Method, device and storage medium for image alignment in wafer defect detection

Abstract

The present application discloses a method, device and storage medium for image alignment in wafer defect detection. The method comprises: selecting a reference grain and a grain to be detected from a key area of ​​a wafer; performing image alignment on the first image and the second image initially captured when the grain to be detected and the reference grain are at the same position, respectively, to obtain an initial offset; performing local alignment on the first image and the second image captured after the grain to be detected and the reference grain are at the same position, respectively, based on the initial offset, and calculating a local alignment score and a local alignment offset; and determining the final alignment result according to the local alignment score, the local alignment offset and a double preset threshold value, so as to achieve image alignment in wafer defect detection. By using the scheme of the present application, the amount of calculation can be reduced, the calculation efficiency can be improved, and rapid alignment of images in wafer defect detection can be achieved.

Description

Technical Field

[0001] The present application generally relates to the field of wafer defect detection technology. More specifically, the present application relates to a method, device and computer-readable storage medium for image alignment in wafer defect detection. Background Art

[0002] The detection of patterned wafer defects in semiconductors is an important part of semiconductor manufacturing. It can detect and locate defects, classify defects, and thus improve semiconductor yield. The usual method for wafer defect detection is D2D (Die to Die), or "Random Mode", which is to detect the difference in the image of the die to be detected and the reference die at the same position. In special cases, since some graphics themselves are periodic, the location of the defect can be obtained by the difference between different adjacent periods (such as a pitch), that is, C2C (Cell to Cell), or array mode. In addition, the location of the defect can be obtained by comparing a single image with the design graphic ("GDS") at that position, that is, D2DB (Die to Database). Among them, no matter which detection method is used, precise alignment operations are required, such as aligning the test image in D2D with the reference image, and aligning the test image in D2DB with the design graphic.

[0003] The so-called image alignment refers to obtaining the deviation of the pattern position between two images. The alignment operation requires comparing the reference image with the image to be detected, so it has a large amount of calculation. The computational complexity of the alignment operation is about O(r^2) of the search area radius r, where r is the search area radius. It can be seen that when the search area is large, the amount of calculation required for the alignment operation is large, resulting in low computational efficiency and long alignment time.

[0004] In view of this, there is an urgent need to provide a solution for image alignment in wafer defect detection, so as to reduce the amount of calculation, improve calculation efficiency, and achieve rapid alignment of images in wafer defect detection. Summary of the invention

[0005] In order to at least solve one or more of the technical problems mentioned above, the present application proposes a solution for image alignment in wafer defect detection in multiple aspects.

[0006] In a first aspect, the present application provides a method for image alignment in wafer defect detection, comprising: selecting a reference grain and a grain to be detected from a key area of ​​a wafer; performing image alignment on a first image and a second image initially captured at the same position of the grain to be detected and the reference grain to obtain an initial offset; based on the initial offset, performing local alignment on a first image and a second image captured after the grain to be detected and the reference grain are respectively at the same position, and calculating a local alignment score and a local alignment offset; and determining a final alignment result based on the local alignment score, the local alignment offset and dual preset thresholds to achieve image alignment in wafer defect detection.

[0007] In one embodiment, locally aligning the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position based on the initial offset includes: selecting a corresponding local search area in the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position based on the initial offset; and locally aligning the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position within the corresponding local search area.

[0008] In another embodiment, wherein the dual preset thresholds include a high threshold and a low threshold, and determining the final alignment result based on the local alignment score, the local alignment offset and the dual preset thresholds includes: in response to the local alignment score being greater than the high threshold, determining the local alignment offset as the final alignment result; in response to the local alignment score being greater than the low threshold and less than the high threshold, performing an expanded search alignment on the first image and the second image captured after the grain to be inspected and the reference grain are respectively in the same position, so as to determine the final alignment result; or in response to the local alignment score being less than the low threshold, performing a global search alignment on the first image and the second image captured after the grain to be inspected and the reference grain are respectively in the same position, so as to determine the final alignment result.

[0009] In yet another embodiment, performing an expanded search alignment on the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position to determine the final alignment result includes: selecting a corresponding expanded search area in the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position based on the initial offset; performing an expanded search alignment on the first image and the second image captured after the grain to be detected and the reference grain are respectively in the same position in the corresponding expanded search area, and calculating an expanded search alignment offset; and comparing the expanded search offset with the local alignment offset to determine the final alignment result.

[0010] In another embodiment, comparing the expanded search offset with the local alignment offset to determine the final alignment result includes: in response to the difference between the expanded search offset and the local alignment offset being less than a preset value, determining the expanded search offset as the final alignment result; or in response to the difference between the expanded search offset and the local alignment offset being greater than a preset value, performing a global search alignment on the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position to determine the final alignment result.

[0011] In yet another embodiment, performing a global search alignment on the first image and the second image captured after the grain to be inspected and the reference grain are respectively at the same position to determine the final alignment result includes: performing a global search alignment on the first image and the second image captured after the grain to be inspected and the reference grain are respectively at the same position in a global search area, and calculating a global search alignment offset; comparing the global search alignment offset with the initial offset to determine the final alignment result.

[0012] In another embodiment, comparing the global search alignment offset with the initial offset to determine the final alignment result includes: in response to the difference between the global search alignment offset and the initial offset being less than a preset value, determining the global search alignment offset as the final alignment result; or in response to the difference between the global search alignment offset and the initial offset being greater than a preset value, replacing the initial offset with the global search alignment offset to perform alignment iterations to determine the final alignment result.

[0013] In yet another embodiment, before determining the final alignment result, the method further includes: selecting a target alignment score for sub-pixel alignment; and determining the sub-pixel alignment result as the final alignment result.

[0014] In a second aspect, the present application provides a device for image alignment in wafer defect detection, comprising: a processor; and a memory on which computer instructions for image alignment in wafer defect detection are stored, and when the computer instructions are executed by the processor, multiple embodiments of the aforementioned first aspect are implemented.

[0015] In a third aspect, the present application provides a computer-readable storage medium having stored thereon computer program instructions for image alignment in wafer defect detection, wherein when the computer program instructions are executed by one or more processors, multiple embodiments of the aforementioned first aspect are implemented.

[0016] Through the scheme for image alignment in wafer defect detection provided above, the embodiment of the present application first performs image alignment on the first image and the second image initially collected at the same position of the inspected grain and the reference grain, and uses the initial offset of the initially collected first image and the second image as a reference to perform local alignment on the first image and the second image collected after the inspected grain and the reference grain are respectively at the same position, that is, a small-range search is performed, which greatly saves the alignment time of the images in the detection, reduces the amount of calculation, and improves the calculation efficiency. Then, by setting a double preset threshold, the final alignment result is determined in combination with the local alignment score and the local alignment offset, thereby improving the accuracy of the alignment. Furthermore, in some embodiments, the embodiment of the present application also adopts sub-pixel alignment, so that the accuracy of the final alignment result is further improved, so as to improve the subsequent accurate detection of wafer defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0018] Figure 1 is an exemplary schematic diagram showing a related wafer;

[0019] Figure 2 is an exemplary flowchart showing a method 200 for image alignment in wafer defect detection according to an embodiment of the present application;

[0020] Figure 3 is an exemplary flowchart showing the overall image alignment for wafer defect detection according to an embodiment of the present application;

[0021] Figure 4 is an exemplary schematic diagram showing several alignment scenarios according to an embodiment of the present application;

[0022] Figure 5 is an exemplary schematic diagram showing an aligned image and a defect detection result according to an embodiment of the present application;

[0023] Figure 6 It is an exemplary structural block diagram of a device for image alignment in wafer defect detection according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0025] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present application indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0027] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0028] For ease of understanding, this application first combines Figure 1 Introduce relevant concepts about wafers.

[0029] Figure 1 FIG. 1 is an exemplary schematic diagram showing a wafer. Figure 1 The circular area in (a) of FIG. 1 is a wafer 101, and the wafer 101 is composed of different dies (such as shown by multiple small squares in the figure), such as the dies Die0 and Die 1 in the figure. When detecting wafer defects, reference dies and dies to be detected are usually selected from the critical area (care area), such as the reference dies R-Die and the dies to be detected Die 0. For example Figure 1FIG. (b) shows a key area 102, which may include multiple groups of detection areas, for example, two detection groups Grop 0 and Grop 1 are shown in the figure. In some implementation scenarios, each detection group includes multiple areas to be detected, for example, detection group Grop 0 includes four areas S to be detected, detection group Grop 1 includes three areas s to be detected, and there are non-detection areas among the areas to be detected, for example, the gray rectangle in the area S to be detected is shown as a non-detection area.

[0030] It is understandable that the area of ​​a critical area is usually large, and multiple consecutive scanning optical images or scanning SEM images are required to cover the entire critical area. Figure 1 As shown in Figure (c). The left and right figures of Figure (c) respectively represent the grain to be tested (for example, Die 0) and the reference grain R-Die. In the figure, 1_1, 1_2, 1_3, 1_4, ..., are multiple images of the grain to be tested that are collected by a single optical or scanning electron microscope multiple times in a continuous sequence. In the figure, 2_1, 2_2, 2_3, 2_4, ..., are multiple images of the reference grain that are collected by a single optical or scanning electron microscope multiple times in a continuous sequence. In actual scenarios, in order to avoid the frequent movement of the sample stage back and forth between different grains, which consumes more time and reduces stability, it is usually adopted to collect multiple images in the same key area in a continuous sequence, and then move to the same position of another reference grain to collect multiple images in a continuous sequence.

[0031] As described in the above background technology, whether D2D or D2DB is used for wafer defect detection, image alignment operations need to be performed in advance. However, when the search area is large, the amount of calculation required for the alignment operation is large, resulting in low calculation efficiency and long alignment time. In addition, since multiple images are generally collected in the same key area in a continuous sequence, and then moved to the same position of another reference grain for the same continuous image collection. There is a certain relationship between these continuously collected images. For example, there is a certain deviation (sx2, sy2) between the positions of the later collected images 1_2 and 2_2. This result is caused by the position deviation (sx1, sy1) between the initially collected images 1_1 and 2_1 in the same key area, plus a small disturbance of the pixels. If this information is not used, a large amount of computing resources will be wasted.

[0032] On the other hand, in order to solve the problem of low detection efficiency, that is, low throughput, the detection algorithm is generally deployed on a small Linux computer cluster, and multi-process calculations are mostly used, and there is no communication between processes. The advantage of this method is that the code is easy to maintain and the parallel efficiency is very high. It grows almost in the same amount as the number of CPU cores, and the CPU utilization rate is very high. However, in fact, it is reasonable for different processes to handle different tasks, and different tasks are independent and uncoupled except for the position deviation between consecutive images of the same key area.

[0033] Based on this, the embodiment of the present application provides a solution for image alignment in wafer defect detection, which utilizes the sequential relationship of image acquisition in key areas during the detection flow, and fully utilizes the image alignment results of the previous acquisition for local alignment, thereby greatly saving the alignment time of the image in the detection, reducing the detection calculation time, and improving the detection efficiency. Furthermore, by setting dual preset thresholds and combining the local alignment score and the local alignment offset to determine the final alignment result, the accuracy of the alignment is improved.

[0034] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0035] Figure 2 FIG. 2 is an exemplary flowchart of a method 200 for image alignment in wafer defect detection according to an embodiment of the present application. Figure 2 As shown in FIG. 1 , at step S201 , reference die and die to be tested are selected from the key area of ​​the wafer. Figure 1 As shown, a reference die R-Die and a to-be-detected die Die 0 are selected. Next, at step S202, the first image and the second image initially captured at the same position of the to-be-detected die and the reference die are aligned to obtain an initial offset. Figure 1 Taking the grain to be inspected and the reference grain shown in FIG. (c) as an example, the first image and the second image initially collected at the same position correspond to 1_1 and 2_1 respectively. That is, the first image 1_1 initially collected from the grain to be inspected and the second image 2_1 initially collected from the reference grain at the same position are image aligned to obtain an initial offset. In some embodiments, the aforementioned image alignment can be achieved by, for example, a standard normalized cross correlation (“NCC”) method.

[0036] Specifically, the target pattern ("pattern") in the first image 1_1 initially collected from the grain to be inspected is moved in the second image 2_1 initially collected from the reference grain at the same position, and the corresponding alignment score is calculated by the NCC method, and the coordinate position at the maximum alignment score is determined. The initial offset is determined by calculating the difference between the coordinate position at the maximum alignment score and the target pattern position in the first image 1_1 initially collected from the grain to be inspected. As an example, assuming that the target pattern position in the first image 1_1 initially collected from the grain to be inspected is (100, 100), and the coordinate position at the maximum alignment score is (110, 110), the initial offset can be obtained as (10, 10). In some embodiments, the aforementioned target pattern position is generally determined based on the coordinates of the upper left corner of the target pattern.

[0037] After obtaining the above initial offset, at step S203, the first image and the second image acquired after the to-be-detected crystal grain and the reference crystal grain are at the same position are locally aligned based on the initial offset, and the local alignment score and the local alignment offset are calculated. That is, using the image alignment result of the previous acquired image (such as the above initial offset) as a reference, a small range search alignment is performed, and the local alignment score and the local alignment offset are calculated.

[0038] In one embodiment, a corresponding local search area can be selected from the first image and the second image acquired after the inspected die and the reference die are at the same position based on the initial offset, and then the first image and the second image acquired after the inspected die and the reference die are at the same position are locally aligned in the corresponding local search area. In an implementation scenario, the first image acquired later (e.g. Figure 1 1_2) and a second image (eg Figure 1 The upper left corner of the target pattern in (c) shown in FIG2_2) is taken as the center, and the initial offset pixel units are searched around to obtain the corresponding local search area. For example, taking the above initial offset (10,10) as an example, the upper left corner of the target pattern on the first image 1_2 acquired later is (95,95), and the upper left corner (95,95) of the target pattern is taken as the center, and 10 pixel units are searched around (up, down, left and right) to obtain the corresponding local search area. Similarly, the corresponding local search area can be obtained in the second image 2_2 acquired later.

[0039] Next, the first image and the second image acquired after the to-be-detected grain and the reference grain are at the same position are locally aligned in the corresponding local search area. Similar to the above-mentioned image alignment of the first image 1_1 initially acquired from the to-be-detected grain and the second image 2_1 initially acquired from the reference grain at the same position, local alignment can be achieved in the corresponding local search area by using, for example, the NCC method, and a local alignment score and a local alignment offset are obtained.

[0040] Further, at step S204, the final alignment result is determined according to the local alignment score, the local alignment offset and the dual preset thresholds to achieve image alignment in wafer defect detection. The embodiment of the present application can improve the accuracy of alignment by setting dual thresholds and combining the local alignment offset and the local alignment score for judgment.

[0041] In one implementation scenario, the dual preset thresholds may include a high threshold and a low threshold. In response to the local alignment score being greater than the high threshold, the local alignment offset is determined as the final alignment result. In response to the local alignment score being greater than the low threshold and less than the high threshold, the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position are subjected to an expanded search alignment to determine the final alignment result. Alternatively, in response to the local alignment score being less than the low threshold, the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position are subjected to a global search alignment to determine the final alignment result.

[0042] That is to say, if the alignment score during local alignment is greater than the high threshold, it means that the target pattern is in the local area, and image alignment can be achieved, so the local alignment offset is determined as the final alignment result. When the alignment score during local alignment is between the high threshold and the low threshold, it means that the target may be in the local area. In this case, the search area can be expanded to search and align to determine the final alignment result. When the alignment score during local alignment is less than the low threshold, it means that the target is not in the local area. In this case, a global search alignment can be performed to determine the final alignment result.

[0043] In one embodiment, for the expanded search, a corresponding expanded search area can be selected from the first image and the second image acquired after the to-be-detected grain and the reference grain are respectively in the same position based on the initial offset, and the first image and the second image acquired after the to-be-detected grain and the reference grain are respectively in the same position are subjected to an expanded search alignment in the corresponding expanded search area, and the expanded search alignment offset is calculated, and then the expanded search offset is compared with the local alignment offset to determine the final alignment result. Preferably, the corresponding expanded search area can be selected by expanding the initial offset by a target multiple (e.g., 2 times, 3 times, etc.). As an example, assuming that the initial offset (10,10) is taken as an example, it is expanded by, for example, 2 times to (20,20). In this scenario, by taking the upper left corner of the target pattern on the first image 1_2 and the second image 2_2 acquired later as the center, 20 pixel units are searched around (up, down, left and right) to obtain the corresponding expanded search area.

[0044] Similarly, in the corresponding expanded search area, the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position are subjected to an expanded search alignment, for example, using the NCC method, to obtain an expanded search alignment offset. Further, the expanded search offset is compared with the local alignment offset to determine the final alignment result. In one embodiment, in response to the difference between the expanded search offset and the local alignment offset being less than a preset value, the expanded search offset is determined as the final alignment result. Alternatively, in response to the difference between the expanded search offset and the local alignment offset being greater than a preset value, a global search alignment is performed on the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position to determine the final alignment result.

[0045] That is, when the alignment score meets the range from the low threshold to the high threshold, it is necessary to compare the previous and next alignment offsets to more accurately determine the final alignment result. If the difference between the previous and next alignment offsets meets the preset value, the next alignment offset (e.g., the expanded search offset) is determined as the final alignment result. If the difference between the previous and next alignment offsets does not meet the preset value, a global search alignment is performed.

[0046] In some embodiments, the preset value may be, for example, 1 or a deviation value t of the previous and next alignment results. In an exemplary scenario, assuming that the local alignment offset is recorded as (x1, y1) and the expanded search offset is recorded as (x2, y2), when |x1-x2| is less than 1 and |y1-y2| is less than 1 (i.e., |x1-x2|<1&&|y1-y2|<1), the expanded search offset (x2, y2) is determined as the final alignment result. When |x1-x2| is greater than 1, or |y1-y2| is greater than 1, or |x1-x2| and |y1-y2| are both greater than 1 (i.e., |x1-x2|>1|||y1-y2|>1), a global search alignment is performed on the first image and the second image captured after the inspected grain and the reference grain are respectively in the same position to determine the final alignment result.

[0047] In one embodiment, a global search alignment is performed on the first image and the second image acquired after the inspected grain and the reference grain are at the same position in the global search area, and a global search alignment offset is calculated to compare the global search alignment offset with the initial offset to determine the final alignment result. That is, a search is performed within the entire image range of the first image and the second image acquired later to determine the final alignment result. In some embodiments, the aforementioned global search alignment can also be performed using, for example, an NCC method to obtain a global search alignment offset.

[0048] In one implementation scenario, in response to the difference between the global search alignment offset and the initial offset being less than a preset value, the global search alignment offset is determined as the final alignment result. Alternatively, in response to the difference between the global search alignment offset and the initial offset being greater than a preset value, the initial offset is replaced with the global search alignment offset to perform alignment iterations to determine the final alignment result.

[0049] In some embodiments, the preset value may also be, for example, 1 or a deviation value t of the previous and next alignment results. In an exemplary scenario, assuming that the global search alignment offset is recorded as (x3, y3) and the initial offset is recorded as (x0, y0), when |x0-x3| is less than t and |y0-y3| is less than t (i.e., |x0-x3|<t&&|y0-y3|<t), the global search alignment offset (x3, y3) is determined as the final alignment result. When |x0-x3| is greater than t, or |y0-y3| is greater than t, or |x0-x3| and |y0-y3| are both greater than t (i.e., |x0-x3|>t|||y0-y3|>t), the global search alignment offset (x3, y3) is used to update the aforementioned initial offset (x0, y0), and the aforementioned operation is repeated until the final alignment result is determined.

[0050] In one embodiment, in order to further improve the alignment accuracy, before determining the final alignment result, the embodiment of the present application also includes selecting a target alignment score for sub-pixel alignment; and determining the final alignment result with the sub-pixel alignment result. Specifically, by selecting, for example, three target alignment scores for curve fitting, a better alignment score is obtained, thereby achieving sub-pixel alignment, and the position corresponding to the better alignment score is the sub-pixel alignment result, that is, the final alignment result. As an example, assuming that the three target alignment scores are 0.97, 0.98, and 0.96, respectively, a curve is fitted through these three points, and assuming that there is a better alignment score of 0.99, the position corresponding to 0.99 is determined as the final alignment result.

[0051] Combined with the above description, it can be seen that the embodiment of the present application first aligns the first image and the second image initially collected at the same position of the grain to be detected and the reference grain to obtain an initial offset, and the initial offset is the relative position relationship between the global images. Then, with the initial offset as a reference, the first image and the second image collected after determining that the grain to be detected and the reference grain are respectively at the same position are preferentially locally aligned, and the target search is performed within a smaller search range to quickly find possible matching position information. In this way, the search efficiency can be improved and the computational burden can be reduced.

[0052] During the local search process, the alignment score and offset corresponding to each local image are calculated by, for example, the NCC method. Among them, the higher the alignment score, the more likely the pixel is to be an alignment point, and the offset represents the position change of the pixel relative to the initial offset. In order to screen out more accurate alignment points, the embodiment of the present application sets a double preset threshold for judgment. Only when the alignment score is greater than the high threshold value will it be regarded as a valid alignment point, so that the accuracy of alignment can be ensured. When the alignment score is between the high threshold value and the low threshold value or is less than the low threshold value, the search range can be expanded or a global search can be performed to find a more accurate alignment point and improve the accuracy of alignment. Further, based on the determined alignment point, the final alignment result can be updated. In addition, in order to further improve the alignment accuracy, sub-pixel level alignment is also adopted to introduce higher accuracy, so that image alignment is more accurate.

[0053] Figure 3 FIG. 1 is an exemplary flow chart showing the overall image alignment for wafer defect detection according to an embodiment of the present application. It should be understood that Figure 3 is the above Figure 2 A specific embodiment of Figure 2 The description also applies to Figure 3 .

[0054] like Figure 3As shown in FIG. 1 , at step S301, the first image (eg, Figure 1 (c) 1_1) and the second image as shown in FIG. Figure 1 The image 1_2 shown in (c) is aligned to obtain the initial offset (x0, y0). As mentioned above, this can be achieved by, for example, the NCC method. Next, at step S302, a local search area is set with the initial offset (x0, y0) as a reference, and the first image (e.g. Figure 1 1_2) and a second image (eg Figure 1 2_2) as shown in FIG. (c) of FIG. 10 performs local search alignment, and at step S303, calculates the local alignment score and the local alignment offset (x1, y1). The corresponding local search area can be obtained by taking the upper left corner of the target pattern in the first image and the second image acquired later as the center and searching upward, downward, left, and right by, for example, x0, y0 pixel units respectively.

[0055] Further, at step S304, first determine whether the local alignment score is greater than the high threshold th_H. If the local alignment score is greater than the high threshold th_H, the local alignment offset (x1, y1) can be determined as the final alignment result at step S314. If the local alignment score is less than the high threshold th_H, at step S305, determine whether the local alignment score is less than the low threshold th_L. When the local alignment score is not less than the low threshold th_L, that is, the local alignment score is in the interval [th_L, h_H], at step S306, the first image and the second image collected after the inspected grain and the reference grain are respectively in the same position are expanded to search and align. For example, the search range is expanded by, for example, 2 times to continue the search alignment, and at step S307, the expanded search offset (x2, y2) is obtained. Next, at step S308, a judgment is made based on the deviation between the local alignment offset (x1, y1) and the expanded search offset (x2, y2), for example, whether |x1-x2|<1&&|y1-y2|<1 holds.

[0056] When |x1-x2|<1&&|y1-y2|<1 holds, the expanded search offset (x2, y2) can be determined as the final alignment result in step S314. When the local alignment score is less than the low threshold th_L or |x1-x2|<1&&|y1-y2|<1 does not hold, jump to step S309, perform global search alignment on the first image and the second image captured after the inspected grain and the reference grain are in the same position, and obtain the global search alignment offset (x3, y3) in step S310. Then, in step S311, determine whether |x0-x3|>t|||y0-y3|>t holds. When |x0-x3|>t|||y0-y3|>t is established, at step S312, the initial offset (x0, y0) is updated using the global search alignment offset (x3, y3), that is, x0=x3, y0=y3, and the return value step S301 repeats the above operation. When |x0-x3|>t|||y0-y3|>t is not established, at step S314, the global search alignment offset (x3, y3) can be determined as the final alignment result.

[0057] The figure further shows that at step S313, sub-pixel level alignment can also be performed to obtain a final alignment result, thereby further improving the alignment accuracy.

[0058] Figure 4 is an exemplary schematic diagram showing several alignment scenarios according to an embodiment of the present application. Figure 4 Figure (a) shows local alignment, where the initial offset is assumed to be (10, 10), with the upper left corner of the target pattern in the image as the center, 10 pixel units are searched upward, downward, left and right to perform local search alignment within the area. Figure 4 Figure (b) shows an expanded search alignment, in which it is assumed that the initial offset (10,10) is expanded by 2 times to form (20,20), and 20 pixel units are searched upward, downward, left and right with the upper left corner of the target pattern in the image as the center to perform an expanded search alignment within the area. Figure 4 Figure (c) shows a global search alignment, that is, it is performed in the entire image area. Based on the final alignment result of the embodiment of the present application, the accuracy of the alignment can be improved, and the defects of the wafer can be detected by subtracting the aligned image of the grain to be detected from the reference grain image.

[0059] Figure 5 is an exemplary schematic diagram showing an aligned image and defect detection results according to an embodiment of the present application. Figure 5 In (a), the upper and lower figures respectively exemplarily show the reference grain image after alignment and the grain image to be detected. Figure 5The upper and lower figures in Figure (b) respectively exemplify the defect detection results without final alignment and the defect detection results after final alignment, where the white dots in the lower figure in Figure (b) indicate the detected defects. It can be seen from the figure that the defect detection result of subtracting the two images after final alignment solves the problem of local alignment failure due to pixel disturbance and improves the accuracy of defect detection.

[0060] Figure 6 FIG. 6 is an exemplary structural block diagram of an apparatus 600 for image alignment in wafer defect detection according to an embodiment of the present application. Figure 6 As shown in , the device 600 of the present application may include a processor 601 and a memory 602, wherein the processor 601 and the memory 602 communicate with each other through a bus. The memory 602 stores program instructions for image alignment in wafer defect detection. When the program instructions are executed by the processor 601, the method steps described in the above text in combination with the accompanying drawings are implemented: selecting a reference grain and a grain to be detected from a critical area of ​​the wafer; aligning the first image and the second image initially captured at the same position of the grain to be detected and the reference grain to obtain an initial offset; based on the initial offset, locally aligning the first image and the second image captured after the grain to be detected and the reference grain are respectively at the same position, and calculating the local alignment score and the local alignment offset; and determining the final alignment result according to the local alignment score, the local alignment offset and the double preset threshold value, so as to achieve image alignment in wafer defect detection.

[0061] According to the above description in combination with the accompanying drawings, those skilled in the art can also understand that the embodiments of the present application can also be implemented by software programs. Therefore, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions for image alignment in wafer defect detection. When the computer-readable instructions are executed by one or more processors, the present application in combination with the accompanying drawings can be implemented. Figure 2 A method for image alignment in wafer defect inspection is described.

[0062] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0063] It should be noted that although the operations of the method of the present application are described in a specific order in the accompanying drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. On the contrary, the steps depicted in the flow chart can be performed in a different order. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0064] It should be understood that when the terms "first", "second", "third" and "fourth" are used in the claims, the specification and the drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0065] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0066] Although the implementation methods of the present application are as above, the contents described are only examples adopted to facilitate the understanding of the present application, and are not intended to limit the scope and application scenarios of the present application. Any technician in the technical field described in the present application can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present application, but the scope of patent protection of the present application shall still be subject to the scope defined in the attached claims.

Claims

1. A method for image alignment in wafer defect detection, comprising: Selecting reference dies and dies to be inspected from critical areas of the wafer; Performing image alignment on a first image and a second image initially captured at the same position of the grain to be inspected and the reference grain to obtain an initial offset; Based on the initial offset, locally align the first image and the second image acquired after the die to be inspected and the reference die are respectively in the same position, and calculate a local alignment score and a local alignment offset; and Determining a final alignment result according to the local alignment score, the local alignment offset and dual preset thresholds to achieve image alignment in wafer defect detection, The dual preset thresholds include a high threshold and a low threshold, and determining a final alignment result according to the local alignment score, the local alignment offset and the dual preset thresholds includes: In response to the local alignment score being greater than the high threshold, determining the local alignment offset as the final alignment result; In response to the local alignment score being greater than the lower threshold and less than the upper threshold, performing an enlarged search alignment on the first image and the second image acquired after the inspected die and the reference die are respectively in the same position, so as to determine the final alignment result; or In response to the local alignment score being less than the lower threshold, a global search alignment is performed on the first image and the second image acquired after the inspected die and the reference die are respectively in the same position, so as to determine the final alignment result.

2. The method according to claim 1, wherein locally aligning the first image and the second image acquired after the to-be-detected die and the reference die are respectively at the same position based on the initial offset comprises: Selecting a corresponding local search area in a first image and a second image acquired after the to-be-detected grain and the reference grain are respectively at the same position based on the initial offset; as well as In the corresponding local search area, the first image and the second image acquired after the to-be-detected crystal grain and the reference crystal grain are respectively in the same position are locally aligned.

3. The method according to claim 1, wherein performing an enlarged search alignment on the first image and the second image acquired after the to-be-detected die and the reference die are respectively in the same position to determine the final alignment result comprises: Selecting a corresponding expanded search area in a first image and a second image acquired after the to-be-detected grain and the reference grain are respectively at the same position based on the initial offset; Performing an enlarged search alignment on the first image and the second image acquired after the to-be-detected crystal grain and the reference crystal grain are respectively in the same position in the corresponding enlarged search area, and calculating an enlarged search offset; and The enlarged search offset is compared with the local alignment offset to determine the final alignment result.

4. The method of claim 3, wherein comparing the enlarged search offset with the local alignment offset to determine the final alignment result comprises: In response to a difference between the enlarged search offset and the local alignment offset being less than a preset value, determining the enlarged search offset as the final alignment result; or In response to the difference between the expanded search offset and the local alignment offset being greater than a preset value, a global search alignment is performed on the first image and the second image captured after the inspected die and the reference die are respectively at the same position to determine the final alignment result.

5. The method according to claim 1 or 4, wherein performing global search alignment on the first image and the second image acquired after the to-be-detected crystal grain and the reference crystal grain are respectively in the same position to determine the final alignment result comprises: Performing a global search alignment on a first image and a second image acquired after the to-be-detected crystal grain and the reference crystal grain are respectively at the same position in a global search region, and calculating a global search alignment offset; The global search alignment offset is compared with the initial offset to determine the final alignment result.

6. The method of claim 5, wherein comparing the global search alignment offset with the initial offset to determine the final alignment result comprises: In response to a difference between the global search alignment offset and the initial offset being less than a preset value, determining the global search alignment offset as the final alignment result; or In response to a difference between the global search alignment offset and the initial offset being greater than a preset value, the initial offset is replaced by the global search alignment offset to perform alignment iterations to determine the final alignment result.

7. The method according to claim 1, wherein before determining the final alignment result, it further comprises: Select the target alignment score for sub-pixel alignment; as well as The sub-pixel level alignment result is determined as the final alignment result.

8. A device for image alignment in wafer defect detection, comprising: processor; as well as A memory having computer instructions for image alignment in wafer defect detection stored thereon, wherein when the computer instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having stored thereon computer program instructions for image alignment in wafer defect detection, wherein when the computer program instructions are executed by one or more processors, the method according to any one of claims 1 to 7 is implemented.

Images