A semiconductor crystal growth defect identification and intelligent sorting method

Abstract

The application discloses a kind of semiconductor long crystal defect identification intelligent sorting methods, it is related to visual identification technical field, solve the problem of current wafer detection to the inaccurate identification of defect type, method is: according to historical defect data to the defect image of wafer is standardized, standard defect image of wafer is obtained by standardization, based on standard defect image detection obtains the defect area of wafer;The binary image of wafer is detected, and the defect type of wafer is determined according to the detection result;Based on defect data, the defect type of wafer and corresponding defect determination rule are constructed;The visual model is trained using the defect type of wafer and corresponding defect determination rule, according to the training result of visual model and the defect type of wafer, corresponding visual model is selected as the preferred visual model, the accurate detection and accurate identification of wafer defect type are realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of visual identification, and specifically relates to a semiconductor long crystal defect identification intelligent sorting method. BACKGROUND

[0002] A wafer is the most basic material in the semiconductor manufacturing process, usually made of high-purity single crystal silicon, and presents a thin and flat disc shape. The wafer is a carrier for integrated circuit chip production, and after multiple complex processes such as photolithography, etching, doping, and thin film deposition on its surface, thousands of tiny electronic devices and circuit structures are formed. The diameter of the wafer is generally from a few inches to twelve inches or even larger, and the thickness is usually only a few hundred microns. Due to the complex and precise manufacturing process, the quality of the wafer surface directly affects the performance and yield of the final chip. Surface defects of the wafer can cause abnormal function or even failure of the chip, so detection and analysis of wafer surface defects are key quality control links in semiconductor manufacturing. By accurately identifying and classifying wafer defects, manufacturing processes can be adjusted in a timely manner to improve product yield and stability. As a basic material for the semiconductor industry, the wafer is an important guarantee for the production of key components of modern electronic equipment, computers, communication equipment, etc.

[0003] In the prior art, the detection of the wafer usually uses optical microscopy, scanning electron microscopy or laser scattering technology, and the traditional detection equipment is difficult to distinguish and regularize the defect types of the wafer. When selecting to identify the defect types of the wafer through the wafer image, there is a problem of insufficient wafer image processing, which leads to poor quality of the input image and affects the accuracy of subsequent defect type identification, making it difficult to distinguish complex defect types, and a single detection model is difficult to simultaneously consider multiple defect types, leading to great difficulty in identifying and determining different defects.

[0004] Therefore, the present application provides a semiconductor long crystal defect identification intelligent sorting method. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application aims to provide a semiconductor long crystal defect identification intelligent sorting method.

[0006] The technical problem to be solved by the present application is:

[0007] How to intelligently select a visual model to detect and identify the defect types of the wafer.

[0008] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0009] A semiconductor long crystal defect identification intelligent sorting method, the method comprising:

[0010] In step S100, the wafer defect image is standardized according to the historical defect data, a standard wafer defect image is obtained by standardization, and a wafer defect area is detected based on the standard wafer defect image;

[0011] In step S200, the binary image of the wafer is detected, and the defect type of the wafer is determined according to the detection result;

[0012] In step S300, the defect type of the wafer and the corresponding defect determination rule are constructed based on the defect data;

[0013] In step S400, the visual model is trained using the defect type of the wafer and the corresponding defect determination rule, and the corresponding visual model is selected as the preferred visual model according to the training result of the visual model and the defect type of the wafer.

[0014] Further, the historical defect data is a flat field defect image and a dark field defect image of the wafer;

[0015] The flat field defect image is a wafer corresponding defect image collected under normal lighting conditions;

[0016] The collection process of the dark field defect image is: the defect image of the wafer is collected in a light-free environment, and the dark field defect image of the wafer is collected.

[0017] Further, the step S100 includes the following sub-steps:

[0018] In step S101, the historical defect data of the wafer is obtained, and all defect images of the historical defect data are cut into a fixed length and a fixed width according to the resolution;

[0019] In step S102, a flat field pixel matrix of the wafer is constructed according to the pixel value of all pixel points in the flat field defect image, and a dark field pixel matrix of the wafer is constructed based on the dark field value of all pixel points in the dark field defect image;

[0020] In step S103, the pixel values of all pixel points in the flat field pixel matrix are summed and averaged to calculate the average pixel value of the flat field defect image;

[0021] In step S104, the corrected pixel value of all pixel points is calculated by a standardization formula according to the pixel value, the dark field value and the average pixel value of the flat field defect image;

[0022] In step S105, a defect image corresponding empty matrix is constructed, the corrected pixel value of all pixel points is filled into the empty matrix according to the standard coordinates, a wafer defect matrix is obtained, and a standard wafer defect image is constructed based on the corrected pixel value of all standard coordinates in the wafer defect matrix;

[0023] Step S106: Cut the wafer portion within the standard defect image, then construct a detection window for the standard defect image, and obtain the average value of the corrected pixel values ​​within the detection window.

[0024] When the corrected pixel value is greater than or equal to the average value of the corrected pixels, the detection value of the corresponding pixel is recorded as one.

[0025] When the corrected pixel value is less than the average corrected pixel value, the detection value of the corresponding pixel is recorded as zero;

[0026] Step S107: Construct a binarized image of the wafer based on the detection values ​​of the pixels, and merge adjacent pixels with a detection value of zero to obtain the defect area of ​​the wafer.

[0027] Further, step S200 includes the following sub-steps:

[0028] Step S201: Using the center of the wafer as the origin, any direction as the positive X-axis, and the direction perpendicular to the X-axis as the positive Y-axis, construct a planar coordinate system for the wafer to obtain the total number of pixels in the defect area of ​​the wafer and the pixel coordinates of all pixels.

[0029] Step S202: Calculate the centroid coordinates corresponding to the centroid of the defect region;

[0030] Step S203: Obtain the radius of the wafer, and calculate the centroid distance between the centroid coordinates of the defect region and the origin using the distance formula;

[0031] Step S204: Construct defect data for the wafer.

[0032] Furthermore, the process of constructing the defect data is as follows:

[0033] Step S2041: Divide the binarized image of the wafer into a fixed number of rings and set the thickness of each ring. Then the distance between the center line formed by the midpoint of the ring and the origin is calculated.

[0034] Step S2042: Compare the centroid distance of all centroids with the inner diameter corresponding to the inner edge of the ring and the outer diameter corresponding to the outer edge of the ring. If the centroid distance is less than or equal to the outer diameter and greater than or equal to the inner diameter, then the centroid belongs to the ring.

[0035] If the distance between the centroids is greater than the outer diameter or less than the inner diameter, then the centroid is determined not to belong to the annulus.

[0036] Step S2043: Inspect all centroids and rings according to the numbering order of the defective areas;

[0037] If the centroid belongs to the first ring, the corresponding count value of the first ring is increased by one, and so on, to obtain the total number of ring centroids corresponding to the first ring, and then the total number of ring centroids of all rings is detected to obtain the total number of ring centroids corresponding to the ring;

[0038] If the centroid does not belong to the first ring, no operation is performed;

[0039] In step S2044, the ring area of the ring is calculated.

[0040] In step S2045, the total number of ring centroids of the ring is divided by the ring area to calculate the ring defect density ρk corresponding to the ring;

[0041] In step S2046, the ring defect densities of all rings are summed and averaged to obtain the average defect density of all rings, and then the standard deviation of the defect density of all rings is calculated;

[0042] In step S2047, the standard deviation of the defect density is divided by the average defect density to calculate the ring defect density distribution value of the wafer;

[0043] In step S2048, the ring defect density, the ring defect density distribution value and the average defect density are summarized as the defect data of the wafer.

[0044] Further, the defect type of the wafer is a central defect, a ring defect and an edge defect.

[0045] Further, the step S300 includes the following sub-steps:

[0046] In step S301, all ring defect density distribution values are traversed to obtain the ring corresponding to the maximum ring defect density distribution value, and the center line of the ring is taken as the defect concentration radius of the wafer;

[0047] In step S302, a detection area of the wafer is constructed with the origin as the center and a fixed length as the radius, and then the total number of region centroids in the detection area is obtained, and the detection defect density of the detection area is calculated by dividing the total number of region centroids by the area of the detection area;

[0048] In step S303, the detection defect density is divided by the ring defect density of the ring to calculate the density value of the detection area, and then a defect judgment rule of the central defect is constructed: if the density value is greater than or equal to a density threshold value, and the total number of region centroids in the detection area is greater than or equal to a minimum centroid number, it is determined that the defect type of the wafer is a central defect;

[0049] If the density value is less than the density threshold value, or the total number of region centroids in the detection area is less than the minimum centroid number, no operation is performed;

[0050] Step S304, judging the ring-shaped defect of the ring belt;

[0051] Step S305, judging the edge defect of the wafer.

[0052] Further, the judging process of the ring-shaped defect is specifically as follows:

[0053] Step S3041, dividing the defect concentration radius of the wafer by the average defect density of the ring belt defect of the ring belt to calculate the ring belt density ratio of the wafer;

[0054] Step S3042, obtaining the area defect density corresponding to the inner side area formed by the inner side edge of the ring belt, dividing the area defect density by the average defect density to calculate the hole value of the inner side area;

[0055] Step S3043, constructing the defect judging rule of the ring-shaped defect:

[0056] When the ring belt density ratio of the wafer is greater than or equal to the ring belt density threshold value, and the hole value of the inner side area is less than or equal to the hole threshold value, the defect type of the wafer is judged as the ring-shaped defect;

[0057] When the ring belt density ratio of the wafer is less than the ring belt density threshold value, or the hole value of the inner side area is greater than the hole threshold value, no operation is performed.

[0058] Further, the judging process of the edge defect is specifically as follows:

[0059] Step S3051, taking the wafer edge as the reference and the preset length as the edge thickness of the wafer to obtain the edge area of the wafer;

[0060] Step S3052, calculating the edge area of the edge area;

[0061] Step S3053, obtaining the edge centroid number of the edge centroid in the edge area, dividing the edge centroid number by the edge area to calculate the edge centroid density of the edge centroid in the edge area;

[0062] Step S3054, dividing the edge centroid density by the average defect density to calculate the edge density ratio of the edge area, and then constructing the defect judging rule of the edge defect:

[0063] When the edge density ratio of the edge area is greater than or equal to the edge density threshold value, and the edge centroid number of the edge area is greater than or equal to the minimum centroid number, the defect type of the wafer is judged as the edge defect;

[0064] When the edge density ratio of the edge area is less than the edge density threshold value, or the edge centroid number of the edge area is less than the minimum centroid number, no operation is performed.

[0065] Further, the step S400 comprises the following sub-steps:

[0066] Step S401, input the defect type of the wafer and the corresponding defect determination rule into the first model and the second model respectively for training, and obtain the first accuracy set corresponding to all defect types after the first model training, and the second accuracy set corresponding to all defect types after the second model training;

[0067] Step S402, match the first accuracy of all defect types in the first accuracy set with the second accuracy of all defect types in the second accuracy set, and then compare the first accuracy and the second accuracy of the same defect type;

[0068] Step S403, when the defect type is a central defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model;

[0069] When the defect type is a ring defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model;

[0070] When the defect type is an edge defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model.

[0071] To sum up, due to the adoption of the above technical scheme, the beneficial effects of the present application are:

[0072] 1. According to the historical defect data, the wafer defect image is standardized to obtain a standard defect image of the wafer, and the defect area of the wafer is detected through the standard defect image, and a binary image of the wafer is also obtained, the binary image of the wafer is detected, and the defect type of the wafer is determined according to the detection result, so that the wafer defect type is accurately monitored;

[0073] 2. The defect type of the wafer and the corresponding defect determination rule are constructed based on the defect data, the defect type of the wafer and the corresponding defect determination rule are used to train the visual model, and the corresponding visual model is selected as the preferred visual model according to the training result of the visual model and the defect type of the wafer, and the accuracy of the wafer defect type detection is further increased by intelligently selecting the visual model of the wafer. BRIEF DESCRIPTION OF DRAWINGS

[0074] For the convenience of those skilled in the art to understand, the present application will be further described below with reference to the accompanying drawings.

[0075] Figure 1 A flow chart of the method of the present application;

[0076] Figure 2 An example of the ring belt and the center of mass of the wafer in the present application;

[0077] Figure 3 An example of the edge region of the wafer in the present application;

[0078] Figure 4 A structural schematic diagram of the computer device in the present application. DETAILED DESCRIPTION

[0079] The technical solutions of the present application will be described below in conjunction with the embodiments, obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0080] Embodiment 1: please refer to Figures 1-3 The technical solutions provided by the present application are as follows: a semiconductor crystal growth defect recognition and intelligent sorting method, the method is as follows:

[0081] Step S100, standardizing the defect image of the wafer according to the historical defect data, obtaining the standard defect image of the wafer by standardization, and detecting the defect area of the wafer based on the standard defect image;

[0082] The historical defect data is specifically the flat field defect image and the dark field defect image of the wafer. Specifically, the flat field defect image is specifically the corresponding defect image of the wafer collected under normal lighting conditions. The collection process of the dark field defect image is as follows: the defect image of the wafer is collected in a lightless environment, and then the defect image of the wafer is collected, and the dark field defect image of the wafer is collected. The historical defect data of the wafer can be obtained through the historical defect library or the wafer defect data set;

[0083] In this embodiment, the step S100 includes the following sub-steps:

[0084] Step S101, obtaining the historical defect data of the wafer, and cutting all the defect images of the historical defect data into fixed length and fixed width according to the resolution;

[0085] In specific implementation, the fixed length of the flat field defect image and the dark field defect image is 512 pixels, and the fixed width is 512 pixels;

[0086] Step S102, construct a flat field pixel matrix I of the wafer according to the pixel values of all pixel points in the flat field defect image, and construct a dark field pixel matrix J of the wafer based on the dark field values of all pixel points in the dark field defect image, and the flat field pixel matrix is specifically as follows:

[0087] wherein (a, b) is specifically the standard coordinates of all pixel points in the flat field defect image, a = 1, 2, …, 512; b = 1, 2, …, 512; I(a, b) is specifically the pixel value corresponding to the pixel point, and the pixel value ranges from 0 to 255;

[0088] The dark field pixel matrix is specifically as follows:

[0089] wherein (a, b) is specifically the standard coordinates of all pixel points in the dark field defect image, and J(a, b) is specifically the dark field value corresponding to the pixel point, and the dark field value ranges from 0 to 255;

[0090] It needs to be specifically explained that since the image size of the flat field defect image and the dark field defect image is the same, the standard coordinates in the image are the same;

[0091] Step S103, add and sum the pixel values of all pixel points in the flat field pixel matrix, and then take the average value to calculate the average pixel value B of the flat field defect image;

[0092] Step S104, according to the pixel value, the dark field value and the average pixel value of the flat field defect image, the corrected pixel value D(a, b) of all pixel points is calculated through a standardization formula, and the formula is specifically as follows:

[0093] D(a, b) = [I(a, b) - J(a, b)] / [B - J(a, b)];

[0094] Step S105, construct a defect image corresponding empty matrix, fill the corrected pixel value of all pixel points into the empty matrix according to the standard coordinates, obtain a defect matrix of the wafer, and construct a standard defect image of the wafer based on the corrected pixel value corresponding to all standard coordinates in the defect matrix;

[0095] wherein the matrix size of the empty matrix is the same as that of the flat field pixel matrix I and the dark field pixel matrix, and the matrix size of the empty matrix is 512*512;

[0096] Step S106, the wafer part in the standard defect image is cut through an edge detection algorithm, and then a detection window of the standard defect image is constructed, and the corrected pixel average value of the corrected pixel value in the detection window is obtained;

[0097] When the corrected pixel value is greater than or equal to the corrected pixel average value, the detection value of the corresponding pixel point is recorded as one;

[0098] When the corrected pixel value is less than the corrected pixel average value, the detection value of the corresponding pixel point is recorded as zero;

[0099] The edge detection algorithm includes a Sobel edge detection algorithm, a Canny edge detection algorithm and a Roberts edge detection algorithm, and the Canny edge detection algorithm is preferred in the embodiment; in specific implementation, the size of the detection window can be 3*3 or 5*5;

[0100] In step S107, a binary image of the wafer is constructed based on the detection value of the pixel point, and adjacent pixel points with a detection value of zero are merged to obtain a defect region of the wafer.

[0101] In step S200, the binary image of the wafer is detected, and the type of the defect of the wafer is determined according to the detection result;

[0102] In the embodiment, the step S200 includes the following sub-steps:

[0103] In step S201, a plane coordinate system of the wafer is constructed with the center of the wafer as the origin (x0, y0), any direction as the positive direction of the X axis, and the direction perpendicular to the X axis as the positive direction of the Y axis, the total number N of pixel points in the defect region of the wafer and the pixel point coordinates (xij, yij) of all pixel points are obtained, wherein i is the number of the defect region, i=1, 2, …, n, n is the total number of the defect regions in the binary image; j is the number of the pixel point in the defect region i, j=1, 2, …, m, m is a positive integer;

[0104] In step S202, the centroid coordinates (xi, yi) of the centroid of the defect region are obtained by calculation, and the formula is as follows:

[0105] ;

[0106] (the centroid calculation formula);

[0107] The centroid is specifically the geometric center of the defect region, the defect region is composed of N pixel points, and the centroid coordinates of the defect region are obtained by averaging the pixel point coordinates of all pixel points;

[0108] In step S203, the radius R of the wafer is obtained, and the centroid distance ri between the centroid coordinates of the defect region and the origin is calculated by the distance formula, and the formula is as follows:

[0109] ;

[0110] Wherein, the value range of ri is [0, R]; the radius is specifically the pixel point number between the origin of the wafer and the wafer boundary;

[0111] Step S204, constructing the defect data of the wafer, and the construction process is specifically as follows:

[0112] Step S2041, as shown in the figure, the binary image of the wafer is equally divided into a fixed number k of annular zones, and the thickness of each annular zone is HD, then the distance Zk between the center line formed by the midpoint of the kth annular zone and the origin is: Figure 2

[0113] Zk=(k-0.5)×HD, k=1, 2, …, o, o is the total number of annular zones;

[0114] Step S2042, comparing the centroid distance of all centroids with the inner side diameter corresponding to the inner side edge of the annular zone and the outer side diameter corresponding to the outer side edge, if the centroid distance is less than or equal to the outer side diameter and greater than or equal to the inner side diameter, it is determined that the centroid belongs to the annular zone;

[0115] If the centroid distance is greater than the outer side diameter or less than the inner side diameter, it is determined that the centroid does not belong to the annular zone;

[0116] Step S2043, detecting all centroids and annular zones according to the numbering order of the defect regions;

[0117] If the centroid belongs to the first annular zone, the corresponding counting value of the first annular zone is increased by one, and so on, to obtain the total number of annular zone centroids corresponding to the first annular zone, and then the total number of annular zone centroids of all annular zones is detected to obtain the total number of annular zone centroids corresponding to the annular zone;

[0118] If the centroid does not belong to the first annular zone, no operation is performed;

[0119] Step S2044, calculating the annular area Ak of the kth annular zone by the formula, and the formula is specifically as follows:

[0120] Ak=π{(k×HD)²-[(k-1)×HD]²};

[0121] Step S2045, dividing the total number of annular zone centroids by the annular area to calculate the annular defect density ρk corresponding to the annular zone;

[0122] Step S2046, adding up the annular defect densities of all annular zones and taking the average to obtain the average defect density PM of all annular zones, and then calculating the defect density standard deviation BC of all annular zones by the standard deviation formula, and the formula is specifically as follows:

[0123] ;

[0124] ​Step S2047, dividing the standard deviation of the defect density by the average defect density to obtain a ring defect density distribution value of the wafer;

[0125] It should be specifically noted that the ring defect density distribution value is used to determine the uniformity of the distribution of the defect region in the wafer, and the ring defect density distribution value is greater than or equal to zero;

[0126] For example, when the ring defect density distribution value of the first wafer is equal to one and the ring defect density distribution value of the second wafer is equal to nine, it is determined that the uniformity of the defect distribution in the first wafer is better than that in the second wafer;

[0127] Step S2048, the ring defect density, the ring defect density distribution value and the average defect density are summarized as the defect data of the wafer.

[0128] Step S300, constructing a defect type of the wafer and a corresponding defect determination rule based on the defect data;

[0129] The defect type of the wafer is specifically a center defect, a ring defect and an edge defect;

[0130] In this embodiment, the step S300 includes the following sub-steps:

[0131] Step S301, traversing all the ring defect density distribution values to obtain a ring corresponding to a maximum ring defect density distribution value, and taking a center line of the ring as a defect concentration radius of the wafer;

[0132] Step S302, constructing a detection area of the wafer with the origin as the center and a fixed length as the radius, and then obtaining a total number of regional centroids in the detection area, dividing the total number of regional centroids by the area of the detection area to obtain a detection defect density of the detection area;

[0133] In specific implementation, the fixed length can be 10% of the radius of the wafer;

[0134] Step S303, dividing the detection defect density by the ring defect density of the ring to obtain a density value of the detection area, and then constructing a defect determination rule of the center defect: if the density value is greater than or equal to a density threshold value, and the total number of regional centroids in the detection area is greater than or equal to a minimum centroid number, it is determined that the defect type of the wafer is a center defect;

[0135] If the density value is less than the density threshold value, or the total number of regional centroids in the detection area is less than the minimum centroid number, no operation is performed;

[0136] Step S304, determining the ring defect of the ring, and the determination process is specifically as follows:

[0137] Step S3041: Divide the annular defect density of the annular zone corresponding to the defect concentration radius of the wafer by the average defect density to calculate the annular density ratio of the wafer.

[0138] Step S3042: Obtain the defect density of the inner region corresponding to the inner edge of the ring, divide the region defect density by the average defect density, and calculate the void value of the inner region.

[0139] The calculation process for regional defect density is the same as that for detection defect density, so it can be obtained directly.

[0140] Step S3043, construct the defect judgment rules for ring defects:

[0141] When the ring density ratio of a wafer is greater than or equal to the ring density threshold, and the void value in the inner region is less than or equal to the void threshold, the defect type of the wafer is determined to be a ring defect.

[0142] No operation is performed when the zone density ratio of the wafer is less than the zone density threshold, or the void value in the inner region is greater than the void threshold.

[0143] Step S305: Determine the edge defects of the wafer. The determination process is as follows:

[0144] Step S3051, as follows Figure 3 As shown, with the wafer edge as the reference, a preset length YC is used as the edge thickness of the wafer to obtain the edge region of the wafer;

[0145] Step S3052: Calculate the area MJ of the edge region using the formula as follows:

[0146] MJ = π[R² - (R - YC)²];

[0147] Step S3053: Obtain the number of edge centroids within the edge region, divide the number of edge centroids by the area of ​​the edge region, and calculate the edge centroid density within the edge region.

[0148] Step S3054: Divide the edge centroid density by the average defect density to calculate the edge density ratio of the edge region, and then construct the defect judgment rule for edge defects:

[0149] When the edge density ratio of the edge region is greater than or equal to the edge density threshold, and the number of edge centroids in the edge region is greater than or equal to the minimum number of centroids, the defect type of the wafer is determined to be an edge defect.

[0150] No operation is performed when the edge density ratio of the edge region is less than the edge density threshold, or when the number of edge centroids in the edge region is less than the minimum number of centroids.

[0151] Step S400, training the visual model using the wafer defect type and the corresponding defect determination rule, and selecting a corresponding visual model as the preferred visual model according to the training result of the visual model and the wafer defect type;

[0152] The visual model is specifically an LCGMM model and a PC model; specifically, the real-time wafer image of the wafer can be collected by a high-definition camera;

[0153] In the embodiment, the step S400 includes the following sub-steps:

[0154] Step S401, inputting the wafer defect type and the corresponding defect determination rule into the first model and the second model respectively for training, and obtaining a first accuracy set corresponding to all defect types after training of the first model, and a second accuracy set corresponding to all defect types after training of the second model;

[0155] The training of the model by using existing rules and data is prior art; it needs to be specifically explained that, for the convenience of description, the LCGMM model is recorded as the first model and the PC model is recorded as the second model in the embodiment;

[0156] Step S402, matching the first accuracy of all defect types in the first accuracy set with the second accuracy of all defect types in the second accuracy set, and then comparing the first accuracy and the second accuracy of the same defect type;

[0157] Step S403, when the defect type is a center defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model;

[0158] When the defect type is a ring defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model;

[0159] When the defect type is an edge defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model.

[0160] Embodiment 2: The embodiment of the application also provides a computer device for running the semiconductor crystal growth defect recognition and intelligent sorting method; see Figure 4The structural schematic diagram of the computer device provided by the embodiment of the application shown in the figure comprises a memory and a processor, wherein the memory is used for storing one or more computer instructions, and the one or more computer instructions are executed by the processor to realize the above-mentioned semiconductor crystal growth defect identification intelligent sorting method.

[0161] Further, Figure 4 The computer device shown further comprises a communication bus and a communication interface, and the processor, the communication interface and the memory are connected through the communication bus;

[0162] The memory can contain a high-speed random access memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface (which can be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used. The communication bus can be an ISA bus, a PCI bus or an EISA bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the figure, but it does not mean that there is only one communication bus or one type of communication bus;

[0163] The processor can be an integrated circuit chip having a processing capability of signals. In the implementation process, the steps of the above method can be completed by the integrated logic circuit of hardware in the processor or the instruction in the form of software. The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and combines the hardware to complete the steps of the method of the above embodiments.

[0164] Embodiment 3: The embodiment of the present application also provides a computer storage medium, the computer storage medium stores computer executable instructions, when the computer executable instructions are called and executed by a processor, the computer executable instructions cause the processor to implement the above-mentioned semiconductor crystal growth defect identification intelligent sorting method, for specific implementation, please refer to the method embodiment, and here is not repeated.

[0165] The computer program product of the semiconductor crystal growth defect identification intelligent sorting method provided by the embodiment of the present application comprises a computer storage medium storing program codes, and the instructions included in the program codes can be used to execute the method in the foregoing method embodiment. For specific implementation, please refer to the method embodiment, and here is not repeated.

[0166] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and / or device described above can refer to the corresponding process in the foregoing method embodiment, and here is not repeated.

[0167] In addition, in the description of the embodiments of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood in a broad sense, for example, can be fixedly connected, or can be detachably connected, or integrally connected; can be mechanically connected, or can be electrically connected; can be directly connected, or indirectly connected through an intermediate medium; can be internal communication of two elements. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0168] The functions described above, if implemented in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part of the present application which essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0169] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A semiconductor crystal growth defect identification and intelligent sorting method, characterized in that, The method comprises: In step S100, the defect image of the wafer is standardized according to historical defect data, a standard defect image of the wafer is obtained by standardization, and a defect area of the wafer is detected based on the standard defect image; the historical defect data is a flat field defect image and a dark field defect image of the wafer; The step S100 comprises the following sub-steps: In step S101, the historical defect data of the wafer is obtained, and all defect images of the historical defect data are cut into a fixed length and a fixed width according to the corresponding resolution; In step S102, a flat field pixel matrix of the wafer is constructed according to the pixel values of all pixel points in the flat field defect image, and a dark field pixel matrix of the wafer is constructed based on the dark field values of all pixel points in the dark field defect image; In step S103, the average value of the flat field defect image is calculated by adding and averaging the pixel values of all pixel points in the flat field pixel matrix; In step S104, the corrected pixel value of each pixel point is calculated by a standardization formula according to the pixel value, the dark field value and the average pixel value of the flat field defect image; In step S105, a corresponding empty matrix of the defect image is constructed, the corrected pixel value of each pixel point is filled into the empty matrix according to the standard coordinates, a defect matrix of the wafer is obtained, and a standard defect image of the wafer is constructed based on the corrected pixel value corresponding to each standard coordinate in the defect matrix; In step S106, the wafer part in the standard defect image is cut, a detection window of the standard defect image is constructed, and the corrected pixel average value of the corrected pixel value in the detection window is obtained; When the corrected pixel value is greater than or equal to the corrected pixel average value, the detection value of the corresponding pixel point is recorded as one; When the corrected pixel value is less than the corrected pixel average value, the detection value of the corresponding pixel point is recorded as zero; In step S107, a binary image of the wafer is constructed based on the detection value of each pixel point, and the pixel points adjacent to each other and having a detection value of zero are merged to obtain a defect area of the wafer; In step S200, the binary image of the wafer is detected, and the defect type of the wafer is determined according to the detection result; In step S300, the defect type of the wafer and the corresponding defect determination rule are constructed based on the defect data; In step S400, the wafer defect type and the corresponding defect determination rule are used to train the visual model, and the corresponding visual model is selected as the preferred visual model according to the training result of the visual model and the defect type of the wafer.

2. The method according to claim 1, wherein the method is characterized by: The flat field defect image is a corresponding defect image of the wafer collected under normal lighting conditions; The collection process of the dark field defect image is as follows: the defect image collection environment of the wafer is set to a dark environment, and then the defect image of the wafer is collected to obtain the dark field defect image of the wafer.

3. The method according to claim 1, wherein the method is characterized by: The step S200 comprises the following sub-steps: In step S201, the center of the wafer is taken as the origin, any direction is taken as the positive direction of the X-axis, and the direction perpendicular to the X-axis is taken as the positive direction of the Y-axis to construct a plane coordinate system of the wafer, and the total number of pixel points in the defect area of the wafer and the pixel point coordinates of all pixel points are obtained; In step S202, the centroid coordinates of the centroid of the defect area are calculated; In step S203, the radius of the wafer is obtained, and the center distance between the center of mass of the defect area and the origin is calculated by a distance formula; In step S204, the defect data of the wafer is constructed.

4. The method according to claim 3, wherein the method is characterized by, The construction process of the defect data is specifically as follows: In step S2041, the binary image of the wafer is equally divided into a fixed number of annular zones, and the thickness of each annular zone is set, and the distance between the center line formed by the midpoint of the annular zone and the origin; In step S2042, the center distance of all centers of mass is compared with the inner side diameter corresponding to the inner side edge of the annular zone and the outer side diameter corresponding to the outer side edge, if the center distance is less than or equal to the outer side diameter and greater than or equal to the inner side diameter, it is determined that the center of mass belongs to the annular zone; If the center distance is greater than the outer side diameter or less than the inner side diameter, it is determined that the center of mass does not belong to the annular zone; In step S2043, all centers of mass and annular zones are detected according to the numbering order of the defect area; If the center of mass belongs to the first annular zone, the corresponding counting value of the first annular zone is increased by one, and the same is true for the other annular zones, so as to obtain the total number of annular zone centers of mass corresponding to the first annular zone, and then the total number of annular zone centers of mass of all annular zones is detected to obtain the total number of annular zone centers of mass corresponding to each annular zone; If the center of mass does not belong to the first annular zone, no operation is performed; In step S2044, the annular zone area of the annular zone is calculated; In step S2045, the total number of annular zone centers of mass of the annular zone is divided by the annular zone area to calculate the annular zone defect density corresponding to the annular zone; In step S2046, the annular zone defect densities of all annular zones are added and averaged to obtain the average defect density of all annular zones, and then the defect density standard deviation of all annular zones is calculated; In step S2047, the defect density standard deviation is divided by the average defect density to calculate the annular zone defect density distribution value of the wafer; In step S2048, the annular zone defect density, the annular zone defect density distribution value and the average defect density are summarized as the defect data of the wafer.

5. The method according to claim 3, wherein the method further comprises: determining the defect type of the defect according to the defect type information of the defect; and sorting the wafer according to the defect type of the defect. The defect types of the wafer are center defects, annular defects and edge defects.

6. The method according to claim 5, wherein the method further comprises: determining the defect type of the defect based on the defect image; and determining the defect type of the defect based on the defect image. The step S300 includes the following sub-steps: In step S301, all annular zone defect density distribution values are traversed to obtain the annular zone corresponding to the maximum annular zone defect density distribution value, and the center line of the annular zone is taken as the defect concentration radius of the wafer; In step S302, a detection area of the wafer is constructed with the origin as the center and a fixed length as the radius, and then the total number of region centers of mass in the detection area is obtained, and the detection defect density of the detection area is calculated by dividing the total number of region centers of mass by the area of the detection area; In step S303, the detection defect density is divided by the annular zone defect density of the annular zone to calculate the density value of the detection area, and then the defect judgment rule of the center defect is constructed: if the density value is greater than or equal to the density threshold value, and the total number of region centers of mass in the detection area is greater than or equal to the minimum center of mass number, it is determined that the defect type of the wafer is the center defect; If the density value is less than the density threshold value, or the total number of region centers of mass in the detection area is less than the minimum center of mass number, no operation is performed; In step S304, the annular defect of the annular zone is determined; In step S305, the edge defect of the wafer is determined.

7. The method according to claim 6, wherein the method further comprises: determining the defect type of the defect based on the defect image and the defect information; and sorting the wafer based on the defect type. The determination process of the annular defect is specifically as follows: Step S3041, the defect concentration radius of the wafer is divided by the average defect density to obtain the ring belt density ratio of the wafer; Step S3042, the inner side area defect density of the inner side area formed by the inner side edge of the ring belt is obtained, and the inner side area defect density is divided by the average defect density to obtain the hole value of the inner side area; Step S3043, the defect determination rule of the ring defect is constructed: When the ring belt density ratio of the wafer is greater than or equal to the ring belt density threshold value, and the hole value of the inner side area is less than or equal to the hole threshold value, it is determined that the defect type of the wafer is a ring defect; When the ring belt density ratio of the wafer is less than the ring belt density threshold value, or the hole value of the inner side area is greater than the hole threshold value, no operation is performed.

8. The method of claim 6, wherein the method further comprises: determining a defect type of the defect based on the defect type information; and determining a defect location of the defect based on the defect location information. The determination process of the edge defect is as follows: Step S3051, the edge area of the wafer is obtained based on the wafer edge and the preset length as the edge thickness of the wafer; Step S3052, the edge area of the edge area is calculated; Step S3053, the edge center density of the edge center in the edge area is calculated by dividing the edge center number by the edge area; Step S3054, the edge density ratio of the edge area is calculated by dividing the edge center density by the average defect density, and then the defect determination rule of the edge defect is constructed: When the edge density ratio of the edge area is greater than or equal to the edge density threshold value, and the edge center number of the edge area is greater than or equal to the minimum center number, it is determined that the defect type of the wafer is an edge defect; When the edge density ratio of the edge area is less than the edge density threshold value, or the edge center number of the edge area is less than the minimum center number, no operation is performed.

9. The method of claim 6, wherein the method further comprises: determining a defect type of the defect based on the defect type information; and determining a defect location of the defect based on the defect location information. The step S400 includes the following sub-steps: Step S401, the defect type of the wafer and the corresponding defect determination rule are input into the first model and the second model for training, and the first accuracy set corresponding to all defect types after the first model is trained and the second accuracy set corresponding to all defect types after the second model is trained are obtained; Step S402, the first accuracy of all defect types in the first accuracy set is matched with the second accuracy of all defect types in the second accuracy set, and then the first accuracy and the second accuracy of the same defect type are compared; Step S403, when the defect type is a center defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model; When the defect type is a ring defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model; When the defect type is an edge defect, if the first accuracy is greater than the second accuracy, the first model is selected as the preferred visual model; if the first accuracy is less than or equal to the second accuracy, the second model is selected as the preferred visual model.

Images