important link to ensure product quality and production efficiency. Image processing technology can automatically extract image features on the surface of silicon wafers, and analyze and process these features through machine learning algorithms, thereby achieving accurate assessment of silicon wafer quality.

[0003] The prior art has the following problems: first, it is difficult to extract and quantify features based on surface images; second, it is difficult to generate feature vectors by feature selection based on feature parameters; then, it is difficult to identify silicon wafer defects based on feature vector analysis, and it is difficult to detect silicon wafer deformation; then, it is difficult to analyze the dopant distribution in the ROI area, and it is difficult to analyze the metal contamination of the silicon wafer; finally, it is difficult to comprehensively analyze the quality evaluation index of the silicon wafer based on various data. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a silicon wafer quality assessment method based on image features, which is used to solve the following technical problems: First, it is difficult to extract and quantify features based on surface images. Second, it is difficult to generate feature vectors by selecting features based on feature parameters. Next, it is difficult to identify silicon wafer defects based on feature vector analysis, and difficult to detect silicon wafer deformation. Then, it is difficult to analyze the dopant distribution in the ROI area, and difficult to analyze the metal contamination of silicon wafers. Finally, it is difficult to comprehensively analyze the quality assessment index of silicon wafers based on various data.

[0005] To solve the above problems, a first aspect of the present invention provides a method for evaluating silicon wafer quality based on image features, comprising the following steps: S1: Collecting surface images and performance data of silicon wafers after initial cleaning, the performance data including: silicon wafer thickness, carrier concentration and mobility, hole concentration and mobility, elementary charge and resistivity, and recording the length of initial cleaning time; S2: preprocessing the collected surface images including: filtering and denoising, image enhancement and grayscale conversion; preprocessing the collected performance data including: cleaning and standardization; S3: extracting features and quantifying features according to the preprocessed surface image, extracting feature parameters; preprocessing the extracted feature parameters, including: cleaning and standardization; performing feature selection according to the preprocessed feature parameters, and generating feature vectors according to the selection results; S4: Identify silicon wafer defects based on feature vector analysis and calculate defect values ​​based on analysis results; perform silicon wafer deformation detection and calculate the deformation amount of silicon wafer based on detection results; analyze dopant distribution in ROI area; calculate silicon wafer performance change rate based on dopant distribution results; analyze metal contamination of silicon wafer and calculate metal contamination rate based on analysis results; S5: Evaluate the quality of silicon wafers based on comprehensive analysis of defect value, deformation, silicon wafer performance change rate and metal contamination rate.

[0006] As a further solution of the present invention: in step S3, feature extraction and quantification of features are performed based on the preprocessed surface image to extract feature parameters, including the following steps: According to the preprocessed surface image, the edge contour of the silicon wafer is extracted by using an edge detection algorithm; according to the edge contour of the silicon wafer, the shape characteristics of the surface image are extracted by morphological processing, and the characteristic parameters of the shape characteristics include: the area within the edge contour area and the total length of the edge contour; through the feature detection algorithm, the ROI area is detected from the preprocessed surface image, and the characteristic parameters of the ROI area extracted according to the detection result include: the coordinates of the four corners of the ROI area, the coordinates of the center point, the width, the height and the area; the texture features are extracted according to the grayscale co-occurrence matrix of the surface image, and the characteristic parameters of the texture features include: contrast, energy, homogeneity, entropy and correlation.

[0007] As a further solution of the present invention: in step S3, feature selection is performed according to the preprocessed feature parameters, and a feature vector is generated according to the selection result, including the following steps: A first feature vector is generated according to the extracted feature parameters. The first feature vector is expressed as ; By analyzing the formula: Get the selected feature parameters ; Where Y represents the selection threshold coefficient, represents the i-th characteristic parameter value in the first characteristic vector, and N represents the total number of characteristic parameters in the first characteristic vector; According to the selected characteristic parameters, a second characteristic vector is generated, and the second characteristic vector is expressed as , where M represents the total number of feature parameters after selection.

[0008] As a further solution of the present invention: the step S4 of identifying silicon wafer defects according to the feature vector analysis and calculating the defect value according to the analysis result includes the following steps: Based on the selected second eigenvector, a convolutional neural network model is used to analyze whether the silicon wafer has defects and identify the defect type; Annotating the second eigenvector and the silicon wafer defect type, wherein the silicon wafer defect type includes: cracks, scratches, dents, edge collapse and bubbles; inputting the annotated second eigenvector and the silicon wafer defect type into a convolutional neural network model for training; According to the trained convolutional neural network model, by collecting and preprocessing the silicon wafer surface image in real time, repeating step S3, inputting the feature vector extracted from the real-time collected silicon wafer surface image into the trained convolutional neural network model to identify and determine whether the silicon wafer has defects, and outputting the defect type; When a defect is identified in a silicon wafer, characteristic parameters of the defect are extracted, including: defect area and defect depth; By analyzing the formula: Get defect value ;in, represents the identified defect area, A represents the area within the edge contour area, represents the identified defect depth, and W represents the thickness of the silicon wafer; When it is identified that the silicon wafer has no defects, a deformation detection of the silicon wafer is performed.

[0009] As a further solution of the present invention: the step S4 of detecting the deformation of the silicon wafer and calculating the deformation amount of the silicon wafer according to the detection result includes the following steps: When it is identified that the silicon wafer does not have defects, the undeformed standard silicon wafer image is collected and preprocessed, the characteristic parameters of the undeformed standard silicon wafer image are extracted and preprocessed for cleaning and standardization, and the standard feature vector is generated through feature selection. ; According to the second eigenvector and the standard eigenvector, a comparative analysis is performed, through the analysis formula: Get the comparison similarity S; where, represents the second eigenvector, represents the standard eigenvector, represents the magnitude of the second eigenvector, represents the magnitude of the standard eigenvector; According to the calculation results of the comparison similarity, a threshold is set to analyze and judge whether the silicon wafer is deformed; When , the silicon wafer does not deform; when When the silicon wafer is deformed; When the silicon wafer is deformed, the following formula is used to analyze it: The deformation B is obtained; where represents the i-th eigenparameter of the second eigenvector, represents the i-th feature parameter of the standard feature vector, and M represents the total number of feature parameters after selection; When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed.

[0010] As a further solution of the present invention: the analysis of the dopant distribution in the ROI region in step S4 comprises the following steps: When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed using a deep learning model based on the extracted characteristic parameters of the ROI area; Marking characteristic parameters of the ROI region and characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution, wherein the dopant distribution includes: uniform distribution, local aggregation and sparse distribution; inputting the marked characteristic parameters of the ROI region and the characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution into a deep learning model for training; According to the trained deep learning model, the characteristic parameters of the ROI area in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing, and the extracted characteristic parameters of the ROI area are input into the trained deep learning model for analysis and identification of the dopant distribution in the ROI area, and the dopant distribution results of the ROI area are output.

[0011] As a further solution of the present invention: the step S4 of analyzing and calculating the silicon wafer performance change rate according to the dopant distribution result includes the following steps: By analyzing the formula: Get the conductivity of the silicon wafer ,in, represents the carrier concentration, represents the elementary charge, represents the carrier mobility, represents the hole concentration, represents the hole mobility; According to the characteristic parameters of the ROI area of ​​the silicon wafer surface image collected in real time, when it is identified that the dopant distribution in the ROI area is locally concentrated or sparsely distributed, the analysis formula is: Get the silicon wafer performance change rate ,in, The conductivity of the silicon wafer when the dopant distribution in the ROI area is uniformly distributed. The conductivity of the silicon wafer indicates whether the dopant distribution in the ROI area is locally concentrated or sparsely distributed.

[0012] As a further solution of the present invention: the step S4 of analyzing the metal contamination of the silicon wafer and calculating the metal contamination rate according to the analysis result comprises the following steps: Acquire a silicon wafer surface image sample containing metal contamination and pre-process the silicon wafer surface image sample; Extracting metal features from the pre-processed silicon wafer surface image sample, including: color features and texture features; the color features include: RGB value, HSV value and gray value; the texture features include: contrast, energy, homogeneity, entropy and correlation; Annotate the metal features in the preprocessed silicon wafer surface image samples, and input the metal features in the silicon wafer surface image samples into the deep learning model for training; According to the trained deep learning model, the metal features in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing of the silicon wafer surface image, and the extracted metal features are input into the trained deep learning model to analyze and identify whether there is metal contamination in the real-time silicon wafer surface image, and the identification result is output; When metal contamination is identified, the analysis formula is: The metal contamination rate G is obtained; where A represents the area within the edge contour region, Represents the area of ​​metal contamination.

[0013] As a further solution of the present invention: the step S5 comprises the following steps: By analyzing the formula: The quality assessment index Q is obtained; among which D represents the defect value, B represents the deformation, R represents the silicon wafer performance change rate, and G represents the metal contamination rate. , , and They represent the weight coefficients of defect value, deformation, silicon wafer performance change rate and metal contamination respectively; Set a threshold H. When the quality evaluation index is greater than or equal to the threshold H, the quality of the silicon wafer is unqualified; when the quality evaluation index is less than the threshold H, the quality of the silicon wafer is qualified; Whether the quality of the silicon wafer after initial cleaning is qualified is analyzed by collecting the surface images and performance data of the silicon wafer after initial cleaning in real time; when the quality of the silicon wafer is unqualified, the silicon wafer is cleaned again, and the time of the Etch etching step is increased by 12 seconds and the time length of the Coat coating step is increased to 60 seconds, and whether the quality of the silicon wafer after the second cleaning is qualified is judged; when the quality of the silicon wafer after the second cleaning is judged to be unqualified, the silicon wafer is cleaned and tested again.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention can accurately identify and quantify defects on the surface of silicon wafers through image processing and feature extraction technology, accurately measure the deformation of silicon wafers through deformation detection technology, evaluate the performance changes of silicon wafers during doping and other processes by analyzing the dopant distribution in the ROI area and the performance change rate of silicon wafers, and accurately measure the metal contamination rate on the surface of silicon wafers through metal contamination detection technology; The present invention selects the most critical features for silicon wafer quality assessment by performing feature selection on the extracted feature parameters, thereby reducing the interference of redundant information on the assessment results and improving the accuracy and reliability of the assessment. The present invention uses a deep learning model to perform intelligent analysis on the surface image of the silicon wafer. This intelligent analysis method not only reduces the possibility of manual intervention and misjudgment, but also improves the efficiency and accuracy of the analysis, thereby achieving rapid and accurate assessment of silicon wafer quality. The present invention comprehensively evaluates the quality of silicon wafers by comprehensively analyzing the surface image and various performance data of silicon wafers, which can more accurately reflect the overall quality status of silicon wafers and improve the accuracy of evaluating the quality of silicon wafers. For silicon wafers initially judged to be unqualified, the quality of silicon wafers is improved by re-cleaning by increasing the time of the Etch etching step and extending the time of the Coat coating step. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 The figure is a flow chart of the system method of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 According to a first aspect of the present invention, a method for evaluating silicon wafer quality based on image features is provided, comprising the following steps: S1: Collecting surface images and performance data of silicon wafers after initial cleaning, the performance data including: silicon wafer thickness, carrier concentration and mobility, hole concentration and mobility, elementary charge and resistivity, and recording the length of initial cleaning time; S2: preprocessing the collected surface images including: filtering and denoising, image enhancement and grayscale conversion; preprocessing the collected performance data including: cleaning and standardization; S3: extracting features and quantifying features according to the preprocessed surface image, extracting feature parameters; preprocessing the extracted feature parameters, including: cleaning and standardization; performing feature selection according to the preprocessed feature parameters, and generating feature vectors according to the selection results; S4: Identify silicon wafer defects based on feature vector analysis and calculate defect values ​​based on analysis results; perform silicon wafer deformation detection and calculate the deformation amount of silicon wafer based on detection results; analyze dopant distribution in ROI area; calculate silicon wafer performance change rate based on dopant distribution results; analyze metal contamination of silicon wafer and calculate metal contamination rate based on analysis results; S5: Evaluate the quality of silicon wafers based on comprehensive analysis of defect value, deformation, silicon wafer performance change rate and metal contamination rate.

[0019] Specifically, obtain the surface image of the silicon wafer after initial cleaning, use the corresponding test instrument, and measure the thickness, carrier concentration, carrier mobility and resistivity performance data of the silicon wafer according to the standard test method. Use Gaussian filtering, mean filtering and other algorithms to remove noise in the image, improve the clarity of the image through contrast enhancement and sharpening algorithms, and convert the color image into a grayscale image to reduce the amount of data and improve the processing speed. Remove outliers or invalid data in the performance data, and convert the performance data into the same dimension to eliminate the dimensional differences between different parameters. Extract features from the pre-processed surface image, and clean and standardize the extracted feature parameters. Select the feature parameters through the formula and generate feature vectors. Identify defects on the surface of the silicon wafer based on the feature vector analysis; detect the deformation of the silicon wafer. In the ROI area, use the image processing algorithm to extract the dopant distribution information; analyze and identify the metal contamination on the surface of the silicon wafer. Comprehensively calculate the various data to obtain the quality evaluation index, and evaluate the quality of the silicon wafer based on the size of the quality evaluation index.

[0020] In one embodiment of the present invention, the step S3 performs feature extraction and quantification based on the preprocessed surface image to extract feature parameters, including the following steps: According to the preprocessed surface image, the edge contour of the silicon wafer is extracted by using an edge detection algorithm; according to the edge contour of the silicon wafer, the shape characteristics of the surface image are extracted by morphological processing, and the characteristic parameters of the shape characteristics include: the area within the edge contour area and the total length of the edge contour; through the feature detection algorithm, the ROI area is detected from the preprocessed surface image, and the characteristic parameters of the ROI area extracted according to the detection result include: the coordinates of the four corners of the ROI area, the coordinates of the center point, the width, the height and the area; the texture features are extracted according to the grayscale co-occurrence matrix of the surface image, and the characteristic parameters of the texture features include: contrast, energy, homogeneity, entropy and correlation.

[0021] Specifically, edge detection algorithms such as Canny, Sobel, and Laplacian are used to apply edge detection algorithms to the preprocessed image, and the edges of the silicon wafers in the silicon wafer surface image are extracted and detected to form the edge contour of the silicon wafer. The edge contour is processed by using morphological operations such as corrosion, expansion, opening operation, and closing operation. These operations can remove small noise points on the edge contour, fill holes, and smooth the edges, and extract the shape features of the surface image. The characteristic parameters of the shape features include: the area within the edge contour area and the total length of the edge contour. Using algorithms such as color threshold, shape matching, and template matching, the ROI detection algorithm is applied to locate the region of interest. The four corner coordinates, center point coordinates, width, height, and area of ​​the ROI region are extracted. The four corner coordinates can be obtained by the vertices of the detected ROI bounding box. The center point coordinates can be obtained by calculating the centroid of the ROI region. The width and height can be directly obtained from the size of the ROI bounding box. The area can be obtained by calculating the number of pixels in the ROI region. The grayscale co-occurrence matrix of the preprocessed image is calculated, and texture feature parameters such as contrast, energy, homogeneity, entropy, and correlation are extracted from the grayscale co-occurrence matrix. These characteristic parameters can be obtained by calculating the statistical characteristics of the gray-level co-occurrence matrix. The contrast is obtained by calculating the variance of the element values ​​in the gray-level co-occurrence matrix. The energy is the sum of the squares of the element values ​​in the gray-level co-occurrence matrix. The homogeneity is obtained by calculating the reciprocal of the sum of the squares of the differences between the element values ​​in the gray-level co-occurrence matrix and the means of their corresponding rows or columns. The entropy is the sum of the negative logarithms of the probability distribution of the element values ​​in the gray-level co-occurrence matrix. The correlation is obtained by calculating the correlation coefficient between the element values ​​in the gray-level co-occurrence matrix and the means of their corresponding rows or columns.

[0022] In one embodiment of the present invention, the step S3 performs feature selection according to the preprocessed feature parameters, and generates a feature vector according to the selection result, including the following steps: A first feature vector is generated according to the extracted feature parameters. The first feature vector is expressed as ; By analyzing the formula: Get the selected feature parameters ; Where Y represents the selection threshold coefficient, represents the i-th characteristic parameter value in the first characteristic vector, and N represents the total number of characteristic parameters in the first characteristic vector; According to the selected characteristic parameters, a second characteristic vector is generated, and the second characteristic vector is expressed as , where M represents the total number of feature parameters after selection.

[0023] Specifically, according to the feature extraction of edge detection, morphological processing, ROI area detection and grayscale co-occurrence matrix texture, the feature parameters extracted from the silicon wafer surface image are arranged in order to form a feature vector. According to the analysis requirements, the threshold coefficient Y is set to 3. This coefficient is used to screen important feature parameters, and is usually adjusted dynamically based on factors such as the significance, correlation or contribution of the feature parameters. According to the results of the formula, the feature parameters that meet the conditions are screened. These screened feature parameters constitute the second feature vector.

[0024] In one embodiment of the present invention, the step S4 of identifying silicon wafer defects according to feature vector analysis and calculating defect values ​​according to the analysis results includes the following steps: Based on the selected second eigenvector, a convolutional neural network model is used to analyze whether the silicon wafer has defects and identify the defect type; Annotating the second eigenvector and the silicon wafer defect type, wherein the silicon wafer defect type includes: cracks, scratches, dents, edge collapse and bubbles; inputting the annotated second eigenvector and the silicon wafer defect type into a convolutional neural network model for training; According to the trained convolutional neural network model, by collecting and preprocessing the silicon wafer surface image in real time, repeating step S3, inputting the feature vector extracted from the real-time collected silicon wafer surface image into the trained convolutional neural network model to identify and determine whether the silicon wafer has defects, and outputting the defect type; When a defect is identified in a silicon wafer, characteristic parameters of the defect are extracted, including: defect area and defect depth; By analyzing the formula: Get defect value ;in, represents the identified defect area, A represents the area within the edge contour area, represents the identified defect depth, and W represents the thickness of the silicon wafer; When it is identified that the silicon wafer has no defects, a deformation detection of the silicon wafer is performed.

[0025] Specifically, the defect type of the silicon wafer is annotated according to the second feature vector extracted from the silicon wafer surface image. The silicon wafer defect types include but are not limited to: cracks, scratches, dents, edge collapse and bubbles. Create a data set, in which each sample contains the second feature vector and its corresponding defect type label. Divide the data set into a training set, a validation set and a test set for model training, validation and testing. Construct a convolutional neural network model, and train the convolutional neural network model using the training set data. During the training process, adjust the model parameters by optimization methods such as back propagation algorithm and gradient descent to minimize the loss function. Monitor indicators such as loss value and accuracy during the training process to evaluate the training effect of the model. Use the validation set data to verify the trained model. Adjust the model structure or hyperparameters according to the verification results to improve the generalization ability of the model. Use an image acquisition device to acquire the silicon wafer surface image in real time, and preprocess the acquired image, including denoising and contrast enhancement. Repeat the steps of extracting feature parameters and selecting features for the preprocessed silicon wafer surface image, extract the second feature vector of the real-time image, and input the extracted second feature vector into the trained convolutional neural network model for recognition. The model will output whether the silicon wafer has defects and the type of defects. When defects are identified in the silicon wafer, the characteristic parameters of the defects are further extracted, including the defect area and defect depth. These parameters can be obtained through image processing technology and geometric measurement. According to the extracted defect characteristic parameters and the relevant parameters of the silicon wafer, the defect value is calculated using an analytical formula. The defect value can be used to assess the severity and impact range of the defect. According to the actual application effect and feedback, the convolutional neural network model is continuously optimized and updated. More data can be collected to enhance the generalization ability of the model, or the model structure can be adjusted to improve recognition accuracy and efficiency.

[0026] In one embodiment of the present invention, the step S4 of detecting the deformation of the silicon wafer and calculating the deformation amount of the silicon wafer according to the detection result includes the following steps: When it is identified that the silicon wafer does not have defects, the undeformed standard silicon wafer image is collected and preprocessed, the characteristic parameters of the undeformed standard silicon wafer image are extracted and preprocessed for cleaning and standardization, and the standard feature vector is generated through feature selection. ; According to the second eigenvector and the standard eigenvector, a comparative analysis is performed, through the analysis formula: Get the comparison similarity S; where, represents the second eigenvector, represents the standard eigenvector, represents the magnitude of the second eigenvector, represents the magnitude of the standard eigenvector; According to the calculation results of the comparison similarity, a threshold is set to analyze and judge whether the silicon wafer is deformed; When , the silicon wafer does not deform; when When the silicon wafer is deformed; When the silicon wafer is deformed, the following formula is used to analyze it: The deformation B is obtained; where represents the i-th eigenparameter of the second eigenvector, represents the i-th feature parameter of the standard feature vector, and M represents the total number of feature parameters after selection; When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed.

[0027] Specifically, when it is identified that there are no defects in the silicon wafer, an image of a standard silicon wafer that is not deformed is collected. The collected standard silicon wafer image is preprocessed, including denoising, contrast enhancement and other steps to ensure image quality. For the preprocessed standard silicon wafer image, the feature extraction step is repeated to extract feature parameters. The extracted feature parameters are cleaned to remove invalid or redundant information. The feature parameters are standardized to ensure that different feature parameters are numerically comparable. Through the feature selection method, the most representative features are selected from the cleaned and standardized feature parameters to generate a standard feature vector. The second feature vector is extracted from the silicon wafer surface image that is collected and preprocessed in real time. A comparative analysis is performed based on the second feature vector and the standard feature vector. The comparison similarity is calculated using the analysis formula. According to the calculation result of the comparison similarity, the threshold is set to , dynamically adjust the threshold according to the actual situation. When the comparison similarity is greater than or equal to the threshold, it is considered that the silicon wafer is not deformed; when the comparison similarity is less than the threshold, it is considered that the silicon wafer is deformed. When the silicon wafer is deformed, the deformation amount is calculated using the analytical formula. When the silicon wafer is not deformed, the dopant distribution in the ROI area is further analyzed.

[0028] In one embodiment of the present invention, the step S4 of analyzing the dopant distribution in the ROI region comprises the following steps: When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed using a deep learning model based on the extracted characteristic parameters of the ROI area; Marking characteristic parameters of the ROI region and characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution, wherein the dopant distribution includes: uniform distribution, local aggregation and sparse distribution; inputting the marked characteristic parameters of the ROI region and the characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution into a deep learning model for training; According to the trained deep learning model, the characteristic parameters of the ROI area in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing, and the extracted characteristic parameters of the ROI area are input into the trained deep learning model for analysis and identification of the dopant distribution in the ROI area, and the dopant distribution results of the ROI area are output.

[0029] Specifically, collect images of the silicon wafer surface to ensure that the images are clear, without blur or distortion, for subsequent processing and analysis. Extract the region of interest ROI from the collected silicon wafer surface image, and the ROI region should cover the part of the silicon wafer that may contain dopants. Annotate the characteristic parameters of the extracted ROI region, depending on the manifestation of the dopant on the silicon wafer. Annotate the ROI region in the silicon wafer surface image containing the dopant distribution based on professional knowledge or actual observation. The dopant distribution includes but is not limited to: uniform distribution, local aggregation and sparse distribution. Combine the annotated ROI region characteristic parameters and dopant distribution into a data set. The data set should contain multiple samples, each sample including the characteristic parameters of an ROI region and the corresponding dopant distribution label. Select a deep learning model based on the complexity of the problem and the size of the data set. Use the annotated data set to train the deep learning model. During the training process, adjust the model parameters through optimization methods such as backpropagation algorithm and gradient descent to minimize the loss function. Monitor indicators such as loss value and accuracy during the training process to evaluate the training effect of the model. Use the validation set data to verify the trained model. Adjust the model structure or hyperparameters based on the verification results to improve the generalization ability of the model. Collect and preprocess the image of the silicon wafer surface in real time, and extract the characteristic parameters of the ROI area from the preprocessed silicon wafer surface image. Input the extracted ROI area characteristic parameters into the trained deep learning model for analysis. In actual applications, collect user feedback data on the results, including but not limited to: the accuracy of the results, false alarm rate, and missed alarm rate indicators. Optimize and improve the deep learning model based on the feedback data. The model structure, hyperparameters, or more training data can be adjusted to improve the performance of the model. Replace the optimized model with the currently used model. Ensure that the new model has better performance and accuracy in actual applications.

[0030] In one embodiment of the present invention, the step S4 of analyzing and calculating the silicon wafer performance change rate according to the dopant distribution result includes the following steps: By analyzing the formula: Get the conductivity of the silicon wafer ,in, represents the carrier concentration, represents the elementary charge, represents the carrier mobility, represents the hole concentration, represents the hole mobility; According to the characteristic parameters of the ROI area of ​​the silicon wafer surface image collected in real time, when it is identified that the dopant distribution in the ROI area is locally concentrated or sparsely distributed, the analysis formula is: Get the silicon wafer performance change rate ,in, The conductivity of the silicon wafer when the dopant distribution in the ROI area is uniformly distributed. The conductivity of the silicon wafer indicates whether the dopant distribution in the ROI area is locally concentrated or sparsely distributed.

[0031] Specifically, the carrier concentration, hole concentration, elementary charge, carrier mobility and hole mobility of the ROI region are obtained through the ROI region of the real-time collected silicon wafer surface image. The collected data is substituted into the above formula to calculate the conductivity of the silicon wafer. First, it is necessary to determine the conductivity of the silicon wafer when the dopant distribution in the ROI region is uniformly distributed. This can be obtained by testing the silicon wafer with uniform dopant distribution under the same conditions. Through the characteristic parameters of the ROI region of the real-time collected silicon wafer surface image, a deep learning model or other image recognition technology is used to identify whether the dopant distribution in the ROI region is locally concentrated or sparsely distributed. After identifying the dopant distribution, the above formula is used again to calculate the conductivity of the current silicon wafer, that is, when the dopant is locally concentrated or sparsely distributed. According to the calculated silicon wafer conductivity and performance change rate, the influence of the dopant distribution on the performance of the silicon wafer is analyzed. If the performance change rate is large, it means that the dopant distribution has a significant impact on the performance of the silicon wafer, and the doping process needs to be further optimized.

[0032] In one embodiment of the present invention, analyzing the metal contamination of the silicon wafer and calculating the metal contamination rate according to the analysis result in step S4 includes the following steps: Acquire a silicon wafer surface image sample containing metal contamination and pre-process the silicon wafer surface image sample; Extracting metal features from the pre-processed silicon wafer surface image sample, including: color features and texture features; the color features include: RGB value, HSV value and gray value; the texture features include: contrast, energy, homogeneity, entropy and correlation; Annotate the metal features in the preprocessed silicon wafer surface image samples, and input the metal features in the silicon wafer surface image samples into the deep learning model for training; According to the trained deep learning model, the metal features in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing of the silicon wafer surface image, and the extracted metal features are input into the trained deep learning model to analyze and identify whether there is metal contamination in the real-time silicon wafer surface image, and the identification result is output; When metal contamination is identified, the analysis formula is: The metal contamination rate G is obtained; where A represents the area within the edge contour region, Represents the area of ​​metal contamination.

[0033] Specifically, obtain silicon wafer surface image samples containing metal contamination through detection equipment in the laboratory or production line to ensure that the samples are representative and can cover different types of metal contamination. Perform denoising on the image to eliminate noise and interference in the image. Perform image enhancement to improve the contrast and clarity of the image and make the metal contamination area more obvious. Perform size normalization on the image to ensure that all images have the same size and resolution. Use image processing software or library to extract the RGB value, HSV value and grayscale value of the image. Use the texture analysis algorithm gray level co-occurrence matrix to extract the image's texture features such as contrast, energy, homogeneity, entropy and correlation. Manually annotate the metal features in the pre-processed silicon wafer surface image samples to ensure the accuracy and completeness of the annotation. Divide the annotated image samples into training sets and test sets for training deep learning models. Input the annotated training set image samples into the deep learning model for training and optimization. Improve the recognition accuracy and generalization ability of the model by adjusting model parameters and hyperparameters such as learning rate. Use real-time acquisition of silicon wafer surface images to pre-process the acquired images to keep them consistent with those during training. The trained deep learning model is used to extract metal features in the real-time silicon wafer surface image. The extracted metal features are compared and analyzed with the features extracted during training. The output results of the deep learning model are used to determine whether there is metal contamination in the real-time silicon wafer surface image. The edge detection algorithm of the image processing algorithm is used to determine the edge contour area of ​​the silicon wafer surface, the identified metal contamination area is segmented and measured, the area of ​​the metal contamination area is calculated, and the metal contamination rate is obtained through the formula.

[0034] In one embodiment of the present invention, the step S5 comprises the following steps: By analyzing the formula: The quality assessment index Q is obtained; among which D represents the defect value, B represents the deformation, R represents the silicon wafer performance change rate, and G represents the metal contamination rate. , , and They represent the weight coefficients of defect value, deformation, silicon wafer performance change rate and metal contamination respectively; Set a threshold H. When the quality evaluation index is greater than or equal to the threshold H, the quality of the silicon wafer is unqualified; when the quality evaluation index is less than the threshold H, the quality of the silicon wafer is qualified; Whether the quality of the silicon wafer after initial cleaning is qualified is analyzed by collecting the surface images and performance data of the silicon wafer after initial cleaning in real time; when the quality of the silicon wafer is unqualified, the silicon wafer is cleaned again, and the time of the Etch etching step is increased by 12 seconds and the time length of the Coat coating step is increased to 60 seconds, and whether the quality of the silicon wafer after the second cleaning is qualified is judged; when the quality of the silicon wafer after the second cleaning is judged to be unqualified, the silicon wafer is cleaned and tested again.

[0035] Specifically, according to the calculation results of defect value, deformation, silicon wafer performance change rate and metal contamination, the quality assessment index is obtained by analyzing the formula. The weight coefficient of the defect value is 0.3, the weight coefficient of the deformation is 0.2, the weight coefficient of the silicon wafer performance change rate is 0.2 and the weight coefficient of the metal contamination is 0.3, and the sum of the weight coefficients is ensured to be 1 to reflect the relative importance of each parameter in the quality assessment. According to the calculation results of the quality assessment index, the threshold is set to 55% to evaluate whether the quality of the silicon wafer is qualified. The weight coefficient and threshold are dynamically set according to the quality requirements of the silicon wafer and the actual application scenario. The quality assessment index after initial cleaning can be calculated by the real-time collected silicon wafer surface image and performance data after initial cleaning. According to the quality assessment index and threshold after initial cleaning, when the quality of the silicon wafer is unqualified, it is necessary to clean it again, including: increasing the time of the Etch etching step by 12 seconds, and increasing the time length of the Coat coating step to 60 seconds, and then calculating the quality assessment index again. When it is judged that the quality of the silicon wafer after the second cleaning is unqualified, the silicon wafer is cleaned and tested again.

[0036] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A silicon wafer quality assessment method based on image features, characterized in that: The following steps are involved: S1: Collecting surface images and performance data of silicon wafers after initial cleaning, the performance data including: silicon wafer thickness, carrier concentration and mobility, hole concentration and mobility, elementary charge and resistivity, and recording the length of initial cleaning time; S2: preprocessing the collected surface images including: filtering and denoising, image enhancement and grayscale conversion; preprocessing the collected performance data including: cleaning and standardization; S3: extracting features and quantifying features according to the preprocessed surface image, extracting feature parameters; preprocessing the extracted feature parameters, including: cleaning and standardization; performing feature selection according to the preprocessed feature parameters, and generating feature vectors according to the selection results; S4: Identify silicon wafer defects based on feature vector analysis and calculate defect values ​​based on analysis results; perform silicon wafer deformation detection and calculate the deformation amount of silicon wafer based on detection results; analyze dopant distribution in ROI area; calculate silicon wafer performance change rate based on dopant distribution results; analyze metal contamination of silicon wafer and calculate metal contamination rate based on analysis results; S5: Evaluate the quality of silicon wafers based on comprehensive analysis of defect value, deformation, silicon wafer performance change rate and metal contamination rate.

2. The method for evaluating silicon wafer quality based on image features according to claim 1, characterized in that: In step S3, feature extraction and quantification of features are performed based on the preprocessed surface image to extract feature parameters, including the following steps: According to the preprocessed surface image, the edge contour of the silicon wafer is extracted by using an edge detection algorithm; according to the edge contour of the silicon wafer, the shape characteristics of the surface image are extracted by morphological processing, and the characteristic parameters of the shape characteristics include: the area within the edge contour area and the total length of the edge contour; through the feature detection algorithm, the ROI area is detected from the preprocessed surface image, and the characteristic parameters of the ROI area extracted according to the detection result include: the coordinates of the four corners of the ROI area, the coordinates of the center point, the width, the height and the area; the texture features are extracted according to the grayscale co-occurrence matrix of the surface image, and the characteristic parameters of the texture features include: contrast, energy, homogeneity, entropy and correlation.

3. The method for evaluating silicon wafer quality based on image features according to claim 2, characterized in that: In step S3, feature selection is performed according to the preprocessed feature parameters, and a feature vector is generated according to the selection result, including the following steps: A first feature vector is generated according to the extracted feature parameters. The first feature vector is expressed as ； By analyzing the formula: Get the selected feature parameters ; Where Y represents the selection threshold coefficient, represents the i-th characteristic parameter value in the first characteristic vector, and N represents the total number of characteristic parameters in the first characteristic vector; According to the selected characteristic parameters, a second characteristic vector is generated, and the second characteristic vector is expressed as , where M represents the total number of feature parameters after selection.

4. The method for evaluating silicon wafer quality based on image features according to claim 1, characterized in that: The step S4 includes the following steps: identifying silicon wafer defects according to the feature vector analysis and calculating defect values ​​according to the analysis results: Based on the selected second eigenvector, a convolutional neural network model is used to analyze whether the silicon wafer has defects and identify the defect type; Annotating the second eigenvector and the silicon wafer defect type, wherein the silicon wafer defect type includes: cracks, scratches, dents, edge collapse and bubbles; inputting the annotated second eigenvector and the silicon wafer defect type into a convolutional neural network model for training; According to the trained convolutional neural network model, by collecting and preprocessing the silicon wafer surface image in real time, repeating step S3, inputting the feature vector extracted from the real-time collected silicon wafer surface image into the trained convolutional neural network model to identify and determine whether the silicon wafer has defects, and outputting the defect type; When a defect is identified in a silicon wafer, characteristic parameters of the defect are extracted, including: defect area and defect depth; By analyzing the formula: Get defect value ;in, represents the identified defect area, A represents the area within the edge contour area, represents the identified defect depth, and W represents the thickness of the silicon wafer; When it is identified that the silicon wafer has no defects, a deformation detection of the silicon wafer is performed.

5. The method for evaluating silicon wafer quality based on image features according to claim 4, characterized in that: The step S4 includes performing a deformation detection on the silicon wafer and analyzing and calculating the deformation amount of the silicon wafer according to the detection result, and comprises the following steps: When it is identified that the silicon wafer does not have defects, the undeformed standard silicon wafer image is collected and preprocessed, the characteristic parameters of the undeformed standard silicon wafer image are extracted and preprocessed for cleaning and standardization, and the standard feature vector is generated through feature selection. ; According to the second eigenvector and the standard eigenvector, a comparative analysis is performed, through the analysis formula: Get the comparison similarity S; where, represents the second eigenvector, represents the standard eigenvector, represents the magnitude of the second eigenvector, represents the magnitude of the standard eigenvector; According to the calculation results of the comparison similarity, a threshold is set to analyze and judge whether the silicon wafer is deformed; When , the silicon wafer does not deform; when When the silicon wafer is deformed; When the silicon wafer is deformed, the following formula is used to analyze it: The deformation B is obtained; where represents the i-th eigenparameter of the second eigenvector, represents the i-th feature parameter of the standard feature vector, and M represents the total number of feature parameters after selection; When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed.

6. The method for evaluating silicon wafer quality based on image features according to claim 5, characterized in that: Analyzing the dopant distribution in the ROI region in step S4 includes the following steps: When the silicon wafer does not deform, the dopant distribution in the ROI area is analyzed using a deep learning model based on the extracted characteristic parameters of the ROI area; Marking characteristic parameters of the ROI region and characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution, wherein the dopant distribution includes: uniform distribution, local aggregation and sparse distribution; inputting the marked characteristic parameters of the ROI region and the characteristic parameters of the ROI region in the silicon wafer surface image containing dopant distribution into a deep learning model for training; According to the trained deep learning model, the characteristic parameters of the ROI area in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing, and the extracted characteristic parameters of the ROI area are input into the trained deep learning model for analysis and identification of the dopant distribution in the ROI area, and the dopant distribution results of the ROI area are output.

7. The method for evaluating silicon wafer quality based on image features according to claim 6, characterized in that: The step S4 includes analyzing and calculating the silicon wafer performance change rate according to the dopant distribution result, and includes the following steps: By analyzing the formula: Get the conductivity of the silicon wafer ,in, represents the carrier concentration, represents the elementary charge, represents the carrier mobility, represents the hole concentration, represents the hole mobility; According to the characteristic parameters of the ROI area of ​​the silicon wafer surface image collected in real time, when it is identified that the dopant distribution in the ROI area is locally concentrated or sparsely distributed, the analysis formula is: Get the silicon wafer performance change rate ,in, The conductivity of the silicon wafer when the dopant distribution in the ROI area is uniformly distributed. The conductivity of the silicon wafer indicates whether the dopant distribution in the ROI area is locally concentrated or sparsely distributed.

8. The method for evaluating silicon wafer quality based on image features according to claim 1, characterized in that: The step S4 includes analyzing the metal contamination of the silicon wafer and calculating the metal contamination rate according to the analysis result, including the following steps: Acquire a silicon wafer surface image sample containing metal contamination and pre-process the silicon wafer surface image sample; Extracting metal features from the pre-processed silicon wafer surface image sample, including: color features and texture features; the color features include: RGB value, HSV value and gray value; the texture features include: contrast, energy, homogeneity, entropy and correlation; Annotate the metal features in the preprocessed silicon wafer surface image samples, and input the metal features in the silicon wafer surface image samples into the deep learning model for training; According to the trained deep learning model, the metal features in the real-time silicon wafer surface image are extracted by real-time acquisition and preprocessing of the silicon wafer surface image, and the extracted metal features are input into the trained deep learning model to analyze and identify whether there is metal contamination in the real-time silicon wafer surface image, and the identification result is output; When metal contamination is identified, the analysis formula is: The metal contamination rate G is obtained; where A represents the area within the edge contour region, Represents the area of ​​metal contamination.

9. The method for evaluating silicon wafer quality based on image features according to claim 1, characterized in that: The step S5 comprises the following steps: By analyzing the formula: The quality assessment index Q is obtained; among which D represents the defect value, B represents the deformation, R represents the silicon wafer performance change rate, and G represents the metal contamination rate. , , and They represent the weight coefficients of defect value, deformation, silicon wafer performance change rate and metal contamination respectively; Set a threshold H. When the quality evaluation index is greater than or equal to the threshold H, the quality of the silicon wafer is unqualified; when the quality evaluation index is less than the threshold H, the quality of the silicon wafer is qualified; Whether the quality of the silicon wafer after initial cleaning is qualified is analyzed by collecting the surface images and performance data of the silicon wafer after initial cleaning in real time; when the quality of the silicon wafer is unqualified, the silicon wafer is cleaned again, and the time of the Etch etching step is increased by 12 seconds and the time length of the Coat coating step is increased to 60 seconds, and whether the quality of the silicon wafer after the second cleaning is qualified is judged; when the quality of the silicon wafer after the second cleaning is judged to be unqualified, the silicon wafer is cleaned and tested again.

Images