Detection method, corresponding equipment and storage medium

Abstract

The present application provides a detection method, and corresponding equipment and storage medium, the method comprising: obtaining an image to be tested of a sample, and dividing the image to be tested into multiple sub-areas; selecting at least part of the sub-areas within the target range of the image to be tested as the focus area, the focus area including one or more sub-areas; obtaining a target image based on the image of the focus area, and detecting the target image to obtain the target to be tested in the sample; wherein the target range is set based on the physical information of the target to be tested in the sample, and the number of targets to be tested in the target range is greater than that in other areas. The present application sets the focus area based on the physical information of different targets to be tested in the sample, and only detects the focus area in the image to be tested, without detecting the entire area of ​​the image, thereby improving the detection speed and preventing noise from non-focus areas from affecting the target detection accuracy of the focus area.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a detection method, and corresponding equipment and storage medium. Background Art

[0002] When surface inspection equipment is used in semiconductor testing, it's common to use an ML (machine learning) model-based inspection method to identify defects on sample surfaces (such as wafers). Specifically, defects are annotated with rectangular boxes, a training set is created, and an ML model is trained. The ML model then inspects the image to be tested, using the rectangular boxes to outline the defects. However, for linear defects such as slip lines, which typically appear at the wafer edge, panoramic inspection of the entire wafer results in slow inspection speeds. Summary of the Invention

[0003] The technical solution of the present application is to provide a detection method, corresponding equipment and storage medium, in order to solve the problem of low detection speed in the existing panoramic detection method for the entire wafer.

[0004] The first aspect of the present application provides a method for detecting a sample, comprising: obtaining an image to be tested of the sample, and dividing the image to be tested into multiple sub-areas; selecting at least part of the sub-areas within a target range of the image to be tested as a focus area, the focus area including one or more sub-areas; obtaining a target image based on the image of the focus area, and detecting the target image to obtain a target to be tested in the target image; wherein the target range is set based on the physical information of the target to be tested in the sample, and the number of targets to be tested in the target range is greater than that in other areas.

[0005] The present application also provides a detection device, including a processor, configured to: obtain an image to be tested of a sample, and divide the image to be tested into multiple sub-areas; select at least part of the sub-areas within the target range of the image to be tested as a focus area, and the focus area includes one or more sub-areas; obtain a target image based on the image of the focus area, and detect the target image to obtain the target to be tested in the target image; wherein the target range is set based on the physical information of the target to be tested in the sample, and the number of targets to be tested in the target range is greater than that in other areas.

[0006] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the detection method described above are implemented.

[0007] The technical solution of the present invention has the following beneficial effects.

[0008] In the invention provided by the technical solution of the present invention, a focus area is set according to the physical information of different targets to be tested in the sample, and only the focus area in the test image is detected, without the need to detect the entire area of ​​the image, thereby improving the detection speed and avoiding the noise of non-focus areas affecting the target detection accuracy of the focus area. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0010] Figure 1 This is a schematic diagram of a first embodiment of the detection method of the present invention;

[0011] Figure 2 Schematic diagram of a first embodiment of the image segmentation to be tested according to the present invention;

[0012] Figure 3 Schematic diagram of a second embodiment of the image segmentation method according to the present invention;

[0013] Figure 4 Schematic diagram of a third embodiment of the image segmentation method according to the present invention;

[0014] Figure 5 Schematic diagram of a second embodiment of the detection method of the present invention;

[0015] Figure 6 This is a schematic diagram of a first embodiment of target image segmentation in the present invention;

[0016] Figure 7 Schematic diagram of a second embodiment of target image segmentation in the present invention;

[0017] Figure 8 Schematic diagram of a third embodiment of target image segmentation in the present invention;

[0018] Figure 9 Schematic diagram of a fourth embodiment of target image segmentation in the present invention;

[0019] Figure 10 Schematic diagram of the third embodiment of the detection method of the present invention;

[0020] Figure 11 A schematic diagram of an embodiment of a sample pattern in the present invention;

[0021] Figure 12 It is a structural diagram of the control system in the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0023] In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, they cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In the present invention, "each" includes one and more than two quantities.

[0024] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0025] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other, and should all be considered to be within the scope of this specification.

[0026] For the sake of convenience, the basic process of the detection method of the present invention is described below. Figure 1 , provides the first embodiment of the detection method in the present invention, which is specifically as follows:

[0027] 110. Acquire a sample image to be tested, and divide the image to be tested into multiple sub-areas;

[0028] In this embodiment, at least one optical imaging system is provided to capture an image of the object to be measured using the optical properties of the object on the sample surface, thereby distinguishing the object to be measured from the sample background in the image. The optical properties include differences in reflectivity between sample defects and the background, and enhanced light scattering at the edges of the object to be measured.

[0029] In this embodiment, the image to be measured can be an image representing the gradient or height of the sample; or the image to be measured can be a dark field image or a bright field image representing the reflectivity of the sample surface. In this embodiment, the image to be measured is a grayscale image; in other embodiments, the image to be measured can be a binary image or a color image.

[0030] Specifically, in an optical imaging system, interference phenomena and phase difference images are used to characterize the differences in optical properties between the target to be measured and the background on the sample surface, thereby highlighting the image features of the target to be measured. For example, differential interference contrast (DIC) can be used in conjunction with interference phenomena such as Wollaston prisms and phase modulation such as piezoelectric ceramic micro-shift prisms to capture multiple frames of phase difference images of the sample and reconstruct a gradient map or height map of the sample surface to highlight the edge contrast of the target to be measured relative to the background. Based on the sample surface gradient map or height map, the continuous phase is restored by integrating the gradient field and converted into a grayscale image, forming a two-dimensional plane image to be measured represented by grayscale.

[0031] If the interference fringes are offset based on a phase difference of multiples of 2π, the target will appear as bright fringes in the image. If the interference fringes are offset based on a phase difference of odd multiples of π, the target will appear as dark fringes in the image. When generating a grayscale image, phase-resolved sample surface height maps or gradient maps generate both bright and dark lines in the image for linear targets.

[0032] In this embodiment, the target to be measured is a defect, and the defect is a chipped edge, a scratch, or a slip line. In other embodiments of the present application, the target to be measured is a metal line or a groove.

[0033] In this embodiment, the required division method can be selected according to the geometric characteristics of the sample or the requirements of defect distribution, such as: 1) Grid division: The image to be tested is evenly divided into several small squares, and each square is an independent area. For example, the image to be tested is divided into a 10x10 grid, with a total of 100 areas (rectangular areas). 2) Ring division: If the sample is circular, it is divided into multiple areas (concentric ring areas). For example, from the center to the edge, it is divided into 5 ring areas in sequence, and the width of each ring is the same. 3) Sector division: The image to be tested is divided into multiple sector areas according to the circular shape of the sample, for example, divided into 12 30-degree areas (sector areas).

[0034] Taking grid division as an example, determining the grid size involves: 1) First, determine the size of each grid cell based on the size and resolution of the image to be measured. For example, for a sample with a diameter of 200 mm and an image resolution of 10,000 x 10,000 pixels, it can be divided into a 100 x 100 pixel grid, for a total of 10,000 regions. 2) Then, using image processing software (such as OpenCV), grid lines are drawn on the image to be measured according to the determined grid size, dividing the image into multiple regions. 3) Each grid cell is then assigned a unique identifier to identify its location within the image to be measured.

[0035] In one embodiment, the image to be tested is divided into multiple overlapping areas, which can further enhance the detection effect of dense defect distribution. Specifically, the image to be tested can be slidingly divided along at least one preset direction according to a preset first size (the description of the division is referred to as the preset size) and a preset step size, so that the areas on both sides of the preset direction are divided to obtain multiple sub-areas, wherein each two adjacent sub-areas partially overlap or touch but do not overlap.

[0036] In this embodiment, before performing the division, the method further includes determining the preset direction based on the extension direction of the target or the position distribution characteristics of the target. Specifically, the preset direction is parallel to the extension direction or has an acute angle plus an included angle with the extension direction; or the area where the target density is greater than a threshold is divided from the area where the target density is less.

[0037] The plurality of sub-regions are arranged in an array after the above sliding division, the preset size determines the size of the sub-region, and the preset step size determines the size of the overlapping portion.

[0038] like Figure 2As shown, in this embodiment, the target to be measured is linear, and the preset direction is one or both of the transverse direction V1 and the longitudinal direction V2, and the preset direction is parallel to the extension direction of the target to be measured. The first image to be measured 210 includes a sample pattern 220, and the preset first direction includes the following two: the transverse direction V1 and the longitudinal direction V2; when the image to be measured is divided by sliding, if the preset size of the transverse direction V1 is S1 and the step length is S1, and the preset size of the longitudinal direction V2 is S2 and the step length is S2, multiple sub-areas (rectangular sub-areas) of size S1*S2 are obtained as shown in the second image to be measured 230, and there is no overlapping part between each two adjacent sub-areas; if the preset size of the transverse direction V1 is S1 and the step length is T1, and the longitudinal direction V2 is S2, the sub-areas (rectangular sub-areas) of size S1*S2 are obtained. The preset size in direction V2 is S2, with a step size of T2. This results in multiple subregions (rectangular subregions) of size S1*S2, as shown in the third test image 240. Each pair of laterally adjacent subregions overlaps by a portion (S1-T1)*S2 (e.g., overlap portion 250), and each pair of longitudinally adjacent subregions overlaps by a portion (S2-T2)*S1. Except for edge subregions, each subregion (e.g., subregion 260) overlaps with both lateral and longitudinal subregions. Expanding each subregion in the third test image 240 yields each subregion as shown in the fourth test image 270, where the expanded size is the sum of the overlapping portions. The test images in this embodiment, based on the division process, can be shown as the first test image 210, the second test image 230, the third test image 240, and the fourth test image 270.

[0039] The preset size S1 and step size T1 in the horizontal direction V1 refer to: setting a segmentation frame with a horizontal length of S1 with the left edge / right edge of the first image to be measured 210 as the starting point, the segmentation frame slides for a length of T1 each time, and the sub-area defined by each sliding is used as the division result in the horizontal direction V1; the preset size S1 and step size T1 in the horizontal direction V1 refer to: the preset size S2 and step size T2 in the longitudinal direction V2 refer to: setting a segmentation frame with a vertical length of S2 with the upper edge / lower edge of the first image to be measured 210 as the starting point, the segmentation frame slides for a length of T2 each time, and the sub-area defined by each sliding is used as the division result in the longitudinal direction V2.

[0040] In addition, in other embodiments, Figure 3 As shown, for the circular sample 320 in the first image to be tested 310, the preset first direction and the corresponding sliding division method may further include (setting the preset step length = the preset size):

[0041] A. Slidingly divide the circular sample into multiple concentric annular sub-regions with a radius of length S3 (i.e., sub-regions such as the first sub-region 340 of the second image to be measured 330 ) along a direction V3 from the edge to the center of the sample, with a preset length S3. Wherein, the target to be measured is located at the edge of the sample, i.e., the sub-region such as the second sub-region 350 of the second image to be measured 330 is the first sub-region of the edge (non-concentric annular), as in the following schemes B and C.

[0042] B. Sliding the circular sample into a plurality of sector-shaped sub-areas (i.e., sub-areas such as the third sub-area 370 of the third image to be tested 360 ) along a direction V4 from the edge of the sample and with a preset dimension of angle S4 (S1-S3 being the lengths);

[0043] C. Simultaneously, the circular sample is divided into a plurality of sector-shaped sub-regions in concentric rings (i.e., sub-regions such as the fourth sub-region 390 of the fourth image to be tested 380) by combining the direction V3 from the edge to the center of the sample + the direction V4 around the edge, the length S3 + the angle S4.

[0044] It should be noted that Figure 3 The concentric annular sub-regions and fan-shaped sub-regions shown are only examples. In the actual division process, the diameter of the concentric annular sub-region with the largest edge is less than or equal to the side length of the image to be measured, that is, the concentric annular sub-region with the largest edge can be tangent to the edge of the image to be measured, or not.

[0045] 120. Select at least a portion of a sub-region within a target range of the image to be measured as a region of interest, where the region of interest includes one or more sub-regions;

[0046] In this embodiment, a target range is set based on the physical information of different test targets within the sample, including at least the edge and center regions. For example, crystal defects caused by stress during crystal growth, crack defects due to mechanical manipulation or stress concentration, and other test targets are identified. The physical information includes one or more of the position, shape, and size of the test targets. Specifically, the positional information of the test targets in the edge region is used as the target range. The number of test targets within the target range is greater than that within other regions.

[0047] In this embodiment, the target range includes at least the edge position of the sample pattern, wherein the edge position includes: a preset number of sub-areas determined from the edge of the sample pattern along the preset direction; wherein the preset direction includes at least the direction toward the inside of the image to be measured.

[0048] Specifically, the sub-area through which the edge of the sample pattern passes and the sub-area determined by extending along a preset direction are combined to form the edge position of the sample pattern. For example, for the edge position set for crystal defects, crack defects, etc., two (two layers) sub-areas determined along the direction from the edge of the sample pattern to the inside of the image to be measured can be used as the area of ​​interest.

[0049] In this embodiment, for the multiple rectangular sub-regions divided from the image to be measured in the aforementioned example, the preset directions include a horizontal direction and a vertical direction, both of which include a first partial direction toward the interior of the image to be measured and a second partial direction toward the exterior of the image to be measured. Selecting at least a portion of the sub-regions within the target range of the image to be measured as the region of interest includes: determining a preset number of sub-regions along the first partial direction with the edge of the image to be measured as the starting point; determining a preset number of sub-regions along the second partial direction with the edge of the image to be measured as the end point; the sub-regions determined by the four edges are combined to form the edge position, and the four edge sub-regions are used as the region of interest. Alternatively, selecting at least a portion of the sub-regions within the target range of the image to be measured as the region of interest includes: extracting the edge contour of the sample pattern; obtaining the sub-region containing the edge contour to obtain the region of interest.

[0050] In this embodiment, for the multiple concentric annular sub-regions divided from the image to be tested in the aforementioned example, the preset direction includes the edge to center direction of the sample; at least part of the sub-region located within the target range of the image to be tested is selected as the area of ​​interest, including: taking the edge of the image to be tested as the starting point, determining a preset number of concentric annular sub-regions along the edge to the center direction, that is, the edge position, and the part of the concentric annular sub-region is used as the area of ​​interest.

[0051] In this embodiment, for the multiple sector-shaped sub-regions divided from the image to be measured in the aforementioned example, the preset directions include the edge-to-center direction and the edge-surrounding direction; at least part of the sub-region located within the target range of the image to be measured is selected as the focus area, including: taking the edge of the image to be measured as the starting point, determining a preset number of concentric annular sub-regions along the edge-to-center direction, the determined concentric annular sub-regions include multiple sector-shaped sub-regions, namely the edge positions, and this part of the sector-shaped sub-region is used as the focus area.

[0052] 130. Acquire a target image according to the image of the area of ​​interest, and detect the target image to obtain a target to be detected in the target image.

[0053] In this embodiment, acquiring the target image according to the image of the area of ​​interest includes: using the image of the area of ​​interest as the target image.

[0054] The image of any of the aforementioned regions of interest is used as the target image. For example, Figure 4 As shown, for example, if the preset number is 2, for the multiple rectangular sub-regions divided from the image to be tested in the aforementioned example, two sub-regions are extended from each of the four edges to obtain a sub-region of interest as shown in the first image to be tested 410; for the multiple concentric annular sub-regions divided from the image to be tested in the aforementioned example, two sub-regions are extended from each of the four edges to obtain a sub-region of interest as shown in the second image to be tested 420, wherein the first sub-region 430 at the outermost edge can be further eliminated; for the multiple sector-shaped sub-regions divided from the image to be tested in the aforementioned example, two sub-regions are extended from each of the four edges to obtain a sub-region of interest as shown in the third image to be tested 440, wherein the second sub-region 450 at the outermost edge can be further eliminated. In this embodiment, the images to be tested are shown as the first image to be tested 410, the second image to be tested 420, and the third image to be tested 440 according to different division methods, and the sub-regions at the outermost edges of the second image to be tested 420 and the third image to be tested 440 are shown as the first sub-region 430 and the second sub-region 450.

[0055] In this embodiment, only the focus area in the image to be tested is detected, and the entire area of ​​the image to be tested does not need to be detected, thereby improving the detection speed and preventing noise in non-focus areas from affecting the target detection accuracy in the focus area.

[0056] The above describes the basic process of the detection method of the present invention. The following describes the specific real-time process of the above detection method when dividing the sub-areas. Please refer to Figure 5 , provides a specific implementation method of dividing the sub-regions in the present invention, as shown below:

[0057] 510. Acquire a sample image to be tested, and divide the image to be tested into multiple sub-areas;

[0058] 520. Determine a sample pattern in the image to be measured, and divide the sample pattern into a plurality of sub-regions according to a positional relationship between the target to be measured and the sample pattern;

[0059] In this embodiment, the sample pattern in the image to be measured can be a circle, an ellipse, or a polygon such as a rectangle or a triangle. For example, the sample of a standard wafer can be a circle, or the sample of a customized wafer can be another polygon. The physical information of the target to be measured includes: the positional relationship between the target to be measured and the sample pattern; the distribution of the target to be measured in the sample is determined based on this positional relationship; the area where the target to be measured is concentrated in the sample is used as the area of ​​interest; most or even all of the target to be measured can be detected from the area of ​​interest.

[0060] In one embodiment, the target to be measured is linear, the sub-area is strip-shaped, the preset direction is parallel to the extension direction of the sub-area, the preset direction is parallel to the extension direction of the target to be measured, or the preset direction has an acute angle with the extension direction of the target to be measured.

[0061] In this embodiment, the positional relationship between the target to be measured and the sample pattern also includes: the size and angle of the target to be measured on the sample pattern, which is used to determine the size of each divided sub-region to accurately extract each target to be measured. For example, if the target to be measured is concentrated in the edge area of ​​the sample pattern, the target to be measured is usually generated in a size of 1*200 pixels on the sample pattern, and the extension direction is parallel to the preset direction or at an acute angle; based on this, the sample pattern is divided into multiple sub-regions with a starting point at the edge, a size of ≥1*200 pixels, and an extension direction parallel to the preset direction.

[0062] Specifically, for example, the preset direction is parallel to the pixel row direction or column direction of the image to be measured. If the linear extension direction of the target to be measured is the pixel column direction, the pixel at the edge of each pixel column of the sample pattern is used as the starting point, and the sub-regions are divided into stripes in the pixel column direction. If the linear extension direction of the target to be measured is the pixel row direction, the pixel at the edge of each pixel row of the sample pattern is used as the starting point, and the sub-regions are divided into stripes in the pixel row direction.

[0063] The subregions divided on the sample pattern based on positional relationships contain the entire image content of the sample. Subregions where the edge of the previous sample pattern passes include at least some non-sample image content, and the target to be measured is necessarily within the sample, thus expanding the detection range of the target to be measured within the subregion. In other words, the subregion division based on positional relationships is more precise and accurate. The acute angle between the preset direction and the extension direction of the target to be measured ensures that each subregion contains the entire target to be measured, avoiding division into different subregions and facilitating subsequent detection of the target to be measured (the target to be measured remains within the acute angle range of the preset direction).

[0064] 530. Select at least a portion of a sub-region within a target range of the image to be measured as a region of interest, where the region of interest includes a plurality of sub-regions;

[0065] In this embodiment, the target to be measured has multiple extension directions, and the number of corresponding preset directions is multiple; corresponding to each preset direction, the number of the regions of interest is multiple. In other embodiments, the preset direction and the region of interest may also be one. Specifically, in this embodiment, each preset direction has two regions of interest, the number of the preset directions is two, and the two preset directions are perpendicular to each other. Therefore, the number of the regions of interest is four, and the four regions of interest are: a first region of interest, a second region of interest, a third region of interest, and a fourth region of interest.

[0066] In another specific implementation, the center coordinates (cx, cy) and radius r of a circular sample pattern are obtained. A preset direction is defined as: translating along pixel rows and columns, and extending a preset number N pixels toward the outside of the image to be measured, starting from the edge of the circular sample pattern. A rectangular region of interest (ROI) is then calculated for each edge: The left-extended region of the pixel column is extended N pixels from the left edge, within the range [cx-r-N, cx-r], and serves as the first ROI. The right-extended region of the pixel column is extended N pixels from the right edge, within the range [cx+r, cx+r+N], and serves as the second ROI. The upper-extended region of the pixel row is extended N pixels from the top edge, within the range [cy-r-N, cy-r], and serves as the third ROI. The lower-extended region of the pixel row is extended N pixels from the bottom edge, within the range [cy+r, cy+r+N], and serves as the fourth ROI. Then, pixels of the first, second, third, and fourth regions of interest in the image to be tested are extracted to generate two regions of interest with a size of H×N and two regions of interest with a size of N*W, respectively, where H is the number of pixels at the left edge and the right edge, and W is the number of pixels at the upper edge and the lower edge.

[0067] In this embodiment, a finer sub-region division allows more sub-regions to be included within the same size of the region of interest. Therefore, based on the positional relationship of the sub-regions, multiple sub-regions need to be selected as the region of interest.

[0068] In one embodiment, for a circular sample pattern and a linear target to be measured, if the image to be measured is divided into sub-areas according to the standard size of the target image, and the sub-areas through which the edges of the sample image pass are used as the areas of interest, there may be incomplete targets to be measured in the sub-areas at the four vertices of the image to be measured.

[0069] For example, Figure 6As shown, the preset direction is parallel to the pixel row and column directions of the image to be tested 610, and the first target to be tested 620 is a linear target parallel to the preset direction; the edge of the sample pattern 630 passes through the sub-region as the focus area for subsequent target detection; wherein, the sub-region 640 for the four vertices of the image to be tested will not be able to cover the complete second target to be tested 650 in the sample pattern 630 due to its short span in the preset direction (pixel row and column directions). In an embodiment of the present invention, the sub-region can include an area extending to one side by a preset length from the edge of the sample pattern along the preset direction, that is, the sub-region at the four vertices of the image to be tested also includes part of the background image. The target to be tested in this embodiment can be shown as the first target to be tested 620 and the second target to be tested 650, depending on whether it is completely covered by the sub-region.

[0070] In other embodiments, at least a portion of the circular sample pattern in the image to be tested can be expanded into a rectangular sample image. For example, the edge or center region of the circular sample pattern can be expanded into a rectangular sample image. This allows the circular sample pattern to be evenly divided into regions large enough to accommodate complete linear defects. Furthermore, to form a standardized rectangular image of each region of interest for input into a preset machine learning model for defect detection, the use of concentric annular or sector-shaped regions is avoided.

[0071] For example, Figure 7 As shown, the edge area of ​​the circular sample pattern is expanded into a rectangular sample image. In one embodiment, the edge area 710 of the circular sample pattern can be expanded into a first rectangular sample image 720. The outer ring edge 711 and the inner ring edge 712 of the edge area 710 are expanded into the bottom side length 721 and the top side length 722 of the first rectangular sample image 720, respectively. The ring width R of the edge area 710 is equal to the wide side of the first rectangular sample image 720. In addition, the inner ring edge 712 can be interpolated to be consistent with the length of the outer ring. In another embodiment, the edge area 710 of the circular sample pattern can be expanded into multiple second rectangular sample images 730. The edge area 710 is divided into multiple sector areas 740. Using the aforementioned expansion method, each sector area 740 is expanded into a different second rectangular sample image 730.

[0072] In this embodiment, the sample pattern of the sample in the image to be measured is circular, and the area of ​​interest is located at the edge of the sample pattern; the target to be measured is linear; the edges of the sample patterns of the multiple areas of interest corresponding to the same preset direction are respectively located on both sides of the dividing line, and the dividing line passes through the center of the sample pattern, and the dividing line is perpendicular to the preset direction or has an acute angle with the preset direction.

[0073] In this embodiment, Figure 8 As shown, a circular sample pattern 811 in a test image 810 is divided into multiple longitudinal sub-regions 812 along a preset direction (the pixel column direction Y of the test image). A transverse dividing line 813 is set perpendicular to the pixel column direction Y (i.e., the transverse dividing line 813 is in the pixel row direction X), and the sub-regions on either side of the transverse dividing line 813 each form two transverse regions of interest 814. Alternatively, the sample pattern can be divided into multiple transverse sub-regions 821 along the pixel row direction X of the test image. A transverse dividing line 822 is set perpendicular to the pixel row direction X (i.e., the transverse dividing line 822 is in the pixel column direction Y), and the sub-regions on either side of the transverse dividing line 822 each form two transverse regions of interest 823.

[0074] 540. Perform a first translation on each sub-region in the region of interest along a preset direction and a preset distance to obtain an intermediate image; segment the intermediate image to obtain multiple target images, each target image having a preset second size and shape.

[0075] In this embodiment, each subregion is divided based on the distribution, size, and angle of the target to be detected on the sample pattern, and may not conform to the standard size for target image detection (preset second size). In this case, each subregion can be first translated along the preset division direction by a preset distance, forming one side length of a preset approximate shape at each starting edge, and obtaining an intermediate image of all subregions at that edge. Finally, the intermediate image is segmented into a target image having the preset second size. At least a portion of the preset approximate shape has the same scale as the preset shape.

[0076] In this embodiment, the sub-areas are of the same size, and the preset distances are the same or different. The edges of the sub-areas are aligned by translating the sub-areas by different preset distances, so that the intermediate image is rectangular. Since the target image is subsequently detected by a machine learning model, the machine learning model needs to input a target image of a specific shape and size. Therefore, in this embodiment, the sub-areas are parallelized to form a rectangular intermediate image, and the target image of the appropriate size is obtained by segmenting the intermediate image. In other embodiments, the target image can also be detected by a traditional threshold comparison algorithm to obtain the target to be measured in the target image, and the sub-areas can be formed into any shape after translation. Alternatively, the detection method does not include the step of translating the sub-areas, and obtaining the target image based on the image of the area of ​​interest includes: directly using the area of ​​interest as the target image.

[0077] For example, Figure 9As shown, for example, the target image is a square with a standard size of 224*224 pixels, the size of the sub-region 910 is divided into 1*224 pixels, and at least a portion of the edge 920 of the sample pattern is arc-shaped; at this time, each sub-region 910 with an arc-shaped distribution needs to be translated to form a partial rectangle with the same scale as the standard size of 224*224 to obtain an intermediate image 930; the intermediate image 930 is divided into two target images 940 with 224*224 pixels.

[0078] In this embodiment, partial edges in each extension direction are determined, and sub-areas starting from the same edge are translated along the extension direction to form an intermediate image, that is, multiple edges form multiple intermediate images respectively, and then each intermediate image is segmented into a target image and detected to obtain the detection results of the target to be measured in each extension direction.

[0079] Specifically, in this embodiment, the target to be measured is linear, and the extension directions of the target to be measured are different; the number of the preset directions is multiple; the detection method includes: repeatedly executing the step of dividing the image to be measured into multiple sub-areas according to each of the preset directions, until the target image is detected and the target to be measured in the sample is obtained. And corresponding to each preset direction, the number of the focus areas is multiple, and the detection method includes: repeatedly executing the step of selecting at least part of the sub-areas within the target range of the image to be measured as the focus area, until the target image is detected and the target to be measured in the sample is obtained. Therefore, in this embodiment, the number of the focus areas is four, and step 540 is executed for each focus area. The two preset directions are respectively along the pixel column Y and pixel row X directions.

[0080] For details, please refer to Figure 8 , each sub-region within the two horizontal regions of interest 814 is translated along pixel column Y to obtain two horizontal intermediate images 830. The horizontal intermediate images 830 are then segmented to obtain multiple horizontal target images 840. Simultaneously, each sub-region within the two vertical regions of interest 823 is translated along pixel row X to obtain two vertical intermediate images 850. The vertical intermediate images 850 are then segmented to obtain multiple vertical target images 860. For example, the target image is a 224*224 pixel square. In this embodiment, the regions of interest are divided along the preset directions of pixel column Y and pixel row X, as shown in the horizontal regions of interest 814 and vertical regions of interest 823. The target images in this embodiment are also shown as the horizontal target images 840 and vertical target images 860.

[0081] In this embodiment, for example, the number of the preset directions is two, and the two preset directions are perpendicular to each other. For the two regions of interest divided along the pixel column direction and the two regions of interest divided along the pixel row direction, the division of linear sub-regions, selection, and determination of the region of interest are repeated for each region of interest, and the intermediate image is further generated, segmented into multiple target images, and the target to be detected is detected in each target image. The method of the present invention is repeated with each region of interest treated as an independent entity until the target to be detected in each region of interest is detected.

[0082] 550. Detect the target image to obtain the target to be detected in the target image.

[0083] In this embodiment, the image to be tested is segmented into target images of preset size and shape to meet the size requirements for image processing of the target detection. At the same time, the target to be tested is completely divided into the same target image to prevent the target from being broken.

[0084] Next, the execution process of the machine learning model applied to the first embodiment of the detection method provided by the present invention is described. Please refer to Figure 10 , as shown below:

[0085] 1010. Acquire an image of a sample to be tested, and divide the image to be tested into a plurality of sub-regions;

[0086] 1020. Select at least a portion of a sub-region within a target range of the image to be measured as a region of interest, where the region of interest includes one or more sub-regions.

[0087] 1030. Acquire a target image according to the image of the area of ​​interest;

[0088] 1040. Input each of the target images into a preset machine learning model, and detect the target to be detected in each of the target images using the preset machine learning model;

[0089] 1050. Map the target to be measured contained in the target image to the image to be measured according to the preset direction to obtain the target to be measured in the sample.

[0090] In this embodiment, the target to be detected may be a linear defect, which may specifically include a scratch, a crack, or a slip line. The following uses a linear defect as an example to illustrate the detection process of the preset machine learning model.

[0091] In this embodiment, the preset machine learning model may be a first model; using the first model, linear defects contained in each target image are detected to obtain angled positioning lines, wherein the angle of the positioning lines is determined by the angle of the linear defects; and based on the angled positioning lines, defect detection results are obtained for each target image. The first model may be an angled target detection model.

[0092] In this embodiment, linear defects refer to defects with a linear shape (exemplarily quantified as: aspect ratio ≥ a preset ratio threshold), such as slip lines (length 100-1000μm, width 0.1-1μm, aspect ratio ≥ 1000), electron migration lines (length 10-100μm, width 10-50nm, aspect ratio ≥ 1000), and laser annealing scan lines (length equal to the full sample size, width 1-5nm, aspect ratio > 1000). Exemplarily, the preset ratio threshold is set to any value between [100, 1000]. Linear defects may include at least one of slip lines, electron migration lines, and laser annealing scan lines.

[0093] Based on this, Figure 11 As shown, it can be determined that when linear defects 1110 are formed in the sample's image 1120 to be tested, they exhibit the following planar characteristics: a very large aspect ratio (extremely small width), resulting in a very small characteristic area; the generated position exhibits random angles, making it difficult to maintain a strict angle parallel or perpendicular to the image to be tested; each type of linear defect has distinct characteristics (linear features) relative to other types of non-linear defects; and different linear defects may be densely distributed or overlap. Using angled positioning lines to identify linear defects helps overcome the problems of small characteristic areas, random angles, and clustered appearance of linear defects, while also leveraging the distinctive characteristics of linear defects to improve detection efficiency.

[0094] Therefore, when locating linear defects by positioning lines, the area range of the linear defects can be accurately covered; in the traditional rectangular bounding box / anchor frame, the linear defect only occupies the diagonal position of the rectangular bounding box / anchor frame, so that the pixel occupancy ratio is low; by replacing the rectangular bounding box / anchor frame with a positioning line, the overlapping of the positioning of different linear defects is avoided, making it impossible to distinguish different linear defects.

[0095] At the same time, due to the extremely large aspect ratio of linear defects, the width of linear defects does not need to be involved in positioning identification. Therefore, when indicating the position of linear defects on the image to be tested, we can mainly focus on two parameters: length and angle. Compared with the rectangular bounding box / anchor box that requires the identification of four vertex coordinates (x1, y1; x2, y2; x3, y3; x4; y4), the positioning line only requires two coordinate points and one angle value (x1, y1; x2, y2; A) to locate the position of the linear defect, reducing the amount of positioning data and dimensions.

[0096] In one embodiment, the sample includes at least two intersecting linear defects, and the preset machine learning model includes a second model; through the second model, instance segmentation is performed on the at least two intersecting linear defects contained in each of the target images to obtain an instance segmentation map, and each linear defect of the at least two intersecting linear defects is represented by a different instance in the instance segmentation map; based on the instance segmentation map, the defect detection result of each of the target images is determined.

[0097] In this embodiment, the second model includes a deep learning model and a non-deep learning model for instance segmentation, wherein the deep learning model includes a model based on a convolutional neural network and a model based on a Transformer; the non-deep learning model may include: a region-based segmentation model, and a computer vision technology model such as a sliding window combined with a classifier model.

[0098] Exemplary models based on convolutional neural networks include Mask R-CNN (Mask Region-based Convolutional Neural Network), YOLACT (You Only LookAt Coefficients, a fully convolutional real-time instance segmentation model), SOLO (Structure of Observed Learning Outcomes), and CenterMas (an anchor-free instance segmentation model); Transformer-based models include Mask2Former (Masked-attention Mask Transformer for Universal Image Segmentation), Mask DINO (Mask Detection Transformer with Improved Denoising Anchor Boxes), and others; region-based segmentation models include the watershed algorithm, GrabCut (graph segmentation algorithm), and the superpixel segmentation algorithm.

[0099] Furthermore, interference fringes migrate according to different phase difference multiples, forming strip defects with different prominent features on the surface of the image to be tested, including prominent features such as bright lines and dark lines. The second model can detect linear defects in the sample in the image to be tested based on bright or dark lines. Regarding the formation of dark lines, dark lines originate from the shadows of concave or convex defects. Therefore, bright lines are always present near convex defects, but not near concave defects. Therefore, dark lines are preferred for characterizing linear defects, eliminating any ambiguity in characterization.

[0100] For at least two linear defects included in the sample; the second model can detect each linear defect from the image to be tested as a different mask, especially in the scene where the linear defects are densely distributed, which can be used to subsequently segment each linear defect into different instances so that each linear defect is separated.

[0101] In this embodiment, the second model uses a deep learning model as an example. Using a convolutional neural network or Transformer, it associates each pixel in the image under test with a global or local spatial relationship, extracting a mask for linear defects with a global or local field of view. With a larger field of view, it is easier to distinguish and predict the area within which each individual linear defect resides.

[0102] Furthermore, if a linear defect, such as a slip line, has a width of 0.1-1μm, the number of pixels occupied by the slip line in the image under test under an optical microscope with a magnification of 100 is 2.9-29. With mask extraction accuracy reaching pixel-level, even slip lines with a very small width can be inspected with more precise pixel occupancy across their width. Furthermore, for densely distributed linear defects, mask separation can also achieve pixel-level accuracy, effectively separating adjacent linear defects.

[0103] When at least two linear defects intersect, the intersecting defects are separated by representing them as different masks. Intersecting linear defects, where two linear defects have a partially overlapping area, can lead to problems such as duplicate predictions of the two linear defects, confusion between the mask boundaries of the predicted linear defects in the overlapping area, difficulties in post-processing, and decreased evaluation metrics for model optimization.

[0104] In this embodiment, the aforementioned second model can be divided into at least three types of instance segmentation models: a bounding box matching-based instance segmentation model, a panoptic segmentation model, and an instance segmentation model. Linear defects in samples are prone to intersecting. Based on this, the mask can clearly identify different instances of the same category within the visible area, making it more reliable to distinguish different linear defects in occluded scenarios.

[0105] Furthermore, regarding the three aforementioned instance segmentation models, the instance segmentation model based on target box matching has higher requirements for the shape of the segmented target. When matching the target box of linear defects, it is easy to include a large amount of background noise and introduce erroneous features (such as introducing the features of another intersecting linear defect), which reduces segmentation accuracy when used for subsequent instance segmentation. The panoramic segmentation model combines semantic segmentation and instance segmentation, which is relatively computationally intensive and reduces detection efficiency. The instance segmentation model directly outputs a mask that just meets the requirements of linear defect and sample detection scenarios, balancing segmentation accuracy and detection efficiency.

[0106] Furthermore, compared to OBB (Oriented Bounding Box) detection, which only outputs the bounding box area, the instance segmentation model outputs both a mask and an instance segmentation map, resulting in more accurate positioning. Compared to semantic segmentation models, which only output pixel categories without masks and instances for each object in the same category, the instance segmentation model outputs both masks and instances for each object in the same category, making post-processing and outputting the detection results for each linear defect much simpler.

[0107] In one embodiment, each of the target images is provided with a division position in the image to be tested; the division positions of multiple target images whose defect forms are linear defects in the defect detection results are determined relative to the image to be tested; based on the division positions, the target images containing the same linear defect are merged to obtain a merged area; the merged area is projected onto the image to be tested to obtain a detection result of the linear defects in the sample.

[0108] Specifically, for example, according to the row number i and column number j of the target image in the image to be tested, the subscript "ij" is used as the dividing position needle; for a long linear defect M, such as a target image with a dividing position 31 and a target image with a dividing position 32, the two target images with dividing positions 31 and 32 are merged, and the merged area 31+32 The linear defect M is projected onto the image to be tested.

[0109] In addition, mapping the target to be measured contained in the target image to the image to be measured includes: translating at least a subregion containing the target to be measured by a preset distance in the opposite direction of the first translation; or mapping the target to be measured contained in the target image to the image to be measured includes: mapping the target to be measured to the image to be measured based on the position coordinates of the target to be measured in the image to be measured. Wherein, the subregion of the aforementioned image to be measured is translated by the first preset distance to obtain the target image, and at this time, the subregion of the target image can be remapped back to the image to be measured by translating the preset distance in the opposite direction of the first translation. It will be clearly understood by those skilled in the art that for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual application, the above-mentioned functions can be assigned to different functional units and modules based on needs, that is, the internal structure of the application device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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 in the form of software functional units. In addition, the specific names of the functional units and modules are only used to facilitate mutual distinction and are not intended to limit the scope of protection of the present invention. The specific working processes of the units and modules in the above-mentioned system can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0110] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0111] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0112] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0113] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the purpose of this embodiment.

[0114] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0115] If the integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the processes in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium.

[0116] The present application also provides a control system. Figure 12 , which shows a schematic diagram of the structure of a control system provided by an embodiment of the present application. Figure 12 As shown, the control system 12000 includes: a processor 1200, a memory 1210, a bus 1220 and a communication interface 1230, and the processor 1200, the communication interface 1230 and the memory 1210 are connected via the bus 1220; the memory 1210 stores computer program instructions that can be executed by the processor 1200, and when the processor 1200 executes the computer program instructions, it executes the detection method provided in any of the aforementioned embodiments of the present application.

[0117] Memory 1210 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between the device network element and at least one other network element is achieved through at least one communication interface 1230 (which may be wired or wireless), and may utilize the Internet, a wide area network, a local area network, a metropolitan area network, or the like.

[0118] The bus 1220 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 1210 is used to store programs. The processor 1200 executes the programs upon receiving execution instructions. The detection method disclosed in any of the aforementioned embodiments of the present application may be applied to the processor 1200 or implemented by the processor 1200.

[0119] The processor 1200 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 1200. The above-mentioned processor 1200 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method 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 mature 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 1210 , and the processor 1200 reads the information in the memory 1210 and completes the steps of the above method in combination with its hardware.

[0120] The control system provided in the embodiment of the present application and the detection method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0121] An embodiment of the present application further provides a computer-readable storage medium corresponding to the detection method provided in the aforementioned embodiment, on which computer program instructions are stored. When the computer program instructions are executed by a processor, they will implement the detection method provided in any of the aforementioned embodiments.

[0122] It should be noted that examples of the computer-readable storage medium may include, but are not limited to, optical discs, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0123] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the detection method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0124] An embodiment of the present application provides a computer program product. When the computer program product is run on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned various method embodiments when executing the computer program product.

[0125] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A detection method, characterized in that: The method comprises: Acquire an image to be tested of a sample, and divide the image to be tested into a plurality of sub-areas; selecting at least a portion of a sub-region within a target range of the image to be measured as a region of interest, wherein the region of interest includes a plurality of the sub-regions; Acquire a target image based on the image of the region of interest, and detect the target image to obtain a target to be detected in the target image, wherein acquiring the target image based on the image of the region of interest comprises: translating subregions of the region of interest starting from the same edge by different preset distances along a preset direction to align the edges of the subregions to form an intermediate image; and segmenting the intermediate image to obtain a plurality of target images; In which, the target range is set based on the physical information of the target to be measured in the sample, the number of targets to be measured in the target range is greater than that in other areas, and the target range includes at least the edge position of the sample pattern, and the edge position includes: a preset number of sub-areas determined from the edge of the sample pattern along the preset direction.

2. The detection method according to claim 1, wherein The multiple sub-regions are arranged in an array; Dividing the image to be measured into multiple sub-areas, including: According to a preset first size and a preset step size, the image to be measured is slidingly divided along at least one preset direction to obtain multiple sub-areas, wherein every two adjacent sub-areas partially overlap or touch but do not overlap; the preset direction is parallel to the extension direction of the target to be measured or has an acute angle with the extension direction of the target to be measured, or the preset direction separates the area where the density of the target to be measured is greater than a threshold from the area where the density of the target to be measured is small.

3. The detection method according to claim 2, characterized in that The preset direction at least includes a direction toward the interior of the image to be measured.

4. The detection method according to claim 1 or 2, characterized in that Before dividing the image to be measured into a plurality of sub-areas, the method further includes: determining a sample pattern in the image to be measured; Dividing the image to be measured into a plurality of sub-areas includes: dividing the sample pattern into a plurality of sub-areas according to a positional relationship between the target to be measured and the sample pattern; The target image has a preset second size and shape.

5. The detection method according to claim 4, characterized in that The target to be measured is linear, the sub-area is strip-shaped, and the preset direction is parallel to the extension direction of the sub-area.

6. The detection method according to claim 5, characterized in that The preset direction is parallel to the row direction or column direction of the pixels of the image to be tested.

7. The detection method according to claim 4, characterized in that There are multiple regions of interest, and the detection method includes: repeatedly selecting at least part of the sub-regions within the target range of the image to be tested as the region of interest, detecting the target image, and obtaining the target to be tested in the sample.

8. The detection method according to claim 7, characterized in that The sample pattern of the sample in the image to be measured is circular, and the area of ​​interest is located at the edge of the sample pattern; the target to be measured is linear; The sample pattern edges of the multiple focus areas are respectively located on both sides of a dividing line, the dividing line passes through the center of the sample pattern, and the dividing line is perpendicular to the preset direction or has an acute angle with the preset direction.

9. The detection method according to claim 4, characterized in that The target to be measured is linear, and the extension directions of the target to be measured are different; The detection method includes: repeatedly dividing the image to be detected into multiple sub-areas according to each of the preset directions, detecting the target image, and obtaining the target to be detected in the sample.

10. The detection method according to claim 4, characterized in that: Detecting the target image to obtain the target to be detected in the sample includes: Each of the target images is input into a preset machine learning model, and the preset machine learning model is used to detect the target to be detected in each of the target images.

11. The detection method according to claim 4, characterized in that: The detection method further comprises: According to the preset direction, the target to be measured contained in the target image is mapped to the image to be measured to obtain the target to be measured in the sample; Mapping the target to be measured contained in the target image to the image to be measured includes: translating at least the sub-area containing the target to be measured by the preset distance in the opposite direction of the preset direction; or mapping the target to be measured contained in the target image to the image to be measured includes: mapping the target to be measured to the image to be measured according to the position coordinates of the target to be measured in the image to be measured.

12. A detection device based on the detection method according to claim 1, characterized in that: comprising a processor configured to: Acquire an image to be tested of a sample, and divide the image to be tested into a plurality of sub-areas; Selecting at least a portion of a sub-region within a target range of the image to be measured as a region of interest, wherein the region of interest includes one or more sub-regions; Acquire a target image based on the image of the region of interest, and detect the target image to obtain a target to be detected in the target image, wherein acquiring the target image based on the image of the region of interest comprises: translating subregions of the region of interest starting from the same edge by different preset distances along a preset direction to align the edges of the subregions to form an intermediate image; and segmenting the intermediate image to obtain a plurality of target images; In which, the target range is set based on the physical information of the target to be measured in the sample, the number of targets to be measured in the target range is greater than that in other areas, and the target range includes at least the edge position of the sample pattern, and the edge position includes: a preset number of sub-areas determined from the edge of the sample pattern along the preset direction.

13. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the detection method according to any one of claims 1 to 11 is implemented.

Images