Semiconductor chip defect detection method and system based on high-precision UV TDI

Abstract

The application provides a semiconductor chip defect detection method and system based on high-precision UV TDI. The method includes: collecting semiconductor chip image data under ultraviolet light through a UV TDI line array camera, preprocessing, and extracting defect features using image processing algorithms. These features are compared with the preset defect database to identify the defect type and severity and classify them. Based on this information, a defect detection report is generated. The UV TDI technology combines the high sensitivity of ultraviolet light with the high resolution and speed of the line array camera. The image processing algorithms used include edge contour recognition and feature extraction. The feature extraction stage constructs a comprehensive feature vector and improves efficiency through dimension reduction processing. The established defect database can identify and classify defects through comparison. When encountering unknown defects, new feature vectors can be added to the database to improve the adaptability and accuracy of detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of chip defect detection, in particular to a semiconductor chip defect detection method and system based on high-precision UV TDI. BACKGROUND

[0002] In the semiconductor manufacturing industry, chip quality control is a key factor to ensure product performance and reliability. With the rapid development of semiconductor technology, the structure of chips is becoming increasingly complex, and the size is continuously shrinking, which puts higher requirements on defect detection in the chip manufacturing process. Traditional defect detection methods, such as manual inspection and visible light-based automatic detection systems, often have low detection accuracy, slow speed, and are easily affected by lighting conditions, which cannot meet the high-precision and high-efficiency requirements of modern semiconductor manufacturing.

[0003] In particular, under ultraviolet (UV) irradiation, certain defects on the surface of a semiconductor chip will exhibit more obvious characteristics that may not be detected under visible light. Therefore, using ultraviolet light for defect detection has potential advantages. However, how to efficiently use ultraviolet image data for defect detection is still a technical problem.

[0004] Traditional image processing algorithms often face problems such as large image noise, blurred edges, and difficult feature extraction when processing semiconductor chip image data. These problems limit the accuracy and reliability of defect detection. SUMMARY

[0005] The present application aims to at least solve the technical problem of low accuracy of defect detection in the prior art, and particularly innovatively provides a semiconductor chip defect detection method and system based on high-precision UV TDI.

[0006] In order to achieve the above-mentioned purpose of the present application, the present application provides a semiconductor chip defect detection method based on high-precision UV TDI, the method comprising:

[0007] S1, acquiring semiconductor chip image data, the semiconductor chip image data being collected by a UV TDI line array camera under ultraviolet light irradiation;

[0008] S2, preprocessing the collected semiconductor chip image data;

[0009] S3, using an image processing algorithm to extract defect features from the preprocessed semiconductor chip image data;

[0010] S4, comparing and analyzing the defect features with a pre-set defect database to identify defect types, judge severity, and classify the defect types;

[0011] S5. Generate a defect detection report based on the defect type and severity.

[0012] As an optional embodiment of the present invention, optionally, the extraction of defect features using an image processing algorithm in step S3 includes:

[0013] S301. Perform image enhancement processing on the semiconductor chip image data and remove noise;

[0014] S302. Identify the edge contours in the semiconductor chip image data;

[0015] S303. Perform morphological processing on the objects in the semiconductor chip image data based on the edge contour;

[0016] S304. Perform feature extraction on the morphologically processed semiconductor chip image data to obtain a feature vector;

[0017] S305. Dimensionally reduce the defect features, and then filter the dimensionality-reduced defect features to obtain defect features.

[0018] As an optional embodiment of the present invention, optionally, identifying the edge contour in step S302 includes:

[0019] S3021. Perform Gaussian filtering on the semiconductor chip image data, the expression is: ,in, This represents the filtered semiconductor chip image data. Represents semiconductor chip image data. This represents the convolution operation. Represents the Gaussian function. This represents the standard deviation of the Gaussian function;

[0020] S3022. Gradient calculation is performed based on the Gaussian-filtered semiconductor chip image data, with the following expression:

[0021] , ,in, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction;

[0022] S3023. Calculate the gradient magnitude and gradient direction based on the gradient, expressed as follows: , ,in, This represents the magnitude of the gradient. Indicates the direction of the gradient;

[0023] S3024. Perform non-maximum suppression based on the gradient magnitude and gradient direction, based on each pixel. Check pixels The values ​​of adjacent pixels along the gradient direction, if the gradient magnitude of the current pixel is not that of the pixel. The local maximum value in the direction is then the pixel point. Set to zero;

[0024] S3025, Set two thresholds and threshold , > Based on the threshold and threshold Traverse the gradient magnitude image; if the pixel value is greater than the threshold... If the pixel value is within the threshold, then it is a strong edge point; if the pixel value is within the threshold... and threshold If the pixel value is between the threshold and the threshold value, then the pixel value is a weak edge point; if the pixel value is less than the threshold value... If so, the pixel is a non-edge point;

[0025] S3026. Obtain the complete edge contour by connecting the weak edge points to the strong edge points.

[0026] As an optional embodiment of the present invention, optionally, performing morphological processing on the object in the semiconductor chip image data based on the edge contour in step S303 includes:

[0027] S30301. Perform an expansion process on the edge contour;

[0028] S30302. Etch the edge contour after the expansion treatment.

[0029] S30303: The edge contour is further optimized using opening and closing operations;

[0030] S30304. Based on the optimized edge contours, segment the objects in the semiconductor chip image data.

[0031] As an optional embodiment of the present invention, optionally, the expression for obtaining the feature vector in step S304 is:

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] in, Represents the eigenvector. The weights represent the texture features. Represents texture features, Weights representing shape features Indicates shape characteristics, The weights representing statistical features Indicates statistical characteristics, Weights representing image energy Represents image energy. Weights representing image contrast Indicates image contrast. Weights representing image relevance Indicates image correlation. Weights representing the homogeneity of images Indicates image homogeneity. The weight representing the area. Represents the total area of ​​an object. This represents the area of ​​the smallest rectangle that encloses the object. The weights representing the perimeter, Represents the total length of the object's boundary. The weights representing the ratio of the object's convex hull. This represents the ratio of the area of ​​the convex hull of an object to the area of ​​the object itself. The weights representing the average gray level of the image. This represents the average gray level of the image. Weights representing the degree of dispersion of the image's grayscale distribution. It indicates the degree of dispersion of the image's grayscale distribution. Weights representing the asymmetry of the image's gray-level distribution. This indicates the asymmetry in the gray-level distribution of an image. Weights representing the sharpness of the image's grayscale distribution. It indicates the sharpness of the grayscale distribution in an image.

[0041] As an optional embodiment of the present invention, optionally, in step S4, a comparative analysis is performed based on the defect features and a preset defect database to identify the defect type, determine the severity, and classify the defect type, including:

[0042] S401. Establish a defect database, which includes known defect types, feature vectors corresponding to the known defect types, severity descriptions of defects, and classification labels.

[0043] S402. The selected defect features are compared one by one with the feature vectors corresponding to the defect types in the defect database using a distance metric to obtain the comparison results.

[0044] S403. Based on the comparison results, determine the defect type that best matches the defect features as the identification result;

[0045] S404. Based on the identified defect type, query the defect database to obtain the severity description of the defect corresponding to the defect type and obtain the severity level;

[0046] S405. Based on the identified defect type and severity level, assign classification labels to the semiconductor chip image data to be inspected from the defect database.

[0047] As an optional embodiment of the present invention, the step S401 of establishing a defect database may include:

[0048] S4010. When the identified defect type is an unknown defect type in the defect database, the feature vector of the unknown defect type is added to the defect database, and the defect database is updated.

[0049] In another aspect, the present invention also provides a semiconductor chip defect detection system based on high-precision UV TDI, the system including a semiconductor chip defect detection method based on high-precision UV TDI;

[0050] The system also includes:

[0051] The image acquisition module is used to acquire image data of semiconductor chips;

[0052] An image preprocessing module, connected to the image acquisition module, is used to preprocess the image data of the semiconductor chip;

[0053] The defect feature extraction module is connected to the image preprocessing module and is used to extract defect features from the preprocessed image data using image processing algorithms.

[0054] The defect identification and classification module, connected to the defect feature extraction module, is used to compare and analyze the extracted defect features with a preset defect database, identify the defect type, determine the severity, and classify the defect type.

[0055] The report generation module, connected to the defect identification and classification module, is used to generate a defect detection report based on the defect type and severity.

[0056] The beneficial effects of this invention are as follows: This invention employs a high-precision UV TDI line scan camera to acquire semiconductor chip image data under ultraviolet light irradiation. UV TDI technology combines the high sensitivity of ultraviolet light with the high resolution and high speed of a line scan camera, enabling the capture of subtle features on the chip surface under ultraviolet light, providing high-quality raw data for subsequent image processing. After acquiring the image data, this solution performs a series of preprocessing operations, including noise reduction and contrast enhancement, to improve image quality and reduce the impact of noise on subsequent processing. This solution employs advanced image processing algorithms, including edge contour recognition, morphological processing, and feature extraction, to accurately extract defect features from semiconductor chip images. In particular, in edge contour recognition, steps such as Gaussian filtering, gradient calculation, non-maximum suppression, and thresholding can accurately identify edge contours in the chip image, providing a reliable foundation for subsequent feature extraction. In the feature extraction stage, this scheme comprehensively considers various features such as texture features, shape features, statistical features, image energy, image contrast, image correlation, image homogeneity, area, perimeter, convex hull ratio, average image grayscale value, dispersion, asymmetry, and sharpness to construct a comprehensive feature vector. Simultaneously, dimensionality reduction removes redundant information, improving the effectiveness and computational efficiency of the feature vectors. This scheme establishes a defect database containing known defect types, feature vectors, severity descriptions, and classification labels. By comparing the extracted defect features with the feature vectors in the database using a distance metric, the defect type can be accurately identified, and its severity determined. Based on the identification results, corresponding classification labels are assigned to the semiconductor chip image data to be inspected. When encountering unknown defect types, this scheme allows the feature vectors of the new defect type to be added to the defect database and the database to be updated. This ensures that the database can be continuously improved with the accumulation of inspection practice, enhancing the adaptability and accuracy of defect detection.

[0057] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0058] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0059] Figure 1 This is a flowchart of the semiconductor chip defect detection method based on high-precision UV TDI in Embodiment 1 of the present invention;

[0060] Figure 2 This is a structural diagram of the semiconductor chip defect detection system based on high-precision UV TDI in Embodiment 2 of the present invention. Detailed Implementation

[0061] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0062] Example 1

[0063] like Figure 1 As shown, a semiconductor chip defect detection method based on high-precision UV TDI is disclosed, the method comprising:

[0064] S1. Acquire semiconductor chip image data, which is obtained by a UV TDI line array camera under ultraviolet light irradiation;

[0065] It should be noted that in step S1, the semiconductor chip image data is acquired using a UV TDI linear array camera under ultraviolet light irradiation. This fully utilizes the high resolution and high speed advantages of UV TDI technology, as well as the high sensitivity of ultraviolet light to subtle features on the chip surface, thereby ensuring high-quality and information-rich image data. In subsequent image processing steps, preprocessing operations such as noise reduction and contrast enhancement can further improve image quality, laying a solid foundation for subsequent edge contour recognition and feature extraction.

[0066] S2. Preprocess the acquired semiconductor chip image data;

[0067] It should be noted that the preprocessing of the acquired semiconductor chip image data in step S2 aims to eliminate image noise and enhance image contrast, providing high-quality image data for subsequent edge detection and feature extraction. By employing appropriate preprocessing algorithms, such as denoising and contrast enhancement algorithms, the signal-to-noise ratio and sharpness of the image can be significantly improved, thereby ensuring the accuracy and reliability of subsequent processing.

[0068] S3. Using image processing algorithms, defect features are extracted from the preprocessed semiconductor chip image data;

[0069] It should be noted that defect feature extraction in step S3 specifically includes multiple stages such as edge contour recognition, morphological processing, and feature vector construction. Edge contour recognition can accurately depict the edges of objects in the chip image, while morphological processing optimizes the edge contours through steps such as dilation, erosion, opening, and closing operations to improve the accuracy of feature extraction. Feature vector construction comprehensively considers various image features, including texture features, shape features, and statistical features, constructing a comprehensive and effective feature vector, providing a reliable basis for subsequent defect identification and classification. By employing advanced image processing algorithms and comprehensive feature vector construction, this embodiment can accurately extract defect features from semiconductor chip images.

[0070] S4. Based on the defect characteristics, perform comparative analysis with a preset defect database to identify defect types, determine severity, and classify defect types.

[0071] It should be noted that in step S4, comparative analysis of defect features is performed using a pre-set defect database, enabling rapid and accurate identification of defect types in semiconductor chip image data. This database contains known defect types and their corresponding feature vectors, defect severity descriptions, and classification labels. This allows the system to compare extracted defect features with information in the database to determine the defect type, severity, and classification label. This step improves the efficiency and accuracy of defect detection. Furthermore, the system possesses self-learning and updating capabilities. When encountering an unknown defect type, it can add its feature vector to the defect database and update the database, thereby continuously improving the system's detection capabilities and adaptability.

[0072] S5. Generate a defect detection report based on the defect type and severity.

[0073] It should be noted that the defect detection report generated in step S5 records in detail the types, severity, and classification labels of defects found during the detection process, providing strong data support for the quality control of semiconductor chip production. This report not only helps production personnel to promptly identify and repair defects in chips, improving product quality, but also provides valuable reference for subsequent process improvements and product design.

[0074] In summary, this embodiment employs a high-precision UV TDI line scan camera to acquire semiconductor chip image data under ultraviolet light irradiation. UV TDI technology combines the high sensitivity of ultraviolet light with the high resolution and high speed of a line scan camera, enabling the capture of subtle features on the chip surface under ultraviolet light, providing high-quality raw data for subsequent image processing. After acquiring the image data, this scheme performs a series of preprocessing operations, including noise reduction and contrast enhancement, to improve image quality and reduce the impact of noise on subsequent processing. This scheme utilizes advanced image processing algorithms, including edge contour recognition, morphological processing, and feature extraction, to accurately extract defect features from the semiconductor chip image. In particular, in edge contour recognition, steps such as Gaussian filtering, gradient calculation, non-maximum suppression, and thresholding accurately identify the edge contours in the chip image, providing a reliable foundation for subsequent feature extraction. In the feature extraction stage, this scheme comprehensively considers various features such as texture features, shape features, statistical features, image energy, image contrast, image correlation, image homogeneity, area, perimeter, convex hull ratio, average image grayscale value, dispersion, asymmetry, and sharpness to construct a comprehensive feature vector. Simultaneously, dimensionality reduction removes redundant information, improving the effectiveness and computational efficiency of the feature vectors. This scheme establishes a defect database containing known defect types, feature vectors, severity descriptions, and classification labels. By comparing the extracted defect features with the feature vectors in the database using a distance metric, the defect type can be accurately identified, and its severity determined. Based on the identification results, corresponding classification labels are assigned to the semiconductor chip image data to be inspected. When encountering unknown defect types, this scheme allows the feature vectors of the new defect type to be added to the defect database and the database to be updated. This ensures that the database can be continuously improved with the accumulation of inspection practice, enhancing the adaptability and accuracy of defect detection.

[0075] As an optional embodiment of the present invention, optionally, the extraction of defect features using an image processing algorithm in step S3 includes:

[0076] S301. Perform image enhancement processing on the semiconductor chip image data and remove noise;

[0077] S302. Identify the edge contours in the semiconductor chip image data;

[0078] S303. Perform morphological processing on the objects in the semiconductor chip image data based on the edge contour;

[0079] S304. Perform feature extraction on the morphologically processed semiconductor chip image data to obtain a feature vector;

[0080] S305. Dimensionally reduce the defect features, and then filter the dimensionality-reduced defect features to obtain defect features.

[0081] It should be noted that in step S301, image enhancement and denoising are performed to improve the signal-to-noise ratio and contrast of the image data, making the defect features in the image more obvious, which facilitates subsequent edge contour recognition and feature extraction. In step S302, edge contour recognition is one of the key steps in defect feature extraction. By accurately depicting the edges of objects in the chip image, an accurate foundation can be provided for subsequent morphological processing and feature extraction. In step S303, morphological processing further improves the accuracy of feature extraction by optimizing the edge contours. In step S304, feature extraction is the key to constructing a comprehensive and effective feature vector, which comprehensively considers multiple image features, providing a reliable basis for subsequent defect recognition and classification. In step S305, dimensionality reduction and feature filtering are performed to remove redundant information, improve the effectiveness and computational efficiency of the feature vector, thereby ensuring the accuracy and real-time performance of defect detection.

[0082] As an optional embodiment of the present invention, optionally, identifying the edge contour in step S302 includes:

[0083] S3021. Perform Gaussian filtering on the semiconductor chip image data, the expression is: ,in, This represents the filtered semiconductor chip image data. Represents semiconductor chip image data. This represents the convolution operation. Represents the Gaussian function. This represents the standard deviation of the Gaussian function;

[0084] S3022. Gradient calculation is performed based on the Gaussian-filtered semiconductor chip image data, with the following expression:

[0085] , ,in, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction;

[0086] S3023. Calculate the gradient magnitude and gradient direction based on the gradient, expressed as follows: , ,in, This represents the magnitude of the gradient. Indicates the direction of the gradient;

[0087] S3024. Perform non-maximum suppression based on the gradient magnitude and gradient direction, based on each pixel. Check pixels The values ​​of adjacent pixels along the gradient direction, if the gradient magnitude of the current pixel is not that of the pixel. The local maximum value in the direction is then the pixel point. Set to zero;

[0088] S3025, Set two thresholds and threshold , > Based on the threshold and threshold Traverse the gradient magnitude image; if the pixel value is greater than the threshold... If the pixel value is within the threshold, then it is a strong edge point; if the pixel value is within the threshold... and threshold If the pixel value is between the threshold and the threshold value, then the pixel value is a weak edge point; if the pixel value is less than the threshold value... If so, the pixel is a non-edge point;

[0089] S3026. Obtain the complete edge contour by connecting the weak edge points to the strong edge points.

[0090] It should be noted that in step S3021, Gaussian filtering, as a linear smoothing filter, can effectively suppress image noise and reduce interference from high-frequency components, making the edge features in the image smoother and more obvious. By selecting an appropriate standard deviation of the Gaussian function, the smoothness of the filter can be controlled to achieve the best filtering effect. In step S3022, gradient calculation is based on the Sobel convolution kernel to perform convolution operations on the image, which can calculate the gradient of the image in the x-axis and y-axis directions, thereby reflecting the changes in the image grayscale values. In step S3023, by calculating the gradient magnitude and direction of each pixel, it can be determined whether the pixel is an edge point and the direction of the edge. In step S3024, non-maximum suppression is an edge thinning technique. By suppressing non-edge pixels and retaining local maxima in the gradient direction, a thinned edge contour is obtained. This step can reduce redundant edge information and improve the continuity and accuracy of the edges. In step S3025, by setting two thresholds, the pixels in the gradient magnitude image can be divided into strong edge points, weak edge points, and non-edge points. Strong edge points have high gradient magnitudes and typically correspond to prominent edges in an image; weak edge points have lower gradient magnitudes but still play an important role in connecting strong edge points; non-edge points have the lowest gradient magnitudes and typically correspond to flat areas in an image. In step S3026, by connecting weak edge points to strong edge points, a complete edge contour can be obtained. This step fills in discontinuities in the edge contour, making the edges more continuous and smooth. By employing the above edge contour recognition method, this embodiment can accurately identify edge contours in semiconductor chip images.

[0091] As an optional embodiment of the present invention, optionally, performing morphological processing on the object in the semiconductor chip image data based on the edge contour in step S303 includes:

[0092] S30301. Perform an expansion process on the edge contour;

[0093] S30302. Etch the edge contour after the expansion treatment.

[0094] S30303: The edge contour is further optimized using opening and closing operations;

[0095] S30304. Based on the optimized edge contours, segment objects in the semiconductor chip image data.

[0096] It should be noted that in step S30301, the dilation process expands the range of the edge contour, filling in small holes and depressions, making the edge contour more complete and continuous. This step helps improve the connectivity of objects in the image. In step S30302, the erosion process shrinks the range of the edge contour, removing small protrusions and making the edge contour smoother and more regular. Alternating dilation and erosion processes allow for fine adjustment of the edge contour, improving its accuracy and reliability. In step S30303, the opening operation removes small objects from the edge contour while maintaining its main structure, helping to eliminate noise and interference in the image. The closing operation fills in discontinuities and small holes on the edge contour, making it more complete and continuous. The combined use of opening and closing operations further optimizes the shape of the edge contour, improving its quality and stability. In step S30304, based on the optimized edge contour, objects in the semiconductor chip image data can be segmented. This step distinguishes different objects in the image, providing an accurate foundation for subsequent feature extraction and defect identification. By employing the morphological processing method described above, this embodiment can optimize the edge contours in semiconductor chip images, thereby improving the accuracy and reliability of feature extraction.

[0097] As an optional embodiment of the present invention, optionally, the expression for obtaining the feature vector in step S304 is:

[0098] ;

[0099] ;

[0100] ;

[0101] ;

[0102] ;

[0103] ;

[0104] ;

[0105] ;

[0106] in, Represents the eigenvector. The weights represent the texture features. Represents texture features, Weights representing shape features Indicates shape characteristics, The weights representing statistical features Indicates statistical characteristics, Weights representing image energy Represents image energy. Weights representing image contrast Indicates image contrast. Weights representing image relevance Indicates image correlation. Weights representing the homogeneity of images Indicates image homogeneity. The weight representing the area. Represents the total area of ​​an object. This represents the area of ​​the smallest rectangle that encloses the object. The weights representing the perimeter, Represents the total length of the object's boundary. The weights representing the ratio of the object's convex hull. This represents the ratio of the area of ​​the convex hull of an object to the area of ​​the object itself. The weights representing the average gray level of the image. This represents the average gray level of the image. Weights representing the degree of dispersion of the image's grayscale distribution. It indicates the degree of dispersion of the image's grayscale distribution. Weights representing the asymmetry of the image's gray-level distribution. This indicates the asymmetry in the gray-level distribution of an image. Weights representing the sharpness of the image's grayscale distribution. It indicates the sharpness of the grayscale distribution in an image.

[0107] As an optional embodiment of the present invention, optionally, in step S4, a comparative analysis is performed based on the defect features and a preset defect database to identify the defect type, determine the severity, and classify the defect type, including:

[0108] S401. Establish a defect database, which includes known defect types, feature vectors corresponding to the known defect types, severity descriptions of defects, and classification labels.

[0109] S402. The selected defect features are compared one by one with the feature vectors corresponding to the defect types in the defect database using a distance metric to obtain the comparison results.

[0110] S403. Based on the comparison results, determine the defect type that best matches the defect features as the identification result;

[0111] S404. Based on the identified defect type, query the defect database to obtain the severity description of the defect corresponding to the defect type and obtain the severity level;

[0112] S405. Based on the identified defect type and severity level, assign classification labels to the semiconductor chip image data to be inspected from the defect database.

[0113] It should be noted that in step S401, the establishment of the defect database is based on a large number of known defect types and their corresponding feature vectors, which are obtained through previous defect detection and analysis processes. The severity descriptions and classification labels of the defects are based on professional assessments and classifications to ensure the accuracy and reliability of the database. In step S402, the distance metric method assesses the similarity between the features of the defect to be detected and the feature vectors of known defect types by calculating the distance between them. The smaller the distance, the higher the similarity, thus determining the best-matching defect type. In step S403, based on the comparison results, the defect type with the smallest distance is selected as the identification result, which helps to accurately determine the defect types present in the semiconductor chip image to be inspected. In step S404, by querying the defect database, the severity level and classification label of the identified defect type can be obtained. Finally, based on step S405, by employing the above defect identification method, this embodiment can achieve accurate identification and classification of defects in semiconductor chip images.

[0114] As an optional embodiment of the present invention, optionally, the expression for the distance metric in step S402 is:

[0115] ;

[0116] in, Represents the selected defect feature vector With the defect database Feature vectors of known defect types The distance between them The dimension of the feature vector. Indicates the first The weights of each feature component, This represents the first defect feature vector selected from the filter. The values ​​of each feature component Represents the first in the defect database In the feature vector of the known defect type, the th The values ​​of each feature component Indicates the first The standard deviation of each characteristic component.

[0117] As an optional embodiment of the present invention, the step S401 of establishing a defect database may include:

[0118] S4010. When the identified defect type is an unknown defect type in the defect database, the feature vector of the unknown defect type is added to the defect database, and the defect database is updated.

[0119] It should be noted that in step S4010, when the identified defect type cannot find a match in the existing defect database, it is considered an unknown defect type. In this case, the feature vector of the unknown defect type can be added to the defect database, and the database record can be updated. This step helps to expand the coverage of the defect database and improve its adaptability and practicality. By continuously updating the defect database, it can be ensured that the defect identification system can accurately identify and classify newly emerging defect types. This also demonstrates the flexibility and scalability of the method of this invention.

[0120] Example 2

[0121] like Figure 2 As shown, a semiconductor chip defect detection system based on high-precision UV TDI is disclosed, the system including a semiconductor chip defect detection method based on high-precision UV TDI;

[0122] The system also includes:

[0123] The image acquisition module is used to acquire image data of the semiconductor chip. In this embodiment, the image acquisition module is an image sensor based on high-precision UV TDI, which has high resolution and high sensitivity, and can achieve accurate acquisition of semiconductor chip images.

[0124] An image preprocessing module, connected to the image acquisition module, is used to preprocess the semiconductor chip image data. In this embodiment, preprocessing includes operations such as noise reduction, contrast enhancement, and image distortion correction to improve image quality and the accuracy of subsequent processing. Through the image preprocessing module, noise and interference factors in the image can be eliminated, and the detailed features of the image can be enhanced, providing high-quality image data for subsequent edge contour recognition and morphological processing.

[0125] The defect feature extraction module, connected to the image preprocessing module, is used to extract defect features from the preprocessed image data using image processing algorithms. Specifically, this module can identify and analyze key features in the image, such as texture, shape, statistical properties, image energy, contrast, correlation, and homogeneity. These features are crucial for defect identification. Through image processing technology, the defect feature extraction module can accurately extract feature vectors related to defects, providing strong support for subsequent defect identification.

[0126] The defect identification and classification module, connected to the defect feature extraction module, is used to compare and analyze the extracted defect features with a preset defect database, identify the defect type, determine the severity, and classify the defect type.

[0127] The report generation module, connected to the defect identification and classification module, is used to generate a defect detection report based on the defect type and severity.

[0128] The defect feature extraction module includes:

[0129] The image enhancement and denoising submodule is used to enhance the image data of the semiconductor chip and remove noise;

[0130] The edge contour recognition submodule is used to identify edge contours in an image. Specifically, this submodule employs image processing algorithms, such as Canny edge detection, to accurately identify edge contours in the image. These edge contours form the basis for subsequent morphological processing and defect feature extraction. Through the processing of the edge contour recognition submodule, key features in the image are accurately captured. Furthermore, the submodule's parameters can be adjusted according to actual needs to adapt to the characteristics and requirements of different semiconductor chip images, improving the system's flexibility and adaptability.

[0131] The morphological processing submodule is used to perform morphological processing on edge contours, including dilation, erosion, opening, and closing operations.

[0132] The feature vector extraction submodule is used to extract feature vectors from the edge contours after morphological processing.

[0133] The defect identification and classification module includes:

[0134] The defect database establishment submodule is used to establish a defect database, which includes known defect types, feature vectors corresponding to the known defect types, severity descriptions of defects, and classification labels.

[0135] The feature comparison submodule is used to compare the selected defect features with the defect types in the defect database one by one using the distance metric method to obtain the comparison results. Specifically, the feature comparison submodule can efficiently process a large amount of defect feature data and quickly find the best matching defect type by comparing it with the feature vectors in the defect database.

[0136] The defect identification and classification submodule is used to determine the most matching defect type based on the comparison results and query the defect database to obtain the severity level and classification label. Specifically, the defect identification and classification submodule can accurately determine the defect type present in the image of the semiconductor chip to be inspected and provide the corresponding severity level and classification label. The implementation of this function depends on the accuracy and completeness of the defect database and the efficient processing capability of the feature comparison submodule.

[0137] The database update submodule is used to add the feature vector of the unknown defect type to the defect database and update the defect database when the identified defect type is an unknown defect type in the defect database.

[0138] The principle of a semiconductor chip defect detection system is as follows:

[0139] The system first acquires image data of the semiconductor chip through the image acquisition module. This data is then passed to the image preprocessing module for necessary preprocessing operations, such as grayscale conversion and filtering, to improve image quality and reduce noise interference. The preprocessed image data is then fed into the defect feature extraction module, which uses image processing algorithms such as image enhancement, denoising, edge contour recognition, and morphological processing to extract defect features from the semiconductor chip image. These feature vectors accurately reflect the defect situation on the chip surface. The extracted feature vectors are then passed to the defect identification and classification module. This module's built-in defect database stores a large number of known defect types and their corresponding feature vectors, severity descriptions, and classification labels. By employing efficient comparison algorithms such as distance metrics, the module can compare the extracted defect features with the known defect types in the database one by one, quickly determining the best-matching defect type. Simultaneously, it can also query the database to obtain the severity level and classification label of the defect type. If the identified defect type has no match in the database, it is considered an unknown defect type, at which point the database update submodule comes into play. It automatically adds the feature vector of the unknown defect type to the defect database and updates the database records, thereby ensuring the continuous accuracy and adaptability of the defect identification system. Finally, the report generation module automatically generates a detailed defect detection report based on the defect type and severity. This report not only provides important evidence for the quality control of semiconductor chip production but also offers valuable reference for subsequent product improvement and optimization. The entire detection system's workflow achieves a high degree of automation and intelligence, greatly improving the efficiency and accuracy of semiconductor chip defect detection.

[0140] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A semiconductor chip defect detection method based on high-precision UV TDI, characterized in that, The method includes: S1. Acquire semiconductor chip image data, which is obtained by a UV TDI line array camera under ultraviolet light irradiation; S2. Use a contrast enhancement algorithm to enhance the contrast of the semiconductor chip image data; S3. Using image processing algorithms, defect features are extracted from the semiconductor chip image data. Defect feature extraction also includes edge contour recognition and morphological processing. The defect feature extraction using image processing algorithms in step S3 includes: S301. Perform image enhancement processing on the semiconductor chip image data and remove noise; S302. Identify the edge contours in the semiconductor chip image data; S303. Perform morphological processing on the objects in the semiconductor chip image data based on the edge contour; S304. Perform feature extraction on the morphologically processed semiconductor chip image data to obtain a feature vector; the expression for obtaining the feature vector in step S304 is: ； ； ； ； ； ； ； ； in, Represents the eigenvector. The weights represent the texture features. Represents texture features, Weights representing shape features Indicates shape characteristics, The weights representing statistical features Indicates statistical characteristics, Weights representing image energy Represents image energy. Weights representing image contrast Indicates image contrast. Weights representing image relevance Indicates image correlation. Weights representing the homogeneity of images Indicates image homogeneity. The weight representing the area. Represents the total area of ​​an object. This represents the area of ​​the smallest rectangle that encloses the object. The weights representing the perimeter, Represents the total length of the object's boundary. The weights representing the ratio of the object's convex hull. This represents the ratio of the area of ​​the convex hull of an object to the area of ​​the object itself. The weights representing the average gray level of the image. This represents the average gray level of the image. Weights representing the degree of dispersion of the image's grayscale distribution. It indicates the degree of dispersion of the image's grayscale distribution. Weights representing the asymmetry of the image's gray-level distribution. This indicates the asymmetry in the gray-level distribution of an image. Weights representing the sharpness of the image's grayscale distribution. Indicates the sharpness of the grayscale distribution in an image; S305. Dimensionally reduce the defect features, and then filter the dimensionality-reduced defect features to obtain defect features; S4. Based on the defect characteristics, perform comparative analysis with a preset defect database to identify defect types, determine severity, and classify the defect types, including: S401. Establish a defect database and update the database. The defect database includes known defect types, feature vectors corresponding to the known defect types, severity descriptions of defects, and classification labels. S402. The selected defect features are compared one by one with the feature vectors corresponding to the defect types in the defect database using a distance metric to obtain the comparison results. S403. Based on the comparison results, determine the defect type that best matches the defect features as the identification result; S404. Based on the identified defect type, query the defect database to obtain the severity description of the defect corresponding to the defect type and obtain the severity level; S405. Based on the identified defect type and severity level, assign classification labels to the semiconductor chip image data to be inspected from the defect database; S5. Generate a defect detection report based on the defect type and severity.

2. The semiconductor chip defect detection method based on high-precision UV TDI as described in claim 1, characterized in that, Identifying edge contours in step S302 includes: S3021. Perform Gaussian filtering on the semiconductor chip image data, the expression is: ,in, This represents the filtered semiconductor chip image data. Represents semiconductor chip image data. This represents the convolution operation. Represents the Gaussian function. This represents the standard deviation of the Gaussian function; S3022. Gradient calculation is performed based on the Gaussian-filtered semiconductor chip image data, with the following expression: , ,in, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction, express Gradient in the axial direction, express Sobel convolution kernels along the axial direction; S3023. Calculate the gradient magnitude and gradient direction based on the gradient, expressed as follows: , ,in, This represents the magnitude of the gradient. Indicates the direction of the gradient; S3024. Perform non-maximum suppression based on the gradient magnitude and gradient direction, based on each pixel. Check pixels The values ​​of adjacent pixels along the gradient direction, if the gradient magnitude of the current pixel is not that of the pixel. The local maximum value in the direction is then the pixel point. Set to zero; S3025, Set two thresholds and threshold , > Based on the threshold and threshold Traverse the gradient magnitude image; if the pixel value is greater than the threshold... If the pixel value is within the threshold, then it is a strong edge point; if the pixel value is within the threshold... and threshold If the pixel value is between the threshold and the threshold value, then the pixel value is a weak edge point; if the pixel value is less than the threshold value... If so, the pixel is a non-edge point; S3026. Obtain the complete edge contour by connecting the weak edge points to the strong edge points.

3. The semiconductor chip defect detection method based on high-precision UV TDI as described in claim 1, characterized in that, Step S303, which involves performing morphological processing on the objects in the semiconductor chip image data based on the edge contour, includes: S30301. Perform an expansion process on the edge contour; S30302. Etch the edge contour after the expansion treatment. S30303: The edge contour is further optimized using opening and closing operations; S30304. Based on the optimized edge contours, segment the objects in the semiconductor chip image data.

4. The semiconductor chip defect detection method based on high-precision UV TDI as described in claim 1, characterized in that, The expression for the distance metric in step S402 is: ； in, Represents the selected defect feature vector With the defect database Feature vectors of known defect types The distance between them The dimension of the feature vector. Indicates the first The weights of each feature component, This represents the first defect feature vector selected from the filter. The values ​​of each feature component Represents the first in the defect database In the feature vector of the known defect type, the th The values ​​of each feature component Indicates the first The standard deviation of each characteristic component.

5. The semiconductor chip defect detection method based on high-precision UV TDI as described in claim 1, characterized in that, Step S401, establishing the defect database, includes: S4011. When the identified defect type is an unknown defect type in the defect database, the feature vector of the unknown defect type is added to the defect database, and the defect database is updated.

6. A semiconductor chip defect detection system based on high-precision UV TDI, characterized in that, The system includes a semiconductor chip defect detection method based on high-precision UV TDI as described in any one of claims 1 to 5; The system also includes: The image acquisition module is used to acquire image data of semiconductor chips; An image preprocessing module, connected to the image acquisition module, is used to preprocess the image data of the semiconductor chip; The defect feature extraction module is connected to the image preprocessing module and is used to extract defect features from the preprocessed image data using image processing algorithms. The defect identification and classification module, connected to the defect feature extraction module, is used to compare and analyze the extracted defect features with a preset defect database, identify the defect type, determine the severity, and classify the defect type. The report generation module, connected to the defect identification and classification module, is used to generate a defect detection report based on the defect type and severity.

7. The semiconductor chip defect detection system based on high-precision UV TDI as described in claim 6, characterized in that, The defect feature extraction module includes: The image enhancement and denoising submodule is used to enhance the image data of the semiconductor chip and remove noise; The edge contour recognition submodule is used to identify edge contours in an image; The morphological processing submodule is used to perform morphological processing on edge contours, including dilation, erosion, opening, and closing operations. The feature vector extraction submodule is used to extract feature vectors from the edge contours after morphological processing. The defect identification and classification module includes: The defect database establishment submodule is used to establish a defect database, which includes known defect types, feature vectors corresponding to the known defect types, severity descriptions of defects, and classification labels. The feature comparison submodule is used to compare the selected defect features with the defect types in the defect database one by one using the distance metric method to obtain the comparison results. The defect identification and classification submodule is used to determine the most matching defect type based on the comparison results, and query the defect database to obtain the severity level and classification label; The database update submodule is used to add the feature vector of the unknown defect type to the defect database and update the defect database when the identified defect type is an unknown defect type in the defect database.

Images