Semiconductor wafer defect detection method and system based on machine vision

Abstract

The invention discloses a semiconductor wafer defect detection method and system based on machine vision, and relates to the technical field of wafer defect detection.The method comprises the steps that a to-be-detected semiconductor wafer image is obtained; extracting a grain image based on the wafer image; based on the grain image and a preset image, a specific position with a defect is obtained, and the preset image is a grain image without a defect; determining morphological characteristics of the defect according to the specific position of the defect; according to the method, the semiconductor wafer defects are detected on the basis of machine vision, human eyes are replaced by a machine for measurement and judgment, the problem that misjudgment is prone to occurring or the detection speed is low when manual detection is conducted through a microscopic system is solved, meanwhile, image processing is conducted on the basis of crystal grain images, and the detection accuracy is improved. Interference of irrelevant areas is avoided, and detection efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wafer defect detection, and more specifically to a method and system for semiconductor wafer defect detection based on machine vision. Background Art

[0002] With the development of the semiconductor industry, the demand for the production and manufacturing of semiconductor wafers is increasing day by day. However, quality defects are inevitable in the production and preparation process of semiconductor wafers, so factories have to use tools such as microscopes to manually observe the products.

[0003] For example, the Chinese invention patent with announcement number CN114160450A provides a semiconductor wafer detection system and detection method thereof. The invention detects the wafer through a microscope system, but when manual detection is performed through a microscope system, it is easy to have problems such as misjudgment or slow detection speed, so the existing technology has shortcomings. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a semiconductor wafer defect detection method and system based on machine vision, which converts the semiconductor wafer into an image signal through a machine vision product, transmits it to a dedicated image processing device, and further processes the image to realize the detection of semiconductor wafer defects.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention provides a method for detecting semiconductor wafer defects based on machine vision, comprising:

[0007] Acquire an image of a semiconductor wafer to be inspected;

[0008] extracting a grain image based on the wafer image;

[0009] Based on the grain image and a preset image, a specific location where a defect exists is obtained, wherein the preset image is an image of a grain without defects;

[0010] Determining the morphological characteristics of the defect according to the specific location of the defect;

[0011] The type of the defect is determined according to the morphological characteristics of the defect.

[0012] As a further improvement of the present invention, the step of extracting a grain image based on the wafer image includes:

[0013] Convolving the wafer image with a Gaussian kernel to obtain a Gaussian pyramid, and subtracting adjacent layers of the Gaussian pyramid to obtain a Gaussian difference pyramid;

[0014] Compare each pixel point in the Gaussian difference pyramid with all its adjacent pixels to determine extreme value points;

[0015] The Gaussian difference pyramid is expanded at the extreme point using the Taylor formula, and the offset of the extreme point is calculated, and the extreme point where the offset is less than a preset value is taken as a feature point;

[0016] A feature vector is constructed according to the feature points, and the feature vector is matched with a preset feature vector to obtain the grain image, wherein the preset feature vector is calculated according to the preset image.

[0017] As a further improvement of the present invention, constructing a feature vector according to the feature points includes:

[0018] Taking the feature point as the center, a circular area within a preset radius is selected, and the circular area is located inside the area where the grain is located;

[0019] Divide the circular area into m circular areas, where m is a positive integer greater than 1;

[0020] Divide the circular area into n intervals, each of which has the same angle, to obtain m×n sub-areas, where n is a positive integer greater than 1;

[0021] In each of the sub-regions, the amplitude of each pixel is calculated, and the amplitude of each pixel is weighted to obtain a characteristic value of the sub-region;

[0022] The feature values ​​of each of the sub-regions are combined to obtain the feature vector of the feature point.

[0023] As a further improvement of the present invention, the calculation formula of the amplitude is specifically: Where (x, y) is the coordinate of the pixel point, σ is the scale space factor, and They are The difference in horizontal and vertical directions, in and are the first-order differences of the Gaussian pyramid along the horizontal direction and the vertical direction respectively.

[0024] As a further improvement of the present invention, obtaining the specific location of the defect based on the grain image and the preset image includes:

[0025] Decomposing the grain image into a light reflection model, and performing logarithmic transformation on the light reflection model;

[0026] Performing Fourier transformation on the result of the logarithmic transformation and performing frequency domain filtering;

[0027] Convert the result of the frequency domain filtering into the spatial domain to obtain a target image;

[0028] The specific location of the defect is obtained according to the target image and the preset image.

[0029] As a further improvement of the present invention, obtaining the specific location of the defect according to the target image and the preset image includes:

[0030] Converting the target image and the preset image into grayscale images to obtain a grayscale target image and a grayscale preset image;

[0031] Subtracting the grayscale target image from the grayscale preset image to obtain a difference between the two images;

[0032] The specific location of the defect is determined based on the difference.

[0033] As a further improvement of the present invention, determining the specific position of the defect according to the difference includes:

[0034] If the difference is a negative number, comparing the negative number with a preset threshold;

[0035] If the negative number is greater than or equal to the preset threshold, the absolute value of the negative number is taken; if the negative number is less than the preset threshold, the negative number is set to zero to obtain an updated difference;

[0036] The specific location of the defect is determined according to the updated difference.

[0037] As a further improvement of the present invention, determining the type of the defect according to the morphological characteristics of the defect includes:

[0038] Determining the complexity and texture information of the defect according to the grayscale entropy of the defect;

[0039] Calculate the class moment of the defect according to the outline size of the defect;

[0040] Calculating the integrity of the defect based on the area and contour length of the defect;

[0041] The type of the defect is determined according to the complexity and texture information, the class regularity and the completeness.

[0042] As a further improvement of the present invention, the calculation formula of the grayscale entropy is specifically: gary_entropy = -∑(P(i)×log2(P(i))), where i is the grayscale value of the defect, and P(i) is the normalized frequency of the grayscale value i.

[0043] The present invention provides a semiconductor wafer defect detection system based on machine vision, comprising:

[0044] Acquisition module: used to acquire the image of the semiconductor wafer to be inspected;

[0045] Matching module: used for extracting a grain image from the wafer image;

[0046] Positioning module: used for obtaining the specific location of defects based on the grain image and a preset image, wherein the preset image is an image of a grain without defects;

[0047] Classification module: used to determine the morphological characteristics of the defect according to the specific location of the defect, and determine the type of the defect according to the morphological characteristics of the defect.

[0048] The present invention acquires images of semiconductor wafers through machine vision products, extracts grain images from the wafer images, and performs image processing on the grain images, thereby realizing detection of wafer defects based on the results of image processing. That is, machines are used instead of human eyes to make measurements and judgments, thereby improving detection efficiency and making up for the shortcomings of manual detection methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of the steps of the method of the present invention;

[0050] Figure 2 It is a schematic diagram of the process of extracting grain images;

[0051] Figure 3 It is a schematic diagram of a circular area;

[0052] Figure 4 It is a schematic diagram of interval angle;

[0053] Figure 5 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations of the technical solution of the present invention.

[0055] The term "and / or" in the following text is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0056] like Figure 1 As shown, the embodiment of the present application provides a method for detecting semiconductor wafer defects based on machine vision, comprising:

[0057] Acquire an image of a semiconductor wafer to be inspected;

[0058] Extracting a grain image based on a wafer image;

[0059] Based on the grain image and the preset image, the specific location of the defect is obtained, and the preset image is a grain image without defects;

[0060] Determine the morphological characteristics of the defects according to the specific location of the defects;

[0061] Determine the type of defect based on the morphological characteristics of the defect.

[0062] The semiconductor wafer defect detection method based on machine vision provided in this embodiment collects images of the semiconductor wafer to be detected through machine vision products, such as CMOS and CCD, converts the images of the semiconductor wafer to be detected into image signals, transmits them to a dedicated image processing device, and obtains wafer defect morphology information through image processing, thereby ultimately realizing the detection of wafer defects. In addition, the detection method based on machine vision can improve the productivity of repetitive tasks, solve the problems of misjudgment or slow detection speed caused by visual fatigue in manual detection, and make up for the shortcomings of the existing technology.

[0063] Wafers are an important component of semiconductors. There are tens of thousands of grains in a wafer. The detection of wafer defects in industrial production is actually the detection of grain defects. There are street background areas around each grain in the wafer. These background areas are irrelevant areas. In the image processing process, if the calculation of irrelevant areas is included, it will lead to large amount of calculation and long calculation time. Therefore, it is necessary to extract the complete grain image while removing the street background. Therefore, further, such as Figure 2 As shown, this embodiment provides a method for extracting a grain image from a wafer image, which improves the computational efficiency while avoiding additional interference from irrelevant areas, specifically including:

[0064] Convolving the wafer image with the Gaussian kernel to obtain a Gaussian pyramid, and subtracting adjacent layers of the Gaussian pyramid to obtain a Gaussian difference pyramid;

[0065] Compare each pixel in the Gaussian difference pyramid with all its adjacent pixels to determine the extreme points;

[0066] The Gaussian difference pyramid is expanded at the extreme point using the Taylor formula, and the offset of the extreme point is calculated. The extreme point with an offset less than the preset value is taken as the feature point.

[0067] A feature vector is constructed according to the feature points, and the feature vector is matched with a preset feature vector to obtain a grain image. The preset feature vector is calculated according to the preset image.

[0068] Specifically, the Gaussian convolution kernel is Where (x, y) is the coordinate of the pixel in the wafer image, σ is the scale space factor, also known as the blur coefficient. The larger the σ value, the blurrier the image. The wafer image is convolved with the Gaussian kernel to obtain the Gaussian pyramid. Where I(x, y) is the wafer image, and the Gaussian difference pyramid is obtained by subtracting the adjacent layers of the Gaussian pyramid (Gaussian images of adjacent scales) in, represents convolution, and k represents the scale.

[0069] All adjacent pixel points include 8 adjacent points of the same scale and 9×2 points corresponding to the upper and lower adjacent scales, a total of 26 points. The extreme point can be determined by the Gaussian difference method (Difference of Gaussian, GOG), that is, by the difference of Gaussian functions, the Gaussian filtering results under different parameters are subtracted to find the extreme point. Then, the Gaussian difference pyramid is expanded at the extreme point using the Taylor formula, and the Taylor expansion is derived and set equal to zero to obtain the offset of the extreme point. When the offset of the extreme point in any dimension (x or y or σ) is less than the preset value, the extreme point is used as a feature point, where there are multiple extreme points and feature points. Preferably, the preset value can be set to 0.5.

[0070] Furthermore, this embodiment provides a method for constructing a feature vector, including:

[0071] Taking the feature point as the center, select a circular area within a preset radius, and the circular area is located inside the area where the grain is located;

[0072] Divide the circular area into m circular areas, where m is a positive integer greater than 1;

[0073] Divide the circular area into n intervals, each with the same angle, to obtain m×n sub-areas, where n is a positive integer greater than 1;

[0074] In each sub-region, the amplitude of each pixel is calculated, and the amplitude of each pixel is weighted to obtain the characteristic value of the sub-region;

[0075] The eigenvalues ​​of each sub-region are combined to obtain the eigenvector of the feature point.

[0076] The preset radius needs to be adjusted according to the size of the grain. In order to avoid interference from irrelevant areas, the circular area should be located inside the area where the grain is located.

[0077] For example, Figure 3 As shown in , a circular area with a radius of 10 can be taken in the 21×21 neighborhood around the feature point with the feature point as the center, and the circular area can be divided into 4 circular ring areas from the inside to the outside. Since the closer the pixels around the feature point are to the feature point, the greater the influence on the calculated eigenvalue, the step size of each circular ring sub-area is 4, 3, 2, and 1 respectively. Figure 4 As shown, the circular area is further divided into 8 intervals, each with an angle of 45 degrees, to obtain 32 sub-areas. In each sub-area, the amplitude of each pixel is calculated.

[0078] The specific calculation formula of the amplitude is: Where (x, y) is the coordinate of the pixel point, σ is the scale space factor, and They are The difference in horizontal and vertical directions, in are the first-order differences of the Gaussian pyramid in the horizontal and vertical directions, S is the SOBEL operator, S x and S y are the partial derivatives of S in the horizontal and vertical directions respectively.

[0079] Then the amplitude of each pixel in each sub-region is weighted, and the weight parameter is Where (m0,n0) represents the feature point, (m,n) represents the pixel point, W m,n is the weight coefficient of the pixel, s is the step size of the circular area where the pixel is located, a is the radius of the circular area where the pixel is located, and k is the scale of the pixel, that is, the number of layers of the feature point in the Gaussian pyramid. The ratio of the step size s and the radius a of the circular area where the pixel is located is used to describe the position of the circular area where the pixel is located relative to the central feature point area, and further refine the influence weight of pixels at different positions on the central feature point. Then, the eigenvalues ​​of each sub-area are combined to obtain a 4×8=32-dimensional feature vector of the feature point, and the feature vector is normalized to obtain the final feature vector.

[0080] The final feature vector is matched with the preset feature vector to obtain the grain image. Since there are multiple feature points, the calculated feature vectors are also multiple. The matching method is to calculate the Euclidean distance between each feature vector and the preset feature vector. If the distance is less than the preset value, the match is successful. The circular area corresponding to the successfully matched feature vector is the extracted grain image.

[0081] The method of generating feature vectors in the prior art is usually to generate a 128-dimensional feature vector using a 16×16 square area in the neighborhood of the feature point. However, the high-dimensional feature vector requires a lot of calculation when matching, which will seriously affect the matching speed and cause large matching errors. Therefore, this embodiment uses a circular area to generate feature vectors. The circle has rotational invariance. While fully retaining the feature information, there is no need to determine the main direction of the feature point, and the dimension of the feature vector can be effectively adjusted as needed. It can meet the extraction needs of different wafer images, and then accurately extract the grain image based on the original wafer image, and perform further image analysis based on the grain image. Compared with the analysis directly on the original image, it avoids the interference of the street background area and can more accurately identify the wafer defect area.

[0082] In the field of machine vision, when using CMOS or CCD to collect images, it is inevitable that they will be affected by lighting, collision with environmental objects and image transmission, so that most of the collected images contain noise. Therefore, image denoising and enhancement are necessary steps in image processing. Existing image denoising and enhancement methods usually use Gaussian filtering, median filtering and other spatial domain-based processing methods, but spatial domain-based processing methods are prone to robustness problems. Therefore, further, this embodiment provides a method for denoising and image enhancement of grain images, which uses a frequency domain-based processing method that focuses more on the overall rather than local details, specifically including:

[0083] Decomposing the grain image into a light reflection model, and performing logarithmic transformation on the light reflection model;

[0084] The result of logarithmic transformation is Fourier transformed and frequency-domain filtered;

[0085] The result of frequency domain filtering is converted into spatial domain to obtain the target image.

[0086] Specifically, for the grain image f(x, y), it can be decomposed into a lighting reflection model:

[0087] f(x, y) = i(x, y) r(x, y)

[0088] Where i(x, y) represents the illumination component at position (x, y) in f(x, y), and r(x, y) represents the reflection component at position (x, y) in f(x, y). Logarithmic transformation of the above illumination reflection model is obtained:

[0089] z(x,y)=lnf(x,y)=lni(x,y)+lnr(x,y)

[0090] Among them, z(x, y) is the illumination reflection model after logarithmic transformation, and then the above results are Fourier transformed to obtain:

[0091] F{z(x, y)}=F{lni(x, y)}+F{lnr(x, y)}

[0092] Wherein, F{z(x, y)} represents z(x, y) after Fourier transformation. The above formula can be simplified as: Z(u, v) = F i (u,v)+F r (u, v), where F i (u, v) is the result of Fourier transform of lni(x, y), F r (u, v) is the result of Fourier transform of lnr(x, y), Z(u, v) is the result of Fourier transform of z(x, y), and frequency domain filtering of Z(u, v) yields:

[0093] B(u,v)=H(u,v)Z(u,v)=H(u,v)F i (u,v)+H(u,v)F r (u, v)

[0094] Where B(u, v) represents the result of frequency domain filtering of Z(u, v), and H(u, v) represents the frequency domain filter, which converts the result of the filtering operation of the above filter H(u, v) from the frequency domain to the spatial domain:

[0095] b(x, y) = F -1 B(u,v)=F -1 (H(u,v)F i (u,v))+F -1 (H(u,v)F r (u,v)

[0096]

[0097] The obtained g(x, y) is the target image.

[0098] The method provided in this embodiment removes noise in the image and enhances image details by performing frequency domain processing on the image. At the same time, the effect of uneven illumination will be reflected in the low-frequency information of the spectrum, but this information is not taken into account during calculation. Therefore, it is not affected by uneven illumination and has higher robustness. The obtained target image will serve as the basis for subsequent wafer defect detection.

[0099] Furthermore, this embodiment provides a method for comparing a target image with a preset image to obtain a specific location of a defect, including:

[0100] Convert the target image and the preset image into grayscale images to obtain a grayscale target image and a grayscale preset image;

[0101] Subtract the grayscale target image from the grayscale preset image to obtain the difference between the two images;

[0102] Based on the difference, the specific location of the defect is determined.

[0103] Specifically, in order to simplify the image storage method while retaining the detailed features of the image, the color image can be gray-scaled. The processing method can adopt the average method, that is, the average value of the three channels in the RGB space at the pixel point is used as the grayscale value of the pixel point.

[0104] The preset image is a template grain image without any defects. If the target image does not contain defects, the grayscale information at the corresponding positions of the two images is the same. If the target image contains defects, except for the grayscale information at the defective position, the grayscale information at other corresponding positions is the same.

[0105] Then, the grayscale target image is subtracted from the grayscale preset image, that is, the two corresponding pixels in the spatial position are subtracted, where the grayscale preset image is used as the subtracted image and the grayscale target image is used as the subtrahend image, and the grayscale value difference is calculated. Since the grayscale value of each pixel is in the interval [0, 255], the grayscale value difference of each corresponding pixel is between [-255, 255]. If the difference result is between [0, 255], it is called a positive difference, and if the difference result is between [-255, 0), it is called a negative difference. That is, when the grayscale value of the pixel of the subtracted image is less than the grayscale value of the subtrahend image, or the former is darker than the latter, a negative difference will appear. The negative difference exceeds the protection range of the grayscale value, which will cause the difference information of the two images to be lost. Therefore, a certain method can be used to convert the negative value into a positive value, thereby retaining the difference information.

[0106] Common conversion methods include the zeroing method and the absolute value method. The zeroing method is to record the negative difference as zero, and the absolute value method is to take the absolute value of the negative difference to obtain the positive difference. However, in actual process production, due to reasons such as the manufacturing process, the pixels at corresponding positions in the two images are not completely consistent, which may cause difference interference when making the difference. Therefore, further, this embodiment provides a method for updating the negative difference, including:

[0107] If the difference is a negative number, the negative number is compared with a preset threshold;

[0108] If the negative number is greater than or equal to the preset threshold, the absolute value of the negative number is taken; if the negative number is less than the preset threshold, the negative number is set to zero to obtain the updated difference;

[0109] Based on the updated difference, the specific location of the defect is determined.

[0110] The preset threshold is determined according to detection requirements.

[0111] The specific formula is:

[0112]

[0113] Where res1(x, y) is the difference obtained by the zeroing method, res2(x, y) is the difference obtained by the absolute value method, res(x, y) is the updated difference, c(x, y) represents the grayscale preset image, g1(x, y) represents the grayscale target image, and K is the preset threshold.

[0114] The method provided in this embodiment determines the specific location of the defect in the grain image by calculating the difference between the two images at the corresponding positions, and provides a method for updating the negative value, which can retain the complete difference information and make the detection result more accurate.

[0115] Furthermore, after the specific location of the defect is determined, the defective pixels at adjacent locations can be connected to obtain the regional morphology of the defect, and the defect type can be determined based on the morphological characteristics of the defect, including:

[0116] Determine the complexity and texture information of the defect based on the grayscale entropy of the defect;

[0117] Calculate the class moment of the defect according to the outline size of the defect;

[0118] Calculate defect completeness based on defect area and contour length;

[0119] Determine the type of defect based on complexity and texture information, regularity and completeness.

[0120] At present, the defect types of grains can be roughly divided into broken chips, arc discharge ablation, foreign dirt, corrosion pits and redundant structures. Among them, the main causes of broken chips are chip fragments, scratches, etc., and their appearance is mostly angular; arc discharge ablation is mostly caused by high concentrations of charged ions accumulating charges in the non-conductive area on the surface of the grains through the arc in the plasma environment, and its appearance is mostly burnt with blurred edges; foreign dirt is generally fine particles, some of which are irregular, and some are strip-shaped or circular; corrosion pit defects generally appear as holes on the metal base or circuit, and are caused by developer corrosion during etching and ACT groove corrosion; redundant structure defects generally appear as abnormal bridging of adjacent structures or circuit structures appearing in the wrong place. The morphology of this defect is similar to that of a normal circuit structure.

[0121] Therefore, it can be concluded that the type of grain defect is strongly correlated with the morphological characteristics of the grain defect. The type of grain defect can be determined based on the morphological characteristics of the grain defect, and then the type of wafer defect can be obtained to detect wafer defects. Since a wafer contains multiple grains, a wafer may contain multiple defects.

[0122] Specifically, the calculation formula of grayscale entropy is:

[0123] gary_entropy=-∑(P(i)×log2(P(i)))

[0124] Where i is the gray value of the grain defect position, P(i) is the normalized frequency of gray value i, and gray entropy is a parameter to measure the gray information of the grain defect area, which can represent the complexity and texture information of the grain defect. A higher gray entropy value indicates that the grain defect has more information and texture changes, while a lower gray entropy value indicates that the grain defect is smoother and has less texture changes. For arc discharge ablation, external dirt, corrosion pits and other types, the gray entropy is generally higher.

[0125] The calculation formula of the class moment is as follows:

[0126] rectangularity=area / rectangularity_area

[0127] Among them, rectangularity is the class squareness, area is the area of ​​the defect, and rectangularity_area represents the minimum rectangular outline size that encloses the grain defect area. Its value range is generally 0-1. The class squareness parameter can effectively represent the similarity between the defect outline and the rectangle. For some defect types, such as redundant structural defects, the class squareness is generally higher.

[0128] The calculation formula for completeness is:

[0129] roundness=4π×area / perimeter 2

[0130] Among them, roundness indicates completeness, perimeter indicates defect contour length, and the completeness value range is usually between 0 and 1. The closer the completeness value is to 1, the closer the contour shape of the grain defect is to a circle. When the edge of the defect is obviously jagged, its completeness is low, indicating that the grain may have a broken chip type defect.

[0131] Several judgment methods can also be combined to determine the defect type. For example, if the integrity of the defect is high and the grayscale entropy is high, the defect is more likely to be a redundant structure. If the integrity is low and the grayscale entropy is also low, it means that the defect is more likely to be a broken chip.

[0132] The method provided in this embodiment can preliminarily determine the type of defect based on the morphological characteristics of the defect, but cannot accurately determine the specific type of the defect. If the exact type needs to be further defined, a more accurate calculation model, such as a neural network, needs to be used.

[0133] Further, such as Figure 5 As shown, the embodiment of the present application provides a semiconductor wafer defect detection system based on machine vision, comprising:

[0134] Acquisition module: used to acquire the image of the semiconductor wafer to be inspected;

[0135] Matching module: used to extract the die image based on the wafer image;

[0136] Positioning module: used to obtain the specific location of defects based on the grain image and the preset image. The preset image is the grain image without defects.

[0137] Classification module: used to determine the morphological characteristics of the defect according to the specific location of the defect, and determine the type of defect based on the morphological characteristics of the defect.

[0138] The acquisition module is specifically a machine vision device, such as CMOS and CCD, and the matching module, comparison module, positioning module and classification module are located in the image processing device. After the acquisition module acquires the image, it transmits the image to the image processing device for image analysis.

[0139] The semiconductor wafer defect detection method and system based on machine vision provided in the embodiments of the present application obtain a target image by extracting grain images and performing image denoising and enhancement processing, and determine the defect position in the target image based on a preset image, and finally preliminarily judge the defect type based on the morphological characteristics of the defect, thereby improving the detection speed and making up for the shortcomings of manual detection.

[0140] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0144] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A semiconductor wafer defect detection method based on machine vision, characterized in that: include: Acquire an image of a semiconductor wafer to be inspected; extracting a grain image based on the wafer image; Based on the grain image and a preset image, a specific location where a defect exists is obtained, wherein the preset image is an image of a grain without defects; Determining the morphological characteristics of the defect according to the specific location of the defect; The type of the defect is determined according to the morphological characteristics of the defect.

2. The method for detecting semiconductor wafer defects based on machine vision according to claim 1, characterized in that: The step of extracting a grain image based on the wafer image comprises: Convolving the wafer image with a Gaussian kernel to obtain a Gaussian pyramid, and subtracting adjacent layers of the Gaussian pyramid to obtain a Gaussian difference pyramid; Compare each pixel point in the Gaussian difference pyramid with all its adjacent pixels to determine extreme value points; The Gaussian difference pyramid is expanded at the extreme point using the Taylor formula, and the offset of the extreme point is calculated, and the extreme point where the offset is less than a preset value is taken as a feature point; A feature vector is constructed according to the feature points, and the feature vector is matched with a preset feature vector to obtain the grain image, wherein the preset feature vector is calculated according to the preset image.

3. The method for detecting semiconductor wafer defects based on machine vision according to claim 2, characterized in that: Constructing a feature vector according to the feature points includes: Taking the feature point as the center, a circular area within a preset radius is selected, and the circular area is located inside the area where the grain is located; Divide the circular area into m circular areas, where m is a positive integer greater than 1; Divide the circular area into n intervals, each of which has the same angle, to obtain m×n sub-areas, where n is a positive integer greater than 1; In each of the sub-regions, the amplitude of each pixel is calculated, and the amplitude of each pixel is weighted to obtain a characteristic value of the sub-region; The feature values ​​of each of the sub-regions are combined to obtain the feature vector of the feature point.

4. The method for detecting semiconductor wafer defects based on machine vision according to claim 3, characterized in that: The calculation formula of the amplitude is specifically: Where (x, y) is the coordinate of the pixel point, σ is the scale space factor, and They are The difference in horizontal and vertical directions, in and are the first-order differences of the Gaussian pyramid along the horizontal direction and the vertical direction respectively.

5. The method for detecting semiconductor wafer defects based on machine vision according to claim 1, characterized in that: The obtaining of a specific location of a defect based on the grain image and the preset image includes: Decomposing the grain image into a light reflection model, and performing logarithmic transformation on the light reflection model; Performing Fourier transformation on the result of the logarithmic transformation and performing frequency domain filtering; Convert the result of the frequency domain filtering into the spatial domain to obtain a target image; The specific location of the defect is obtained according to the target image and the preset image.

6. The method for detecting semiconductor wafer defects based on machine vision according to claim 5, characterized in that: According to the target image and the preset image, the specific location of the defect is obtained, including: Converting the target image and the preset image into grayscale images to obtain a grayscale target image and a grayscale preset image; Subtracting the grayscale target image from the grayscale preset image to obtain a difference between the two images; The specific location of the defect is determined based on the difference.

7. The method for detecting semiconductor wafer defects based on machine vision according to claim 6, characterized in that: Determining the specific location of the defect according to the difference includes: If the difference is a negative number, comparing the negative number with a preset threshold; If the negative number is greater than or equal to the preset threshold, the absolute value of the negative number is taken; if the negative number is less than the preset threshold, the negative number is set to zero to obtain an updated difference; The specific location of the defect is determined according to the updated difference.

8. The method for detecting semiconductor wafer defects based on machine vision according to claim 1, characterized in that: Determining the type of the defect according to the morphological characteristics of the defect includes: Determining the complexity and texture information of the defect according to the grayscale entropy of the defect; Calculate the class moment of the defect according to the outline size of the defect; Calculating the integrity of the defect based on the area and contour length of the defect; The type of the defect is determined according to the complexity and texture information, the class regularity and the completeness.

9. The method for detecting semiconductor wafer defects based on machine vision according to claim 8, characterized in that: The calculation formula of the grayscale entropy is specifically: gary_entropy = -∑(P(i)×log2(P(i))), where i is the grayscale value of the defect, and P(i) is the normalized frequency of the grayscale value i.

10. A semiconductor wafer defect detection system based on machine vision, characterized in that: include: Acquisition module: used to acquire the image of the semiconductor wafer to be inspected; Matching module: used for extracting a grain image from the wafer image; Positioning module: used for obtaining the specific location of defects based on the grain image and a preset image, wherein the preset image is an image of a grain without defects; Classification module: used to determine the morphological characteristics of the defect according to the specific location of the defect, and determine the type of the defect according to the morphological characteristics of the defect.

Images