Wafer edge structure offset detection method, device, equipment and storage medium

Abstract

The specification discloses a method, apparatus, device, and storage medium for detecting wafer edge structure offset. The method includes: generating candidate rings corresponding to the wafer edge and at least one edge ring structure in the wafer edge region based on image data of the wafer under test; dividing each candidate ring into multiple ring regions along its radial direction, and constructing multi-channel radial profile features of the candidate rings based on pixel features within each ring region; determining the matching score and matching position information corresponding to each candidate ring; determining the center position based on the matching score, and determining a radius parameter based on the matching score and matching position information; and calculating the relative position between the wafer edge and at least one edge ring structure based on the center position and radius parameter. This enables accurate identification of ring structures and calculation of their offset information under complex imaging conditions, achieving effective monitoring and closed-loop control of edge processes in semiconductor manufacturing.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of semiconductor technology, and more particularly to a method, apparatus, device, and storage medium for detecting wafer edge structure offset. Background Technology

[0002] In semiconductor manufacturing, the relative positional accuracy between the wafer edge and various edge ring structures within its edge region is a critical process affecting patterning quality, preventing particle contamination, and ensuring overall wafer yield. As examples, the Edge Bead Removal (EBR) and Edge Protection Ring (EPR) are two typical edge ring structures.

[0003] EBR (Extended Boundary Removal) is a ring-shaped area without photoresist formed by removing the photoresist around the edge of the wafer using chemical solvents after photoresist coating. If the EBR removal is insufficient, the residual photoresist at the edge will detach as contaminant particles during subsequent processes, affecting the cleanliness of the wafer surface; if the EBR removal is too wide, it may damage the effective chip area at the wafer edge.

[0004] EPR (Epoxy Reflection Plate) is a ring-shaped protective structure formed during the photolithography exposure stage through mask pattern transfer. It is used to protect the wafer edges during subsequent etching, deposition, and other processes. If the EPR is not aligned correctly, the edge structure will lose its protection, leading to cracks or uneven film layers, directly affecting the reliability of chip edge devices.

[0005] With the development of advanced manufacturing processes, EPR structures are becoming increasingly complex, exhibiting special morphologies such as fractures and discontinuities. Simultaneously, factors such as wafer edge chamfering, optical interference effects, and process noise degrade imaging quality, making image distortion in edge regions more pronounced. Traditional edge detection and feature recognition methods rely on clear edge point extraction, making it difficult to stably and accurately complete wafer edge structure detection tasks under conditions of poor imaging quality and structural discontinuities. Summary of the Invention

[0006] In view of this, this specification provides a method and apparatus for detecting wafer edge structure offset.

[0007] Specifically, this specification is implemented through the following technical solution: According to a first aspect of one or more embodiments of this specification, a method for detecting wafer edge structure offset is provided, comprising: Based on the image data of the wafer under test, candidate rings corresponding to the wafer edge and at least one edge ring structure are generated in the wafer edge region. For each candidate ring, multiple annular regions are divided along the radial direction of the candidate ring, and a multi-channel radial profile feature of the candidate ring is constructed based on the pixel features in each annular region. The multi-channel radial profile features of each candidate ring are matched with the corresponding reference features to determine the matching score and matching position information of each candidate ring. The center positions of the wafer edge and at least one edge ring structure are determined based on the matching score. The matching position information is then optimized based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure. Based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure, the offset information between the wafer edge and the at least one edge ring structure is calculated.

[0008] Optionally, dividing the candidate annulus into multiple annular regions along its radial direction includes: Using the candidate annulus as a reference circumference, the annulus is extended equidistantly inward and outward to obtain multiple concentric annular regions. Wherein, the maximum radial extension length when constructing the multi-channel radial profile feature corresponding to the candidate annulus is less than the maximum radial extension length when constructing the reference feature.

[0009] Optionally, constructing the multi-channel radial profile feature of the candidate annulus based on pixel features within each annular region includes: The original image and the corresponding gradient feature image of the wafer under test are obtained respectively; In the original image of the wafer under test and the gradient feature image, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the candidate annulus.

[0010] Optionally, generating candidate rings in the wafer edge region corresponding to the wafer edge and at least one edge ring structure respectively includes: The wafer region is segmented from the image of the wafer to be tested, and the initial center is determined based on the wafer region; Multiple candidate center points are generated within a preset tolerance error radius, with the initial center point as the center. Using each candidate circle center as the center and the reference radius corresponding to the wafer edge and at least one edge ring structure as the radius, candidate rings corresponding to the wafer edge and at least one edge ring structure are generated.

[0011] Optionally, determining the initial center based on the wafer region includes: A circle fitting algorithm is used to fit the contour points of the wafer region to determine the initial circle center.

[0012] Optionally, the baseline features are pre-constructed in the following manner: In the reference wafer image, determine the wafer edge and the reference ring corresponding to the at least one edge ring structure, respectively; For each of the aforementioned reference rings, multiple annular regions are divided along its radial direction; In the original image and the corresponding gradient feature image of the reference wafer, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the reference ring as the reference features.

[0013] Optionally, the gradient feature image is obtained by calculating the Sobel gradient magnitude on the original image.

[0014] Optionally, extracting pixel features for each of the annular regions includes: Calculate multiple image statistics within each of the aforementioned annular regions; The image statistics corresponding to each annular region are combined radially to form multiple image feature vectors, thereby constructing the multi-channel radial profile feature.

[0015] Optionally, the step of performing a matching analysis between the multi-channel radial profile features of each candidate ring and the corresponding reference features to determine the matching score and matching position information of each candidate ring includes: The multi-channel radial profile features of each candidate ring are cross-correlated with the corresponding reference features to obtain the matching score and matching position information. The matching position information is used to characterize the radial offset of the candidate ring relative to the corresponding reference ring.

[0016] Optionally, the step of cross-correlation calculation of the multi-channel radial profile features of each candidate annulus with the corresponding reference features to obtain the matching score and matching position information includes: Each image feature vector is cross-correlated with the corresponding reference image feature vector in the reference features to obtain multiple sub-matching scores and multiple sub-matching position information; The matching score is calculated based on the multiple sub-matching scores, and the matching location information is calculated based on the multiple sub-matching location information.

[0017] Optionally, determining the center position of the wafer edge and at least one of the edge ring structures based on the matching score includes: For each candidate ring, a three-dimensional spatial distribution map is constructed using the candidate center coordinates of the candidate ring as planar coordinates and the corresponding matching score as the height value. After smoothing the height value, the highest peak point is located; The highest peak point is determined as the center position of the wafer edge or edge ring structure corresponding to the candidate ring.

[0018] Optionally, the step of optimizing the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one of the edge ring structures includes: For each candidate ring, the matching position information of the candidate ring is subjected to anomaly filtering to remove abnormal matching position information; The remaining matching position information is weighted and averaged using the matching score as the weight to obtain optimized matching position information; The radius parameter is calculated based on the optimized matching position information and the reference radius of the wafer edge or edge ring structure corresponding to the candidate ring.

[0019] Optionally, at least one of the edge ring structures includes an edge photoresist removal region (EBR) and / or an edge protection ring (EPR).

[0020] According to a second aspect of one or more embodiments of this specification, a wafer edge structure offset detection device is provided, comprising: The candidate ring generation unit generates candidate rings in the wafer edge region based on the image data of the wafer to be tested, which correspond to the wafer edge and at least one edge ring structure, respectively. The feature construction unit divides each candidate ring into multiple annular regions along the radial direction of the candidate ring, and constructs a multi-channel radial profile feature of the candidate ring based on the pixel features in each annular region. The feature matching unit performs matching analysis on the multi-channel radial profile features of each candidate ring and the corresponding reference features to determine the matching score and matching position information of each candidate ring. The center radius determination unit determines the center position of the wafer edge and at least one edge ring structure based on the matching score, and optimizes the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure. The offset calculation unit calculates the offset information between the wafer edge and the at least one edge ring structure based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure.

[0021] According to a third aspect of one or more embodiments of this specification, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor implements the steps of the aforementioned wafer edge structure offset detection method by executing the executable instructions.

[0022] According to a fourth aspect of one or more embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the aforementioned wafer edge structure offset detection method.

[0023] According to a fifth aspect of one or more embodiments of this specification, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the aforementioned method for detecting wafer edge structure offset.

[0024] Using the above implementation method, candidate rings corresponding to the wafer edge and at least one edge ring structure are first generated in the wafer edge region. Then, for each candidate ring, a multi-channel radial profile feature corresponding to the candidate ring is constructed based on the pixel features within each ring region. Replacing the single continuous boundary point relied upon by traditional methods with pixel features within the ring region, stable characterization can still be obtained even if the edge ring structure has breaks, discontinuities, or local defects. This solves the problems of circle fitting failure due to edge ring structure breaks and edge detection instability due to image quality degradation. Next, the multi-channel radial profile features of each candidate ring are matched with the corresponding reference features to determine the matching score and matching position information, thereby determining the center position and radius parameters. Offset information is then calculated based on the center position and radius parameters corresponding to each structure. Thus, under complex imaging conditions, the ring structure can be accurately identified and its offset information calculated, achieving effective monitoring and closed-loop control of semiconductor manufacturing edge processes. Attached Figure Description

[0025] Figure 1 This is a schematic diagram illustrating an exemplary embodiment of the detection process for wafer edge structure offset.

[0026] Figure 2 This is a schematic flowchart illustrating an exemplary embodiment of the method for generating a candidate ring in this specification.

[0027] Figure 3 This is a schematic diagram illustrating a manually annotated example of an exemplary embodiment of this specification.

[0028] Figure 4 This is a schematic diagram illustrating an exemplary embodiment of the present specification, showing the division of a ring-shaped region.

[0029] Figure 5This is a flowchart illustrating an exemplary embodiment of the present specification of a method for constructing a candidate annular multi-channel radial profile feature.

[0030] Figure 6 This is a flowchart illustrating an exemplary embodiment of a method for determining matching scores and matching location information.

[0031] Figure 7 This is a flowchart illustrating an exemplary embodiment of a method for determining the center position of a circle.

[0032] Figure 8 This is a flowchart illustrating an exemplary embodiment of a method for determining a radius parameter.

[0033] Figure 9 This is a schematic diagram of the structure of a device provided in an exemplary embodiment of this specification.

[0034] Figure 10 This is a block diagram illustrating an exemplary embodiment of a wafer edge structure offset detection device. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] In semiconductor manufacturing processes, the wafer edge region is one of the key areas affecting overall yield. This region typically involves the wafer edge and at least one edge ring structure. As examples, EBR and EPR are two typical edge ring structures that are interconnected.

[0038] The physical edge of the wafer is the boundary of the wafer substrate itself, which is chamfered to form an arc-shaped profile. It serves as the geometric reference for all subsequent edge processes and will be referred to as the wafer edge in this specification. The EBR (Extended Boundary Resin) is a ring-shaped area formed by selectively removing the photoresist around the wafer edge using chemical solvents after the photoresist coating process. Its purpose is to prevent the edge photoresist from detaching and causing particulate contamination during subsequent processes due to mechanical friction or thermal stress. The EPR (Extended Peripheral Resin) is a ring-shaped protective structure formed during the photolithography exposure stage through mask pattern transfer. It is used to protect the wafer edge area during etching, deposition, and other processes, preventing structural damage or uneven film layers caused by process environment erosion.

[0039] Under ideal process conditions, the three ring structures described above should have a consistent center position. However, in actual production, due to limitations in the edge positioning accuracy of the coating equipment, the alignment deviation of the lithography machine, and the reference transfer error between multiple devices, the centers of the EBR and EPR often shift relative to the center of the wafer edge. This shift can lead to numerous process risks. For example, an EBR center shift may cause uneven removal of the photoresist at the edge. If the removal is insufficient, the residual photoresist will detach and become contaminant particles in subsequent processes, contaminating the effective area of ​​the wafer surface. If the removal is too extensive, it may erode the effective chip area at the wafer edge. As another example, an EPR center shift may prevent the edge protection ring from accurately covering the edge structure that needs protection. This results in the wafer edge area losing protection during etching, deposition, and other processes, leading to cracks, uneven film thickness, or edge chip pattern distortion, affecting the reliability and yield of chip edge devices.

[0040] Therefore, accurately measuring the offset information (such as relative offset and offset direction) between the wafer edge and various edge ring structures in the edge region is a key step in realizing closed-loop control of edge processes and ensuring the overall yield of the wafer.

[0041] Traditional methods for detecting wafer edge structure offsets primarily rely on two-dimensional vision imaging technology. Edge detection algorithms extract boundary points of each ring structure, and then algorithms such as circle fitting are used to calculate the center position and offset. Acceptable detection results can be obtained when the imaging quality is ideal and the ring structure is intact. However, with the evolution of advanced manufacturing processes, these traditional methods are no longer able to reliably and accurately complete the detection task. For example, in advanced processes, to optimize stress distribution and process compatibility in edge regions, EPR patterns are becoming increasingly complex, exhibiting special structures such as breaks, discontinuities, and segmented arcs. Traditional edge detection algorithms rely on continuous boundary point sets for circle fitting. When the EPR pattern has breaks or missing points, the number of extractable boundary points is significantly reduced, or even unable to form effective arc segments, leading to circle fitting failure or a sharp decrease in center positioning accuracy.

[0042] Therefore, there is an urgent need for a wafer edge structure offset method that can reliably and accurately identify broken ring structures and calculate their offset information under complex imaging conditions, so as to achieve effective monitoring of semiconductor manufacturing edge processes.

[0043] This specification provides a detection scheme for wafer edge structure offset, which performs multi-channel radial profile matching analysis between the annular region image of the wafer under test and the reference feature template to achieve high-precision and high-reliability center positioning and offset calculation, overcoming the limitations of traditional edge detection methods under complex imaging conditions.

[0044] The term "wafer edge" as used in this specification refers to the physical boundary of the wafer substrate itself, which is chamfered to form an arc-shaped profile and serves as the geometric reference for all subsequent edge processes.

[0045] The “edge ring structure” described in this specification includes the aforementioned edge photoresist removal region (EBR) and edge protection ring (EPR).

[0046] The “multi-channel radial profile feature” mentioned in this specification refers to a set of feature vectors composed of various statistical measures (such as mean, median, interquartile range, range, etc.) extracted from the original grayscale image and gradient feature image respectively in multiple concentric annular regions along the radial direction of the annulus. It is used to characterize the grayscale distribution and edge distribution pattern of the annulus at different radial distances.

[0047] The “offset information” mentioned in this specification refers to the positional deviation of the center of an edge ring structure such as EBR or EPR relative to the center of the wafer edge. The positional deviation includes the relative offset amount and the offset direction, which can be used to quantify the degree of deviation in the edge process.

[0048] The wafer edge structure offset detection scheme provided in this manual can be applied to electronic devices, such as PCs, mobile phones, tablets, laptops, and PDAs (Personal Digital Assistants).

[0049] The wafer edge structure offset detection scheme provided in this manual can also be applied to CS (Client-Server) or BS (Browser-Server) architectures. Taking the CS architecture as an example, the client can upload image data of the wafer under test, and the server can perform offset detection on the edge structure of the wafer under test based on the image data and return the detection results to the client for user viewing, etc. This manual does not impose any special limitations on this.

[0050] Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of the present specification for detecting wafer edge structure offset.

[0051] Please refer to Figure 1 The method for detecting wafer edge structure offset may include the following steps: Step 102: Based on the image data of the wafer to be tested, generate candidate rings in the wafer edge region that correspond to the wafer edge and at least one edge ring structure, respectively.

[0052] In some embodiments, the image data of the wafer under test can be obtained in the following manner: the wafer under test is placed on a rotating stage, the camera position is fixed, the wafer is controlled to rotate around its center for one revolution, and the camera continuously acquires a set of unfolded rectangular images in polar coordinate system; then, through inverse polar coordinate transformation, the rectangular images are converted back to Cartesian coordinate system to reconstruct a complete image of the annular region of the wafer edge, and the annular region image can be used as the image data of the wafer under test.

[0053] In some embodiments, the edge ring structure includes EBR and / or EPR. Of course, the edge ring structure described in this embodiment is not limited to EBR and EPR, and can also be extended to other process structures with ring shapes in the wafer edge region.

[0054] In some embodiments, a wafer region can first be segmented from the image of the wafer to be tested, and a coarse positioning center can be obtained based on the wafer region. Then, multiple candidate centers can be generated within a preset tolerance radius, using the coarse positioning center as the center. Next, using each candidate center as the center, and taking the reference radius of the reference ring corresponding to the wafer edge and at least one edge ring structure as the radius, candidate rings corresponding one-to-one with each edge ring structure in the wafer edge and at least one edge ring structure are generated. The reference rings and their reference radii can be determined and stored based on the reference wafer image.

[0055] The reference wafer image clearly shows the wafer edge and at least one edge ring structure feature. For example, the reference wafer image is an image of the first batch or the first wafer under the same process technology, or it can be an image of any wafer under the same process technology. This embodiment does not limit its source.

[0056] Step 104: For each candidate ring, divide multiple annular regions along the radial direction of the candidate ring, and construct multi-channel radial profile features of the candidate ring based on the pixel features in each annular region.

[0057] In some embodiments, for each candidate annulus, using the candidate annulus as a reference circumference, multiple concentric annular regions are obtained by radially expanding at equal distances inward and outward. For each of the resulting annular regions, statistical features of the pixels within that annular region are extracted.

[0058] For example, within each annular region, various gray-level statistics of the original image and the corresponding gradient feature map can be calculated separately. These gray-level statistics include, but are not limited to, mean, median, and standard deviation, and are used to describe the distribution characteristics of pixel values ​​within the annular region from different dimensions.

[0059] In some embodiments, for the same statistic, the feature values ​​extracted from all annular regions are arranged sequentially in a specified radial order (e.g., from smallest to largest radius) to construct an image feature vector. Multiple image feature vectors constructed from various statistical measures together constitute the multi-channel radial profile feature of the candidate annulus.

[0060] Step 106: Perform a matching analysis between the multi-channel radial profile features of each candidate ring and the corresponding baseline features to determine the matching score and matching position information of each candidate ring.

[0061] In some embodiments, the reference features may be pre-extracted and stored from a reference wafer image, and the construction method of the reference features is the same as in step 104.

[0062] In some embodiments, when performing matching analysis on the multi-channel radial profile features of the candidate ring and the corresponding reference features, algorithms such as cross-correlation can be used. The obtained matching score can characterize the similarity between the candidate ring and the corresponding reference ring. The higher the score, the closer the center of the candidate ring is to the true center. The obtained matching position information characterizes the radial offset of the candidate ring relative to the corresponding reference ring.

[0063] Step 108: Determine the center position of the wafer edge and at least one edge ring structure based on the matching score, and optimize the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure.

[0064] In some embodiments, a three-dimensional peak positioning method can be used to determine the center position, which can reflect the actual center position of the corresponding ring structure on the wafer and is the basis for subsequent calculation of offset information.

[0065] In some embodiments, a weighted optimization method can be used to determine the radius parameter, which is typically a radius value obtained by adding the optimized matching position to the reference radius. This radius parameter can reflect the actual radial size of the corresponding ring structure and can be used to determine whether there is a radial deviation in at least one edge ring structure.

[0066] It should be noted that the process of determining the center position and radius parameters described above is performed independently for each edge ring structure in the wafer edge and at least one edge ring structure. In other words, for the wafer edge, the center position and radius parameters of the wafer edge are obtained according to the above method based on its corresponding candidate ring set; taking at least one edge ring structure including EBR and EPR as an example, for EBR, the center position and radius parameters of EBR are obtained independently based on its corresponding candidate ring set; for EPR, the center position and radius parameters of EPR are obtained similarly.

[0067] Step 110: Calculate the offset information between the wafer edge and at least one edge ring structure based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure.

[0068] In some embodiments, the offset information includes a relative offset (such as an offset distance) and an offset direction.

[0069] For the center of a circle, the Euclidean distance between the centers can be calculated as the offset distance to quantify the magnitude of the offset. For example, the Euclidean distance between the center of a certain edge ring structure and the center of the wafer edge can be calculated as the offset distance, and the direction from the wafer edge center to the center of the edge ring structure can be used as the offset direction to indicate the orientation of the offset. The relative offset of the center can reflect the overall deviation of the center and can be used for equipment calibration and offset trend analysis.

[0070] For ring-shaped structures, the radial distance difference at the boundary can be calculated at a specified azimuth angle, taking into account the difference in center offset and radius parameters. For example, to calculate the radial distance difference between the boundary of a certain edge ring-shaped structure and the wafer boundary, the radial distance difference can be calculated at four azimuth angles: 0°, 90°, 180°, and 270°, with the wafer notch direction as a reference. The relative offset of the ring reflects the boundary deviation in a specific direction, which can be used to analyze the directional characteristics of process deviations.

[0071] Of course, the above offset information is merely illustrative and is not intended to limit practical applications. Those skilled in the art can flexibly select or expand the method of offset information calculation according to specific process monitoring needs and equipment calibration requirements.

[0072] As described above, the scheme provided in this specification first generates candidate rings corresponding to the wafer edge and at least one edge ring structure in the wafer edge region. Then, for each candidate ring, a multi-channel radial profile feature is constructed based on the pixel features within each ring region. The pixel features within the ring region replace the single continuous boundary points relied upon by traditional methods. Even if the edge ring structure has breaks, discontinuities, or local defects, a stable characterization can still be obtained, solving the problems of circle fitting failure due to edge ring structure breaks and edge detection instability due to image quality degradation. Next, the multi-channel radial profile features of each candidate ring are matched with the corresponding reference features to determine the matching score and matching position information, thereby determining the center position and radius parameters. Offset information is then calculated based on the center position and radius parameters corresponding to each structure. Thus, under complex imaging conditions, the ring structure can be accurately identified and its offset information calculated, achieving effective monitoring and closed-loop control of semiconductor manufacturing edge processes.

[0073] In some embodiments, in order to perform offset detection on the wafer image under test, candidate rings corresponding to the wafer edge and at least one edge ring structure (such as EBR and / or EPR) are first generated. Please refer to Figure 2 The process of generating the candidate ring includes the following steps: Step 202: Segment the wafer region from the image of the wafer to be tested, and determine the initial center based on the wafer region.

[0074] In some embodiments, an image of the wafer to be tested can be obtained first. The image of the wafer to be tested can be an annular region image in Cartesian coordinate system. Its reconstruction method can adopt the aforementioned polar coordinate transformation and inverse transformation methods, which will not be elaborated here.

[0075] Then, the wafer region can be segmented from the wafer image under test. For example, the Otsu method (Maximum Between-Class Variance Method) can be used to perform binary segmentation on the wafer image under test, dividing the image into a wafer region and a background region, and extracting the contour of the wafer region. The Otsu method can adaptively determine the segmentation threshold without manual intervention and is suitable for wafer images under different lighting conditions.

[0076] Next, an initial circle center is determined based on the wafer region. Specifically, a circle fitting algorithm can be used to fit the contour points of the wafer region to determine the initial circle center. For example, the Tukey (Tukey's Biweight Method) robust circle fitting algorithm can be used to fit the contour points. This algorithm reduces the influence of abnormal contour points (such as edge defects, noise points, etc.) on the fitting results through iterative weighting, thus obtaining a stable and reliable initial circle center.

[0077] Step 204: Using the initial circle center as the center, generate multiple candidate circle centers within a preset tolerance error radius.

[0078] In some embodiments, the initial center determined in step 202 is used as the center, and within a preset tolerance error radius... Multiple candidate center points are generated within the range. The tolerance error radius represents the upper limit of the allowable offset of the center point, which can be preset according to process requirements and equipment accuracy, such as 0.5 micrometers, 0.6 micrometers, 0.7 micrometers, 0.85 micrometers, 1.0 micrometers, etc.

[0079] The candidate circle center can be generated using methods such as grid sampling, random sampling, or uniform angle sampling. This specification does not impose any special restrictions on this method.

[0080] Therefore, multiple candidate centers are generated based on the initial center within the tolerance error radius. These candidate centers are then confined to a reasonable local area, ensuring that the true center is not missed while controlling the computational load, thus achieving a balance between efficiency and accuracy.

[0081] Step 206: Using each candidate circle center as the center and the reference radius corresponding to the wafer edge and at least one edge ring structure as the radius, generate candidate rings corresponding to the wafer edge and at least one edge ring structure.

[0082] In some embodiments, for each candidate circle center generated in step 204, a candidate ring is constructed using the reference radius corresponding to the wafer edge and the reference radius corresponding to at least one edge ring structure as the radius. For example, a candidate ring corresponding to the wafer edge is generated using the candidate circle center as the center and the reference radius corresponding to the wafer edge as the radius. As another example, a candidate ring corresponding to a certain edge ring structure is generated using the candidate circle center as the center and the reference radius corresponding to that edge ring structure as the radius.

[0083] The reference radius corresponding to the wafer edge and the reference radius corresponding to at least one edge ring structure are obtained from the aforementioned reference ring.

[0084] For example, on a reference wafer image, at least three points on the wafer edge and at least one edge ring structure boundary can be manually marked, and then circle fitting can be performed to obtain corresponding reference rings. Each reference ring includes a reference center position and a reference radius parameter, which serve as a subsequent geometric reference. Taking at least one edge ring structure as an example, such as EBR and EPR, please refer to... Figure 3 Example, Figure 3 In the diagram, red marks indicate wafer edge locations, green marks indicate EBR boundary locations, and blue marks indicate EPR boundary locations. By manually marking at least three locations, and using algorithms such as least squares to perform circle fitting on each set of locations, wafer edge reference rings, EBR reference rings, and EPR reference rings can be obtained respectively.

[0085] In some embodiments, for each candidate center, a set of candidate rings corresponding to the wafer edge and each edge ring structure are generated. When subsequent steps process different structures independently, their respective corresponding candidate rings can be called.

[0086] In some embodiments, when constructing a multi-channel radial profile feature for a candidate ring, multiple annular regions can be first divided along the radial direction of each candidate ring.

[0087] In some embodiments, each candidate annulus can be used as a reference circumference, and the annulus can be extended radially inward and outward at equal distances to obtain multiple concentric annular regions.

[0088] Please refer to Figure 4 , Figure 4 The red ring in the diagram is a candidate ring. Using this red ring as a reference circumference, two rings are extended radially inward and radially outward at equal distances, resulting in four extended rings, totaling four concentric annular regions. Each annular region is formed by two adjacent rings.

[0089] Assuming that when constructing the baseline feature, the number of annular regions obtained by dividing the region into annular areas is N1, and the spacing between the annular regions is fixed at d1, then the total expansion spacing along the radial direction towards the inside and outside is N1×d1. When constructing the multi-channel radial profile feature for the candidate ring, the spacing between the divided annular regions is fixed at d2, which is the same as the spacing d1 used when expanding the baseline ring. The number of annular regions obtained after expansion is denoted as N2, where N2 is less than N1. That is, the number of annular regions generated for the candidate ring is less than the number of annular regions generated for the baseline ring. Thus, using the same spacing, the maximum radial expansion spacing (N2×d2) for the candidate ring is less than the maximum radial expansion spacing (N1×d1) when constructing the baseline feature. Subsequently, when performing matching analysis between the multi-channel radial profile feature of the candidate ring and the corresponding baseline feature, it is ensured that the multi-channel radial profile feature of the candidate ring has a certain sliding range on the baseline feature vector, thereby calculating the subsequent matching position information.

[0090] In some embodiments, after dividing multiple annular regions along the radial direction of each candidate annulus, please refer to... Figure 5 The process of constructing the multi-channel radial profile feature of the candidate annulus may include the following steps: Step 502: Obtain the original image and the corresponding gradient feature image of the wafer to be tested.

[0091] In some embodiments, an original image of the wafer under test can be acquired. This original image can be reconstructed using the aforementioned polar coordinate transformation and inverse transformation methods, and is typically a grayscale image of a ring-shaped region in Cartesian coordinates. Then, the Sobel gradient magnitude is calculated on the original image to obtain the corresponding gradient feature image. This gradient feature image enhances the edge information of the wafer edges and at least one edge ring structure boundary, improving the response capability to blurred boundaries.

[0092] Step 504: Extract pixel features of each annular region from the original image of the wafer to be tested and the gradient feature image to construct multi-channel radial profile features of the candidate annulus.

[0093] In some embodiments, for each annular region obtained by division, the pixel features of that region can be calculated separately to construct the multi-channel radial profile features of the candidate annulus. The "multi-channel" refers to extracting features from two information channels, the original image and the gradient feature image, at the same time, thereby enhancing the adaptability of the reference features to complex imaging conditions.

[0094] The pixel features can be obtained by calculating various image statistics within each annular region.

[0095] For example, various statistical measures of pixel values ​​within a ring-shaped region can be calculated, including the mean, median, interquartile range, and range. The interquartile range is the difference between the upper and lower quartiles, and the range is the difference between the maximum and minimum values. These statistical measures describe the distribution characteristics of pixel values ​​within the ring-shaped region from different dimensions. For example, the mean reflects the overall brightness level, the median can resist the interference of outliers, and the interquartile range and range can reflect the degree of distribution dispersion.

[0096] Next, the image statistics corresponding to each annular region can be combined in radial order (from inside to outside or from outside to inside) to form multiple image feature vectors. Specifically, for each statistic, the statistical values ​​of all annular regions arranged radially constitute an image feature vector. The length of this vector is equal to the number of annular regions N2, and less than the length N1 of the baseline feature vector. For example, the mean of the first annular region is used as the first element of the vector, the mean of the second annular region is used as the second element, and so on, to obtain the mean image feature vector. Similarly, the median image feature vector, the interquartile range image feature vector, and the range image feature vector can be obtained.

[0097] The above statistical process is performed on both the original image and the gradient feature image of the wafer under test. Thus, four image feature vectors (e.g., mean, median, interquartile range, and range) can be constructed from the original image, and four more image feature vectors can be generated from the gradient feature image, for a total of eight image feature vectors. These eight image feature vectors together constitute the multi-channel radial profile feature of the candidate annulus, which is then matched and analyzed with the corresponding baseline features.

[0098] Therefore, by employing the aforementioned multi-channel radial profile feature construction method, which integrates edge information from the original image and gradient feature image, the radial profile distribution pattern of the annular region can be characterized from multiple dimensions. Even if the annular structure in the subsequent test image has broken, discontinuous, or locally missing edges, stable matching can still be achieved as long as the overall grayscale distribution characteristics within the annular region are similar to the baseline features. This effectively solves the problem of structural recognition under conditions of structural discontinuity, structural deformation, and complex imaging environments. Simultaneously, the baseline features constructed using a combination of multiple statistical measures describe the pixel distribution characteristics of the annular region from different dimensions. The complementary nature of these statistical measures enhances the feature's representational ability and robustness to local noise.

[0099] It should be noted that the gradient feature images described above are merely illustrative examples. In other examples, texture feature images, variance feature images, etc., can also be used as feature extraction channels. Similarly, the statistics described above are also merely illustrative examples. In other examples, variance, standard deviation, quantiles, etc., can also be calculated as statistics, and this specification does not impose any special restrictions on this.

[0100] In some embodiments, the reference features corresponding to the multi-channel radial profile features of the candidate ring can be pre-constructed, and the construction method can refer to the multi-channel radial profile features of the candidate ring. Specifically, for each reference ring, multiple annular regions can be divided along its radial direction. In the original image of the reference wafer and the corresponding gradient feature image, the pixel features of each annular region are extracted respectively to construct the multi-channel radial profile features of the reference ring as the reference features.

[0101] Therefore, this embodiment uses the same method to construct the radial profile feature vector and the reference feature vector of the candidate ring, so that the radial profile feature vector of the candidate ring and the reference feature vector have the same physical meaning in the feature space, which provides a semantic alignment basis for subsequent matching calculation, thereby ensuring the accuracy and reliability of subsequent matching analysis.

[0102] In some embodiments, a matching analysis is performed based on the multi-channel radial profile features of the candidate rings constructed above and the pre-constructed reference features to determine the matching score and matching position information corresponding to each candidate ring. For example, the multi-channel radial profile features of each candidate ring can be cross-correlated with the corresponding reference features to obtain the matching score and matching position information.

[0103] Please refer to Figure 6 The determination of the matching score and matching position includes the following steps: Step 602: Perform cross-correlation calculation on each image feature vector and the corresponding reference image feature vector in the reference features to obtain multiple sub-matching scores and multiple sub-matching position information.

[0104] In some embodiments, for each candidate annulus, its multi-channel radial profile features are composed of multiple image feature vectors, and the corresponding reference features are composed of multiple reference image feature vectors, with both constructed in the same way. This embodiment can perform matching analysis between the two using a cross-correlation algorithm. Specifically, for each image feature vector of the candidate annulus, cross-correlation is performed with the corresponding reference image feature vector in the reference features to obtain the sub-matching score and sub-matching position information corresponding to that image feature vector.

[0105] In some embodiments, the image feature vector corresponding to the reference ring is denoted as B. The length of vector B is N1, and the image feature vector corresponding to the candidate ring is denoted as T. The length of vector T is N2, where N2 is less than N1. Since N2 is less than N1, when performing cross-correlation calculation, the shorter vector T can slide on the longer vector B, and the similarity between the two can be calculated at each sliding position to obtain the cross-correlation value R(k) at each sliding position k.

[0106] In some embodiments, for each sliding position k, the cross-correlation value R(k) is obtained, and the maximum cross-correlation value is selected as the sub-matching score of the corresponding image feature vector, denoted as . , Simultaneously, based on the sliding position corresponding to this maximum cross-correlation... The distance d between the ring region and the annular region can be used to calculate the corresponding sub-matching position information P. .

[0107] In some embodiments, taking the multi-channel radial profile features of the candidate ring as an example, which include 8 image feature vectors, for each of these 8 image feature vectors, cross-correlation calculation is performed with the corresponding image feature vector in the baseline features, thereby obtaining 8 sub-matching scores and 8 word matching position information.

[0108] Step 604: Calculate the matching score based on multiple sub-matching scores, and calculate the matching position information based on multiple sub-matching position information.

[0109] In some embodiments, the average of the plurality of sub-matching scores can be calculated as the matching score corresponding to the candidate ring, and the average of the plurality of matching position information can be calculated as the matching position information corresponding to the candidate ring, etc.

[0110] Of course, the above average values ​​are only illustrative examples. In other examples, the matching score and matching position information corresponding to the candidate ring can also be calculated by weighted average, maximum value method, median method, etc. This specification does not impose any special restrictions on this.

[0111] Therefore, this embodiment adopts a multi-feature vector fusion method, performs cross-correlation calculation on each image feature vector to obtain sub-matching scores and sub-matching position information, and then summarizes them to obtain the final matching score and matching position information. This allows the information between the original image channel and the gradient feature image channel, as well as various statistical measures, to complement each other. When a single feature vector produces a deviation, other feature vectors can still provide correct matching results, thereby improving the reliability and accuracy of matching analysis.

[0112] In some embodiments, the matching scores of each candidate ring obtained based on the aforementioned matching analysis can be used to determine the center positions of the wafer edge and at least one edge ring structure, respectively. Please refer to... Figure 7 The process of determining the center position may include the following steps: Step 702: For each candidate ring, construct a three-dimensional spatial distribution map using the coordinates of the candidate ring's center as the planar coordinates and the corresponding matching score as the height value.

[0113] In some embodiments, taking the wafer edge as an example, assuming that a total of M candidate circle centers are generated, each candidate circle center corresponds to a candidate ring, and after the aforementioned matching analysis, each candidate ring can obtain a matching score. The matching score The matching score represents the degree of similarity between the candidate ring and the reference ring. The larger the value, the closer the candidate center is to the true center. The value of i ranges from 1 to M.

[0114] In some embodiments, the candidate center coordinates of each candidate annulus are used. Using planar coordinates, with corresponding matching scores Using the height value, construct a three-dimensional spatial distribution map to obtain the point set in three-dimensional space. This three-dimensional spatial distribution map is a confidence-based topographic map, where the planar coordinates represent the hypothetical center location, and the altitude represents the confidence level of that hypothetical location. Theoretically, the true center location should correspond to the highest peak on this topographic map.

[0115] Step 704: After smoothing the height value, locate the highest peak point.

[0116] In some embodiments, due to factors such as image noise and process fluctuations, the matching score may have local random fluctuations. In order to obtain a more stable and accurate center position, the height value of the three-dimensional spatial distribution map can be smoothed first, and then the highest peak point after smoothing can be located.

[0117] The specific algorithm for smoothing can be Gaussian filtering, mean filtering, median filtering, etc., and this specification does not impose any special restrictions on it.

[0118] Step 706: Determine the highest peak point as the center position of the wafer edge or edge ring structure corresponding to the candidate ring.

[0119] In some embodiments, after the smoothing and peak location in step 704, the planar coordinates corresponding to the highest peak point are obtained. The coordinates are then used to determine the center position of the wafer edge and the edge ring structure.

[0120] In some embodiments, the above steps are performed on the wafer edge and each of the at least one edge ring structure to obtain the center position corresponding to the wafer edge and each edge ring structure.

[0121] Therefore, this embodiment constructs a three-dimensional spatial distribution map based on the matching score and locates the highest peak point after smoothing. This effectively suppresses the interference of residual outliers and noise on the results, thereby obtaining a more stable and accurate center position.

[0122] In some embodiments, the matching position information is optimized based on the matching scores of each candidate ring obtained from the aforementioned matching analysis, which can determine the radius parameters of the wafer edge and at least one edge ring structure, respectively. Please refer to Figure 8 The process of determining the radius parameter may include the following steps: Step 802: For each candidate ring, perform anomaly filtering on the matching position information of the candidate ring to remove abnormal matching position information.

[0123] In some embodiments, the aforementioned matching analysis calculates a matching position information P for each candidate ring, which represents the radial offset of the candidate ring relative to the reference ring. Due to factors such as image noise, local defects, or process variations, the matching position information of some candidate rings may deviate from the normal range, becoming outliers. If these outliers are directly used for subsequent radius calculations, it will severely affect the accuracy of the radius parameter.

[0124] In this process, the matching position information of all candidate rings is filtered for anomalies to remove outliers that deviate from the normal range. Various methods can be used, such as the standard deviation method, the median absolute deviation method, and the fixed threshold method. This manual does not impose any special restrictions on this.

[0125] Taking the fixed threshold method as an example, a reasonable radial offset range can be set based on prior knowledge, and matching position information that exceeds this range can be judged as anomalies.

[0126] Taking the standard deviation method as an example, the arithmetic mean and standard deviation of all matching position information can be calculated, the difference between each matching position information and the mean can be calculated, and then the matching position information whose difference exceeds a preset multiple of the standard deviation can be judged as an outlier.

[0127] Step 804: The remaining matching position information is weighted and averaged using the matching score as the weight to obtain the optimized matching position information.

[0128] In some embodiments, after anomaly removal of the matching location information, the remaining matching location information can be weighted and averaged to obtain optimized matching location information.

[0129] The weight of each remaining matching position information is its corresponding matching score. This weighted average method makes the matching position information with a higher matching score contribute more to the final result, and the matching position information with a lower matching score contribute less to the final result, thereby improving the accuracy of the radius parameter.

[0130] Step 806: Calculate the radius parameter based on the optimized matching position information and the reference radius of the wafer edge or edge ring structure corresponding to the candidate ring.

[0131] In some embodiments, the reference radius of the wafer edge or edge ring structure is determined based on a reference wafer image. By adding the optimized matching position information and the reference radius, the radius parameter of the corresponding structure (wafer edge or edge ring structure) can be obtained, which can represent the actual radius of the corresponding result.

[0132] In some embodiments, the above steps are performed on the wafer edge and each edge ring structure in at least one edge ring structure to obtain the radius parameters corresponding to the wafer edge and each edge ring structure.

[0133] Therefore, this embodiment removes out-of-normal matching position information by filtering out anomalies, avoiding interference from outliers caused by local noise, image artifacts or process defects on the radius parameter calculation results, thus improving the robustness of radius parameter determination. At the same time, the matching position information after anomaly filtering is weighted by the matching score, thereby obtaining a more accurate actual radius parameter.

[0134] Figure 9 This is a schematic structural diagram of a device provided in an exemplary embodiment. Please refer to... Figure 9 At the hardware level, the device includes a processor 902, an internal bus 904, a network interface 906, memory 908, and non-volatile memory 910, and may also include other hardware required for its functions. One or more embodiments of this specification can be implemented in software, for example, the processor 902 reads the corresponding computer program from the non-volatile memory 910 into memory 908 and then runs it. Of course, besides software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0135] Figure 10 This is a block diagram illustrating an exemplary embodiment of a wafer edge structure offset detection device.

[0136] Please refer to Figure 10 The wafer edge structure offset detection device 1000 can be applied to... Figure 9 The electronic device shown implements the technical solution of this specification. The wafer edge structure offset detection device 1000 includes: The candidate ring generation unit 1001 generates candidate rings in the wafer edge region based on the image data of the wafer to be tested, which correspond to the wafer edge and at least one edge ring structure, respectively. The feature construction unit 1002 divides multiple annular regions along the radial direction of each candidate annular ring, and constructs a multi-channel radial profile feature of the candidate annular ring based on the pixel features in each annular region. The feature matching unit 1003 performs matching analysis on the multi-channel radial profile features of each candidate ring and the corresponding reference features to determine the matching score and matching position information of each candidate ring. The center radius determination unit 1004 determines the center position of the wafer edge and at least one edge ring structure based on the matching score, and optimizes the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure. The offset calculation unit 1005 calculates the offset information between the wafer edge and the at least one edge ring structure based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure.

[0137] Optionally, the feature construction unit 1002 divides the candidate annulus into multiple annular regions along the radial direction, including: Using the candidate annulus as a reference circumference, the annulus is extended equidistantly inward and outward to obtain multiple concentric annular regions. Wherein, the maximum radial extension length when constructing the multi-channel radial profile feature corresponding to the candidate annulus is less than the maximum radial extension length when constructing the reference feature.

[0138] Optionally, the feature construction unit 1002 constructs a multi-channel radial profile feature of the candidate annulus based on pixel features within each annular region, including: The original image and the corresponding gradient feature image of the wafer under test are obtained respectively; In the original image of the wafer under test and the gradient feature image, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the candidate annulus.

[0139] Optionally, the candidate ring generation unit 1001 generates candidate rings in the wafer edge region that correspond to the wafer edge and at least one edge ring structure, respectively, including: The wafer region is segmented from the image of the wafer to be tested, and the initial center is determined based on the wafer region; Multiple candidate center points are generated within a preset tolerance error radius, with the initial center point as the center. Using each candidate circle center as the center and the reference radius corresponding to the wafer edge and at least one edge ring structure as the radius, candidate rings corresponding to the wafer edge and at least one edge ring structure are generated.

[0140] Optionally, the candidate annulus generation unit 1001 determines the initial center based on the wafer region, including: A circle fitting algorithm is used to fit the contour points of the wafer region to determine the initial circle center.

[0141] Optionally, the baseline features are pre-constructed in the following manner: In the reference wafer image, determine the wafer edge and the reference ring corresponding to the at least one edge ring structure, respectively; For each of the aforementioned reference rings, multiple annular regions are divided along its radial direction; In the original image and the corresponding gradient feature image of the reference wafer, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the reference ring as the reference features.

[0142] Optionally, the gradient feature image is obtained by calculating the Sobel gradient magnitude on the original image.

[0143] Optionally, the feature construction unit 1002 extracts pixel features for each of the annular regions, including: Calculate multiple image statistics within each of the aforementioned annular regions; The image statistics corresponding to each annular region are combined radially to form multiple image feature vectors, thereby constructing the multi-channel radial profile feature.

[0144] Optionally, the feature matching unit 1003 performs matching analysis on the multi-channel radial profile features of each candidate ring with the corresponding reference features to determine the matching score and matching position information corresponding to each candidate ring, including: The multi-channel radial profile features of each candidate ring are cross-correlated with the corresponding reference features to obtain the matching score and matching position information. The matching position information is used to characterize the radial offset of the candidate ring relative to the corresponding reference ring.

[0145] Optionally, the feature matching unit 1003 performs cross-correlation calculations on the multi-channel radial profile features of each candidate annulus with the corresponding reference features to obtain matching scores and matching position information, including: Each image feature vector is cross-correlated with the corresponding reference image feature vector in the reference features to obtain multiple sub-matching scores and multiple sub-matching position information; The matching score is calculated based on the multiple sub-matching scores, and the matching location information is calculated based on the multiple sub-matching location information.

[0146] Optionally, the center radius determination unit 1004 determines the center position of the wafer edge and at least one of the edge ring structures based on the matching score, including: For each candidate ring, a three-dimensional spatial distribution map is constructed using the candidate center coordinates of the candidate ring as planar coordinates and the corresponding matching score as the height value. After smoothing the height value, the highest peak point is located; The highest peak point is determined as the center position of the wafer edge or edge ring structure corresponding to the candidate ring.

[0147] Optionally, the center radius determination unit 1004 optimizes the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one of the edge ring structures, including: For each candidate ring, the matching position information of the candidate ring is subjected to anomaly filtering to remove abnormal matching position information; The remaining matching position information is weighted and averaged using the matching score as the weight to obtain optimized matching position information; The radius parameter is calculated based on the optimized matching position information and the reference radius of the wafer edge or edge ring structure corresponding to the candidate ring.

[0148] Optionally, at least one of the edge ring structures includes an edge photoresist removal region (EBR) and / or an edge protection ring (EPR).

[0149] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0150] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. 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 the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0151] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0152] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0153] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0154] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, 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 memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0155] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0156] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0157] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0158] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0159] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for detecting wafer edge structure offset, characterized in that, include: Based on the image data of the wafer under test, candidate rings corresponding to the wafer edge and at least one edge ring structure are generated in the wafer edge region. For each candidate ring, multiple annular regions are divided along the radial direction of the candidate ring, and a multi-channel radial profile feature of the candidate ring is constructed based on the pixel features in each annular region. The multi-channel radial profile features of each candidate ring are matched with the corresponding reference features to determine the matching score and matching position information of each candidate ring. The center positions of the wafer edge and at least one edge ring structure are determined based on the matching score. The matching position information is then optimized based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure. Based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure, the offset information between the wafer edge and the at least one edge ring structure is calculated.

2. The method according to claim 1, characterized in that, The process of dividing the candidate annulus into multiple annular regions along its radial direction includes: Using the candidate annulus as a reference circumference, the annulus is extended equidistantly inward and outward to obtain multiple concentric annular regions. Wherein, the maximum radial extension length when constructing the multi-channel radial profile feature corresponding to the candidate annulus is less than the maximum radial extension length when constructing the reference feature.

3. The method according to claim 1, characterized in that, The construction of the multi-channel radial profile feature of the candidate annulus based on pixel features within each annular region includes: The original image and the corresponding gradient feature image of the wafer under test are obtained respectively; In the original image of the wafer under test and the gradient feature image, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the candidate annulus.

4. The method according to claim 1, characterized in that, The step of generating candidate rings in the wafer edge region corresponding to the wafer edge and at least one edge ring structure includes: The wafer region is segmented from the image of the wafer to be tested, and the initial center is determined based on the wafer region; Multiple candidate center points are generated within a preset tolerance error radius, with the initial center point as the center. Using each candidate circle center as the center and the reference radius corresponding to the wafer edge and at least one edge ring structure as the radius, candidate rings corresponding to the wafer edge and at least one edge ring structure are generated.

5. The method according to claim 4, characterized in that, Determining the initial center based on the wafer region includes: A circle fitting algorithm is used to fit the contour points of the wafer region to determine the initial circle center.

6. The method according to claim 1, characterized in that, The baseline features are pre-constructed in the following manner: In the reference wafer image, determine the wafer edge and the reference ring corresponding to the at least one edge ring structure, respectively; For each of the aforementioned reference rings, multiple annular regions are divided along its radial direction; In the original image and the corresponding gradient feature image of the reference wafer, the pixel features of each of the annular regions are extracted to construct the multi-channel radial profile features of the reference ring as the reference features.

7. The method according to claim 3 or 6, characterized in that, The gradient feature image is obtained by calculating the Sobel gradient magnitude from the original image.

8. The method according to claim 3 or 6, characterized in that, The extraction of pixel features for each of the annular regions includes: Calculate multiple image statistics within each of the aforementioned annular regions; The image statistics corresponding to each annular region are combined radially to form multiple image feature vectors, thereby constructing the multi-channel radial profile feature.

9. The method according to claim 1, characterized in that, The step of matching the multi-channel radial profile features of each candidate ring with the corresponding reference features to determine the matching score and matching position information of each candidate ring includes: The multi-channel radial profile features of each candidate ring are cross-correlated with the corresponding reference features to obtain the matching score and matching position information. The matching position information is used to characterize the radial offset of the candidate ring relative to the corresponding reference ring.

10. The method according to claim 9, characterized in that, The step of cross-correlation calculation between the multi-channel radial profile features of each candidate annulus and the corresponding reference features to obtain the matching score and matching position information includes: Each image feature vector is cross-correlated with the corresponding reference image feature vector in the reference features to obtain multiple sub-matching scores and multiple sub-matching position information; The matching score is calculated based on the multiple sub-matching scores, and the matching location information is calculated based on the multiple sub-matching location information.

11. The method according to claim 9, characterized in that, Determining the center position of the wafer edge and at least one of the edge ring structures based on the matching score includes: For each candidate ring, a three-dimensional spatial distribution map is constructed using the candidate center coordinates of the candidate ring as planar coordinates and the corresponding matching score as the height value. After smoothing the height value, the highest peak point is located; The highest peak point is determined as the center position of the wafer edge or edge ring structure corresponding to the candidate ring.

12. The method according to claim 9, characterized in that, The optimization processing of the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one of the edge ring structures includes: For each candidate ring, the matching position information of the candidate ring is subjected to anomaly filtering to remove abnormal matching position information; The remaining matching position information is weighted and averaged using the matching score as the weight to obtain optimized matching position information; The radius parameter is calculated based on the optimized matching position information and the reference radius of the wafer edge or edge ring structure corresponding to the candidate ring.

13. The method according to claim 1, characterized in that, At least one of the edge ring structures includes an edge photoresist removal region (EBR) and / or an edge protection ring (EPR).

14. A device for detecting wafer edge structure misalignment, characterized in that, include: The candidate ring generation unit generates candidate rings in the wafer edge region based on the image data of the wafer to be tested, which correspond to the wafer edge and at least one edge ring structure, respectively. The feature construction unit divides each candidate ring into multiple annular regions along the radial direction of the candidate ring, and constructs a multi-channel radial profile feature of the candidate ring based on the pixel features in each annular region. The feature matching unit performs matching analysis on the multi-channel radial profile features of each candidate ring and the corresponding reference features to determine the matching score and matching position information of each candidate ring. The center radius determination unit determines the center position of the wafer edge and at least one edge ring structure based on the matching score, and optimizes the matching position information based on the matching score to determine the radius parameters of the wafer edge and at least one edge ring structure. The offset calculation unit calculates the offset information between the wafer edge and the at least one edge ring structure based on the center position and radius parameters corresponding to the wafer edge and at least one edge ring structure.

15. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor implements the method as described in any one of claims 1 to 13 by executing the executable instructions.

16. A computer-readable storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 13.

Images