Semiconductor crystal growth defect identification intelligent sorting method

Abstract

The invention discloses an intelligent sorting method for identifying semiconductor crystal growth defects, relates to the technical field of visual identification, and solves the problem of inaccurate identification of defect types in current wafer detection.The method comprises the steps that a defect image of a wafer is standardized according to historical defect data, and a standard defect image of the wafer is obtained through standardization; detecting a defect area of the wafer based on the standard defect image; detecting the binary image of the wafer, and judging the defect type of the wafer according to a detection result; constructing a defect type of the wafer and a corresponding defect judgment rule based on the defect data; the visual model is trained by using the defect type of the wafer and the corresponding defect judgment rule, 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 accurate detection and accurate identification of the defect type of the wafer are realized.

Description

Technical Field

[0001] The present invention belongs to the field of visual recognition technology, and in particular is a method for intelligent sorting of semiconductor crystal growth defects. Background Art

[0002] Wafers are the most fundamental material in semiconductor manufacturing. Typically made of high-purity single-crystal silicon, they appear as thin, flat discs. Wafers serve as the substrate for integrated circuit chip production. Through a complex array of processes, including photolithography, etching, doping, and thin-film deposition, thousands of tiny electronic devices and circuit structures are formed on their surfaces. Wafer diameters typically range from a few inches to twelve inches or even larger, and thicknesses are typically only a few hundred microns. Due to the complex and precise manufacturing process, the quality of the wafer surface directly impacts the performance and yield of the final chip. Surface defects on wafers can cause chip malfunction or even failure, making the detection and analysis of wafer surface defects a critical quality control step in semiconductor manufacturing. Accurately identifying and classifying wafer defects enables timely adjustments to the manufacturing process, improving product yield and stability. As the fundamental material of the semiconductor industry, wafers are crucial for the production of key components in modern electronic devices, computers, communications equipment, and more.

[0003] In existing technologies, wafer inspection typically uses optical microscopes, scanning electron microscopes, or laser scattering techniques. Traditional inspection equipment struggles to distinguish and normalize wafer defect types. When wafer images are used for defect type identification, insufficient wafer image processing can result in poor input image quality, impacting the accuracy of subsequent defect type identification and making it difficult to distinguish complex defect types. Furthermore, a single inspection model struggles to simultaneously account for multiple defect types, leading to significant difficulties in identifying and determining different defects. To this end, the present invention proposes a semiconductor crystal growth defect identification and intelligent sorting method. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a semiconductor crystal growth defect identification and intelligent sorting method.

[0005] The technical problems to be solved by the present invention are: How to intelligently select visual models to detect and identify wafer defect types.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A semiconductor crystal growth defect identification and intelligent sorting method, the method comprising: Step S100, standardizing the defect image of the wafer according to the historical defect data to obtain a standard defect image of the wafer, and detecting the defect area of ​​the wafer based on the standard defect image; Step S200 , inspecting the binary image of the wafer and determining the defect type of the wafer based on the inspection result; Step S300, constructing a defect type of the wafer and a corresponding defect determination rule based on the defect data; Step S400 , using the defect type of the wafer and the corresponding defect judgment rules to train the visual model, and selecting the corresponding visual model as the preferred visual model based on the training result of the visual model and the defect type of the wafer.

[0007] Furthermore, the historical defect data are flat-field defect images and dark-field defect images of the wafer; The flat field defect image is the wafer defect image acquired under normal lighting conditions; The acquisition process of the dark field defect image is as follows: the acquisition environment corresponding to the defect image of the wafer is set to a no-light environment, and then the defect image of the wafer is acquired to obtain the dark field defect image of the wafer.

[0008] Furthermore, the step S100 includes the following sub-steps: Step S101, obtaining historical defect data of a wafer, and cropping all defect images of the historical defect data into fixed length and fixed width at corresponding resolutions; Step S102, constructing a flat-field pixel matrix of the wafer based on the pixel values ​​of all pixels in the flat-field defect image, and constructing a dark-field pixel matrix of the wafer based on the dark-field values ​​of all pixels in the dark-field defect image; Step S103, adding up the pixel values ​​of all pixels in the flat-field pixel matrix and taking the average value to calculate the average pixel value of the flat-field defect image; Step S104, calculating the corrected pixel values ​​of all pixels using a standardized formula based on the pixel values, dark field values, and average pixel values ​​of the flat field defect image; Step S105, constructing an empty matrix corresponding to the defect image, filling the empty matrix with the corrected pixel values ​​of all pixels according to the standard coordinates to obtain a defect matrix of the wafer, and constructing a standard defect image of the wafer based on the corrected pixel values ​​corresponding to all standard coordinates in the defect matrix; Step S106, cutting the wafer portion within the standard defect image, then constructing a detection window for the standard defect image, and obtaining a corrected pixel average of corrected pixel values ​​within the detection window; When the corrected pixel value is greater than or equal to the corrected pixel average, the detection value of the corresponding pixel point is recorded as one; When the corrected pixel value is less than the corrected pixel average, the detection value of the corresponding pixel point is recorded as zero; Step S107 , constructing a binary image of the wafer based on the detection values ​​of the pixel points, merging adjacent pixel points with zero detection values ​​to obtain the defective area of ​​the wafer.

[0009] Furthermore, the step S200 includes the following sub-steps: Step S201: construct a plane coordinate system of the wafer with the center of the wafer as the origin, 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, and obtain the total number of pixels in the defect area of ​​the wafer and the pixel coordinates of all pixels; Step S202, calculating the centroid coordinates corresponding to the centroid of the defect area; Step S203, obtaining the radius of the wafer, and calculating the distance between the centroid coordinates of the defect area and the origin using a distance formula; Step S204: constructing wafer defect data.

[0010] Furthermore, the defect data construction process is specifically as follows: Step S2041, dividing the binary image of the wafer into a fixed number of annular zones, and setting the thickness of each annular zone, and the distance between the center line formed by the midpoint of the annular zone and the origin; Step S2042: Compare the centroid distances of all centroids with the inner diameter corresponding to the inner edge of the annulus and the outer diameter corresponding to the outer edge of the annulus. If the centroid distance is less than or equal to the outer diameter and greater than or equal to the inner diameter, the centroid is determined to belong to the annulus. If the distance between the centroids is greater than the outer diameter or less than the inner diameter, the centroid is determined not to belong to the annulus; Step S2043, inspecting all centroids and annular zones according to the numbering order of the defective areas; If the centroid belongs to the first ring zone, the count value corresponding to the first ring zone is increased by one, and so on, the total number of ring zone centroids corresponding to the first ring zone is obtained, and then the total number of ring zone centroids of all ring zones is detected to obtain the total number of ring zone centroids corresponding to the ring zones; If the centroid does not belong to the first ring, no operation is performed; Step S2044, calculating the annular zone area; Step S2045, dividing the total number of the ring zone centroids of the ring zone by the ring zone area to calculate the ring zone defect density ρk of the corresponding ring zone; Step S2046, summing up the defect densities of all the ring zones and taking the average value to obtain the average defect density of all the ring zones, and then calculating the standard deviation of the defect density of all the ring zones; Step S2047, dividing the defect density standard deviation by the average defect density to calculate the ring defect density distribution value of the wafer; In step S2048 , the ring-shaped defect density, the ring-shaped defect density distribution value, and the average defect density are summarized as defect data of the wafer.

[0011] Furthermore, the defect types of the wafer are center defect, ring defect and edge defect.

[0012] Furthermore, the step S300 includes the following sub-steps: Step S301, traversing all the ring-zone defect density distribution values, obtaining the ring-zone corresponding to the maximum ring-zone defect density distribution value, and taking the center line of the ring-zone as the defect concentration radius of the wafer; Step S302 , constructing a wafer inspection area with the origin as the center and a fixed length as the radius, then obtaining the total number of regional centroids within the inspection area, dividing the total number of regional centroids by the area of ​​the inspection area, and calculating the inspection defect density of the inspection area; Step S303: Divide the detected defect density by the ring-shaped defect density of the ring-shaped zone to calculate the density value of the detection area, and then establish a defect judgment rule for center defects: if the density value is greater than or equal to the density threshold, and the total number of regional centroids in the detection area is greater than or equal to the minimum number of centroids, then the defect type of the wafer is determined to be a center defect; If the density value is less than the density threshold, or the total number of regional centroids in the detection area is less than the minimum number of centroids, no operation is performed; Step S304, determining the annular defect of the annular zone; Step S305 , determining edge defects of the wafer.

[0013] Furthermore, the determination process of the ring defect is specifically as follows: Step S3041, dividing the ring zone defect density of the ring zone corresponding to the defect concentration radius of the wafer by the average defect density to calculate the ring zone density ratio of the wafer; Step S3042, obtaining the defect density of the inner region corresponding to the inner edge of the annular zone, dividing the regional defect density by the average defect density, and calculating the void value of the inner region; Step S3043: Constructing defect determination rules for ring defects: When the ring density ratio of the wafer is greater than or equal to the ring density threshold, and the void value of the inner area is less than or equal to the void threshold, the defect type of the wafer is determined to be a ring defect; When the ring density ratio of the wafer is less than the ring density threshold, or the void value of the inner area is greater than the void threshold, no operation is performed.

[0014] Furthermore, the edge defect determination process is specifically as follows: Step S3051, taking the wafer edge as a reference and a preset length as the wafer edge thickness, obtaining the wafer edge region; Step S3052, calculating and obtaining the edge area of ​​the edge region; Step S3053, obtaining the number of edge centroids of the centroids in the edge region, dividing the number of edge centroids by the area of ​​the edge region, and calculating the edge centroid density of the centroids in the edge region; Step S3054: Divide the edge centroid density by the average defect density to calculate the edge density ratio of the edge area, and then construct the defect judgment rule for edge defects: When the edge density ratio of the edge area is greater than or equal to the edge density threshold, and the number of edge centroids in the edge area is greater than or equal to the minimum number of centroids, the defect type of the wafer is determined to be an edge defect; When the edge density ratio of the edge area is less than the edge density threshold, or the number of edge centroids in the edge area is less than the minimum number of centroids, no operation is performed.

[0015] Furthermore, the step S400 includes the following sub-steps: Step S401: Input the defect types of the wafer and the corresponding defect judgment rules into the first model and the second model for training, respectively, and obtain a first accuracy set corresponding to all defect types after the first model is trained, and a second accuracy set corresponding to all defect types after the second model is trained; Step S402 , matching the first accuracy rates of all defect types in the first accuracy rate set with the second accuracy rates of all defect types in the second accuracy rate set, and then comparing the first accuracy rate and the second accuracy rate of the same defect type; Step S403: When the defect type is a center defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary visual model; When the defect type is a ring defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary 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.

[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. The present invention standardizes the defect image of the wafer based on historical defect data to obtain a standard defect image of the wafer. The defect area of ​​the wafer is detected by the standard defect image. 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 based on the detection result, thereby achieving accurate monitoring of the defect type of the wafer. 2. Construct the defect type of the wafer and the corresponding defect judgment rules based on the defect data, use the defect type of the wafer and the corresponding defect judgment rules to train the visual model, and select the corresponding visual model as the preferred visual model based on the training results of the visual model and the defect type of the wafer. The present invention further increases the accuracy of wafer defect type detection by intelligently selecting the visual model of the wafer. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0018] Figure 1 is a flow chart of the method of the present invention; Figure 2 An example diagram of the ring zone and center of mass of the wafer in the present invention; Figure 3 This is an example diagram of the edge area of ​​the wafer in the present invention; Figure 4 It is a structural diagram of the computer device in the present invention. DETAILED DESCRIPTION

[0019] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] Example 1: Please refer to Figure 1-Figure 3 As shown, the technical solution provided by the present invention is: a method for intelligent sorting of semiconductor crystal growth defects, the method is specifically as follows: Step S100, standardizing the defect image of the wafer according to the historical defect data to obtain a standard defect image of the wafer, and detecting the defect area of ​​the wafer based on the standard defect image; The historical defect data specifically includes flat-field defect images and dark-field defect images of the wafer. Specifically, the flat-field defect image is a defect image of the wafer acquired under normal lighting conditions. The acquisition process of the dark-field defect image is as follows: the acquisition environment corresponding to the defect image of the wafer is set to a no-light environment, and then the defect image of the wafer is acquired to obtain the dark-field defect image of the wafer. The historical defect data of the wafer can be obtained through a historical defect library or a wafer defect dataset. In this embodiment, step S100 includes the following sub-steps: Step S101, obtaining historical defect data of a wafer, and cropping all defect images of the historical defect data into fixed length and fixed width at corresponding resolutions; In a specific implementation, the fixed length of the flat field defect image and the dark field defect image are both 512 pixels, and the fixed width is 512 pixels; Step S102: construct a flat-field pixel matrix I of the wafer based on the pixel values ​​of all pixels 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 pixels in the dark-field defect image. The flat-field pixel matrix is ​​specifically as follows: , where (a, b) is the standard coordinate of all pixels in the flat field defect image, a=1, 2, ..., 512; b=1, 2, ..., 512; I (a, b) is the pixel value corresponding to the pixel, and the pixel value range is [0, 255]. The dark field pixel matrix is ​​as follows: , where (a, b) is the standard coordinate of all pixels in the dark field defect image, J (a, b) is the dark field value corresponding to the pixel, and the range of the dark field value is [0, 255]; It should be specifically noted that, since the image sizes of the flat-field defect image and the dark-field defect image are the same, the standard coordinates within the images are the same; Step S103, adding up the pixel values ​​of all pixels in the flat-field pixel matrix and taking the average value to calculate the average pixel value B of the flat-field defect image; In step S104, the corrected pixel values ​​D(a, b) of all pixels are calculated using a standardized formula based on the pixel values, dark field values, and the average pixel value of the flat field defect image. The formula is as follows: D(a,b)=[I(a,b)-J(a,b)] / [BJ(a,b)]; Step S105, constructing an empty matrix corresponding to the defect image, filling the empty matrix with the corrected pixel values ​​of all pixels according to the standard coordinates to obtain a defect matrix of the wafer, and constructing a standard defect image of the wafer based on the corrected pixel values ​​corresponding to all standard coordinates in the defect matrix; 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; Step S106, cutting the wafer portion within the standard defect image using an edge detection algorithm, then constructing a detection window for the standard defect image, and obtaining a corrected pixel average of corrected pixel values ​​within the detection window; When the corrected pixel value is greater than or equal to the corrected pixel average, the detection value of the corresponding pixel point is recorded as one; When the corrected pixel value is less than the corrected pixel average, the detection value of the corresponding pixel point is recorded as zero; Among them, the edge detection algorithm includes the Sobel edge detection algorithm, the Canny edge detection algorithm and the Roberts edge detection algorithm. In this embodiment, the Canny edge detection algorithm is preferred. In a specific implementation, the size of the detection window can be 3×3 or 5×5; Step S107 , constructing a binary image of the wafer based on the detection values ​​of the pixel points, merging adjacent pixel points with zero detection values ​​to obtain the defective area of ​​the wafer.

[0021] Step S200 , inspecting the binary image of the wafer and determining the defect type of the wafer based on the inspection result; In this embodiment, step S200 includes the following sub-steps: Step S201: 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, a plane coordinate system of the wafer is constructed to obtain the total number N of pixels in the defective area of ​​the wafer and the pixel coordinates (xij, yij) of all pixels, where i is the number of the defective area, i=1, 2, ..., n, and n is the total number of defective areas in the binary image; j is the number of the pixel in the defective area i, j=1, 2, ..., m, and m is a positive integer; In step S202, the centroid coordinates (xi, yi) corresponding to the centroid of the defect area are calculated using the formula. The formula is as follows: ; (Center of mass calculation formula); The centroid is specifically the geometric center of the defect area. The defect area consists of N pixels. The average value of the corresponding pixel coordinates of all pixels is taken to obtain the centroid coordinates of the defect area. Step S203: Obtain the radius R of the wafer, and calculate the distance ri between the centroid coordinates of the defect area and the origin using the distance formula. The formula is as follows: ; The value range of ri is [0, R]; the radius is the number of pixels between the origin of the wafer and the wafer boundary; Step S204: construct defect data of the wafer. The construction process is as follows: Step S2041, as Figure 2 As shown, the binary image of the wafer is divided into a fixed number k of rings, and the thickness of each ring is HD. Then the distance Zk between the center line formed by the midpoint of the kth ring and the origin is: Zk=(k-0.5)×HD, k=1, 2, …, o, o is the total number of rings; Step S2042: Compare the centroid distances of all centroids with the inner diameter corresponding to the inner edge of the annulus and the outer diameter corresponding to the outer edge of the annulus. If the centroid distance is less than or equal to the outer diameter and greater than or equal to the inner diameter, the centroid is determined to belong to the annulus. If the distance between the centroid and the outer diameter is greater than or less than the inner diameter, the centroid is determined not to belong to the annulus; Step S2043, inspecting all centroids and annular zones according to the numbering order of the defective areas; If the centroid belongs to the first ring zone, the count value corresponding to the first ring zone is increased by one, and so on, the total number of ring zone centroids corresponding to the first ring zone is obtained, and then the total number of ring zone centroids of all ring zones is detected to obtain the total number of ring zone centroids corresponding to the ring zones; If the centroid does not belong to the first ring, no operation is performed; In step S2044, the area Ak of ​​the kth annulus is calculated using the formula, which is as follows: Ak=π{(k×HD)²-[(k-1)×HD]²}; Step S2045, dividing the total number of the ring zone centroids of the ring zone by the ring zone area to calculate the ring zone defect density ρk of the corresponding ring zone; In step S2046, the defect densities of all the ring zones are summed and averaged to obtain the average defect density PM of all the ring zones. The standard deviation BC of the defect density of all the ring zones is then calculated using the standard deviation formula. The formula is as follows: ; Step S2047, dividing the defect density standard deviation by the average defect density to calculate the ring defect density distribution value of the wafer; It should be specifically noted that the ring-shaped defect density distribution value is used to determine the uniformity of the distribution of defect areas within the wafer, and the ring-shaped defect density distribution value is greater than or equal to zero; Exemplarily, when the ring-shaped defect density distribution value of the first wafer is equal to one and the ring-shaped defect density distribution value of the second wafer is equal to nine, it is determined that the defect distribution uniformity within the first wafer is better than the defect distribution uniformity within the second wafer; In step S2048 , the ring-shaped defect density, the ring-shaped defect density distribution value, and the average defect density are summarized as defect data of the wafer.

[0022] Step S300, constructing a defect type of the wafer and a corresponding defect determination rule based on the defect data; Among them, the defect types of wafers are specifically center defects, ring defects and edge defects; In this embodiment, step S300 includes the following sub-steps: Step S301, traversing all the ring-zone defect density distribution values, obtaining the ring-zone corresponding to the maximum ring-zone defect density distribution value, and taking the center line of the ring-zone as the defect concentration radius of the wafer; Step S302 , constructing a wafer inspection area with the origin as the center and a fixed length as the radius, then obtaining the total number of regional centroids within the inspection area, dividing the total number of regional centroids by the area of ​​the inspection area, and calculating the inspection defect density of the inspection area; In a specific implementation, the fixed length may be 10% of the wafer radius; Step S303: Divide the detected defect density by the ring-shaped defect density of the ring-shaped zone to calculate the density value of the detection area, and then establish a defect judgment rule for center defects: if the density value is greater than or equal to the density threshold, and the total number of regional centroids in the detection area is greater than or equal to the minimum number of centroids, then the defect type of the wafer is determined to be a center defect; If the density value is less than the density threshold, or the total number of regional centroids in the detection area is less than the minimum number of centroids, no operation is performed; Step S304: determine the annular defect of the annular zone. The determination process is as follows: Step S3041, dividing the ring zone defect density of the ring zone corresponding to the defect concentration radius of the wafer by the average defect density to calculate the ring zone density ratio of the wafer; Step S3042, obtaining the defect density of the inner region corresponding to the inner edge of the annular zone, dividing the regional defect density by the average defect density, and calculating the void value of the inner region; The calculation process of regional defect density is the same as that of detection defect density, so it can be obtained directly; Step S3043: Constructing defect determination rules for ring defects: When the ring density ratio of the wafer is greater than or equal to the ring density threshold, and the void value of the inner area is less than or equal to the void threshold, the defect type of the wafer is determined to be a ring defect; When the ring density ratio of the wafer is less than the ring density threshold, or the void value of the inner area is greater than the void threshold, no operation is performed; Step S305: determining the edge defects of the wafer. The specific determination process is as follows: Step S3051, as Figure 3 As shown, the edge of the wafer is used as a reference, and the preset length YC is used as the edge thickness of the wafer to obtain the edge area of ​​the wafer; Step S3052: Calculate the edge area MJ of the edge area using the formula. The formula is as follows: MJ=π[R²-(R-YC)²]; Step S3053, obtaining the number of edge centroids of the centroids in the edge region, dividing the number of edge centroids by the area of ​​the edge region, and calculating the edge centroid density of the centroids in the edge region; Step S3054: Divide the edge centroid density by the average defect density to calculate the edge density ratio of the edge area, and then construct the defect judgment rule for edge defects: When the edge density ratio of the edge area is greater than or equal to the edge density threshold, and the number of edge centroids in the edge area is greater than or equal to the minimum number of centroids, the defect type of the wafer is determined to be an edge defect; When the edge density ratio of the edge area is less than the edge density threshold, or the number of edge centroids in the edge area is less than the minimum number of centroids, no operation is performed.

[0023] Step S400: training a visual model using the defect type of the wafer and the corresponding defect determination rules, and selecting the corresponding visual model as a preferred visual model based on the training result of the visual model and the defect type of the wafer; Among them, the visual models are specifically the LCGMM model and the PC model; specifically, a real-time wafer image of the wafer can be collected by a high-definition camera; In this embodiment, step S400 includes the following sub-steps: Step S401: Input the defect types of the wafer and the corresponding defect judgment rules into the first model and the second model for training, respectively, and obtain a first accuracy set corresponding to all defect types after the first model is trained, and a second accuracy set corresponding to all defect types after the second model is trained; Among them, training the model using existing rules and data is an existing technology; it should be specifically noted that, for the convenience of description, in this embodiment, the LCGMM model is recorded as the first model and the PC model is recorded as the second model; Step S402 , matching the first accuracy rates of all defect types in the first accuracy rate set with the second accuracy rates of all defect types in the second accuracy rate set, and then comparing the first accuracy rate and the second accuracy rate of the same defect type; Step S403: When the defect type is a center defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary visual model; When the defect type is a ring defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary 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.

[0024] Example 2: The present invention also provides a computer device for running the semiconductor crystal growth defect identification and intelligent sorting method; see Figure 4 The structure diagram of a computer device provided by an embodiment of the present invention is shown, the computer device including a memory and a processor, wherein the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the above-mentioned semiconductor crystal growth defect identification and intelligent sorting method; Furthermore, Figure 4 The computer device shown further includes a communication bus and a communication interface, and the processor, the communication interface and the memory are connected via the communication bus; The memory may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is achieved through at least one communication interface (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used. The communication bus can be an ISA bus, PCI bus or 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, but it does not mean that there is only one communication bus or one type of communication bus; The processor may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above-mentioned method may be completed by hardware integrated logic circuits within the processor or by software instructions. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may 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 gate or transistor logic devices, or discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention may be directly executed by a hardware decoding processor or by a combination of hardware and software modules within the decoding processor. The software modules may be located in storage media well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other such storage media. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the method of the above embodiment in combination with its hardware.

[0025] Embodiment 3: The embodiment of the present invention further provides a computer storage medium, which 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 and intelligent sorting method. The specific implementation can be found in the method embodiment and will not be repeated here. An embodiment of the present invention provides a computer program product for a semiconductor crystal growth defect identification and intelligent sorting method, including a computer storage medium storing program code. The instructions included in the program code can be used to execute the method in the previous method embodiment. The specific implementation can be found in the method embodiment and will not be repeated here.

[0026] Those skilled in the art will 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 aforementioned method embodiment and will not be repeated here.

[0027] In addition, in the description of the embodiments of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0028] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0029] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A semiconductor crystal growth defect identification and intelligent sorting method, characterized in that: Methods include: Step S100, standardizing a defect image of a wafer based on historical defect data to obtain a standard defect image of the wafer, and detecting a defect area of ​​the wafer based on the standard defect image; the historical defect data includes a flat-field defect image and a dark-field defect image of the wafer; Wherein, the step S100 includes the following sub-steps: Step S101, obtaining historical defect data of a wafer, and cropping all defect images of the historical defect data into fixed length and fixed width at corresponding resolutions; Step S102, constructing a flat-field pixel matrix of the wafer based on the pixel values ​​of all pixels in the flat-field defect image, and constructing a dark-field pixel matrix of the wafer based on the dark-field values ​​of all pixels in the dark-field defect image; Step S103, adding up the pixel values ​​of all pixels in the flat-field pixel matrix and taking the average value to calculate the average pixel value of the flat-field defect image; Step S104, calculating the corrected pixel values ​​of all pixels using a standardized formula based on the pixel values, dark field values, and average pixel values ​​of the flat field defect image; Step S105, constructing an empty matrix corresponding to the defect image, filling the empty matrix with the corrected pixel values ​​of all pixels according to the standard coordinates to obtain a defect matrix of the wafer, and constructing a standard defect image of the wafer based on the corrected pixel values ​​corresponding to all standard coordinates in the defect matrix; Step S106, cutting the wafer portion within the standard defect image, then constructing a detection window for the standard defect image, and obtaining a corrected pixel average of corrected pixel values ​​within the detection window; When the corrected pixel value is greater than or equal to the corrected pixel average, the detection value of the corresponding pixel point is recorded as one; When the corrected pixel value is less than the corrected pixel average, the detection value of the corresponding pixel point is recorded as zero; Step S107, constructing a binary image of the wafer based on the detection values ​​of the pixel points, merging adjacent pixels with zero detection values ​​to obtain a defective area of ​​the wafer; Step S200 , inspecting the binary image of the wafer and determining the defect type of the wafer based on the inspection result; Step S300, constructing a defect type of the wafer and a corresponding defect determination rule based on the defect data; Step S400 , using the defect type of the wafer and the corresponding defect judgment rules to train the visual model, and selecting the corresponding visual model as the preferred visual model based on the training result of the visual model and the defect type of the wafer.

2. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 1, characterized in that: The flat field defect image is the wafer defect image acquired under normal lighting conditions; The acquisition process of the dark field defect image is as follows: the acquisition environment corresponding to the defect image of the wafer is set to a no-light environment, and then the defect image of the wafer is acquired to obtain the dark field defect image of the wafer.

3. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 1, characterized in that: The step S200 includes the following sub-steps: Step S201: construct a plane coordinate system of the wafer with the center of the wafer as the origin, 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, and obtain the total number of pixels in the defect area of ​​the wafer and the pixel coordinates of all pixels; Step S202, calculating the centroid coordinates corresponding to the centroid of the defect area; Step S203, obtaining the radius of the wafer, and calculating the distance between the centroid coordinates of the defect area and the origin using a distance formula; Step S204: constructing wafer defect data.

4. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 3, characterized in that: The construction process of the defect data is as follows: Step S2041, dividing the binary image of the wafer into a fixed number of annular zones, and setting the thickness of each annular zone, and the distance between the center line formed by the midpoint of the annular zone and the origin; Step S2042: Compare the centroid distances of all centroids with the inner diameter corresponding to the inner edge of the annulus and the outer diameter corresponding to the outer edge of the annulus. If the centroid distance is less than or equal to the outer diameter and greater than or equal to the inner diameter, the centroid is determined to belong to the annulus. If the distance between the centroids is greater than the outer diameter or less than the inner diameter, the centroid is determined not to belong to the annulus; Step S2043, inspecting all centroids and annular zones according to the numbering order of the defective areas; If the centroid belongs to the first ring zone, the count value corresponding to the first ring zone is increased by one, and so on, the total number of ring zone centroids corresponding to the first ring zone is obtained, and then the total number of ring zone centroids of all ring zones is detected to obtain the total number of ring zone centroids corresponding to the ring zones; If the centroid does not belong to the first ring, no operation is performed; Step S2044, calculating the annular zone area; Step S2045, dividing the total number of the ring zone centroids of the ring zone by the ring zone area to calculate the ring zone defect density of the corresponding ring zone; Step S2046, summing up the defect densities of all the ring zones and taking the average value to obtain the average defect density of all the ring zones, and then calculating the standard deviation of the defect density of all the ring zones; Step S2047, dividing the defect density standard deviation by the average defect density to calculate the ring defect density distribution value of the wafer; In step S2048 , the ring-shaped defect density, the ring-shaped defect density distribution value, and the average defect density are summarized as defect data of the wafer.

5. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 3, characterized in that: The defect types of wafers are center defects, ring defects and edge defects.

6. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 5, characterized in that: The step S300 includes the following sub-steps: Step S301, traversing all the ring-zone defect density distribution values, obtaining the ring-zone corresponding to the maximum ring-zone defect density distribution value, and taking the center line of the ring-zone as the defect concentration radius of the wafer; Step S302 , constructing a wafer inspection area with the origin as the center and a fixed length as the radius, then obtaining the total number of regional centroids within the inspection area, dividing the total number of regional centroids by the area of ​​the inspection area, and calculating the inspection defect density of the inspection area; Step S303: Divide the detected defect density by the ring-shaped defect density of the ring-shaped zone to calculate the density value of the detection area, and then establish a defect judgment rule for center defects: if the density value is greater than or equal to the density threshold, and the total number of regional centroids in the detection area is greater than or equal to the minimum number of centroids, then the defect type of the wafer is determined to be a center defect; If the density value is less than the density threshold, or the total number of regional centroids in the detection area is less than the minimum number of centroids, no operation is performed; Step S304, determining the annular defect of the annular zone; Step S305 , determining edge defects of the wafer.

7. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 6, characterized in that: The determination process of the ring defect is as follows: Step S3041, dividing the ring zone defect density of the ring zone corresponding to the defect concentration radius of the wafer by the average defect density to calculate the ring zone density ratio of the wafer; Step S3042, obtaining the defect density of the inner region corresponding to the inner edge of the annular zone, dividing the regional defect density by the average defect density, and calculating the void value of the inner region; Step S3043: Constructing defect determination rules for ring defects: When the ring density ratio of the wafer is greater than or equal to the ring density threshold, and the void value of the inner area is less than or equal to the void threshold, the defect type of the wafer is determined to be a ring defect; When the ring density ratio of the wafer is less than the ring density threshold, or the void value of the inner area is greater than the void threshold, no operation is performed.

8. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 6, characterized in that: The determination process of the edge defect is as follows: Step S3051, taking the wafer edge as a reference and a preset length as the wafer edge thickness, obtaining the wafer edge region; Step S3052, calculating and obtaining the edge area of ​​the edge region; Step S3053, obtaining the number of edge centroids of the centroids in the edge region, dividing the number of edge centroids by the area of ​​the edge region, and calculating the edge centroid density of the centroids in the edge region; Step S3054: Divide the edge centroid density by the average defect density to calculate the edge density ratio of the edge area, and then construct the defect judgment rule for edge defects: When the edge density ratio of the edge area is greater than or equal to the edge density threshold, and the number of edge centroids in the edge area is greater than or equal to the minimum number of centroids, the defect type of the wafer is determined to be an edge defect; When the edge density ratio of the edge area is less than the edge density threshold, or the number of edge centroids in the edge area is less than the minimum number of centroids, no operation is performed.

9. The semiconductor crystal growth defect identification and intelligent sorting method according to claim 6, characterized in that: The step S400 includes the following sub-steps: Step S401: Input the defect types of the wafer and the corresponding defect judgment rules into the first model and the second model for training, respectively, and obtain a first accuracy set corresponding to all defect types after the first model is trained, and a second accuracy set corresponding to all defect types after the second model is trained; Step S402 , matching the first accuracy rates of all defect types in the first accuracy rate set with the second accuracy rates of all defect types in the second accuracy rate set, and then comparing the first accuracy rate and the second accuracy rate of the same defect type; Step S403: When the defect type is a center defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary visual model; When the defect type is a ring defect, if the first accuracy rate is greater than the second accuracy rate, the first model is selected as the primary visual model; if the first accuracy rate is less than or equal to the second accuracy rate, the second model is selected as the primary 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