Wafer defect detection method, system, and storage medium

Abstract

This disclosure provides a wafer defect detection method, system, and storage medium. The system controls an imaging unit to perform stepping motion along a vertical direction perpendicular to the wafer surface, acquiring at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range; candidate defects are identified from each frame of the image; for each frame, the straight-line distance between the candidate defect and the center of the image is determined; based on the straight-line distance, the target defect is determined from the candidate defects in the image, and the detection result of the target defect is output. Through the above stepping imaging and distance determination, the target defect can be accurately captured.

Description

Technical Field

[0001] This disclosure relates to the field of silicon wafer inspection technology, and in particular to a wafer defect detection method, system and storage medium. Background Technology

[0002] With the increase in 12-inch wafer production capacity, the semiconductor manufacturing industry has increasingly stringent requirements for yield control. During the wafer pulling process, due to the influence of temperature gradient and pulling speed fluctuations, bubbles are easily generated inside the ingot, resulting in holes or pits on or near the surface of the finished wafer.

[0003] Currently, machine vision is the mainstream solution for defect detection in the industry. However, in actual inspection processes, 12-inch wafers, due to their extremely high aspect ratio, experience significant center sag under the influence of gravity on the supporting structure. Furthermore, the nanoscale morphological undulations on the wafer surface cause different areas to not be on the same horizontal plane in the vertical direction. Existing optical inspection equipment, when performing high-speed step-scan scanning, often cannot cover this centimeter-scale vertical displacement deviation in its imaging system's depth of field. This results in easily localized defocusing and blurring of the acquired images, making it difficult to distinguish between real defects and background noise or particle interference. Summary of the Invention

[0004] This disclosure provides a wafer defect detection method, system, and storage medium; it can accurately capture defect targets even when there are fluctuations in the physical morphology of the wafer.

[0005] The technical solution disclosed herein is implemented as follows: In a first aspect, this disclosure provides a wafer defect detection method, including: The imaging unit is controlled to move in a stepping motion along a vertical direction perpendicular to the wafer surface, and at least two frames of images to be detected are acquired within a preset vertical travel range for the same horizontal coordinate position. Identify candidate defects in each frame of the image to be detected; For each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected; Based on the straight-line distance, the target defect is determined from the candidate defects in the image to be detected, and the detection result of the target defect is output.

[0006] Secondly, this disclosure provides a wafer defect detection system, including: The image acquisition module is configured to: control the imaging unit to perform stepping motion along a vertical direction perpendicular to the wafer surface, and acquire at least two frames of images to be detected for the same horizontal coordinate position within a preset vertical travel range. The defect screening module is configured to identify candidate defects from each frame of the image to be detected; The defect detection module is configured to: for each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected; and, based on the straight-line distance, determine the target defect from the candidate defects of the image to be detected, and output the detection result of the target defect.

[0007] Thirdly, this disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the wafer defect detection method described in the first aspect.

[0008] This disclosure provides a wafer defect detection method, system, and storage medium. The system controls an imaging unit to perform stepping motion along a vertical direction perpendicular to the wafer surface, acquiring at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range. This stepping acquisition method enables full-coverage monitoring of different height layers of the wafer, ensuring that the defect in at least one frame of the image is precisely within the optimal focal plane of the lens. After preliminary interference removal, candidate defects can be identified from each frame of the image. For each frame, the target defect is determined from the candidate defects based on the straight-line distance between the candidate defect and the center of the image. The smaller the straight-line distance, the closer the candidate defect is to the center of the image, resulting in higher detection reliability and more accurate target defect detection. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the wafer defect detection equipment provided in this disclosure.

[0010] Figure 2 This is a schematic diagram illustrating the sag of a wafer under gravity provided in this disclosure.

[0011] Figure 3 This is a flowchart of the wafer defect detection method provided in this disclosure.

[0012] Figure 4 This is a flowchart for acquiring the image to be detected, as provided in this disclosure.

[0013] Figure 5 This is a schematic diagram of the imaging unit provided in this disclosure performing stepping motion in the vertical direction.

[0014] Figure 6 This is a schematic diagram of optical path correction provided in this disclosure.

[0015] Figure 7 A flowchart for identifying candidate defects provided in this disclosure.

[0016] Figure 8 This is a schematic diagram comparing the same defect in images acquired from different vertical positions provided in this disclosure.

[0017] Figure 9 This is a flowchart for determining the target defect from candidate defects, as provided in this disclosure.

[0018] Figure 10 This is a schematic diagram of the wafer defect detection system provided in this disclosure. Detailed Implementation

[0019] The technical solutions in this disclosure will now be clearly and completely described with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the wafer defect detection equipment provided in this disclosure. Figure 1 As shown, the device includes a support platform 110, a motion control subsystem 120, an imaging unit 130, and a main control computer 140.

[0021] The support platform 110 is used to adsorb and fix the wafer 100 to be tested. In the disclosure, the wafer 100 is a standard 300mm (12-inch) single-crystal silicon polished wafer, and the standard thickness of a wafer of this size is typically 775μm ± 25μm. Due to the extremely high aspect ratio of the wafer, it exhibits the physical characteristics of a flexible thin plate. When the wafer 100 is transferred by the robotic arm to the edge support structure or three-point support position of the support platform 110, the central region of the wafer will naturally sag due to its own gravity. Finite element analysis and experimental data show that the central gravity sag of a 300mm silicon wafer under conventional support is typically between 150μm and 250μm, with the specific value depending on the support method and the wafer material (such as the slight influence of doping concentration on Young's modulus).

[0022] Imaging unit 130 is positioned above support platform 110, with its optical axis aligned with the wafer surface along the Z-axis (i.e., the direction perpendicular to the wafer surface). To eliminate imaging perspective errors and ensure constant magnification across the entire field of view, imaging unit 130 preferably employs a dual telecentric lens in conjunction with a high-resolution industrial camera (such as a 12MP or 25MP Global Shutter CMOS camera). Simultaneously, to highlight the morphological characteristics of hole defects, a coaxial point light source is configured in the optical system to emit incident light onto the wafer surface. Due to the scattering characteristics of their inner walls, hole defects typically appear as dark spots with a specific grayscale gradient under bright-field illumination, while a flat wafer surface presents a bright background.

[0023] The motion control subsystem 120 is the actuator for implementing this disclosure. It includes an air-bearing guide rail or linear motor platform for driving the support platform 110 to move in the XY plane, and a high-precision linear motor module for driving the imaging unit 130 to move along the Z-axis. The Z-axis linear motor is equipped with a grating ruler with a resolution of 0.1 μm as a position feedback sensor, thus forming a fully closed-loop control system capable of precisely executing micron-level stepping movements. In this disclosure, the XY plane is the horizontal support surface where the wafer 100 is located. Specifically, the movement direction of the support platform 110 along the first horizontal guide rail is set as the X-axis, and the movement direction of the support platform 110 along the second horizontal guide rail is set as the Y-axis; the X-axis and Y-axis are perpendicular to each other and jointly define the horizontal support surface where the wafer 100 is located. The direction perpendicular to the XY plane is set as the vertical direction (Z-axis), which is the stepping movement direction of the imaging unit 130.

[0024] The main control computer 140 is connected to the above modules and is responsible for sending motion commands, triggering camera exposure, receiving image data, and executing the steps of this disclosure.

[0025] Figure 2 This is a schematic diagram illustrating the sag of a wafer under gravity, as provided in this disclosure. The vertical axis represents height (typically in μm), and the horizontal axis represents the wafer's position in the X-direction. For example... Figure 2 As shown, as the wafer position changes from "wafer edge - wafer center - wafer edge", its height change exhibits a distinct "U"-shaped parabola. This indicates that when the wafer is placed horizontally, the central region collapses downward (or arches upward) due to gravity or stress, forming a concave surface. This disclosure aims to perform defect detection on wafers considering wafer sag.

[0026] Figure 3 The flowchart of the wafer defect detection method provided in this disclosure is as follows: Figure 3 As shown, the method includes: Step S310: Control the imaging unit to move in a stepping motion along a vertical direction perpendicular to the wafer surface, and acquire at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range.

[0027] In this disclosure, wafer 100 is placed on carrier platform 110, and motion control subsystem 120 drives carrier platform 110 to move in the XY plane, moving the horizontal coordinate position of wafer 100 directly below imaging unit 130.

[0028] In this disclosure, during semiconductor metrology, the imaging unit 130 is preferably a 25MP camera equipped with dual telecentric lenses, whose depth of field (DOF) is typically limited to between 20μm and 45μm. For a 300mm wafer, the center sag can be as high as 200μm. To overcome this physical limitation, this disclosure does not seek a "single focus point" but performs a "scan search." The main control computer 140 controls the motion control subsystem 120 to drive the imaging unit 130 to perform a stepping motion along the vertical direction (Z-axis). Within a preset vertical travel range (e.g., from 0.5mm above to 0.5mm below the nominal height of the wafer), an exposure is triggered at preset intervals. Through this process, a sequence of images to be inspected with different degrees of defocus can be obtained for the same defect coordinate point on the XY plane. This acquisition method ensures that regardless of the undulation of the wafer surface, there will always be at least one frame of the image to be inspected in the sequence, where the defect is exactly within the optimal focal plane of the lens.

[0029] The image to be detected acquired by the imaging unit 130 is transmitted back to the main control computer 140 in real time via a high-speed data link so as to facilitate defect detection in subsequent steps.

[0030] Step S320: Identify candidate defects from each frame of the image to be detected.

[0031] The main defect in this disclosure refers to hole defects. After obtaining the image sequence to be detected, hole defects exhibit a light trapping effect under coaxial bright-field illumination, appearing as dark spots with specific contrast. This disclosure employs a preset feature extraction algorithm to extract all potential abnormal regions whose geometric features meet a preliminary threshold from each frame of the image to be detected, and marks them as candidate defects.

[0032] Step S330: For each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected.

[0033] The lens of imaging unit 130 suffers from field curvature and distortion. Typically, the modulation transfer function is highest and the aberration is lowest in the central region of the lens, resulting in better image quality and measurement accuracy than edge regions. During inspection, the wafer moves continuously relative to the lens. Defects located at the edge of the current image to be inspected may move to the center of the field of view after the next movement step. To avoid repeated detection of the same defect, this disclosure focuses only on defects in the central region of the image to be inspected. Specifically, the centroid coordinates of each candidate defect are calculated. Center point of the image to be detected The straight-line distance between This straight-line distance is used as a key indicator to measure whether a defect is located in the center region of the image to be detected.

[0034] Step S340: Based on the straight-line distance, determine the target defect from the candidate defects in the image to be detected, and output the detection result of the target defect.

[0035] When there are multiple candidate defects in the same frame of the image to be detected, this disclosure selects the candidate defect that is closer to the center of the image to be detected as the target defect by comparing the straight-line distance, and outputs its size, coordinates, grayscale and other parameters.

[0036] In this disclosure, the imaging unit is controlled to move in a stepping motion along a direction perpendicular to the wafer surface, acquiring at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range. This stepping acquisition method enables full-coverage monitoring of different height layers of the wafer, ensuring that the defect in at least one frame of the image is precisely within the optimal focal plane of the lens. After preliminary interference removal, candidate defects can be identified from each frame of the image. For each frame, the target defect is determined from the candidate defects based on the straight-line distance between the candidate defect and the center of the image. The smaller the straight-line distance, the closer the candidate defect is to the center of the image, resulting in higher detection reliability and more accurate target identification, thus improving the accuracy of defect detection results.

[0037] Figure 4 This is a flowchart for acquiring the image to be detected, as provided in this disclosure. Figure 4 As shown, the imaging unit is controlled to move in a stepping motion along a direction perpendicular to the wafer surface, acquiring at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range, including: Step S410: Take the highest point of the wafer surface in the vertical direction as the mechanical origin.

[0038] Specifically, a high-precision ranging sensor is used to scan the edge region of wafer 100 to find the highest point on the wafer surface in the vertical direction. This highest point is defined as the mechanical origin and denoted as Z0.

[0039] Step S420: Using the mechanical origin as a reference, within the vertical travel range including the preset wafer gravity sag range, drive the imaging unit to perform step-by-step imaging at a preset step distance to obtain at least two frames of images to be detected.

[0040] In this disclosure, the vertical travel range is set based on the dimensions of wafer 100 and the expected sag. For example, if the wafer thickness is 0.8 mm, considering its thickness tolerance and sag, the vertical travel range is set to extend downwards by 1 mm from the initial acquisition point (i.e., the mechanical origin). The height distance corresponding to the vertical travel range is much greater than the sum of the physical thickness fluctuation of the wafer and the gravitational sag, providing sufficient safety redundancy.

[0041] The logic origin for each detection action is set at a fixed distance above the mechanical origin (e.g., Z0 + 0.5 mm). This allows for an acceleration and stabilization range for the imaging unit 130, ensuring that when the imaging unit 130 moves downward to reach its true vertical travel range, it has entered a uniform and stable operating state, thus avoiding image blurring caused by startup vibration.

[0042] When inspecting a specific field of view of the wafer, the XY plane remains stationary, locking the horizontal coordinate position. The imaging unit 130, starting from the logical origin, moves downwards along the Z-axis at a preset step interval under the drive of the motion control subsystem 120. The step interval is less than or equal to the optical depth of field of the imaging unit 130, for example, 50 μm.

[0043] Whenever the grating ruler feedback imaging unit 130 descends by 50μm, the grating ruler sends a positioning signal to the motion control subsystem 120. The motion control subsystem 120 then transmits the corresponding position coordinates to the main control computer 140, triggering the camera to turn on its light source and sending a photographing command to the imaging unit 130. Upon receiving the photographing command, the imaging unit 130 triggers the photographing action, acquiring a frame of the image to be detected. This process continues until the imaging unit 130 has completed its entire vertical travel range.

[0044] Taking a step size of 50μm and a vertical distance of 1mm corresponding to the vertical travel range as an example, a sequence of 20 images with different focal planes will be acquired for the same horizontal coordinate position. In this sequence, there will inevitably be one or more images that are precisely focused on the defect on the wafer surface, thus physically eliminating the possibility of missed detection due to defocus.

[0045] Figure 5 This is a schematic diagram illustrating the stepping motion of the imaging unit provided in this disclosure along the vertical direction. For example... Figure 5 As shown, the imaging unit 130 moves vertically downwards from the highest point, taking one picture of the wafer 100 with each movement to obtain an image of the target at the corresponding vertical height. Through the stepping images taken by the imaging unit 130, a continuous sequence of images of the target can be obtained.

[0046] Figure 6 This is a schematic diagram of optical path correction provided in this disclosure. Figure 6 (a) shows the optical path of the point light source without optical path correction, with the light rays radiating outwards from the top of the point light source 610. Due to the large divergence angle of the beam, when it illuminates a wafer surface with undulations or sag, the center position of the beam spot will undergo a drastic geometric shift with changes in height, causing the wafer defect detection equipment to be unable to obtain a stable coordinate reference. Figure 6(b) illustrates the optical path of the point light source after optical path correction in this disclosure. A lens group 620 is added to the emitting end of the point light source 610. After the light emitted by the light source passes through the lens group 620, the divergence angle is significantly reduced, and the beam tends to be parallel or collimated. Under the illumination of the beam corrected by the lens group 620 in this disclosure, regardless of how the wafer is drooping, the projection position of the defect on the wafer surface remains relatively stable, thereby making the brightness of the image to be detected acquired by the imaging unit 130 more uniform, and thus improving the accuracy of defect detection.

[0047] Figure 7 This is a flowchart for identifying candidate defects provided in this disclosure. The candidate defect is a hole defect. For example... Figure 7 As shown, candidate defects are identified from each frame of the image to be detected, including: Step S710: Binarize each frame of the image to be detected and extract the connected regions representing the defects.

[0048] Specifically, the image to be detected is first denoised. Considering that holes and defects are usually small targets at the micrometer level, in order to preserve the high-frequency information of the hole and defect edges while denoising, this disclosure preferably uses a 5×5 Gaussian filter for denoising. The standard deviation of the Gaussian filter is... Set to 1.0 to 1.5. Compared to mean filtering, Gaussian filtering removes noise more smoothly and is less likely to blur defect boundaries.

[0049] The color or multi-channel image is then converted into a single-channel grayscale image with a pixel grayscale value range of 0-255.

[0050] To separate hole defects from the background, this disclosure employs an adaptive binarization algorithm to calculate the average gray level within the neighborhood (e.g., a 31×31 pixel block) of each pixel in the grayscale image, and subtracts a constant C (e.g., C=5) as the local threshold for that pixel. If the gray level of a pixel is lower than its local threshold, it is determined to be a "dark spot," representing a potential hole region, because holes cannot reflect coaxial light and appear dark, and can be marked as "0"; otherwise, it is marked as "1."

[0051] After the binarization process described above, the grayscale image is converted into a black-and-white mask. Then, a connected component labeling (CCL) algorithm is used to aggregate adjacent black pixels in the black-and-white mask into an independent connected region. This connected region may be a circular hole or a non-circular defect such as a scratch, requiring further screening in subsequent steps.

[0052] Step S720: Based on preset roundness threshold, shape factor threshold and grayscale threshold, the connected regions are filtered to obtain hole defects.

[0053] Among them, roundness is a dimensionless parameter used to measure the similarity between a connected region and a standard circle, and shape factor is a key geometric feature parameter used to quantify the complexity or smoothness of the contour of a connected region.

[0054] Specifically, connected regions are filtered based on preset roundness threshold, shape factor threshold, and grayscale threshold to obtain hole defects, including: First-level filtering: By comparing the roundness of each connected region with a roundness threshold, the first target region is selected.

[0055] Among these, scratches on the wafer surface are typically elongated strips, while pore defects formed by bubbles exhibit highly regular circular or elliptical cross-sections after slicing. Therefore, this disclosure filters out particle interference and scratch interference with abnormal boundary contours by determining whether the roundness of the connected region is greater than a roundness threshold. The roundness can be obtained by calculating the ratio of the area of ​​the connected region to the area of ​​its smallest circumcircle. The formula for calculating roundness is, for example: ; In the formula, Area represents the area (number of pixels) of the connected region. This represents the maximum distance from the center of the connected region to its boundary contour. Roundness The closer the value is to 1, the more rounded the connected region. In this disclosure, a roundness threshold of 0.8 is assumed. If the roundness of the connected region... C If the value is less than 0.8, it will be filtered out as an interference item, thus selecting the first target area that is roughly circular.

[0056] Second-level screening: By comparing the shape factor of each first target region with the shape factor threshold, second target regions are selected; the shape factor is calculated from the perimeter and area of ​​the first target region.

[0057] Particulate contamination on wafer surfaces is typically caused by environmental dust or process residue. Its edges are often irregular and jagged, while voids and defects exhibit highly regular circular or elliptical cross-sections after slicing. Therefore, this disclosure filters out particulate interference by determining whether the shape factor of the connected region is greater than a shape factor threshold. The formula for calculating the shape factor is, for example: ; In the formula, Area is the area of ​​the connected region (number of pixels), Perimeter is the perimeter of the connected region, and for a perfect circle, the shape factor is... The value is 1.0. The shape factor threshold set in this disclosure is, for example, 0.85. To further filter out noise points with very few pixels, an area limit can be added, for example, Area ≥ 5 pixels. If the shape factor of a connected region is less than 0.85, or its area is less than 5 pixels, it is considered noise or a non-critical defect and is removed, thereby filtering out a second target region with normal boundary contours.

[0058] The third level of screening: By comparing the average gray value of each second target area with the gray value threshold, hole defects are screened out.

[0059] Under coaxial bright-field illumination, true hole defects have extremely low internal grayscale values ​​(close to black) due to the light trapping effect. However, some surface pits or metallic contaminants, although circular, may have high reflectivity and brighter grayscale. This disclosure sets a grayscale threshold range, for example, [0, 100]. For each second target area, its average grayscale value is calculated. If the average grayscale value falls within the grayscale threshold range, it is confirmed as a hole defect; if the grayscale value is too high (e.g., >100), it may be a surface spot and is excluded.

[0060] Through the above three-level screening, a set of candidate defects with high credibility can be extracted from each frame of the image to be detected.

[0061] Figure 8 This is a comparative illustration of the same defect in images acquired from different vertical positions provided in this disclosure. Figure 8 (a) is the image to be detected acquired at the first vertical position, where the defect is relatively clear; while Figure 8 (b) is the image to be detected acquired at the second vertical position, where the defect is relatively blurry. For ease of analysis, this disclosure requires selecting a clearer defect as the target defect to improve the accuracy of subsequent defect detection.

[0062] based on Figure 8 The problem shown is that the clarity of the same defect varies in different images to be detected. Figure 9 This is a flowchart for determining the target defect from candidate defects, as provided in this disclosure. Figure 9 As shown, the target defect is determined from candidate defects in the image to be detected based on straight-line distance, including: Step S910: For each frame of the image to be detected, the candidate defect with the smallest straight-line distance is determined as the corresponding alternative target. If there is no candidate defect in the frame of the image to be detected, the frame of the image to be detected is marked as empty.

[0063] In this disclosure, for each frame of the image to be detected, only the defect closest to the center point is considered. In a continuous sequence of images to be detected, the probability of the same defect appearing consecutively in the center region of two adjacent frames is relatively small, thereby reducing the occurrence of repeated detections.

[0064] Step S920: Compare the sharpness of the candidate targets corresponding to each frame of the image to be detected, and select the candidate target with the highest sharpness as the target defect.

[0065] After identifying the candidate targets for each frame of the image to be detected, since these candidate targets may physically represent the same defect (e.g., the same defect may appear consecutively in the center region of two adjacent frames of the image to be detected), it is necessary to select the clearest candidate defect as the target defect.

[0066] This disclosure allows the extraction of edge gradient values ​​or contrast values ​​of candidate targets from all images to be detected as a sharpness evaluation metric. The sharpness evaluation function can employ the Laplacian variance or the Tenengrad function. The formula for calculating sharpness using the Laplacian variance is as follows: ; In the formula, I(x,y) is the image brightness function, which represents the gray value (brightness) of the pixel with coordinates (x,y) in the image to be detected; This refers to the Laplacian operator, specifically the second derivative operator. In the image to be detected, the second derivative reflects the drastic nature of gray-level changes; at sharp edges, the abrupt changes in gray-level are extremely drastic. The values ​​are relatively large; while in blurred or smooth areas, the grayscale changes slowly. The value is close to zero.

[0067] In this disclosure, the clarity score of the candidate targets corresponding to all the images to be detected in the image sequence is compared, and the candidate target with the highest score (i.e. the clearest outline and the sharpest edge) is selected as the target defect for this detection action.

[0068] This method consistently selects the central region data with the best optical imaging quality in the horizontal direction (XY plane), improving the accuracy of dimensional measurements. In the vertical direction (Z-axis), it automatically "autofocuses" from multi-layer scans using an algorithm, eliminating mechanical positioning errors and defocusing blur caused by gravity sagging. Even if wafer sagging causes a certain shift in the optimal focal plane, this logic can automatically find the clear image after the shift in the image sequence to be detected.

[0069] In this disclosure, at least two frames of images to be detected are acquired for each horizontal coordinate position, and candidate defects with the closest straight-line distance are identified. The image with the clearest features among the candidate defects is selected as the target defect at the corresponding horizontal coordinate position. Target defects at multiple horizontal coordinate positions can be obtained by continuously scanning the wafer.

[0070] During continuous wafer scanning, if two adjacent defects are very close together, the wafer defect detection equipment may experience micron-level deviations due to thermal drift or mechanical hysteresis. This causes the original target defect to move away from the center of the image being inspected, while a spurious target defect (noise or a minor defect) may move closer to the center of the image due to the deviation. This can easily lead to missed detections or misjudgments. Furthermore, there is a possibility of repeated detection of the same target defect.

[0071] To avoid missing detections or recording the same target defect multiple times, in this disclosure, after determining the target defect from candidate defects in the image to be detected, the wafer defect detection method further includes: For the image to be detected at the current horizontal coordinate position, a judgment region is established based on the target defect; If the determination region in the image to be detected corresponding to the current horizontal coordinate position has the same features as the determination region in the image to be detected corresponding to the adjacent horizontal coordinate position, then the target defect in the image to be detected corresponding to the current horizontal coordinate position is filtered out, and the target defect is re-determined from the candidate defects.

[0072] Specifically, when at the horizontal coordinate position Determining from the k-th frame of the acquired image to be detected The clearest target defect T k Then, using the coordinates of the target defect... Based on this, a determination region is constructed. In this disclosure, the determination region is a local feature window formed by extending a specified distance in at least one preset vector direction from the centroid of the target defect as the origin, defining a reference point, and centered on the reference point with a specified side length. For example, using... Using the centroid as the reference point, eight sampling points are determined in eight discrete geometric directions (usually 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°) with a preset step size L. A rectangular local feature window with a side length of 10 pixels is selected with the eight sampling points as the center.

[0073] This disclosure involves backtracking the images to be detected at 1-2 adjacent horizontal coordinate positions, for example... and The detection results for the corresponding image to be detected. For example, if The features of the corresponding detection region in the image to be detected and or If the features of the corresponding region in the image to be detected highly match (e.g., similarity > 0.95), then it is determined that... The corresponding image to be detected actually captures or The corresponding defect in the image to be detected has already been recorded (e.g., defect A). However, when mechanical displacement occurs, In the corresponding image to be detected, defect A is closest to the center due to drift, while defect B, which should also be closest to the center, may now be the second closest. If only simple filtering is performed, defect B will be... The corresponding images to be detected are completely ignored, resulting in missed detections. This disclosure automatically removes... In the corresponding image to be detected, defect A is identified as the target defect, and defect B, which is the second closest in the straight-line distance from the candidate defect list, is re-identified as the target defect. The corresponding target defect in the image to be detected.

[0074] like The features of the corresponding detection region in the image to be detected and or If the features of the corresponding region in the image to be detected do not match, it indicates that it is a real defect that has been captured for the first time, and it is confirmed as such. The final target defect in the corresponding image to be detected is identified and the result is output.

[0075] Figure 10 This is a schematic diagram of the wafer defect detection system provided in this disclosure. Figure 10 As shown, the wafer defect detection system 1000 includes: The image acquisition module 1010 is configured to: control the imaging unit to perform stepping motion along a vertical direction perpendicular to the wafer surface, and acquire at least two frames of images to be detected for the same horizontal coordinate position within a preset vertical travel range. The defect screening module 1020 is configured to identify candidate defects from each frame of the image to be detected; The defect detection module 1030 is configured to: for each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected; and, based on the straight-line distance, determine the target defect from the candidate defects of the image to be detected, and output the detection result of the target defect.

[0076] The wafer defect detection system disclosed herein can implement the wafer defect detection method provided herein, and their implementation principles are similar. The actions performed by each module and unit in the wafer defect detection system disclosed herein correspond to the steps of the wafer defect detection method disclosed herein. For a detailed description of each module and unit in the wafer defect detection system, please refer to the description of the corresponding wafer defect detection method shown above, which will not be repeated here.

[0077] This disclosure also provides a computer-readable storage medium storing at least one instruction that is executed by a processor to implement the wafer defect detection method as described in the above embodiments.

[0078] This disclosure also provides a computer program product including computer instructions stored in a computer-readable storage medium; a processor of a computing device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computing device to perform the wafer defect detection method described in the above embodiments.

[0079] Those skilled in the art will recognize that the functions described in this disclosure in one or more of the examples above can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium accessible to a general-purpose or special-purpose computer.

[0080] It should be noted that the technical solutions described in this disclosure can be combined arbitrarily as long as they do not conflict.

[0081] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A wafer defect detection method, characterized by, include: The imaging unit is controlled to move in a stepping motion along a vertical direction perpendicular to the wafer surface, and at least two frames of images to be detected are acquired within a preset vertical travel range for the same horizontal coordinate position. Candidate defects are identified from each frame of the image to be detected; For each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected; Based on the straight-line distance, the target defect is determined from the candidate defects in the image to be detected, and the detection result of the target defect is output.

2. The wafer defect detection method of claim 1, wherein The controlled imaging unit moves in a stepping motion along a direction perpendicular to the wafer surface, acquiring at least two frames of images to be detected at the same horizontal coordinate position within a preset vertical travel range, including: The highest point of the wafer surface in the vertical direction is taken as the mechanical origin; Using the mechanical origin as a reference, within a vertical travel range including a preset wafer gravity sag range, the imaging unit is driven to perform step-by-step imaging at a preset step distance to obtain at least two frames of the image to be detected.

3. The wafer defect detection method of claim 1, wherein The candidate defect is a void defect; The step of identifying candidate defects from each frame of the image to be detected includes: Each frame of the image to be detected is binarized, and the connected regions representing defects are extracted. The connected regions are filtered based on preset roundness threshold, shape factor threshold and grayscale threshold to obtain the hole defects.

4. The wafer defect detection method of claim 1, wherein The step of determining the target defect from the candidate defects in the image to be detected based on the straight-line distance includes: For each frame of the image to be detected, the candidate defect with the smallest straight-line distance is determined as the corresponding alternative target; The sharpness of the candidate targets corresponding to each frame of the image to be detected is compared, and the candidate target with the highest sharpness is selected as the target defect.

5. The wafer defect detection method according to claim 1, characterized in that, After determining the target defect from the candidate defects in the image to be detected based on the straight-line distance, the wafer defect detection method further includes: For the image to be detected corresponding to the current horizontal coordinate position, a judgment region is established based on the target defect; If the determination region in the image to be detected corresponding to the current horizontal coordinate position has the same features as the determination region in the image to be detected corresponding to the adjacent horizontal coordinate position, then the target defect in the image to be detected corresponding to the current horizontal coordinate position is filtered out, and the target defect is re-determined from the candidate defects.

6. The wafer defect detection method according to claim 5, characterized in that, The determination region is a local feature window formed by taking the centroid of the target defect as the origin, extending a specified distance in at least one preset vector direction to determine a reference point, and selecting the reference point as the center with a specified side length.

7. The wafer defect detection method according to claim 3, characterized in that, The process of filtering the connected regions based on preset roundness thresholds, shape factor thresholds, and grayscale thresholds to obtain the hole defects includes: The first target region is selected by comparing the roundness of each connected region with the roundness threshold. Second target regions are selected by comparing the shape factor of each first target region with the shape factor threshold; the shape factor is calculated from the perimeter and area of ​​the first target region. The hole defects are screened out by comparing the average gray value of each of the second target regions with the gray value threshold.

8. The wafer defect detection method according to claim 1, characterized in that, The vertical travel range is set according to the wafer size and the expected droop.

9. A wafer defect detection system, characterized in that, include: The image acquisition module is configured to: control the imaging unit to perform stepping motion along a vertical direction perpendicular to the wafer surface, and acquire at least two frames of images to be detected for the same horizontal coordinate position within a preset vertical travel range. The defect screening module is configured to identify candidate defects from each frame of the image to be detected; The defect detection module is configured to: for each frame of the image to be detected, determine the straight-line distance between the candidate defect and the center of the image to be detected; and, based on the straight-line distance, determine the target defect from the candidate defects of the image to be detected, and output the detection result of the target defect.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the wafer defect detection method as described in any one of claims 1 to 8.

Images