Defect classification method and device, equipment and storage medium

Abstract

The present application relates to a defect classification method and device, equipment and storage medium, wherein the method includes: reading a defect image, a reference image and a defect position to be currently classified; obtaining a difference image according to the defect image and the reference image, and extracting specific features from the difference image; wherein the specific features correspond to the defect type to be currently classified; inputting the extracted specific features into a trained classifier, and the classifier classifies the defect image according to the specific features. It achieves accurate distinction of defect types by extracting specific features corresponding to the defect type from the difference image during defect classification, and performing corresponding defect classification based on the extracted specific features, thereby enabling accurate defect cause analysis, thereby improving yield and increasing production.

Description

Technical Field

[0001] The present application relates to the field of semiconductor device detection technology, and in particular to a defect classification method and apparatus, equipment and storage medium. Background Art

[0002] In the semiconductor manufacturing process, defect detection and classification are used to perform defect tracing analysis to adjust the process and achieve the purpose of yield. In related technologies, defect detection on wafers is usually divided into two major steps: defect detection and ADC (Automatic Defect Classification). Among them, in the defect detection process, the detection image is compared with the reference image to detect and locate the defects, and the defect image, reference image and defect position are obtained. In the ADC process, after manually extracting features, defect classification is performed through classifiers such as SVM (Support Vector Machine) and KNN (K Nearest Neighbor). However, based on traditional feature extraction, when using SVM classifiers for classification, BVC (Bright Voltage Contrast) and DVC (Dark Voltage Contrast) are often misclassified, which affects the accuracy of the final classification result. Summary of the invention

[0003] In view of this, the present application proposes a defect classification method, which can effectively improve the accuracy of classification results.

[0004] According to one aspect of the present application, a defect classification method is provided, comprising:

[0005] Read the defect image, reference image and defect position to be classified;

[0006] Obtaining a difference image according to the defect image and the reference image, and extracting specific features from the difference image; wherein the specific features correspond to the defect type to be currently classified;

[0007] The extracted specific features are input into a trained classifier, and the classifier classifies the defect image according to the specific features.

[0008] In a possible implementation manner, when obtaining the difference image according to the defect image and the reference image, the difference image is obtained by performing a difference operation on the defect image and the reference image.

[0009] In a possible implementation manner, extracting a specific feature from the difference image includes:

[0010] Performing noise evaluation on the difference image to obtain a range of noise grayscale values;

[0011] The specific feature is extracted from the difference image based on the range of the noise grayscale value and the defect area in the difference image.

[0012] In a possible implementation manner, extracting the specific feature from the difference image based on the range of the noise grayscale value and the defect area in the difference image includes:

[0013] Based on the range of the noise grayscale value, counting pixel information of the defective area in the difference image, and determining corresponding specific features according to the pixel information of the defective area obtained by counting;

[0014] The pixel information includes at least one of the number of bright pixels, the number of dark pixels and the total number of pixels in the defective area;

[0015] The specific features include at least one of the ratio of the bright pixels to the total pixels in the defect area, the ratio of the dark pixels to the total pixels in the defect area, and the ratio between the bright pixels and the dark pixels in the defect area.

[0016] In a possible implementation manner, determining corresponding specific features based on statistically obtained pixel information of the defective area includes:

[0017] According to the formula: Calculating and obtaining the specific feature;

[0018] Among them, feature_val1 is the ratio of the bright pixels to the total pixels in the defect area, num_bright is the number of the bright pixels; sum_pixel is the total number of pixels in the defect area.

[0019] In a possible implementation manner, determining corresponding specific features based on statistically obtained pixel information of the defective area includes:

[0020] According to the formula: Calculating and obtaining the specific feature;

[0021] Among them, feature_val2 is the ratio of the dark pixels in the defect area to the total pixels, num_dark is the number of dark pixels; sum_pixel is the total number of pixels in the defect area.

[0022] In a possible implementation manner, determining corresponding specific features based on statistically obtained pixel information of the defective area includes:

[0023] According to the formula:

[0024]

[0025] Calculating and obtaining the specific feature;

[0026] Among them, feature_val3 is the ratio between the bright pixels and the dark pixels in the defect area, num_dark is the number of pixels lower than the second value; num_bright is the number of pixels higher than the first value; sum_pixel is the total number of pixels in the defect area.

[0027] In a possible implementation manner, the number of the bright pixels and the number of the dark pixels are determined based on a median value and a noise grayscale value of the difference image.

[0028] In a possible implementation, the number of bright pixels is determined according to the formula: medina+noise_band / 2;

[0029] The number of dark pixels is determined according to the formula: medina-noise_band / 2;

[0030] Among them, medina is the median of the difference image, and noise_band is the range of the noise grayscale value.

[0031] According to another aspect of the present application, there is also provided a defect classification device, comprising: a data reading module, a feature extraction module and a defect classification module;

[0032] The data reading module is configured to read the defect image to be currently classified, the reference image and the defect position;

[0033] The feature extraction module is configured to obtain a difference image according to the defect image and the reference image, and extract a specific feature from the difference image; wherein the specific feature corresponds to a defect type to be currently classified;

[0034] The defect classification module is configured to input the extracted specific features into a trained classifier, and the classifier classifies the defect image according to the specific features.

[0035] According to another aspect of the present application, a defect classification device is also provided, including:

[0036] processor;

[0037] a memory for storing processor-executable instructions;

[0038] Wherein, the processor is configured to implement any of the above-mentioned methods when executing the executable instructions.

[0039] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, any of the above-mentioned methods is implemented.

[0040] After reading the defect image, reference image and defect position to be currently classified, the corresponding difference image is obtained according to the determined defect image and reference image, and the specific features corresponding to the defect type currently being classified are extracted from the difference image, and then the defect image is classified based on the extracted specific features, thereby realizing defect classification in the semiconductor manufacturing process. During defect classification, the specific features corresponding to the defect type are extracted from the difference image, and the corresponding defect classification is performed based on the extracted specific features, thereby realizing accurate distinction of defect types, thereby being able to accurately analyze the causes of defects, thereby improving yield and increasing production.

[0041] Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the present application and, together with the description, serve to explain the principles of the present application.

[0043] Figure 1 A flow chart showing a defect classification method according to an embodiment of the present application;

[0044] Figure 2 A framework diagram showing the operation steps of a defect classification method according to an embodiment of the present application;

[0045] Figure 3 The difference image of DVC obtained in the defect classification method according to one embodiment of the present application is shown;

[0046] Figure 4 The difference image of BVC obtained in the defect classification method according to one embodiment of the present application is shown;

[0047] Figure 5 A structural block diagram of a defect classification device according to an embodiment of the present application is shown;

[0048] Figure 6 A structural block diagram of a defect classification device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0049] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0050] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0051] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0052] Figure 1 FIG. 1 is a flow chart showing a defect classification method according to an embodiment of the present application. Figure 1 As shown, the method includes: step S100, reading the defect image, reference image and defect position. It should be pointed out here that the defect image, reference image and defect position can be obtained by performing defect detection on the manufactured semiconductor during the semiconductor manufacturing process. Among them, those skilled in the art can understand that the defect detection method used when performing defect detection on a semiconductor refers to comparing the detection image with the reference image during the defect detection process to detect and locate the defects, thereby determining the defect image, reference image and defect position of the image currently being detected. The detection image can be an image of a semiconductor device captured by any imaging device, and the image can be a high-resolution image or a low-resolution image based on the SEM output. The reference image is a defect-free image corresponding to the semiconductor device. It should also be noted here that the detection image and the reference image can be acquired by the same imaging device.

[0053] After reading the defect image, reference image and defect position to be currently classified, step S200 can be executed to obtain a difference image based on the defect image and the reference image, and specific features can be extracted from the difference image. It should be pointed out here that in the method of the embodiment of the present application, the specific features correspond to the defect type to be currently classified. That is to say, in the present application, the specific features extracted are different for different defect types currently being classified. By extracting different specific features for different types of defect classification, the matching degree of the feature data based on which the defect classification is performed is higher.

[0054] Finally, through step S300, the extracted specific features are input into the trained classifier, and the classifier classifies the defect image according to the specific features.

[0055] Therefore, the defect classification method of the embodiment of the present application obtains the corresponding difference image according to the determined defect image and reference image after reading the defect image, reference image and defect position to be defect classified, and extracts the specific features corresponding to the defect type currently being classified from the difference image, and then classifies the defect image based on the extracted specific features, thereby realizing defect classification in the semiconductor manufacturing process. It realizes accurate distinction of defect types by extracting specific features corresponding to the defect type from the difference image during defect classification, and performing corresponding defect classification based on the extracted specific features, so that the cause of the defect can be accurately analyzed, thereby improving the yield and increasing production.

[0056] Here, it should also be noted that the specific features extracted are different for different defect types to be classified. For example, when the defect types to be classified are BVC and DVC, the specific features extracted at this time can be at least one of the ratio of bright pixels in the defect area to the total pixels in the defect area, the ratio of dark pixels in the defect area to the total pixels in the defect area, and the ratio between bright pixels and dark pixels in the defect area. For other defect types, other features can be selected as specific features in a targeted manner. Examples are not given here one by one.

[0057] That is, see Figure 2 , after reading out each defect data (including defect image, defect location, reference image and other data information) from the defect classification data set through step S100, then through step S210, based on the read defect data, the defect image and the reference image are subtracted to obtain a corresponding difference image, and the corresponding specific features are extracted from the difference image. At this time, for the defect data in the defect classification data set, a variety of different defect types are usually included when performing defect classification, so there will be multiple specific features extracted, not just corresponding to one defect type. Therefore, at this time, it is also necessary to select corresponding specific features from the extracted features for different defect types through step S220, and then execute step S300 to input the selected specific features into the corresponding trained classifier, and the classifier performs corresponding defect classification.

[0058] It should be noted that, in step S230, when selecting corresponding specific features for different defect types, it can be implemented based on the defect type and specific feature mapping data that has been statistically sorted in advance, or it can be implemented by other methods, which are not specifically limited here. At the same time, it should also be noted that when determining the corresponding specific features for different defect types, it can be determined based on actual experience, or it can be determined by other methods, which are not specifically limited here.

[0059] In addition, in the defect classification method of the embodiment of the present application, after extracting the specific features, the extracted specific features are input into the trained classifier, and when the classifier classifies the defect image according to the specific features, the classifier used can be a linear SVM classifier. Automatic defect classification can be achieved using a linear SVM classifier. However, it should be noted that when using a linear SVM classifier for defect classification, the linear SVM classifier needs to be trained first.

[0060] See also Figure 2 , when training the linear SVM classifier, it can be mainly achieved in the following ways. Specifically, first, mark the existing defect samples and extract features. Here, those skilled in the art can understand that the way to mark the existing defect samples can adopt conventional technical means in this field, which is not specifically limited here. Then, each feature value of the defect sample is counted, and the features corresponding to the defect type to be currently classified are selected from the extracted features based on the statistical information. Then, the features selected from the marked samples are input into the linear SVM classifier for training to obtain a classification model.

[0061] After the linear SVM classifier is trained in the above manner, the corresponding features of the defect image to be classified can be extracted, and after selecting specific features from the extracted features, the selected specific features are input into the trained linear classifier model for classification.

[0062] In a possible implementation, when obtaining the difference image based on the defect image and the reference image, the difference image can be obtained by performing a difference operation on the defect image and the reference image. Figure 3 and Figure 4 , which are the difference image of DVC defect image and the difference image of BVC defect image respectively.

[0063] After performing a difference operation on the defect image and the reference image to obtain a difference image, specific features can be extracted from the difference image. According to the above description, different specific features are extracted, and the corresponding feature extraction methods are also different.

[0064] In one possible implementation, when the defect types are BVC and DVC, specific features can be extracted from the difference image based on the noise grayscale value of the difference image and the defect area in the difference image (i.e., the area of ​​interest in the defect image of the semiconductor device collected and detected). That is, firstly, the difference image is subjected to noise evaluation to obtain the range of noise grayscale values ​​noise_band. Then, based on the range of noise grayscale values ​​and the defect area in the difference image, specific features are extracted from the difference image. Here, it should be pointed out that the noise evaluation of the difference image can be implemented by conventional technical means in the field, which is not specifically limited here. At the same time, it should also be noted that the range of noise grayscale values ​​obtained by performing noise estimation on the difference image is specifically a numerical value.

[0065] Furthermore, when extracting specific features from the difference image based on the range of noise grayscale values ​​and the defective area of ​​the difference image, the pixel information of the defective area in the difference image can be first counted based on the range of noise grayscale values ​​noise_band, and then the corresponding specific features can be determined based on the pixel information of the defective area obtained by statistics.

[0066] It should be noted here that, according to the foregoing, when the defect types are BVC and DVC, the specific feature can be at least one of the ratio of bright pixels to total pixels in the defect area, the ratio of dark pixels to total pixels in the defect area, and the ratio between bright pixels and dark pixels in the defect area.

[0067] Correspondingly, the pixel information may include at least one of the number of bright pixels, the number of dark pixels, and the total number of pixels in the defective area. Bright pixels refer to pixels in the image whose pixel values ​​are higher than a certain value, and dark pixels refer to pixels whose pixel values ​​are lower than a certain value. It should be noted here that the statistical determination of bright pixels and dark pixels can use the same set pixel value as a reference, or different pixel values ​​as a reference, which is not specifically limited here.

[0068] For example, in the method of the embodiment of the present application, the number of bright pixels can be determined according to the formula: medina+noise_band / 2; the number of dark pixels can be determined according to the formula: medina-noise_band / 2, where medina is the median of the difference image and noise_band is the range of the noise grayscale value.

[0069] That is, by calculating the median of the defective area in the difference image and counting the number of pixels in the defective area that are higher than medina+noise_band / 2, the number of bright pixels num_bright can be obtained. By counting the number of pixels in the defective area that are lower than medina-noise_band / 2, the number of dark pixels num_dark can be obtained. At the same time, the total number of pixels in the defective area sum_pixel can also be obtained by conventional pixel calculation methods.

[0070] After the pixel information of the defective area in the difference image is statistically obtained by any of the above methods, the corresponding specific features can be obtained based on the statistically obtained pixel information of the defective area.

[0071] Specifically, when the specific feature is at least one of the ratio of bright pixels in the defect area to the total pixels, the ratio of dark pixels in the defect area to the total pixels, and the ratio between bright pixels and dark pixels in the defect area, the acquisition method can be implemented according to the following formula:

[0072]

[0073]

[0074]

[0075] Among them, feature_val1, feature_val2 and feature_val3 are three different specific features determined, corresponding to the ratio of bright pixels to total pixels in the defect area, the ratio of dark pixels to total pixels in the defect area, and the ratio between bright pixels and dark pixels in the defect area, respectively.

[0076] After the specific features for defect type classification are determined in the above manner, the specific features are input into a trained classifier, and the classifier determines the defect type of the defect image currently being classified based on the specific features.

[0077] Therefore, the defect classification method of the embodiment of the present application targets two common types of defects in the semiconductor manufacturing process (i.e., BVC and DVC). By extracting specific features specifically for these two defect types and performing corresponding defect classification based on the extracted specific features, it not only realizes the automatic classification of defects in the semiconductor manufacturing process, but also effectively improves the results of defect classification. While ensuring a high classification accuracy, it can also perform correct defect cause analysis, thereby improving yield and increasing production.

[0078] Among them, it should be pointed out that the above method can be applied to manufacturing tools, such as manufacturing tools for semiconductor factories, packaging factories, printed circuit manufacturing factories, solar factories, panel factories, mask factories and light-emitting diode factories. Moreover, these manufacturing tools can be executed in devices with computing capabilities. Compared with traditional methods, the above method can increase manufacturing efficiency and output several times, and reduce the pass rate learning cycle and manufacturing costs. At the same time, it should also be pointed out that the method of the embodiment of the present application can be applied to defect detection and classification of various semiconductor devices. The semiconductor device can be a processed and prepared wafer, or a prepared integrated circuit, etc., and the semiconductor device is not specifically limited here.

[0079] Accordingly, based on any of the above-mentioned defect classification methods, the present application also provides a defect classification device. Since the working principle of the defect classification device provided by the present application is the same or similar to the principle of the defect classification method provided by the present application, the repeated parts will not be repeated.

[0080] See also Figure 5 The defect classification device 100 provided in the present application includes a data reading module 110, a feature extraction module 120 and a defect classification module 130. The data reading module 110 is configured to read the defect image, the reference image and the defect position to be currently classified. The feature extraction module 120 is configured to obtain a difference image based on the defect image and the reference image, and extract specific features from the difference image; wherein the specific features correspond to the defect type to be currently classified. The defect classification module 130 is configured to input the extracted specific features into a trained classifier, and the classifier classifies the defect image according to the specific features.

[0081] Furthermore, according to another aspect of the present application, a defect classification device 200 is also provided. Figure 6 The defect classification device 200 of the embodiment of the present application includes a processor 210 and a memory 220 for storing executable instructions of the processor 210. The processor 210 is configured to implement any of the above-mentioned defect classification methods when executing the executable instructions.

[0082] Here, it should be noted that the number of processors 210 may be one or more. Meanwhile, in the defect classification device 200 of the embodiment of the present application, an input device 230 and an output device 240 may also be included. The processor 210, the memory 220, the input device 230 and the output device 240 may be connected via a bus or in other ways, which are not specifically limited here.

[0083] The memory 220 is a computer-readable storage medium that can be used to store software programs, computer executable programs, and various modules, such as the program or module corresponding to the defect classification method of the embodiment of the present application. The processor 210 executes various functional applications and data processing of the defect classification device 200 by running the software program or module stored in the memory 220.

[0084] The input device 230 may be used to receive input numbers or signals. The signals may be key signals related to user settings and function control of the device / terminal / server. The output device 240 may include display devices such as display screens.

[0085] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by the processor 210, any of the above-mentioned defect classification methods is implemented.

[0086] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A defect classification method, characterized in that: include: Read the defect image, reference image and defect position to be classified; Obtaining a difference image according to the defect image and the reference image, and extracting specific features from the difference image; wherein the specific features correspond to the defect type to be currently classified; Inputting the extracted specific features into a trained classifier, and allowing the classifier to classify the defect image according to the specific features; Specific features are extracted from the difference image, including: Performing noise evaluation on the difference image to obtain a range of noise grayscale values; Based on the range of the noise grayscale value, counting pixel information of the defective area in the difference image, and determining corresponding specific features according to the pixel information of the defective area obtained by counting; The pixel information includes at least one of the number of bright pixels, the number of dark pixels and the total number of pixels in the defective area; The specific features include at least one of the ratio of the bright pixels to the total pixels in the defect area, the ratio of the dark pixels to the total pixels in the defect area, and the ratio between the bright pixels and the dark pixels in the defect area.

2. The method according to claim 1, characterized in that When obtaining the difference image according to the defect image and the reference image, the difference image is obtained by performing a difference operation on the defect image and the reference image.

3. The method according to claim 1, characterized in that Based on the statistically obtained pixel information of the defective area, the corresponding specific features are determined, including: According to the formula: Calculating and obtaining the specific feature; Among them, feature_val1 is the ratio of the bright pixels to the total pixels in the defect area, num_bright is the number of the bright pixels; sum_pixel is the total number of pixels in the defect area.

4. The method according to claim 1, characterized in that: Based on the statistically obtained pixel information of the defective area, the corresponding specific features are determined, including: According to the formula: Calculating and obtaining the specific feature; Among them, feature_val2 is the ratio of the dark pixels in the defect area to the total pixels, num_dark is the number of dark pixels; sum_pixel is the total number of pixels in the defect area.

5. The method according to claim 1, characterized in that: Based on the statistically obtained pixel information of the defective area, the corresponding specific features are determined, including: According to the formula: Calculating and obtaining the specific feature; Among them, feature_val3 is the ratio between the bright pixels and the dark pixels in the defect area, num_ dark is the number of pixels lower than the second value; num_bright is the number of pixels higher than the first value; sum_pixel is the total number of pixels in the defect area.

6. The method according to claim 1, characterized in that The pixel number of the bright pixels and the pixel number of the dark pixels are determined based on a median value and a noise grayscale value of the difference image.

7. The method according to claim 6, characterized in that The number of bright pixels is determined according to the formula: medina+noise_band / 2; The number of dark pixels is determined according to the formula: medina-noise_band / 2; Among them, medina is the median of the difference image, and noise_band is the range of the noise grayscale value.

8. A defect classification device, characterized in that: include: Data reading module, feature extraction module and defect classification module; The data reading module is configured to read the defect image to be currently classified, the reference image and the defect position; The feature extraction module is configured to obtain a difference image according to the defect image and the reference image, and extract specific features from the difference image, including: performing noise evaluation on the difference image to obtain a range of noise grayscale values; based on the range of noise grayscale values, counting pixel information of the defect area in the difference image, and determining corresponding specific features according to the pixel information of the defect area obtained by counting; wherein the pixel information includes at least one of the number of bright pixels, the number of dark pixels and the total number of pixels in the defect area; the specific features include at least one of the ratio of the bright pixels in the defect area to the total pixels, the ratio of the dark pixels in the defect area to the total pixels, and the ratio between the bright pixels and the dark pixels in the defect area; wherein the specific features correspond to the defect type to be currently classified; The defect classification module is configured to input the extracted specific features into a trained classifier, and the classifier classifies the defect image according to the specific features.

9. A defect classification device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the method described in any one of claims 1 to 7 when executing the executable instructions.

10. A non-volatile computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images