Wafer image detection method, device, electronic device and storage medium

Abstract

The present invention provides a method, device, electronic device and storage medium for detecting wafer images. The above detection method includes: determining the character area corresponding to each numbering character of the wafer number of the wafer to be tested in the image to be tested; for each numbering character, inputting the area image corresponding to the character area corresponding to the numbering character in the image to be tested into the detection model corresponding to the numbering character to obtain the first detection result of the numbering character; based on the first detection results of all numbering characters, determining whether the image to be tested of the wafer to be tested has an abnormality. On the one hand, the above scheme can identify each numbering character through the corresponding detection model, which is conducive to improving the detection accuracy of the numbering characters; on the other hand, the above scheme can also determine whether the image to be tested of the wafer has an abnormality based on the detection results related to the wafer number, which is conducive to assisting in improving the processing and detection process of the wafer.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically to a wafer image detection method, a wafer image detection device, an electronic device, a storage medium and a computer program product. Background Art

[0002] In the industrial field, with the continuous development of production technology, people have put forward higher requirements for the production quality and production efficiency of industrial products. With the development of image processing, more and more product inspections are based on images. If the inspection accuracy is low, the practicality of the image-based inspection method will also be reduced.

[0003] During the wafer manufacturing process, the wafer number is usually engraved on the wafer. Through the wafer number, the wafer can be tracked during the processing and testing process. Therefore, the accuracy of image recognition of the wafer number directly affects whether the wafer can be tracked and located. If the accuracy is low, the efficiency of the wafer processing and testing process will be reduced, greatly reducing the number of wafers shipped. Therefore, how to accurately detect and identify the wafer number is a technical problem that technicians in this field urgently need to solve. Summary of the invention

[0004] The present invention is proposed in view of the above problems. The present invention provides a wafer image detection method, a wafer image detection device, an electronic device, a storage medium and a computer program product.

[0005] According to one aspect of the present invention, a method for detecting a wafer image is provided, the method comprising:

[0006] Determine the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested;

[0007] For each numbered character, an area image corresponding to a character area corresponding to the numbered character in the image to be tested is input into a detection model corresponding to the numbered character to obtain a first detection result of the numbered character, wherein the first detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested;

[0008] Based on the first detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

[0009] Exemplarily, the detection model is trained through the following steps:

[0010] For each training image of the training wafer,

[0011] Determine the character area corresponding to each training character of the reference number of the training wafer in the training image;

[0012] Performing feature extraction operations on the regional images corresponding to the character regions corresponding to each training character in the training image, respectively, to obtain training image features corresponding to each training character;

[0013] For each training character, the training image features corresponding to the training character are used as a first feature set of the training character, and outlier features in the first feature set are filtered out to obtain a second feature set of the training character;

[0014] For each training character, the detection model corresponding to the training character is trained using the second feature set of the training character.

[0015] Exemplarily, for each training image of the training wafer, determining the character area corresponding to each training character of the reference number of the training wafer in the training image includes:

[0016] For each training image of the training wafer, the training image is input into an optical character recognition model to obtain a character region corresponding to each training character of the reference number of the training wafer in the training image;

[0017] For each training image of the training wafer, feature extraction operations are performed on the regional images corresponding to the character regions corresponding to each training character in the training image, including:

[0018] For the character area corresponding to each training character in the training image, the confidence of the character area determined by the optical character recognition model is compared with the preset threshold, and the feature extraction operation is performed only on the area image corresponding to the target character area in all character areas in the training image, wherein the confidence of the target character area is higher than the preset threshold.

[0019] Exemplarily, the above detection method further includes:

[0020] Determine the wafer number of the wafer to be tested based on the image to be tested;

[0021] Determine whether there is an abnormality in the wafer number to obtain a determination result;

[0022] When the judgment result indicates that the wafer number is abnormal, it is determined that the image to be tested is abnormal;

[0023] Wherein, determining the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested is performed when the judgment result indicates that there is no abnormality in the wafer number.

[0024] Exemplarily, determining whether there is an abnormality in the wafer number includes at least one of the following:

[0025] The wafer number is compared with the preset number of the wafer to be tested. If they are inconsistent, it is determined that there is an abnormality in the wafer number;

[0026] The total number of characters in the wafer number is compared with the preset total number. If they are inconsistent, it is determined that there is an abnormality in the wafer number.

[0027] Exemplarily, the above detection method further includes:

[0028] For each numbered character, based on a region image corresponding to a character region corresponding to the numbered character in the image to be tested, determining a tilt angle of the numbered character;

[0029] Based on the first detection results of all the number characters, determining whether the image to be tested of the wafer to be tested has an abnormality includes:

[0030] For each numbered character, based on the first detection result of the numbered character and the tilt angle of the numbered character, determine a second detection result of the numbered character, wherein the second detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested;

[0031] Based on the second detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

[0032] Exemplarily, the above detection method further includes:

[0033] For each numbered character, determine the area size of the character area corresponding to the numbered character;

[0034] Based on the first detection results of all the number characters, determining whether the image to be tested of the wafer to be tested has an abnormality includes:

[0035] For each numbered character, based on the first detection result of the numbered character and the area size of the character area corresponding to the numbered character, determine a third detection result of the numbered character, wherein the third detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested;

[0036] Based on the third detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

[0037] According to yet another aspect of the present invention, a wafer image detection device is provided. The detection device includes: a character area determination module, a first detection result determination module and a to-be-detected image abnormality determination module.

[0038] The character region determination module is used to determine the character region corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested;

[0039] The first detection result determination module is used for inputting, for each numbered character, a region image corresponding to a character region corresponding to the numbered character in the image to be tested into a detection model corresponding to the numbered character to obtain a first detection result of the numbered character, wherein the first detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested;

[0040] The test image abnormality determination module is used to determine whether the test image of the test wafer has an abnormality based on the first detection results of all the numbered characters.

[0041] According to another aspect of the present invention, an electronic device is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and the computer program instructions are used to execute the above-mentioned wafer image detection method when the processor is running.

[0042] According to another aspect of the present invention, a storage medium is provided, on which program instructions are stored, and the program instructions are used to execute the above-mentioned wafer image detection method when running.

[0043] According to another aspect of the present invention, a computer program product is further provided, wherein the computer program product comprises computer program instructions, and the computer program instructions are used to execute the above-mentioned wafer image detection method when running.

[0044] According to the above scheme of the present invention, the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested can be determined. Then, for each number character, the area image corresponding to the character area corresponding to the number character in the image to be tested is input into the detection model corresponding to the number character to obtain the first detection result of the number character. Finally, based on the first detection results of all number characters, it is determined whether the image to be tested of the wafer to be tested is abnormal. On the one hand, the above scheme can be identified by the corresponding detection model for each number character, and targeted detection of different number characters can be realized, which is conducive to improving the detection accuracy of number characters; on the other hand, the above scheme can also determine whether the image to be tested of the wafer is abnormal based on the detection results related to the wafer number, which is conducive to assisting in improving the processing and detection process of the wafer. For example, if the wafer is scratched during the processing stage, the phantom of the wafer number is generated, resulting in an abnormality in the image to be tested. Then the user can adjust the relevant processing instruments in time to avoid damage to the wafer that has not undergone the processing stage. For another example, if the image to be tested is abnormal due to the skewness of the image acquisition device or the skewness of the platform carrying the wafer. The user can then adjust the posture of the image acquisition device or platform in a timely manner to avoid low accuracy of subsequent defect detection on the wafer due to the skewness of the image acquisition device or platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0046] Figure 1 A schematic flow chart of a wafer image detection method according to an embodiment of the present invention is shown;

[0047] Figure 2 A schematic diagram of an image to be tested according to an embodiment of the present invention is shown;

[0048] Figure 3 A schematic diagram showing a method for detecting a wafer image according to an embodiment of the present invention is shown;

[0049] Figure 4 A schematic block diagram showing a wafer image detection device according to an embodiment of the present invention; and

[0050] Figure 5 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical scheme and advantages of the present invention more obvious, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the protection scope of the present invention.

[0052] In the wafer manufacturing process, optical character recognition technology (OCR) is usually used to identify the wafer number in the wafer image to track and locate the wafer during the processing and inspection process. However, the recognition accuracy in existing technologies is usually low, which leads to reduced efficiency in the wafer processing and inspection process, greatly reducing the wafer shipment.

[0053] In order to at least partially solve the above problems, an embodiment of the present invention provides a method for detecting a wafer image. Figure 1 FIG. 1 is a schematic flow chart of a method for detecting a wafer image according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps S110 to S130.

[0054] In step S110 , character regions corresponding to each character of the wafer number of the wafer to be tested in the image to be tested are determined.

[0055] The image to be tested of the wafer to be tested may be a static image or any video frame in a dynamic video. The image to be tested of the wafer to be tested may be an original image captured by an image acquisition device (for example, an original image captured by an image sensor in a camera), or an image obtained after preprocessing the original image (such as digitization, normalization, smoothing, etc.). It should be understood that the image to be tested may also be an image including a portion of the wafer to be tested.

[0056] The wafer to be tested can be placed on the carrying platform manually or by a robotic arm. Then, a wafer number is generated on the wafer to be tested by laser marking, electroplating, etc. The wafer number can be related to the processing batch of the wafer to be tested and is set by the developer according to preset rules. For example, the wafer numbers of the same batch of wafers to be tested can be continuous. For another example, the characters of specific digits in the wafer numbers of different batches of wafers to be tested can be different. In the above way, the wafer number can be generated on the wafer to be tested.

[0057] See also Figure 2 As shown, Figure 2 FIG. 1 is a schematic diagram of an image to be tested according to an embodiment of the present invention. Figure 2 , the image of the wafer to be tested may be an image without abnormality or an image with abnormality. Figure 2 In the example shown in FIG. 1 , the wafer number in the abnormal image is ghosted.

[0058] The image to be tested can be processed by template matching, feature extraction, the optical character recognition technology mentioned above, etc. to obtain the character area corresponding to each numbered character in the image to be tested. Figure 3 , Figure 3 FIG. 5 is a schematic diagram showing a method for detecting a wafer image according to an embodiment of the present invention. The above steps can be used to implement Figure 3The positioning and identification of numbered characters in the application stage. For example, if the wafer number is "B-AP0319-25", then "B", "-", "A", "P", "0", "3", "1", "9", "-", "2", "5" may be the numbered characters of the wafer number. The above-mentioned character area may be represented based on the coordinate system of the image to be tested. For example, the character area may be a rectangle, and the character area may be determined by the coordinates of the four vertices of the rectangle in the image to be tested. For another example, the character area may be determined by the center of the rectangle in combination with a preset height and length. It should be understood that the shape of the character area may be determined according to the actual needs of the developer, and the embodiment of the present invention is not limited here. Furthermore, for each numbered character, the regional image of the character area corresponding to the numbered character in the image to be tested may only include the numbered character.

[0059] In step S120, for each numbered character, an area image corresponding to a character area corresponding to the numbered character in the image to be tested is input into a detection model corresponding to the numbered character to obtain a first detection result of the numbered character.

[0060] Step S120 can be used to implement Figure 3 For each numbered character, an abnormality detection is performed through the corresponding detection model; the first detection result obtained can be used to determine whether the numbered character has defects. The character area can be an area in the image to be tested, so the image in the area can be used as the area image corresponding to the character area. It should be understood that the area image corresponding to the numbered character can also be further processed. Then, it is input into the detection model corresponding to the numbered character. For example, the area image can be expanded or cropped. For another example, the area image can be feature enhanced. For example, the area image can be subjected to feature enhancement processing such as histogram equalization (Histogram Equalization), adaptive histogram equalization (Adaptive Histogram Equalization), edge detection (Edge Detection). Then, the area image after feature enhancement processing is input into the detection model corresponding to the numbered character to obtain the first detection result of the numbered character.

[0061] Different numbered characters may correspond to different detection models to achieve targeted detection of different numbered characters. The specific structure of the detection model can refer to the architecture of machine learning models such as convolutional neural networks and recurrent neural networks in related technologies, and the embodiments of the present invention are not limited here.

[0062] The first detection result can be used to indicate whether the numbered character is abnormal in the image to be tested. For example, the numbered character may be blocked or have a ghost image in the image to be tested. For each numbered character, the training process of the detection model corresponding to the numbered character may be as follows: input the training image into the untrained detection model to obtain a prediction result. Among them, the training image includes the numbered character, and the training image may correspond to a labeling result to mark whether the numbered character in the training image is abnormal. Based on the difference between the prediction result and the labeling result, the model parameters of the untrained detection model are adjusted, and the detection model is iteratively trained according to the above steps through different training images, and the trained detection model is used as the detection model used in step S120. The above training completion conditions may include that the accuracy of the untrained detection model is higher than the preset accuracy, and the training images have been input into the untrained detection model. It should be understood that the developer can also set the loss function of the detection model according to actual needs, and the embodiment of the present invention is not limited here. The detection model trained in the above manner can generate a first detection result for indicating whether the numbered character is abnormal for the regional image of the numbered character.

[0063] In step S130 , based on the first detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

[0064] The first detection results of all numbered characters can be combined to determine whether there is an abnormality in the image to be tested. For example, the first detection results of all numbered characters can be counted, and if the total number of numbered characters with abnormalities in the first detection result is higher than the preset total number of characters, it can be considered that there is an abnormality in the image to be tested. The specific value of the above-mentioned preset total number of characters can be set by the developer according to actual conditions, and the embodiment of the present invention is not limited here. For another example, if the first detection result of any numbered character is that there is an abnormality, it can be considered that there is an abnormality in the image to be tested. For another example, if the corresponding confidence level of the first detection result of any numbered character in the detection model is lower than the preset confidence level, it can be considered that there is an abnormality in the image to be tested.

[0065] Combined with the actual scenario, if there is an abnormality in the numbering characters, the defect detection result of the wafer may also be abnormal. The defect detection result of the wafer may include whether the wafer has defects, the specific defect type of the wafer, etc. If there is an abnormality in the numbering characters, the cause of the abnormality may also cause the wafer to be tested to be abnormal. For example, if the abnormality of the numbering characters of the wafer is caused by being blocked by foreign matter, the foreign matter may also cause the defect detection result for the wafer to be tested to be that there is a foreign matter in the wafer to be tested. Therefore, it can be determined whether the wafer is defective in combination with whether there is an abnormality in the image to be tested and the defect detection result. For example, if there is an abnormality in the image to be tested, and the defect detection result is that there is no defect in the wafer to be tested, it can be regarded as an inaccurate defect detection result, and it is necessary to focus on the defect detection result of the wafer to be tested. For another example, if there is an abnormality in the image to be tested, and the defect detection result is that there is a foreign matter in the wafer to be tested, it can be regarded as a high probability of defects in the wafer to be tested.

[0066] In another scenario, if there is a problem in the wafer processing flow, resulting in the wafer being scratched, the number characters may have a ghost image. Therefore, if there is an abnormality in the image to be tested, it can be regarded as a possible problem in the processing flow of the wafer to be tested, so that the user can improve the processing flow of the wafer to be tested in time to avoid more wafers being scratched.

[0067] In another scenario, if due to the image acquisition device, a ghost image (such as a scratched lens) or an obstruction (such as a contaminant on the lens) appears in the image to be tested, it may cause the image to be tested to be abnormal. If the above image acquisition device is also used to generate defect detection results of the wafer to be tested (for example, after the image is acquired by the image acquisition device, the defect detection result is generated based on the image), it may cause ghost images or obstructions to appear in the image used to generate the defect detection result. The above situation will reduce the accuracy of the defect detection results of the wafer. For example, a non-defective wafer may be identified as a defective wafer. In addition, if there is a deviation in the angle of the base supporting the wafer, it may cause the image to be tested to be abnormal, and it may also reduce the accuracy of the defect detection results. For example, the defect detection model used for the defect detection result is usually trained based on the training image of the wafer. The training image of the wafer is usually an image taken at a preset angle, so the trained defect detection model is usually less robust to images with angle deviations, thereby reducing the accuracy of the defect detection results. Therefore, if there is an abnormality in the image to be tested, the user can promptly maintain the image acquisition device or the base that carries the wafer to avoid the low accuracy of the defect detection result. Especially in the assembly line scenario, if the image acquisition device or the base that carries the wafer is not maintained in time, the defect detection results of all wafers may be defective, which is not in line with the actual situation, greatly increasing the difficulty of wafer re-inspection and not conducive to increasing the wafer shipment.

[0068] According to the above scheme of the embodiment of the present invention, the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested can be determined. Then, for each number character, the area image corresponding to the character area corresponding to the number character in the image to be tested is input into the detection model corresponding to the number character to obtain the first detection result of the number character. Finally, based on the first detection results of all number characters, it is determined whether the image to be tested of the wafer to be tested is abnormal. On the one hand, the above scheme can be identified by the corresponding detection model for each number character, and targeted detection of different number characters can be realized, which is conducive to improving the detection accuracy of the number characters; on the other hand, the above scheme can also determine whether the image to be tested of the wafer is abnormal based on the detection results related to the wafer number, which is conducive to assisting in improving the processing and detection process of the wafer. For example, if the wafer is scratched during the processing stage, the phantom of the wafer number is generated, resulting in an abnormality in the image to be tested. Then the user can adjust the relevant processing instruments in time to avoid damage to the wafer that has not undergone the processing stage. For another example, if the image to be tested is abnormal due to the skewness of the image acquisition device or the skewness of the base supporting the wafer. The user can then adjust the posture of the image acquisition device or the base in a timely manner to avoid low accuracy of subsequent defect detection on the wafer due to the tilt of the image acquisition device or the base.

[0069] Exemplarily, the detection model used in step S120 can be obtained by training through steps S210 to S240.

[0070] In step S210 , for each training image of the training wafer, character regions corresponding to each training character of the reference number of the training wafer in the training image are determined.

[0071] The training wafers may be wafers from different processing batches than the wafers to be tested. The training image includes the reference number of the training wafer. The training image may be processed by the template matching, feature extraction, optical character recognition technology, etc. in step S110 above to obtain the character regions corresponding to each training character of the reference number of the training wafer in the training image. The embodiments of the present invention will not be described in detail here. Figure 3 , the positioning and recognition of the training characters can be achieved through step S210.

[0072] In step S220, for each training image of the training wafer, feature extraction operations are performed on the regional images corresponding to the character regions corresponding to each training character in the training image to obtain training image features corresponding to each training character.

[0073] Based on the training image, the regional image corresponding to the character region corresponding to each training character in the training image can be obtained by generating the relevant regional image in the above step S120, which will not be described in detail in the embodiment of the present invention.

[0074] The feature extraction operation can be performed on the regional image by a feature extraction algorithm or a feature extraction model to obtain the training image features. For example, the feature extraction operation can be performed on the regional image by using a Haar-like Features extraction algorithm in the related art, a feature extraction module in a convolutional neural network, a feature extraction module in a recurrent neural network, etc. to obtain the training image features.

[0075] In step S230, for each training character, the training image features corresponding to the training character are used as a first feature set of the training character, and outlier features in the first feature set are filtered out to obtain a second feature set of the training character.

[0076] Combination Figure 3 , step S230 can realize the removal of outlier features from the first feature set of each training character respectively, so as to obtain the second feature set of each training character respectively. In one example, a training character may correspond to multiple training images. For example, the training character "6" may correspond to a training image 1 including a reference number "1116" and a training image 2 including a reference number "6544". In another example, a training character may also correspond to multiple character areas in a training image. For example, a training image 3 including a reference number "616263" may include three character areas of the training character "6". It should be understood that the number character of the above-mentioned wafer to be tested may be one of the training characters.

[0077] For each training character, the training image features corresponding to the training character in different training images or the same training image are integrated to obtain the above-mentioned first feature set. The above-mentioned outlier features can be used to represent training image features that have a large difference in feature distribution from most training image features. For example, the outlier features in the first feature set can be determined by Z-score screening method, clustering algorithm, local outlier factor (LOF) algorithm, etc. In one example, the difference between each training image feature and other training image features can also be calculated, for example, the difference can be represented by Euclidean distance, cosine similarity, etc. Then, the training image features with the largest difference and a preset proportion are screened out. For example, if the differences between the five training image features a are small, and there is another training image feature b that has a large difference from the five training image features a, then the training image feature b can be used as an outlier feature. Since the outlier features in the first feature set are screened out, the training image features in the obtained second feature set are more representative.

[0078] In step S240, for each training character, a detection model corresponding to the training character is trained using the second feature set of the training character.

[0079] Combination Figure 3 , step S240 can implement the second feature set of each training character for training the corresponding detection model. The training image features in the second feature set can be sequentially input into the untrained detection model, so that the untrained detection model can output prediction results for the training image features respectively. Then, based on the difference between the prediction result and the annotation result corresponding to the training image feature, the model parameters of the untrained detection model are adjusted until the training is completed. The above annotation result can be used to indicate whether there is an abnormality in the training character corresponding to the training image feature. Optionally, the model parameters of the untrained detection model corresponding to different training characters can be the same to reduce the difficulty of training initialization. Combined with Figure 3 Steps S210 to S240 can be applied in the preparation stage to implement the training of the detection model, and then the trained detection model can be used in the application stage for abnormality detection on the image to be tested.

[0080] According to the above scheme of the embodiment of the present invention, for each training image of the training wafer, the character area corresponding to each training character of the reference number of the training wafer in the training image can be determined. Then, the feature extraction operation is performed on the regional image corresponding to the character area corresponding to each training character in the training image to obtain the training image features corresponding to each training character. For each training character, the training image features corresponding to the training character are used as the first feature set of the training character, and the outlier features in the first feature set are screened out to obtain the second feature set of the training character. Finally, for each training character, the detection model corresponding to the training character is trained using the second feature set of the training character. On the one hand, the above scheme can train the corresponding detection model for the training character, so it is more targeted and can improve the recognition accuracy of the numbered characters in the image to be tested; on the other hand, the above scheme can screen out the outlier features of the training image features to improve the representativeness of the training image features in the second feature set, which is conducive to improving the accuracy of the first detection result of the wafer to be tested.

[0081] Exemplarily, step S210 determines, for each training image of the training wafer, the character area corresponding to each training character of the reference number of the training wafer in the training image, and may include: for each training image of the training wafer, inputting the training image into an optical character recognition model to obtain the character area corresponding to each training character of the reference number of the training wafer in the training image.

[0082] The optical character recognition model may be a model capable of realizing an optical character recognition function. The optical character recognition model may divide the character area in the training image and recognize the corresponding training characters for the character area. The specific content of the optical character recognition model may refer to the relevant technology, and the embodiments of the present invention will not be described in detail here.

[0083] In step S220, for each training image of the training wafer, feature extraction operations are performed on the regional image corresponding to the character area corresponding to each training character in the training image, which may include: for the character area corresponding to each training character in the training image, the confidence of the character area determined by the optical character recognition model is compared with a preset threshold, and feature extraction operations are performed only on the regional image corresponding to the target character area among all character areas in the training image.

[0084] Combination Figure 3, the above steps can be used to collect training characters with confidence levels higher than a preset threshold, and form a first feature set after cropping. The cropped image can be a regional image corresponding to the character area of ​​the training character mentioned above. By performing a feature extraction operation on the regional image, the training image features in the first feature set can be obtained. When outputting the character area and the corresponding training characters, the above optical character recognition model can simultaneously output the confidence level of the training characters in the character area. The higher the confidence level, the more confident the optical character recognition model is in the accuracy of the training characters generated in the character area. It should be understood that the higher the confidence level, the more it can be regarded that there is no abnormality in the training image features of the training character in the training image, that is, there is no abnormality in the training character.

[0085] The confidence of the target character area may be higher than a preset threshold. The specific value of the above preset threshold is not limited in the embodiment of the present invention. It should be understood that the higher the value of the preset threshold, the fewer the total number of training image features, and the worse the robustness of the detection model may be; the lower the value of the preset threshold, the lower the representativeness of the training image features, and the lower the learning efficiency of the detection model may be.

[0086] It should be understood that the confidence level of the target character area is relatively high, so the training characters in the target character area usually do not have abnormalities in the image to be tested. In addition, since the defects that may appear on the wafer are complex and diverse, the feature distance between the training image features with abnormalities is usually large, so the total number of training image features without abnormalities in the obtained second feature set may be greater. In summary, there may be a large number of training image features without abnormalities in the second feature set. In one example, the training images may all be images without abnormalities, so that the training image features in the second feature set do not have abnormalities, and the detection model can then focus on learning the feature distribution of training image features without abnormalities. After the training is completed, it can be determined whether the numbered characters of the wafer to be tested have abnormalities. Optionally, if the training images are all images without abnormalities, the prediction results can be directly compared with the label of "no abnormality" to save computing power.

[0087] According to the above scheme of the embodiment of the present invention, for each training image of the training wafer, the training image can be input into the optical character recognition model to obtain the character area corresponding to each training character of the reference number of the training wafer in the training image. Then, for the character area corresponding to each training character in the training image, the confidence of the character area determined by the optical character recognition model can be compared with the preset threshold, and only the regional image corresponding to the target character area in all character areas in the training image is subjected to feature extraction operation. The above scheme can perform feature extraction operation on the regional image with higher confidence. On the one hand, it can reduce the number of executions of feature extraction operation, thereby saving computing power, which is beneficial to the training efficiency of the detection model; on the other hand, it can improve the representativeness of the training image features, which is beneficial to improve the accuracy of the first detection result output by the detection model. In addition, the defects on the wafer are usually complex and numerous, and the corresponding training images cannot be exhaustively listed. Therefore, the accuracy of the first detection result output by the directly trained detection model is unstable for different defect scenarios. In the above scheme, most of the training image features in the second training set are training image features without abnormalities, which can make the detection model focus on learning the feature distribution of training image features without abnormalities, which is conducive to stabilizing the accuracy of the first detection results under different defect scenarios.

[0088] Exemplarily, the above detection method may further include: step S310 to step S330.

[0089] In step S310 , the wafer number of the wafer to be tested is determined based on the image to be tested.

[0090] Optionally, the wafer number of the wafer to be tested in the image to be tested can be determined by the optical character recognition model described above. Alternatively, the wafer number of the wafer to be tested in the image to be tested can be determined by matching the image to be tested with a template of a preset character, which is not limited in this embodiment of the present invention.

[0091] In step S320, it is determined whether there is any abnormality in the wafer number to obtain a determination result.

[0092] In one example, it is possible to determine whether the wafer number is complete based on the optical character recognition model, and then determine the judgment result. In another example, the overall confidence of the wafer number can be determined based on the optical character recognition model to determine the judgment result. For example, if it is determined that the wafer number is incomplete or the overall confidence of the wafer number is too low, the judgment result may indicate that there is an abnormality in the wafer number. The above judgment result can be used to indicate whether there is an abnormality in the wafer number.

[0093] In step S330, when the determination result indicates that the wafer number is abnormal, it is determined that the image to be tested is abnormal.

[0094] The above steps can be used to implement Figure 3 The step S110 of determining the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested can be performed when the judgment result indicates that there is no abnormality in the wafer number.

[0095] According to the above scheme of the embodiment of the present invention, the wafer number of the wafer to be tested can be determined based on the image to be tested. Then, it is determined whether there is an abnormality in the wafer number to obtain a judgment result. Finally, when the judgment result indicates that there is an abnormality in the wafer number, it is determined that there is an abnormality in the image to be tested. The above scheme can realize the preliminary screening of the wafer to be tested. If the judgment result of the wafer to be tested is that there is an abnormality in the wafer number, it can be directly determined that there is an abnormality in the image to be tested, which is beneficial to improve the detection efficiency of the first detection result of the wafer to be tested. If the judgment result of the wafer to be tested is that there is no abnormality in the wafer number, the first detection result can be determined through the relevant steps above to further determine whether there is an abnormality in the image to be tested, which is beneficial to improve the accuracy of the first detection result of the wafer to be tested.

[0096] Exemplarily, determining whether there is an abnormality in the wafer number in step S320 may include comparing the wafer number with a preset number of the wafer to be tested. If they are inconsistent, it is determined that there is an abnormality in the wafer number.

[0097] The above-mentioned preset number can be the actual number of the wafer to be tested. Combined with the actual scenario, the wafer number of the wafer to be tested is related to the processing batch, so the preset number of the processing batch can be generated on the wafer to be tested by laser marking, electroplating, etc. By comparing the wafer number in the image to be tested with the preset number, it can be determined whether there is an abnormality in the wafer number in the image to be tested, and then whether there is an abnormality in the image to be tested. In one example, if the wafer number is consistent with the preset number of the wafer to be tested, it can be determined that there is no abnormality in the wafer number. In another example, if the wafer number is consistent with the preset number of the wafer to be tested, it can also be determined by other methods whether there is an abnormality in the wafer number.

[0098] According to the above scheme of the embodiment of the present invention, the wafer number can be compared with the preset number of the wafer to be tested to determine whether the wafer number is abnormal. The implementation of the above scheme is a simple character comparison based on string matching, which can be implemented without consuming a lot of computing power, which is conducive to improving the abnormality detection efficiency of the image to be tested.

[0099] Exemplarily, determining whether there is an abnormality in the wafer number in step S320 may include comparing the total amount of characters in the wafer number with a preset total amount. If they are inconsistent, it is determined that there is an abnormality in the wafer number.

[0100] The above-mentioned preset total amount may be the total amount of characters of the preset number. Combined with the actual scenario, the total amount of characters of the wafer numbers of different wafers is usually the same, so the total amount of characters can be used as the preset total amount, and then compared with the total amount of characters of the wafer number of the wafer to be tested, it can be determined whether the wafer number of the wafer to be tested is abnormal, and then whether the image to be tested is abnormal. In one example, if the total amount of characters of the wafer number is consistent with the preset total amount, it can be determined that there is no abnormality in the wafer number. In another example, if the total amount of characters of the wafer number is consistent with the preset total amount, it can be determined by other means whether there is an abnormality in the wafer number. For example, if the total amount of characters of the wafer number is consistent with the preset total amount, and the wafer number is inconsistent with the preset number of the wafer to be tested, it can be determined that there is an abnormality in the wafer number.

[0101] Exemplarily, the detection method may further include determining, for each numbered character, a tilt angle of the numbered character based on a region image corresponding to a character region corresponding to the numbered character in the image to be detected.

[0102] The tilt angle can be determined using a tilt angle detection algorithm in the related art. For example, the tilt angle of the numbered character can be determined using a tilt angle detection algorithm based on a text line baseline, a text image tilt angle detection algorithm based on geometric constraints, etc., which is not limited in the embodiment of the present invention.

[0103] Further, step S130 determines whether there is an abnormality in the image to be tested of the wafer to be tested based on the first detection results of all the numbered characters, and may include: step S131a and step S132a.

[0104] In step S131a, for each numbered character, a second detection result of the numbered character is determined based on the first detection result of the numbered character and the tilt angle of the numbered character.

[0105] The second detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested. If the tilt angle is greater than the preset angle, it can be considered that the numbered character is tilted, which may be caused by a change in the shooting angle of the image acquisition device, a tilt of the base supporting the wafer, etc. The above preset angle can be set by the developer according to the actual situation, and the embodiment of the present invention is not limited here.

[0106] By combining the first detection result and the inclination angle of the numbered character, it can be more accurately determined whether there is an abnormality in the numbered character. For example, if the inclination angle of the numbered character is too large and the first detection result indicates that there is an abnormality in the numbered character in the image to be tested, the second detection result may indicate that there is an abnormality in the numbered character in the image to be tested. For another example, if the inclination angle of the numbered character is too large and the first detection result indicates that there is no abnormality in the numbered character in the image to be tested, the second detection result may indicate that there is an abnormality in the numbered character in the image to be tested. For another example, if the inclination angle of the numbered character does not exceed the preset angle and the first detection result indicates that there is an abnormality in the numbered character in the image to be tested, the second detection result may indicate that there is an abnormality in the numbered character in the image to be tested.

[0107] In step S132a, based on the second detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has any abnormality.

[0108] The second detection results of all numbered characters can be combined to determine whether there is an abnormality in the image to be tested. For example, the second detection results of all numbered characters can be counted, and if the total number of numbered characters with abnormalities as shown in the second detection result is higher than the preset total number of characters, it can be considered that there is an abnormality in the image to be tested. The specific value of the above-mentioned preset total number of characters can be set by the developer according to actual conditions, and the embodiment of the present invention is not limited here. For another example, if the second detection result of any numbered character is that there is an abnormality, it can be considered that there is an abnormality in the image to be tested. For another example, if the corresponding confidence level of the second detection result of any numbered character in the detection model is lower than the preset confidence level, it can be considered that there is an abnormality in the image to be tested.

[0109] According to the above scheme of the embodiment of the present invention, for each numbered character, the inclination angle of the numbered character can be determined based on the regional image corresponding to the character area corresponding to the numbered character in the image to be tested. Then, for each numbered character, the second detection result of the numbered character is determined based on the first detection result of the numbered character and the inclination angle of the numbered character. Finally, it can be determined whether the image to be tested of the wafer to be tested has an abnormality based on the second detection results of all numbered characters. The above scheme can comprehensively determine the second detection result of the numbered character in combination with the inclination angle of the numbered character, thereby improving the accuracy of the second detection result of the numbered character, which is conducive to improving the accuracy of determining whether the wafer to be tested has an abnormality.

[0110] Exemplarily, the above detection method may further include: for each numbered character, determining the area size of the character area corresponding to the numbered character.

[0111] The area size corresponding to the character area can be directly calculated by the coordinates of the vertices of the character area. For example, if the character area is a rectangle, the area size can be expressed as the length or width of the rectangle (which can be obtained by subtracting the coordinates of the vertices), or as the area of ​​the rectangle, which is not limited in the embodiment of the present invention.

[0112] Further, step S130 determines whether there is an abnormality in the image to be tested of the wafer to be tested based on the first detection results of all the numbered characters, and may include: step S131b and step S132b.

[0113] In step S131b, for each numbered character, a third detection result of the numbered character is determined based on the first detection result of the numbered character and the area size of the character area corresponding to the numbered character.

[0114] The third detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested. If the area size is larger than the preset maximum size, the numbered character can be regarded as being too large; if the area size is smaller than the preset minimum size, the numbered character can be regarded as being too small. The above results may be caused by the image acquisition device or the base supporting the wafer being offset. It should be understood that it is also possible to set only the normal size, and then determine the difference between the normal size and the area size. If the difference is higher than the preset value, the numbered character is regarded as being too large or too small. The specific values ​​of the above-mentioned preset maximum size, preset minimum size, normal size and preset value can be set by the developer according to actual conditions, and the embodiments of the present invention are not limited here.

[0115] By combining the first detection result and the area size of the character area corresponding to the numbered character, it can be more accurately determined whether there is an abnormality in the numbered character. For example, if the difference between the area size of the character area and the normal size is greater than a preset value, and the first detection result indicates that there is an abnormality in the numbered character in the image to be tested, then the third detection result may indicate that there is an abnormality in the numbered character in the image to be tested. For another example, if the difference between the area size of the character area and the normal size is greater than a preset value, and the first detection result indicates that there is no abnormality in the numbered character in the image to be tested, then the third detection result may indicate that there is an abnormality in the numbered character in the image to be tested. For another example, if the difference between the area size of the character area and the normal size is less than a preset value, and the first detection result indicates that there is an abnormality in the numbered character in the image to be tested, then the third detection result may indicate that there is an abnormality in the numbered character in the image to be tested.

[0116] In step S132b, based on the third detection results of all the numbered characters, it is determined whether the image to be tested of the wafer to be tested has any abnormality.

[0117] The third detection results of all numbered characters can be combined to determine whether the image to be tested has an abnormality. For example, the third detection results of all numbered characters can be counted, and if the total number of numbered characters with abnormalities as shown in the third detection result is higher than the total number of preset characters, it can be regarded as an abnormality in the image to be tested. The specific value of the above-mentioned preset total number of characters can be set by the developer according to actual conditions, and the embodiment of the present invention is not limited here. For another example, if the third detection result of any numbered character is an abnormality, it can be regarded as an abnormality in the image to be tested. For another example, if the corresponding confidence level of the third detection result of any numbered character in the detection model is lower than the preset confidence level, it can be regarded as an abnormality in the image to be tested. It should be understood that the second detection result above can also be combined to determine whether the image to be tested has an abnormality, and the specific rules may include: if any one of the first detection result, the second detection result, and the third detection result is used to indicate that the image to be tested has an abnormality, it can be regarded as an abnormality in the image to be tested.

[0118] According to the above scheme of the embodiment of the present invention, for each numbered character, the area size of the character area corresponding to the numbered character can be determined. Then, for each numbered character, based on the first detection result of the numbered character and the area size of the character area corresponding to the numbered character, the third detection result of the numbered character is determined. Finally, based on the third detection results of all numbered characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality. The above scheme can comprehensively determine the third detection result of the numbered character in combination with the area size of the character area corresponding to the numbered character, thereby improving the accuracy of the third detection result of the numbered character, which is conducive to improving the accuracy of determining whether the wafer to be tested has an abnormality.

[0119] An embodiment of the present invention also provides a wafer image detection device. Figure 4 FIG. 4 is a schematic block diagram of a wafer image detection device 400 according to an embodiment of the present invention. Figure 4 As shown, the detection device 400 may include: a character area determination module 410, a first detection result determination module 420 and a detected image abnormality determination module 430.

[0120] The character area determination module 410 is used to determine the character area corresponding to each numbered character of the wafer number of the wafer to be tested in the image to be tested. The first detection result determination module 420 is used to input the area image corresponding to the character area corresponding to the numbered character in the image to be tested into the detection model corresponding to the numbered character for each numbered character, so as to obtain the first detection result of the numbered character, wherein the first detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested. The image to be tested abnormality determination module 430 is used to determine whether there is an abnormality in the image to be tested of the wafer to be tested based on the first detection results of all numbered characters.

[0121] Exemplarily, the detection device 400 may include a training character area determination module, a training image feature generation module, a second feature set generation module and a model training module.

[0122] The training character area determination module is used to determine, for each training image of the training wafer, the character area corresponding to each training character of the reference number of the training wafer in the training image. The training image feature generation module is used to perform feature extraction operations on the area image corresponding to the character area corresponding to each training character in the training image for each training image of the training wafer, so as to obtain the training image features corresponding to each training character. The second feature set generation module is used to use, for each training character, the training image features corresponding to the training character as the first feature set of the training character, and filter out the outlier features in the first feature set to obtain the second feature set of the training character. The model training module is used to train the detection model corresponding to each training character using the second feature set of the training character.

[0123] Exemplarily, the training character region determining module may include a training character region first determining module.

[0124] The first training character region determination module is used for inputting each training image of the training wafer into the optical character recognition model to obtain the character region corresponding to each training character of the reference number of the training wafer in the training image.

[0125] The training image feature generation module may include a threshold comparison module.

[0126] The threshold comparison module is used to compare the confidence of the character area determined by the optical character recognition model with a preset threshold for the character area corresponding to each training character in the training image, and only perform feature extraction operations on the area image corresponding to the target character area in all character areas in the training image, wherein the confidence of the target character area is higher than the preset threshold.

[0127] Exemplarily, the detection device 400 may further include: a wafer number determination module, a judgment result generation module and an abnormality determination module.

[0128] The wafer number determination module is used to determine the wafer number of the wafer to be tested based on the image to be tested. The judgment result generation module is used to determine whether there is an abnormality in the wafer number to obtain a judgment result. The abnormality determination module is used to determine whether there is an abnormality in the image to be tested when the judgment result indicates that there is an abnormality in the wafer number. Wherein, determining the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested is executed when the judgment result indicates that there is no abnormality in the wafer number.

[0129] Exemplarily, the judgment result generation module may include at least one of the following: a preset number comparison module and a preset total amount comparison module.

[0130] The preset number comparison module is used to compare the wafer number with the preset number of the wafer to be tested. If they are inconsistent, it is determined that the wafer number is abnormal. The preset total amount comparison module is used to compare the total amount of characters in the wafer number with the preset total amount. If they are inconsistent, it is determined that the wafer number is abnormal.

[0131] Exemplarily, the detection device 400 may further include a tilt angle determination module.

[0132] The tilt angle determination module is used to determine the tilt angle of each numbered character based on the area image corresponding to the character area corresponding to the numbered character in the image to be tested;

[0133] The image to be tested abnormality determination module 430 may include: a second detection result determination module and a first abnormality determination module.

[0134] The second detection result determination module is used to determine, for each numbered character, a second detection result of the numbered character based on the first detection result of the numbered character and the tilt angle of the numbered character, wherein the second detection result is used to indicate whether the numbered character is abnormal in the image to be tested. The first abnormality determination module is used to determine whether the image to be tested of the wafer to be tested is abnormal based on the second detection results of all numbered characters.

[0135] Exemplarily, the above detection device may further include a region size determination module.

[0136] The area size determination module is used to determine the area size of the character area corresponding to each numbered character.

[0137] The image to be tested abnormality determination module 430 may include: a third detection result generation module and a second abnormality determination module.

[0138] The third detection result generating module is used to determine, for each numbered character, a third detection result of the numbered character based on the first detection result of the numbered character and the area size of the character area corresponding to the numbered character, wherein the third detection result is used to indicate whether the numbered character is abnormal in the image to be tested. The second abnormality determining module is used to determine whether the image to be tested of the wafer to be tested is abnormal based on the third detection results of all numbered characters.

[0139] According to another aspect of the present invention, an electronic device is also provided. Figure 5 FIG. 5 shows a schematic block diagram of an electronic device 500 according to an embodiment of the present invention. Figure 5As shown, the electronic device 500 includes a processor 510 and a memory 520. The memory 520 stores a computer program. When the computer program instructions are executed by the processor 510, they are used to execute the above-mentioned wafer image detection method.

[0140] In addition, according to another aspect of the present invention, a storage medium is also provided, on which program instructions are stored, and when the program instructions are executed by a computer or a processor, the computer or the processor executes the corresponding steps of the above-mentioned wafer image detection method of the embodiment of the present invention, and is used to implement the corresponding modules in the above-mentioned wafer image detection device or the corresponding modules in the above-mentioned electronic device according to the embodiment of the present invention. The storage medium may include, for example, a memory card of a smart phone, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above-mentioned storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media. According to another aspect of the present invention, a computer program product is also provided, including computer program instructions, and when the above-mentioned computer program instructions are executed by a computer or a processor, the computer or the processor executes the corresponding steps of the above-mentioned wafer image detection method.

[0141] A person skilled in the art may understand the specific implementation scheme of the above-mentioned electronic device and storage medium by reading the above-mentioned description of the wafer image detection method, which will not be described here for the sake of brevity.

[0142] Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present invention thereto. Various changes and modifications may be made therein by one of ordinary skill in the art without departing from the scope and spirit of the present invention. All such changes and modifications are intended to be included within the scope of the present invention as required by the appended claims.

[0143] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0144] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0145] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0146] Similarly, it should be understood that in order to streamline the present invention and help understand one or more of the various inventive aspects, in the description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present invention should not be interpreted as reflecting the following intention: the claimed invention requires more features than the features explicitly stated in each claim. More specifically, as reflected in the corresponding claims, the inventive point is that the corresponding technical problem can be solved with less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present invention.

[0147] It will be understood by those skilled in the art that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0148] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments may be used in any combination.

[0149] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used in practice to implement some or all of the functions of some modules in a wafer image detection device according to an embodiment of the present invention. The present invention may also be implemented as a device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0150] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc., does not indicate any order. These words may be interpreted as names.

[0151] The above is only a specific embodiment of the present invention or an explanation of a specific embodiment. The protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A wafer image detection method, characterized in that: The method comprises: Determine the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested; For each numbered character, inputting a region image corresponding to a character region corresponding to the numbered character in the image to be tested into a detection model corresponding to the numbered character to obtain a first detection result of the numbered character, wherein the first detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested; Based on the first detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

2. The method according to claim 1, characterized in that The detection model is trained through the following steps: For each training image of the training wafer, Determine the character area corresponding to each training character of the reference number of the training wafer in the training image; Performing feature extraction operations on the regional images corresponding to the character regions corresponding to each training character in the training image, respectively, to obtain training image features corresponding to each training character; For each training character, the training image features corresponding to the training character are used as a first feature set of the training character, and outlier features in the first feature set are filtered out to obtain a second feature set of the training character; For each training character, the detection model corresponding to the training character is trained using the second feature set of the training character.

3. The method according to claim 2, characterized in that The step of determining, for each training image of the training wafer, a character region corresponding to each training character of the reference number of the training wafer in the training image, comprises: For each training image of the training wafer, the training image is input into an optical character recognition model to obtain a character region corresponding to each training character of the reference number of the training wafer in the training image; For each training image of the training wafer, the feature extraction operation is performed on the regional image corresponding to the character region corresponding to each training character in the training image, including: For the character area corresponding to each training character in the training image, the confidence of the character area determined by the optical character recognition model is compared with a preset threshold, and a feature extraction operation is performed only on the area image corresponding to the target character area in all character areas in the training image, wherein the confidence of the target character area is higher than the preset threshold.

4. The method according to claim 1, characterized in that The method further comprises: Based on the image to be tested, determining the wafer number of the wafer to be tested; Determine whether the wafer number is abnormal to obtain a determination result; When the judgment result indicates that the wafer number is abnormal, determining that the image to be tested is abnormal; The step of determining the character area corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested is performed when the judgment result indicates that there is no abnormality in the wafer number.

5. The method according to claim 4, characterized in that The determining whether the wafer number is abnormal includes at least one of the following: Compare the wafer number with the preset number of the wafer to be tested, and if they are inconsistent, determine that the wafer number is abnormal; The total amount of characters in the wafer number is compared with the preset total amount. If they are inconsistent, it is determined that there is an abnormality in the wafer number.

6. The method according to claim 1, characterized in that The method further comprises: For each numbered character, based on a region image corresponding to a character region corresponding to the numbered character in the image to be tested, determining a tilt angle of the numbered character; The determining whether the image to be tested of the wafer to be tested is abnormal based on the first detection result of all the numbered characters comprises: For each numbered character, based on the first detection result of the numbered character and the tilt angle of the numbered character, determine a second detection result of the numbered character, wherein the second detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested; Based on the second detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

7. The method according to claim 1, characterized in that The method further comprises: For each numbered character, determine the area size of the character area corresponding to the numbered character; The determining whether the image to be tested of the wafer to be tested is abnormal based on the first detection result of all the numbered characters comprises: For each numbered character, based on the first detection result of the numbered character and the area size of the character area corresponding to the numbered character, determine a third detection result of the numbered character, wherein the third detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested; Based on the third detection results of all the number characters, it is determined whether the image to be tested of the wafer to be tested has an abnormality.

8. A wafer image detection device, characterized in that: The device comprises: A character region determination module is used to determine the character region corresponding to each number character of the wafer number of the wafer to be tested in the image to be tested; A first detection result determination module is used to input, for each numbered character, a region image corresponding to a character region corresponding to the numbered character in the image to be tested into a detection model corresponding to the numbered character to obtain a first detection result of the numbered character, wherein the first detection result is used to indicate whether there is an abnormality in the numbered character in the image to be tested; The test image abnormality determination module is used to determine whether the test image of the test wafer has an abnormality based on the first detection results of all the numbered characters.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein: the memory is used to store a computer program; the processor is used to execute the computer program to implement the wafer image detection method as described in any one of claims 1 to 7.

10. A storage medium storing a computer program / instruction, characterized in that: The computer program / instructions are used to execute the wafer image detection method as described in any one of claims 1 to 7 when running.

11. A computer program product comprising computer program instructions, characterized in that The computer program instructions are used to execute the wafer image detection method as described in any one of claims 1 to 7 when running.

Images