Machine vision identification method and device for silicon wafer surface defect, and medium

Abstract

Embodiments of the present application relate to the technical field of semiconductor manufacturing, and disclose a machine vision identification method and device for a silicon wafer surface defect, and a medium. The method comprises: on the basis of an acquired image of a local defect on the surface of a silicon wafer, identifying a defect area to be classified; according to a ratio of a length value to a width value of said defect area, determining an aspect ratio of said defect area, wherein the length value is a value of the maximum distance between edge pixels of said defect area, and the width value is a value of the maximum distance between edge pixels of said defect area in a direction perpendicular to the straight line where the length value is present; and according to the aspect ratio, identifying a defect type represented by said defect area.

Description

A method, device and medium for machine vision recognition of silicon wafer surface defects

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Chinese patent application No. 202310684968.X filed in China on June 9, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The embodiments of the present application relate to the field of semiconductor manufacturing technology, and in particular to a method, device, and medium for machine vision recognition of silicon wafer surface defects. Background Art

[0004] In the semiconductor industry, silicon wafers are generally the raw material for integrated circuits. Various semiconductor devices can be fabricated through processes such as photolithography and ion implantation. The silicon wafer production process typically includes various mechanical surface treatments, such as slicing, grinding, and polishing. These processes inevitably damage the wafer surface, introducing defects such as cracks, scratches, and dislocations on the surface and subsurface of the wafer. These defects affect the performance of the resulting semiconductor product, and consequently, the device lifespan and yield. Therefore, surface defects in silicon wafers have become a key factor in measuring product quality.

[0005] With the development of advanced manufacturing processes, the characteristic line width of integrated circuits continues to decrease, the size of defects on the surface of silicon wafers is smaller and the number is lower. The requirements for defect control on the surface of silicon wafers are getting higher and higher, which leads to a sharp increase in the difficulty of control. Scanning Electron Microscope (SEM) measurement can give an intuitive morphology of defects, but the measurement speed is slow, and it is not fully inspected in the current mass products. However, the next high-end products require more stringent control, which will inevitably lead to an increase in the number of full inspections or sampling of SEMs. Therefore, classifying and counting defects according to the measurement results of SEM is an inevitable requirement for high-end products. However, the current defect classification of SEM results relies only on manual visual inspection, and the efficiency of this solution is far from meeting production needs. In order to control the large number of nanoscale defects on silicon wafers, it is urgent to develop a defect machine automatic recognition system to realize the rapid machine recognition of the main surface defect types, and to achieve efficient classification, quantity and density distribution calculation and other requirements.

[0006] Summary of the Invention

[0007] In view of this, the embodiments of the present application hope to provide a machine vision recognition method for silicon wafer surface defects; it can convert the collected defect image into a grayscale matrix, identify the defect area based on the pixel grayscale value, and accurately extract and judge the aspect ratio of the defect area's external features, thereby realizing rapid identification and classification of defects.

[0008] The technical solution of the embodiment of the present application is implemented as follows:

[0009] In a first aspect, an embodiment of the present application provides a method for machine vision recognition of silicon wafer surface defects, comprising:

[0010] Identify defect areas to be classified based on the collected images of local defects on the silicon wafer surface;

[0011] determining an aspect ratio of the defect area to be classified according to a ratio of a length value to a diameter value of the defect area to be classified, wherein the length value of the defect area to be classified is a value at which the distance between edge pixels of the defect area to be classified is the maximum, and the diameter value of the defect area to be classified is a value at which the distance between edge pixels of the defect area to be classified is the maximum in a direction perpendicular to a straight line on which the length value lies;

[0012] The defect type represented by the defect region to be classified is identified according to the aspect ratio.

[0013] In a second aspect, an embodiment of the present application provides a machine vision device for identifying surface defects of a silicon wafer, the device comprising:

[0014] Acquisition part, feature extraction part, classification and recognition part; among them,

[0015] The acquisition part is configured to identify the defect area to be classified based on the collected image of the local defect on the silicon wafer surface;

[0016] The feature extraction portion is configured to determine an aspect ratio of the defect region to be classified based on a ratio of a length value to a diameter value of the defect region to be classified, wherein the length value of the defect region to be classified is a value at which a distance between edge pixels of the defect region to be classified is the largest, and the diameter value of the defect region to be classified is a value at which a distance between edge pixels of the defect region to be classified is the largest in a direction perpendicular to a straight line on which the length value lies;

[0017] The classification and identification portion is configured to identify the defect type represented by the defect region to be classified according to the aspect ratio.

[0018] In a third aspect, an embodiment of the present application provides a computer storage medium, which stores a machine vision recognition program for silicon wafer surface defects. When the machine vision recognition program for silicon wafer surface defects is executed by at least one processor, it implements the steps of the machine vision recognition method for silicon wafer surface defects described in the first aspect.

[0019] The embodiments of the present application provide a method, device, and medium for machine vision recognition of silicon wafer surface defects. By using the aspect ratio as an external feature of the defect area, defects can be more accurately identified and classified, further reducing the steps for final defect identification, thereby achieving rapid classification of major defects on the silicon wafer surface and improving the monitoring level of high-end products. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG1 is a schematic flow chart of a method for machine vision identification of silicon wafer surface defects provided in an embodiment of the present application;

[0021] FIG2 is an example of a common silicon wafer surface defect image provided by an embodiment of the present application;

[0022] FIG3( a ) is a grayscale image of a raised defect provided in an embodiment of the present application;

[0023] FIG3( b ) is a grayscale variation curve of a convex defect provided in an embodiment of the present application;

[0024] FIG3( c ) is a grayscale image of a concave defect provided in an embodiment of the present application;

[0025] FIG3( d ) is a grayscale variation curve of a concave defect provided in an embodiment of the present application;

[0026] FIG4 is a schematic diagram of the positions of linear defective pixels in a matrix according to an embodiment of the present application;

[0027] FIG5 is a flow chart of silicon wafer surface defect classification and identification provided by an embodiment of the present application;

[0028] FIG6 is a schematic diagram of a machine vision recognition device for silicon wafer surface defects provided in an embodiment of the present application;

[0029] FIG7 is a schematic diagram of the hardware structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0031] Optional embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the accompanying drawings show optional embodiments of the present application, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0032] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0033] It should be understood that although the terms "first", "second", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0034] Statistics show that among small-scale defects on silicon wafer surfaces, the majority are fat juts, line juts, polish-induced defects (PID), scratches, particles, pits, residuals, and metal contamination (MID). Each of these defects has distinct characteristics and stems from different process failures. Identifying the type, quantity, and density distribution of various defects on silicon wafers can serve as a control parameter and provide feedback for process improvement.

[0035] Referring to FIG1 , an embodiment of the present application provides a method for machine vision identification of silicon wafer surface defects. By extracting aspect ratio features of silicon wafer surface defects, the defects are quickly machine identified and efficiently classified. This supports in-depth analysis of defect type, quantity, and density distribution, and provides feedback for process improvement, including:

[0036] S101: identifying defect areas to be classified based on the collected images of local defects on the silicon wafer surface;

[0037] S102: determining an aspect ratio of the defect region to be classified according to a ratio of a length value to a diameter value of the defect region to be classified, wherein the length value of the defect region to be classified is a value at which the distance between edge pixels of the defect region to be classified is the largest, and the diameter value of the defect region to be classified is a value at which the distance between edge pixels of the defect region to be classified is the largest in a direction perpendicular to a straight line on which the length value is located;

[0038] S103: Identify the defect type represented by the defect region to be classified according to the aspect ratio.

[0039] It should be noted that images of local defects on the silicon wafer surface can be acquired using an SEM or other acquisition method that meets the requirements for capturing small-scale defects. The aspect ratio, as a dimensional characteristic of the defect area, can more accurately identify and classify defects, further reducing the number of steps required for final defect identification. For example, when identifying the dimensional characteristics of a linear defect with a certain curvature, it is easier to identify it as a non-linear defect using the aspect ratio of the minimum enclosing rectangle.

[0040] Regarding the technical solution shown in FIG1 , in some embodiments, the step of identifying the defective area to be classified based on the collected image of the local defect on the silicon wafer surface includes:

[0041] Convert the collected image of local defects on the silicon wafer surface from a red, green, blue (RGB) pixel matrix to a grayscale matrix;

[0042] Setting a first grayscale threshold and a second grayscale threshold according to the average grayscale value of the grayscale matrix, wherein the first grayscale threshold is smaller than the average grayscale value, and the second grayscale threshold is larger than the average grayscale value;

[0043] Pixels in the grayscale matrix whose grayscale values ​​are less than the first grayscale threshold or greater than the second grayscale threshold are defective pixels, and an area formed by all the defective pixels is used as a defective area to be classified.

[0044] It should be noted that in the grayscale image matrix of a defective silicon wafer, the color of the defective area appears too dark or too bright compared to the normal area, and its pixel grayscale value will be correspondingly lower or higher than the grayscale value of the normal area. Therefore, a reasonable grayscale threshold is set based on the grayscale average value, and the defective area is identified by the abnormal pixel grayscale value.

[0045] Therefore, for the technical solution shown in FIG1 , in some embodiments, identifying the defect type represented by the defect region to be classified according to the aspect ratio includes:

[0046] Corresponding to the aspect ratio of the defect region to be classified being greater than a first aspect ratio threshold, determining that the defect type represented by the defect region to be classified is a first type of defect, where the first type of defect includes a linear protrusion defect, a scratch defect, and a polishing damage defect;

[0047] Corresponding to the aspect ratio of the defect region to be classified being between the first aspect ratio threshold and the second aspect ratio threshold, determining that the defect type represented by the defect region to be classified is a wide convex defect, and the first aspect ratio threshold is greater than the second aspect ratio threshold;

[0048] Corresponding to the aspect ratio of the defect area to be classified being less than the second aspect ratio threshold, it is determined that the defect type represented by the defect area to be classified is a second type of defect, and the second type of defects includes pit defects, granular defects, empty defects, residue defects, metal contamination defects, and crystal native defects.

[0049] In the above implementation manner, in some examples, the setting range of the first aspect ratio threshold is [15, 20], and the setting range of the second aspect ratio threshold is [0, 3].

[0050] It should be noted that, in the defect grayscale image shown in Figure 2, L represents the length value of the defect area to be classified, and W represents the diameter value of the defect area to be classified. The appearance of a linear raised defect is a linear raised defect. Scratch defects generally occur during the grinding or polishing process, when larger hard particles or piece fragments are mixed into the grinding powder or polishing powder, and the mechanical polishing asphalt disk is locally too hard, resulting in obvious scratch marks. The appearance is a linear depression with a certain curvature along the grinding or polishing track. Polishing damage defects are defects introduced by polishing, which appear as linear raised protrusions with a certain curvature. The aspect ratio of these three defects is significantly greater than that of other types of defects. A wide raised defect is a raised defect with a certain width, which can be directly identified based on the aspect ratio. Granular defects are convex and generally round, and may be formed by foreign matter such as silicon slag falling on the surface of the silicon wafer; pit defects are concave pits, and their appearance is also relatively round; residue defects have irregular morphology, generally discontinuous convexities, which may be caused by the residue of grinding slurry etc. on the surface of the silicon wafer due to incomplete cleaning; metal contamination defects are generally honeycomb-shaped pits, and the defects are caused by metal contamination on the surface of the silicon wafer, and the morphology formed by the aggregation of metal impurities; empty defects are attempts to amplify the current area when SEM is looking for defects, and operations such as increasing voltage are used to look for defects, which eventually result in a high-voltage electron beam leaving a clear mark on the surface of the silicon wafer; crystal native defects are pore-shaped depressions formed by the polymerization of vacancies in the silicon crystal during the crystal pulling process and the temperature drop process. These holes are octahedral in shape, and their top view is rhombus. Therefore, the above common defects are classified based on the aspect ratio provided in the embodiment of the present application, and can avoid identifying linear defects with a certain curvature as wide convex defects or second-class defects, and can directly identify wide convex defects by only one feature of the aspect ratio.

[0051] The first type of defects and the second type of defects include multiple types of defects. Furthermore, when the defect type represented by the defect area to be classified is the first type of defect, the method further includes:

[0052] Determine whether the morphological feature of the defect area to be classified is a convex type or a concave type:

[0053] If the morphological feature of the defect region to be classified is a convex type, then the defect type represented by the defect region to be classified is determined to be a third type defect, including a linear convex defect and a polishing damage defect;

[0054] If the morphological feature of the defective region to be classified is a concave type, it is determined that the defect type represented by the defective region to be classified is a scratch defect.

[0055] It should be noted that since scratch defects are concave defects, while linear convex defects and polishing damage defects are convex defects, the convex or concave features of the first type of defects are extracted and further identified. Since the first type of defects are convex, the defect type represented by the defect area to be classified is determined to be a third type of defect, including linear convex defects and polishing damage defects. Since the first type of defects are concave, they are identified as scratch defects. Therefore, the final defect type can be determined for scratch defects through the two feature extraction and judgment.

[0056] The identification of convex or concave features may be determined by the grayscale value of the defective area. Optionally, in the above example, the determination of whether the morphological feature of the defective area to be classified is convex or concave includes:

[0057] Selecting at least one row from the rows containing defective pixels in the grayscale matrix, wherein the leftmost defective pixel in the row is a left edge pixel and the rightmost defective pixel in the row is a right edge pixel;

[0058] For the left edge pixels and the right edge pixels, respectively calculating an average grayscale change rate k1 of the left edge pixels and an average grayscale change rate k2 of the right edge pixels according to a Sobel operator;

[0059] Calculate the grayscale change rate characteristic value deltak=k1-λ1×k2, where λ1 is a positive real number;

[0060] Corresponding to the grayscale change rate characteristic value being greater than 0, it is determined that the morphological feature of the defective area to be classified is a convex type;

[0061] Corresponding to the grayscale change rate characteristic value being less than 0, it is determined that the morphological feature of the defective area to be classified is a concave type.

[0062] Specifically, Figure 3(a) shows a grayscale image of a raised defect. Generally speaking, the grayscale value range of a gray image is from 0 to 255, with white being 255 and black being 0. Raised areas are generally lighter in color, while recessed areas are darker in color. Figure 3(b) shows the change curve of the grayscale array of a row containing defective pixels in the grayscale matrix corresponding to Figure 3(a). It can be seen that the grayscale value of the raised defect area is significantly higher than that of other normal areas. Figure 3(c) shows a grayscale image of a recessed defect. Figure 3(d) shows the change curve of the grayscale value array of a row containing defective areas in the grayscale matrix corresponding to Figure 3(c). It can be seen that the grayscale value of the recessed defect area is significantly lower than that of other normal areas. Based on this phenomenon, the embodiment of the present application uses the method of calculating the grayscale change rate of the edge of the defect area to identify convex or concave features. For the grayscale curve shown in Figure 3(b), the grayscale change rate k1 of the left edge of the defect area is greater than 0, and the grayscale change rate k2 of the right edge of the defect area is less than 0. It can be seen that the grayscale change rate characteristic value deltak = k1-λ1×k2 is greater than 0, and the defect shown in Figure 3(a) is a convex defect. For the grayscale curve shown in Figure 3(d), the grayscale change rate k1 of the left edge of the defect area is less than 0, and the grayscale change rate k2 of the right edge of the defect area is greater than 0. It can be seen that the grayscale change rate characteristic value is less than 0, and the defect shown in Figure 3(c) is a concave defect. The grayscale change rate of the edge of the defect area can be calculated using the Sobel operator. The Sobel operator is a discrete difference operator that takes the weighted difference of the grayscale values ​​of the four areas of the top, bottom, left, and right of each pixel in the image to obtain the first-order gradient of the digital image.

[0063] For the above example, further, corresponding to the defect type represented by the defect area to be classified being a third type defect, the method further includes:

[0064] Uniformly sampling in the length direction of the defect area to be classified to obtain at least three sampling points;

[0065] Based on the row index value and the column index value of the sampling point in the grayscale matrix, respectively calculating the slopes of any two sampling points based on the matrix position to obtain a slope array;

[0066] If the variance of the slope array is less than the variance threshold, it is determined that the defect type represented by the defect area to be classified is a linear convex defect;

[0067] If the variance of the slope array is greater than or equal to a variance threshold, it is determined that the defect type represented by the defect region to be classified is a polishing damage defect.

[0068] Optionally, the variance threshold can be set to 10 -1 .

[0069] Optionally, the number of sampling points is in the range of [5, 8].

[0070] Specifically, when calculating the slope, you can traverse and calculate the slope between any two sampling points to form a slope array, or you can only calculate the slope between adjacent sampling points to form a slope array. ij 、a st 、a pq Represents three sampling points on the defect area, and the slopes between any two sampling points are The variance of the slope array is D. Theoretically, when sampling uniformly on an ideal straight line, D is 0, that is, In actual engineering applications, when D is less than the variance threshold, it is considered to have linear characteristics.

[0071] It should also be noted that metal contamination defects usually have a certain curvature, and the identification of defect linear features can distinguish linear protrusion defects from metal contamination defects.

[0072] For the above embodiment, in some examples, when the defect type represented by the defect region to be classified is a second type defect, the method further includes:

[0073] Calculate and obtain the radius R of the minimum circumscribed circle of the defect area to be classified;

[0074] The number of pixels contained in the defective area to be classified is S;

[0075] Calculate the circularity of the defect area to be classified curve = |π×R 2 -S|;

[0076] Corresponding to the fact that the near-circularity rate is less than the near-circularity rate threshold, it is determined that the defect type represented by the defect area to be classified is a fourth type of defect, including a granular defect and a pit defect;

[0077] Corresponding to the fact that the near-circularity is greater than or equal to the near-circularity threshold, it is determined that the defect type represented by the defect region to be classified is a fifth type of defect, including void defects, residue defects, metal contamination defects, and crystal native defects.

[0078] Optionally, the near-circularity threshold may be set to 0.1.

[0079] It should be noted that the minimum circumscribed circle is a circle that can contain all pixels in the defect area and has the smallest radius.

[0080] For the above example, further, corresponding to the defect type represented by the defect area to be classified being a fourth type defect, the method further includes:

[0081] Determine whether the morphological characteristics of the defect area to be classified are convex or concave:

[0082] If the morphological feature of the defect region to be classified is a concave type, then determining that the defect type represented by the defect region to be classified is a pit defect;

[0083] If the morphological feature of the defect region to be classified is a convex type, it is determined that the defect type represented by the defect region to be classified is a granular defect.

[0084] It should be noted that for granular and pit defects, final defect type identification can be achieved through three-step feature extraction: aspect ratio, near-circularity, and convex-concave. Void defects, residue defects, metal contamination defects, and native crystal defects account for a relatively small proportion of defects encountered in high-end product production and are not further identified for now. Instead, they are statistically analyzed as a single category, the fifth category of defects.

[0085] As shown in Figure 5, the embodiment of the present application uses four features: the aspect ratio of the defect area, the characteristic value of the grayscale change rate of the pixel in the defect area, the multi-point slope variance statistics of the defect area, and the near-circularity of the defect area to perform step-by-step classification and identification, thereby achieving rapid classification of the main defects on the silicon wafer surface. The actual length, width, height (depth) of the defect area can be calculated by the magnification ratio of the SEM to obtain the defect size, thereby improving the monitoring level of high-end products. In some embodiments, defects can also be classified and identified in different orders based on the above four features. The method steps shown in Figure 5 are optional implementation methods that can achieve rapid identification of the final defect type. For example, for granular defects, pit defects, linear protrusion defects, and polishing damage defects, three features can be used for extraction and judgment. For wide protrusion defects, only one feature needs to be extracted to complete the identification. Although the first, second, third, fourth, and fifth types of defects are not the final defect types, as defect types with one or more common characteristics, their statistical characteristics are used to monitor and analyze defects, and feedback guidance can be provided for defect analysis and product production processes.

[0086] Based on the same inventive concept as the above technical solution, referring to FIG6 , a machine vision recognition device 60 for silicon wafer surface defects provided in an embodiment of the present application is shown. The machine vision recognition device 60 for silicon wafer surface defects includes: an acquisition part 601 , a feature extraction part 602 , and a classification and recognition part 603 ; wherein,

[0087] The acquisition part 601 is configured to identify defect areas to be classified based on the collected images of local defects on the surface of the silicon wafer;

[0088] The feature extraction portion 602 is configured to determine the aspect ratio of the defect region to be classified based on a ratio of a length value to a diameter value of the defect region to be classified, wherein the length value of the defect region to be classified is a value at which the distance between edge pixels of the defect region to be classified is the maximum, and the diameter value of the defect region to be classified is a value at which the distance between edge pixels of the defect region to be classified is the maximum in a direction perpendicular to a straight line on which the length value lies;

[0089] The classification and identification part 603 is configured to identify the defect type represented by the defect region to be classified according to the aspect ratio.

[0090] In some examples, the acquiring portion 601 is configured to identify a defective region to be classified based on the collected image of local defects on the silicon wafer surface, including:

[0091] Convert the collected image of local defects on the silicon wafer surface from an RGB pixel matrix to a grayscale matrix;

[0092] Setting a first grayscale threshold and a second grayscale threshold according to the average grayscale value of the grayscale matrix, wherein the first grayscale threshold is smaller than the average grayscale value, and the second grayscale threshold is larger than the average grayscale value;

[0093] Pixels in the grayscale matrix whose grayscale values ​​are less than the first grayscale threshold or greater than the second grayscale threshold are defective pixels, and an area formed by all the defective pixels is used as a defective area to be classified.

[0094] In some examples, the classification and identification portion 603 is configured to identify the defect type represented by the defect region to be classified according to the aspect ratio, including:

[0095] Corresponding to the aspect ratio of the defect region to be classified being greater than a first aspect ratio threshold, determining that the defect type represented by the defect region to be classified is a first type of defect, where the first type of defect includes a linear protrusion defect, a scratch defect, and a polishing damage defect;

[0096] Corresponding to the aspect ratio of the defect region to be classified being between the first aspect ratio threshold and the second aspect ratio threshold, determining that the defect type represented by the defect region to be classified is a wide convex defect, and the first aspect ratio threshold is greater than the second aspect ratio threshold;

[0097] Corresponding to the aspect ratio of the defect area to be classified being less than the second aspect ratio threshold, it is determined that the defect type represented by the defect area to be classified is a second type of defect, and the second type of defects includes pit defects, granular defects, empty defects, residue defects, metal contamination defects, and crystal native defects.

[0098] In some examples, the classification and identification portion 603 is configured such that the first aspect ratio threshold is set in a range of [15, 20] and the second aspect ratio threshold is set in a range of [0, 3].

[0099] In some examples, the classification and identification portion 603 is configured to correspond to the defect type represented by the defect area to be classified being a first type defect, and the method further includes:

[0100] Determine whether the morphological feature of the defect area to be classified is a convex type or a concave type:

[0101] If the morphological feature of the defect region to be classified is a convex type, then the defect type represented by the defect region to be classified is determined to be a third type defect, including a linear convex defect and a polishing damage defect;

[0102] If the morphological feature of the defective region to be classified is a concave type, it is determined that the defect type represented by the defective region to be classified is a scratch defect.

[0103] In some examples, the classification and identification portion 603 is configured to correspond to the defect type represented by the defect area to be classified being a third type defect, and the method further includes:

[0104] Uniformly sampling in the length direction of the defect area to be classified to obtain at least three sampling points;

[0105] Based on the row index value and the column index value of the sampling point in the grayscale matrix, respectively calculating the slopes of any two sampling points based on the matrix position to obtain a slope array;

[0106] If the variance of the slope array is less than the variance threshold, it is determined that the defect type represented by the defect area to be classified is a linear convex defect;

[0107] If the variance of the slope array is greater than or equal to a variance threshold, it is determined that the defect type represented by the defect region to be classified is a polishing damage defect.

[0108] In some examples, the classification identification part 603 is configured such that the variance threshold can be set to 10 -1 .

[0109] In some examples, the classification identification portion 603 is configured so that the number of sampling points is in the range of [5, 8].

[0110] In some examples, the classification and identification portion 603 is configured to correspond to the defect type represented by the defect area to be classified being a second type defect, and the method further includes:

[0111] Calculate and obtain the radius R of the minimum circumscribed circle of the defect area to be classified;

[0112] The number of pixels contained in the defective area to be classified is S;

[0113] Calculate the circularity of the defect area to be classified curve = |π×R 2 -S|;

[0114] Corresponding to the fact that the near-circularity rate is less than the near-circularity rate threshold, it is determined that the defect type represented by the defect area to be classified is a fourth type of defect, including a granular defect and a pit defect;

[0115] Corresponding to the fact that the near-circularity is greater than or equal to the near-circularity threshold, it is determined that the defect type represented by the defect region to be classified is a fifth type of defect, including void defects, residue defects, metal contamination defects, and crystal native defects.

[0116] In some examples, the classification and identification part 603 is configured so that the near-circularity threshold can be set to 0.1.

[0117] In some examples, the classification and identification portion 603 is configured to correspond to the defect type represented by the defect area to be classified being a fourth type defect, and the method further includes:

[0118] Determine whether the morphological characteristics of the defect area to be classified are convex or concave:

[0119] If the morphological feature of the defect region to be classified is a concave type, then determining that the defect type represented by the defect region to be classified is a pit defect;

[0120] If the morphological feature of the defect region to be classified is a convex type, it is determined that the defect type represented by the defect region to be classified is a granular defect.

[0121] In some examples, the classification and identification portion 603 is configured to determine whether the morphological feature of the defective area to be classified is a convex or concave type, including:

[0122] Selecting at least one row from the rows containing defective pixels in the grayscale matrix, wherein the leftmost defective pixel in the row is a left edge pixel and the rightmost defective pixel in the row is a right edge pixel;

[0123] For the left edge pixels and the right edge pixels, respectively calculating the average grayscale change rate k1 of the left edge pixels and the average grayscale change rate k2 of the right edge pixels according to the Sobel operator;

[0124] Calculate the grayscale change rate characteristic value deltak=k1-λ1×k2, where λ1 is a positive real number;

[0125] Corresponding to the grayscale change rate characteristic value being greater than 0, it is determined that the morphological feature of the defective area to be classified is a convex type;

[0126] Corresponding to the grayscale change rate characteristic value being less than 0, it is determined that the morphological feature of the defective area to be classified is a concave type.

[0127] It can be understood that in this embodiment, "part" can be part of a circuit, part of a processor, part of a program or software, etc., and of course it can also be a unit, a module, or a non-modular one.

[0128] In addition, the components in this embodiment may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The above-mentioned integrated units may be implemented in the form of hardware or software functional modules.

[0129] If the integrated unit is implemented as a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, or the portion that contributes to the prior art, or all or part 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.) or a processor to execute all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0130] Therefore, this embodiment provides a computer storage medium, which stores a machine vision recognition program for silicon wafer surface defects. When the machine vision recognition program for silicon wafer surface defects is executed by at least one processor, it implements the steps of the machine vision recognition method for silicon wafer surface defects described in the above technical solution.

[0131] According to the above-mentioned machine vision recognition device 60 for silicon wafer surface defects and computer storage medium, see Figure 7, which shows the specific hardware structure of a computing device 70 provided by an embodiment of the present application that can implement the above-mentioned machine vision recognition device 60 for silicon wafer surface defects. The computing device 70 can be a wireless device, a mobile or cellular phone (including a so-called smart phone), a personal digital assistant (PDA), a video game console (including a video display, a mobile video game device, a mobile video conferencing unit), a laptop computer, a desktop computer, a TV set-top box, a tablet computing device, an e-book reader, a fixed or mobile media player, etc. The computing device 70 includes: a communication interface 701, a memory 702 and a processor 703; the various components are coupled together through a bus system 704. It can be understood that the bus system 704 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 704 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the various buses are labeled as bus system 704 in Figure 5. Among them,

[0132] The communication interface 701 is used to receive and send signals during the process of sending and receiving information between other external network elements;

[0133] The memory 702 is used to store computer programs that can be run on the processor 703;

[0134] The processor 703 is used to execute the steps of the machine vision identification method for silicon wafer surface defects in the aforementioned technical solution when running the computer program, which will not be described in detail here.

[0135] It is understood that the memory 702 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM). The memory 702 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0136] The processor 703 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits or software instructions in the processor 703. The processor 703 may be a general-purpose processor, 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 this application can 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 this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory 702, and the processor 703 reads the information in the memory 702 and, in conjunction with its hardware, completes the steps of the above method.

[0137] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or a combination thereof.

[0138] For software implementation, the techniques described herein can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0139] Specifically, the processor 703 is further configured to execute the steps of the machine vision identification method for silicon wafer surface defects in the aforementioned technical solution when running the computer program, which will not be repeated here.

[0140] It is understandable that the exemplary technical solutions of the aforementioned machine vision recognition device 60 for silicon wafer surface defects and the computing device 70 are based on the same concept as the technical solution of the aforementioned machine vision recognition method for silicon wafer surface defects. Therefore, any details not described in detail in the aforementioned technical solutions of the machine vision recognition device 60 for silicon wafer surface defects and the computing device 70 can be referred to the description of the technical solution of the aforementioned machine vision recognition method for silicon wafer surface defects. This embodiment of the present application does not elaborate on this.

[0141] It should be noted that the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0142] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for machine vision recognition of silicon wafer surface defects, comprising: Identify defect areas to be classified based on the collected images of local defects on the surface of the silicon wafer; Determining the aspect ratio of the defect area to be classified according to the ratio of the length value to the diameter value of the defect area to be classified, wherein the length value of the defect area to be classified is the value of the maximum distance between edge pixels of the defect area to be classified, and the diameter value of the defect area to be classified is the value of the maximum distance between edge pixels of the defect area to be classified in a direction perpendicular to the straight line where the length value is located; The defect type represented by the defect area to be classified is identified according to the aspect ratio.

2. The method according to claim 1, wherein: The method of identifying defect areas to be classified based on the collected images of local defects on the surface of the silicon wafer includes: Convert the collected image of local defects on the silicon wafer surface from an RGB pixel matrix to a grayscale matrix; Setting a first grayscale threshold and a second grayscale threshold according to an average grayscale value of the grayscale matrix, wherein the first grayscale threshold is less than the average grayscale value, and the second grayscale threshold is greater than the average grayscale value; Pixels in the grayscale matrix whose grayscale values ​​are less than the first grayscale threshold or greater than the second grayscale threshold are defective pixels, and an area formed by all the defective pixels is used as a defective area to be classified.

3. The method according to claim 2, wherein: The step of identifying the defect type represented by the defect region to be classified according to the aspect ratio includes: Corresponding to the aspect ratio of the defect area to be classified being greater than a first aspect ratio threshold, determining that the defect type represented by the defect area to be classified is a first type of defect, wherein the first type of defect includes a linear protrusion defect, a scratch defect, and a polishing damage defect; Corresponding to the aspect ratio of the defect region to be classified being between the first aspect ratio threshold and the second aspect ratio threshold, determining that the defect type represented by the defect region to be classified is a wide convex defect, and the first aspect ratio threshold is greater than the second aspect ratio threshold; Corresponding to the aspect ratio of the defect area to be classified being less than the second aspect ratio threshold, it is determined that the defect type represented by the defect area to be classified is a second type of defect, and the second type of defects includes pit defects, granular defects, empty defects, residue defects, metal contamination defects, and crystal native defects.

4. The method according to claim 3, wherein: Corresponding to the defect type characterized by the defect area to be classified being a first type of defect, the method further includes: Determine whether the morphological feature of the defect area to be classified is a convex type or a concave type: If the morphological feature of the defect region to be classified is a convex type, it is determined that the defect type represented by the defect region to be classified is a third type defect, including a linear convex defect and a polishing damage defect; If the morphological feature of the defect region to be classified is a concave type, it is determined that the defect type represented by the defect region to be classified is a scratch defect.

5. The method according to claim 4, wherein: Corresponding to the defect type characterized by the defect area to be classified being a third type defect, the method further includes: Uniformly sampling in the length direction of the defect area to be classified to obtain at least three sampling points; Based on the row index values ​​and column index values ​​of the sampling points in the grayscale matrix, respectively calculate the slopes of any two sampling points based on the matrix positions to obtain a slope array; Corresponding to the variance of the slope array being less than the variance threshold, it is determined that the defect type represented by the defect area to be classified is a linear convex defect; If the variance of the slope array is greater than or equal to the variance threshold, it is determined that the defect type represented by the defect region to be classified is a polishing damage defect.

6. The method according to claim 3, wherein: Corresponding to the defect type characterized by the defect area to be classified being a second type defect, the method further includes: Calculate and obtain the radius R of the minimum circumscribed circle of the defect area to be classified; The number of pixels contained in the defective area to be classified is S; Calculate the circularity curve of the defect area to be classified = |π×R 2 -S|; Corresponding to the fact that the near-circularity rate is less than the near-circularity rate threshold, it is determined that the defect type represented by the defect area to be classified is a fourth type of defect, including a granular defect and a pit defect; Corresponding to the fact that the near-circularity is greater than or equal to the near-circularity threshold, it is determined that the defect type represented by the defect area to be classified is a fifth type of defect, including empty defects, residue defects, metal contamination defects, and crystal native defects.

7. The method according to claim 6, wherein: Corresponding to the defect type characterized by the defect area to be classified being a fourth type defect, the method further includes: Determine whether the morphological characteristics of the defect area to be classified are convex or concave: Corresponding to the morphological feature of the defect region to be classified being a concave type, determining that the defect type represented by the defect region to be classified is a pit defect; If the morphological feature of the defect region to be classified is a convex type, it is determined that the defect type represented by the defect region to be classified is a granular defect.

8. The method according to claim 4 or 7, wherein: The determining whether the morphological feature of the defective area to be classified is a convex type or a concave type comprises: Selecting at least one row from the rows containing defective pixels in the grayscale matrix, wherein the leftmost defective pixel in the row is a left edge pixel, and the rightmost defective pixel in the row is a right edge pixel; For the left edge pixels and the right edge pixels, respectively calculate and obtain the average grayscale change rate k1 of the left edge pixels and the average grayscale change rate k2 of the right edge pixels according to the Sobel operator; Calculate the grayscale change rate characteristic value deltak=k1-λ1×k2, where λ1 is a positive real number; Corresponding to the grayscale change rate characteristic value being greater than 0, it is determined that the morphological feature of the defect area to be classified is a convex type; Corresponding to the grayscale change rate characteristic value being less than 0, it is determined that the morphological feature of the defective region to be classified is a concave type.

9. A machine vision recognition device for silicon wafer surface defects, the device comprising: Acquisition part, feature extraction part, classification and recognition part; among them, The acquisition part is configured to identify the defect area to be classified based on the collected image of the local defect on the surface of the silicon wafer; The feature extraction part is configured to determine the aspect ratio of the defect area to be classified according to the ratio of the length value to the diameter value of the defect area to be classified, wherein the length value of the defect area to be classified is the value of the maximum distance between the edge pixels of the defect area to be classified, and the diameter value of the defect area to be classified is the value of the maximum distance between the edge pixels of the defect area to be classified in the direction perpendicular to the straight line where the length value is located; The classification and identification part is configured to identify the defect type represented by the defect area to be classified according to the aspect ratio.

10. A computer storage medium, wherein the computer storage medium stores a program for machine vision recognition of silicon wafer surface defects, wherein the program for machine vision recognition of silicon wafer surface defects, when executed by at least one processor, implements the steps of the method for machine vision recognition of silicon wafer surface defects as claimed in any one of claims 1 to 8.

Images