Image processing method and wafer inspection apparatus for wafer inspection
By using group correction and staged seam elimination methods, the problem of poor image stitching quality in rotating silicon wafer inspection was solved, achieving efficient and reliable silicon wafer inspection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI YUWEI SEMICON TECH CO LTD
- Filing Date
- 2026-02-11
- Publication Date
- 2026-05-29
AI Technical Summary
In existing technologies, when inspecting silicon wafers by rotating images, the image stitching quality is poor, resulting in inconsistent grayscale and obvious stitching marks, which affects the accuracy and stability of defect detection.
Images are acquired by rotating and scanning a silicon wafer. Flat field correction is performed based on the coordinates of the image center point. Linear orientation stitching and annular seam elimination are carried out in stages. Gray-scale fusion processing is used to eliminate gray-scale differences at the stitching points, ensuring the integrity and continuity of the image.
It improves the efficiency and accuracy of silicon wafer inspection, avoids false detections and missed detections, enhances image quality and reliability, and is adapted to the high-efficiency imaging characteristics of rotating imaging.
Smart Images
Figure CN122115379A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of semiconductor inspection technology, specifically relating to an image processing method and silicon wafer inspection equipment for silicon wafer inspection. Background Technology
[0002] In the field of semiconductor inspection technology, the imaging method for silicon wafers often employs a step-by-step triggering approach using an area array camera. For example, patent CN121007906A discloses an image acquisition control method, device, equipment, storage medium, and product. After center alignment is achieved using a wafer edge finder, the wafer is transferred to a defect detection platform using a wafer transfer fork. Under various lighting fields such as bright field, dark field, and PL field, a line scan camera or area array camera traverses and captures images according to a preset coordinate grid, acquiring images containing features such as surface defects, scratches, and deep growth defects. Finally, the images captured in segments are stitched together based on their coordinate markers to form a complete wafer defect image. However, this step-by-step triggering imaging path is complex and has low imaging efficiency.
[0003] The applicant sought to improve the imaging path and efficiency, and developed a rotating imaging method based on an area scan camera for efficient scanning and detection. However, during further development, the applicant discovered that images from different angles and positions are affected by factors such as uneven lighting, differences in camera response, and changes in imaging conditions caused by rotating scanning, resulting in significant differences in image grayscale characteristics. If a fixed number of images or randomly selected images are used to generate a grayscale correction template during the flat-field correction process, the corrected image may still exhibit grayscale inconsistencies. Furthermore, when directly stitching images based on the acquisition coordinates, grayscale differences and geometric deviations between different image segments can easily create obvious stitching marks, leading to a decrease in the quality of the stitched image. The presence of grayscale inconsistencies and stitching seams in the stitched image can interfere with subsequent defect recognition algorithms, reducing the accuracy and stability of defect detection, and even causing false detections or missed detections. Summary of the Invention
[0004] This invention provides an image processing method and a silicon wafer inspection device for silicon wafer inspection, solving the technical problem of poor quality of stitched images obtained by rotating photography during silicon wafer defect inspection.
[0005] The technical solution adopted in this invention is as follows: This invention provides an image processing method for silicon wafer inspection, comprising the following steps: S1, obtain all images of the silicon wafer and the coordinates of the center points of each image by rotating and scanning the silicon wafer; S2, group the images according to the coordinates of the center points of each image, perform flat field correction on each image in the same group, and obtain the corrected image of each image; S3, perform linear orientation stitching on the corrected images, extract the stitching images of adjacent corrected images, eliminate the stitching seams of adjacent corrected images to obtain linear stitched images, and perform inverse polar coordinate transformation on the linear stitched images to obtain the annular stitched images of each group. S4, stitch the adjacent ring-shaped stitched images together, extract the stitching area of the adjacent ring-shaped stitched images, eliminate the ring-shaped stitching seams of the adjacent ring-shaped stitched images, and obtain a complete stitched image.
[0006] The present invention provides an image processing method for silicon wafer inspection, which uses rotating scanning of silicon wafers to achieve rapid image acquisition. In step S2, flat field correction is performed by grouping images according to the size of the center point coordinates. This avoids the insufficient correction targeting caused by traditional fixed number or random selection of images to generate correction templates. It enables the correction process to adapt to similar imaging conditions of images within the same group, making flat field correction more targeted and accurate, and can accurately offset grayscale deviations generated by images within the same group. After flat-field correction, steps S3 to S4 achieve phased stitching and seam elimination. First, linear directional stitching and seam elimination ensure a smooth transition of images within a single group. Then, inverse polar coordinate transformation yields a ring-shaped stitched image. Finally, the ring seams between adjacent ring images are specifically eliminated, avoiding geometric deviations and grayscale discontinuities that are easily caused by direct coordinate stitching. This effectively ensures image integrity and continuity, ultimately guaranteeing high quality of the complete stitched image. It not only avoids interference from grayscale inconsistencies and seams on the defect recognition algorithm, preventing false detections and missed detections, but also leverages the efficient scanning foundation of rotating imaging to improve detection efficiency while ensuring detection accuracy, balancing both efficiency and quality. In summary, this invention, through the combination of coordinate grouping correction and phased seam elimination, fundamentally solves the problems of uneven grayscale and obvious stitching marks. It not only adapts to the efficient imaging characteristics of rotating imaging but also significantly improves the reliability and practicality of silicon wafer inspection.
[0007] In a preferred embodiment, step S4 includes: S41, stitching adjacent annular stitched images together to obtain a preliminary combined image; S42, transforming the preliminary combined image into a linear combined image through polar coordinate transformation; S43, extracting the stitching area images of adjacent annular stitched images based on the linear combined image to eliminate the annular stitching seams between adjacent annular stitched images; S44, obtaining a complete stitched image.
[0008] Step S41 completes the initial combination of adjacent annular stitched images, providing a foundation for subsequent seam processing. Step S42 then converts the initial combined image into a linear combined image, resulting in more comprehensive and refined processing. Since the subsequent stitched images and seams are linear, seam extraction and elimination are easier, enabling precise location and extraction of the annular stitched seam region. Step S43, based on the linear combined image, accurately extracts the stitched images of adjacent annular stitched images, achieving targeted seam elimination. This avoids the problem of unprocessed seam residue from direct stitching or general processing. The grayscale transition processing is thorough, ensuring the continuity of the stitched image, ultimately resulting in a complete stitched image. In summary, this preferred embodiment decomposes the seam elimination process of adjacent annular stitched images into a step-by-step process of initial stitching, polar coordinate transformation, targeted seam elimination, and complete image acquisition, significantly improving the accuracy and effectiveness of annular seam elimination, thereby ensuring the quality of the complete stitched image of the silicon wafer and the reliability of subsequent inspection.
[0009] In a preferred embodiment, eliminating the annular seam between adjacent annular stitched images includes: performing grayscale fusion processing on the images at the stitching point to eliminate the annular seam between adjacent annular stitched images.
[0010] Gray-scale fusion processing enables a smooth transition of gray values at the stitching point, avoiding abrupt gray-scale changes and obvious stitching marks caused by direct stitching. This significantly improves the consistency and continuity of the complete stitched image. Gray-scale fusion processing can optimize the stitching point more efficiently without changing the features of other areas of the image, making the processing efficient and accurate, and further ensuring the reliability of the rotating silicon wafer inspection method.
[0011] In a preferred embodiment, step S43, which involves extracting the stitching image of adjacent ring-shaped stitched images based on the linear combination image, includes: dividing the linear combination image into multiple regions to be processed according to a preset pixel width, and extracting the stitching image of adjacent ring-shaped stitched images for each region to be processed.
[0012] By dividing the linear composite image into multiple regions to be processed according to the width of a preset pixel, the originally continuous and complex ring seam can be decomposed into several local, regular and easy-to-process sub-regions, making the seam extraction process more accurate, controllable and efficient, thereby significantly improving the targeting of subsequent grayscale fusion processing and the seam elimination effect.
[0013] In a preferred embodiment, step S3, which involves extracting the stitching image of adjacent corrected images and eliminating the stitching seam between adjacent corrected images, includes: extracting the edge images of each of the two adjacent corrected images, creating a mask image based on the edge images, and using average weighted fusion to eliminate the stitching seam between adjacent corrected images.
[0014] By extracting edge images from adjacent corrected images, the area where the seam is located can be accurately located. Creating a mask image based on the edge image can provide precise guidance for the scope of application of average weighted fusion, so that the fusion process only applies to the seam and surrounding transition area, ensuring the targeting and accuracy of the fusion. Furthermore, by using average weighted fusion, the grayscale transition between adjacent corrected images can be achieved at the seam, completely eliminating obvious seam marks caused by grayscale differences.
[0015] In a preferred embodiment, step S3, which involves linearly oriented stitching the corrected images, includes stitching the corrected images vertically in the scanning order.
[0016] By stitching the corrected images vertically in the scanning order, the orderliness of the stitching process and the rationality of the image spatial distribution are ensured, the positioning error in the stitching process is reduced, omissions are avoided, a high-quality foundation is laid for subsequent seam elimination and circular stitching, the integrity of the silicon wafer image is guaranteed, and the stitching path is optimized.
[0017] In a preferred embodiment, step S2, grouping the images according to the coordinates of the center points of each image, includes: determining the distance between the coordinates of each center point and the coordinates of the silicon wafer center point based on the coordinates of the center points of each image, and grouping the images at the same distance into one group.
[0018] Images at the same distance are grouped together. During the rotational scanning process, images in the same group are at the same radial position and are more consistent with environmental factors such as lighting conditions and camera imaging angle. Based on this distance grouping, flat field correction can be applied to images with similar imaging conditions to accurately offset common grayscale deviations. However, whether images are selected randomly or grouped by a specific number, there are obvious defects. In particular, when selecting by a specific number, images from different annular regions are mixed in the same correction group, which destroys the spatial grayscale consistency of the silicon wafer image.
[0019] In a preferred embodiment, step S2, which involves performing flat-field correction on each image in the same group to obtain a corrected image for each image, includes: splitting each image in the same group into three channels (R, G, and B) according to color; calculating the pixel mean of all images in a single channel; generating a correction template image corresponding to a single channel; performing flat-field correction on each image using the correction template image; and fusing the images from each channel to obtain a corrected image for each image.
[0020] By generating a correction template by calculating the pixel mean of all images in a single channel, the overall grayscale characteristics of that channel under specific imaging conditions can be accurately reflected, making the correction template image more representative and accurate. Using this correction template image to perform flat-field correction on each image effectively eliminates systematic deviations such as uneven illumination and inconsistent sensor response within the same channel, ensuring that the grayscale baseline of each image in the same group remains consistent at the channel level. Finally, fusing the images from each channel yields the corrected image, achieving overall grayscale uniformity while maintaining the integrity of color information, making the corrected image closer to the actual surface features of a silicon wafer. Therefore, the above-mentioned flat-field correction method fully considers the differences in illumination, camera response, and noise characteristics of different color channels during imaging, enabling more precise grayscale compensation at the channel level, thereby significantly improving the color consistency and grayscale uniformity of the corrected image.
[0021] In a preferred embodiment, in step S2, before performing flat field correction on each image in the same group, the field of view of the gripper region of the silicon wafer is filtered out.
[0022] By filtering out the field of view image of the clamping area of the silicon wafer, interference from non-effective areas of the silicon wafer can be effectively eliminated, improving the accuracy of the flat field correction template and the quality of the corrected image, thus providing a reliable image basis for subsequent stitching and defect detection.
[0023] The present invention also provides a silicon wafer inspection device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-described image processing method for silicon wafer inspection.
[0024] The silicon wafer inspection equipment provided by this invention can realize coordinate grouping correction and staged seam elimination, fundamentally solving the problems of uneven grayscale and obvious stitching marks in images obtained by rotational shooting. It not only adapts to the high-efficiency imaging characteristics of rotational shooting, but also significantly improves the reliability and practicality of silicon wafer inspection. Attached Figure Description
[0025] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of an image processing method for silicon wafer inspection in one embodiment of the present invention; Figure 2 This is a flowchart of step S4 of the image processing method for silicon wafer inspection in one embodiment of the present invention; Figure 3This is a schematic diagram showing the positional relationship between the array camera and the silicon wafer in a rotating scanning silicon wafer state according to one embodiment of the present invention; Figure 4 This is a schematic diagram showing the coordinate positions of the center points of each image on a silicon wafer in one embodiment of the present invention; Figure 5 This is a schematic diagram showing the coordinate positions of the center points of the same group of images after grouping, according to one embodiment of the present invention; Figure 6 This is a partial schematic diagram of a linearly stitched image according to one embodiment of the present invention; Figure 7 This is a schematic diagram of a linearly stitched image after seam elimination at the stitching point, according to one embodiment of the present invention. Figure 8 This is a schematic diagram of a circularly stitched image according to one embodiment of the present invention; Figure 9 This is a test result image of a complete stitched image according to one embodiment of the present invention; Figure 10 This is a test result diagram of the complete stitched image obtained from the test example of the present invention. Detailed Implementation
[0026] One or more embodiments of the present invention provide an image processing method for silicon wafer inspection, which can be executed by a silicon wafer inspection device provided by one or more embodiments of the present invention. Certain input parameters or intermediate results in the inspection method can be manually adjusted to help improve accuracy.
[0027] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.
[0028] like Figure 1As shown, in one embodiment of the present invention, an image processing method for silicon wafer inspection is provided, comprising the following steps: S1, obtaining all images of the silicon wafer and the coordinates of the center points of each image by rotating and scanning the silicon wafer; S2, grouping the images according to the coordinates of the center points of each image, performing flat field correction on each image in the same group to obtain a corrected image for each image; S3, linearly oriented stitching the corrected images, extracting the stitching images at the junctions of adjacent corrected images, eliminating the stitching seams of adjacent corrected images to obtain a linearly stitched image, and performing inverse polar coordinate transformation on the linearly stitched image to obtain a ring-shaped stitched image for each group; S4, stitching adjacent ring-shaped stitched images, extracting the stitching images at the junctions of adjacent ring-shaped stitched images, eliminating the ring-shaped stitching seams of adjacent ring-shaped stitched images to obtain a complete stitched image.
[0029] The present invention provides an image processing method for silicon wafer inspection, which uses rotating scanning of silicon wafers to achieve rapid image acquisition. In step S2, flat field correction is performed by grouping images according to the size of the center point coordinates. This avoids the insufficient correction targeting caused by traditional fixed number or random selection of images to generate correction templates. It enables the correction process to adapt to similar imaging conditions of images within the same group, making flat field correction more targeted and accurate, and can accurately offset grayscale deviations generated by images within the same group. After flat-field correction, steps S3 to S4 achieve phased stitching and seam elimination. First, linear directional stitching and seam elimination ensure a smooth transition of images within a single group. Then, inverse polar coordinate transformation yields a ring-shaped stitched image. Finally, the ring seams between adjacent ring images are specifically eliminated, avoiding geometric deviations and grayscale discontinuities that are easily caused by direct coordinate stitching. This effectively ensures image integrity and continuity, ultimately guaranteeing high quality of the complete stitched image. It not only avoids interference from grayscale inconsistencies and seams on the defect recognition algorithm, preventing false detections and missed detections, but also leverages the efficient scanning foundation of rotating imaging to improve detection efficiency while ensuring detection accuracy, balancing both efficiency and quality. In summary, this invention, through the combination of coordinate grouping correction and phased seam elimination, fundamentally solves the problems of uneven grayscale and obvious stitching marks. It not only adapts to the efficient imaging characteristics of rotating imaging but also significantly improves the reliability and practicality of silicon wafer inspection.
[0030] It should be noted that the present invention does not limit the specific implementation means of the above-mentioned grouping correction and the means of eliminating splicing seams at different stages. The following provides different implementation examples of the present invention to describe in detail the image processing method for silicon wafer inspection.
[0031] Implementation Example 1 like Figure 1-9 As shown, in a preferred embodiment, an image processing method for silicon wafer inspection includes: S1, by rotating and scanning the silicon wafer, obtains a complete image of the silicon wafer and the coordinates of the center point of each image.
[0032] Optionally, such as Figure 3 As shown, Figure 3 This diagram illustrates the positional relationship between the area scan camera and the silicon wafer during rotational scanning. By keeping the silicon wafer 200 stationary (fixed by grippers), the area scan camera 100 rotates to achieve rotational scanning of the silicon wafer 200, acquiring all images of the silicon wafer 200 and the coordinates of the center points of each image. For example... Figure 4 As shown, Figure 4 This is a schematic diagram showing the coordinate positions of the center points of each image on the silicon wafer.
[0033] More specifically, before scanning, the number of rotations required by the area scan camera is calculated based on the silicon wafer diameter and the field of view width of the area scan camera. The path length for each rotation is then calculated. Based on the camera's field of view length, the circular path is divided into equal parts, and the positional interval angle and the camera's shooting angle are calculated. This yields all the shooting coordinates along the scanning path. The optical system then rotates the area scan camera to completely capture the silicon wafer along the circular trajectory, obtaining an image of the entire wafer. To capture the entire wafer completely, overlapping areas must be considered when planning the coordinates to cover the entire wafer and obtain a complete image.
[0034] S2, group the images according to the coordinates of their center points, and perform flat-field correction on the images within the same group to obtain corrected images for each image. Preferably, grouping the images according to the coordinates of their center points includes: determining the distance between the coordinates of each center point and the coordinates of the silicon wafer center point, and grouping images at the same distance into a group. Combined with... Figure 4 , 5 ,like Figure 4 This is a schematic diagram showing the coordinates of the center points of each image; for example... Figure 5 This is a schematic diagram showing the coordinate positions of the center points of the same group of images after grouping.
[0035] Preferably, the field images of the gripper area of the silicon wafer are filtered out before flat field correction is performed on each image in the same group.
[0036] Preferably, flat-field correction is performed on each image in the same group to obtain a corrected image for each image, including: splitting each image in the same group into three channels (R, G, and B) according to color, calculating the pixel mean of all images in a single channel, generating a correction template image corresponding to a single channel, performing flat-field correction on each image using the correction template image, and fusing the images of each channel to obtain a corrected image for each image.
[0037] S3, perform linear orientation stitching on the corrected images, extract the stitching area between adjacent corrected images, eliminate the stitching seams between adjacent corrected images, and obtain a linearly stitched image. For example... Figure 6 As shown, Figure 6 This illustrates a portion of a linearly stitched image.
[0038] Preferably, by stitching the corrected images vertically in scanning order, extracting the edge images of each of two adjacent corrected images, creating a mask image based on the edge images, and using average weighted fusion, the stitching seams between adjacent corrected images are eliminated. Figure 7 As shown, Figure 7 This illustrates the effect of eliminating seams at the seams of a linearly stitched image.
[0039] Specifically, the stitched images obtained by vertical stitching are cropped with overlapping portions of 10 pixels to obtain an upper edge image group and a lower edge image group. These two image groups are used to create a mask image at the stitching point between adjacent images. The upper edge image group and the mask image created using the upper edge image are converted into bitmaps and multiplied together to obtain a set of images. The lower edge image group is processed in the same way to obtain a set of images. The two sets of images are added together with a weight of 0.5 to convert them into pixel images, thus eliminating the stitching seams between adjacent corrected images.
[0040] Next, the linearly stitched images are transformed in reverse polar coordinates to obtain the circular stitched images for each group. Understandably, this is achieved by repeating steps S2 and S3 to obtain the circular stitched images for each group. For example... Figure 8 As shown, Figure 8 This is a schematic diagram of a circular stitched image.
[0041] It should also be explained that, as will be understood by those skilled in the art, in this application... Figure 7 , Figure 8 The color differences between the images, and Figure 8 The color difference between the inner and outer rings did not affect the seam elimination effect of this application. The color difference was only affected by the product back sealing process, and there was a color transition in the edge area.
[0042] S4, stitch adjacent ring-shaped images together, extract the image at the stitching point of adjacent ring-shaped images, and eliminate the ring-shaped seam between adjacent ring-shaped images to obtain a complete stitched image. Preferably, this step specifically includes: S41, stitch adjacent ring-shaped images together to obtain a preliminary combined image; that is, combine two adjacent ring-shaped images.
[0043] S42, the preliminary combined image is transformed into a linear combined image through polar coordinate transformation; that is, the stitched ring is changed into a straight line.
[0044] S43, based on the linear combination image, extract the splicing image of adjacent ring-shaped spliced images and eliminate the ring-shaped splicing seam of adjacent ring-shaped spliced images; More preferably, extracting the stitching area image of adjacent ring-shaped stitched images based on the linear combination image includes: dividing the linear combination image into multiple regions to be processed according to a preset pixel width, and extracting the stitching area image of adjacent ring-shaped stitched images for each region to be processed. The ring-shaped stitching seam between adjacent ring-shaped stitched images is eliminated by performing grayscale fusion processing on the stitching area image.
[0045] Specifically, the initial combined image is converted into a horizontally linearly combined image. The image is then divided horizontally into multiple regions with a width of 5 pixels each. For each region, a grayscale projection is performed vertically to obtain a grayscale projection curve. The maximum and minimum values of the curve are obtained using differentiation, thus acquiring the grayscale coordinates of the extreme points. The difference in coordinate spacing between adjacent extreme points is calculated, and one-third of this spacing is taken as the interval for modifying the pixel value at the vertical coordinate of the extreme point. The average grayscale value within this interval is used as the target grayscale value for modification, and then the pixel value is modified. This process is repeated to modify the grayscale of all regions of the linearly combined image, thereby eliminating the circular seams between adjacent circularly stitched images.
[0046] S44, to obtain the complete stitched image.
[0047] Understandably, in this embodiment, the rotating area array camera scans along a set trajectory in step S1, which improves scanning efficiency and provides a more comprehensive and clearer image of the circular silicon wafer.
[0048] In step S2, by filtering out the field-of-view image of the clamping area of the silicon wafer, interference from non-effective areas of the silicon wafer can be effectively eliminated, improving the accuracy of the flat-field correction template and the quality of the corrected image, thus providing a reliable image basis for subsequent stitching and defect detection. Furthermore, grouping images at the same distance ensures that the images in the same group are at the same radial position during rotational scanning, resulting in greater consistency in the influence of environmental factors such as lighting conditions and camera imaging angles. Based on this distance grouping, flat-field correction can be specifically applied to images with similar imaging conditions, accurately offsetting common grayscale deviations. However, both randomly selecting images and grouping images by a specific number have significant drawbacks. Specifically, when selecting by a specific number, images from different annular regions may be mixed in the same correction group, disrupting the spatial grayscale consistency of the silicon wafer image.
[0049] During flat-field correction, a correction template is generated by calculating the pixel mean of all images in a single channel. This template accurately reflects the overall grayscale characteristics of that channel under specific imaging conditions, making it more representative and accurate. Using this template to perform flat-field correction on each image effectively eliminates systematic biases such as uneven illumination and inconsistent sensor response within the same channel, ensuring consistent grayscale references across images in the same group at the channel level. Finally, the images from each channel are fused to obtain the corrected image, achieving overall grayscale uniformity while maintaining color information integrity. This makes the corrected image closer to the actual surface features of a silicon wafer. Therefore, the flat-field correction method described above fully considers the differences in illumination, camera response, and noise characteristics of different color channels during imaging, enabling more precise grayscale compensation at the channel level and significantly improving the color consistency and grayscale uniformity of the corrected image.
[0050] In step S3, by stitching the corrected images vertically in the scanning order, the orderliness of the stitching process and the rationality of the image spatial distribution are ensured. This reduces positioning errors during the stitching process, avoids omissions, lays a solid foundation for subsequent seam elimination and circular stitching, ensures the integrity of the silicon wafer image, and optimizes the stitching path. Furthermore, by extracting edge images from adjacent corrected images, the seam area can be accurately located. Creating a mask image based on the edge images provides precise guidance for the effective range of average weighted fusion, ensuring that the fusion processing only targets the seam and surrounding transition areas, guaranteeing the targeting and accuracy of the fusion. The average weighted fusion method allows for a smooth grayscale transition between adjacent corrected images at the seam, completely eliminating obvious seam marks caused by grayscale differences.
[0051] In step S4, the initial combination of adjacent annular stitched images is completed in S41, providing a foundation for subsequent seam processing. Then, in S42, the initial combined image is converted into a linear combined image, making the processing more comprehensive and precise. Since the subsequent stitched images and seams are linear, seam extraction and elimination are easier, enabling accurate location and extraction of the annular stitched seam region. Step S43, based on the linear combined image, accurately extracts the stitched images of adjacent annular stitched images, achieving targeted elimination of the annular seams. This avoids the problem of unprocessed seam residue from direct stitching or general processing. The grayscale transition processing is thorough, ensuring the continuity of the stitched image, ultimately resulting in a complete stitched image. In this preferred embodiment, the seam elimination process of adjacent annular stitched images is broken down into a step-by-step process of initial stitching, polar coordinate transformation, targeted seam elimination, and complete image acquisition, significantly improving the accuracy and effectiveness of annular seam elimination, thereby ensuring the quality of the complete stitched image of the silicon wafer and the reliability of subsequent inspection. Gray-scale fusion processing enables a smooth transition of gray values at the stitching point, avoiding abrupt gray-scale changes and obvious stitching marks caused by direct stitching. This significantly improves the consistency and continuity of the complete stitched image. Gray-scale fusion processing can optimize the stitching point more efficiently without changing the features of other areas of the image, making the processing efficient and accurate, and further ensuring the reliability of the rotating silicon wafer inspection method.
[0052] By dividing the linear composite image into multiple regions to be processed according to the width of a preset pixel, the originally continuous and complex ring seam can be decomposed into several local, regular and easy-to-process sub-regions, making the seam extraction process more accurate, controllable and efficient, thereby significantly improving the targeting of subsequent grayscale fusion processing and the seam elimination effect.
[0053] Figure 9 This is a test result image of the complete stitched image obtained using the image processing method for silicon wafer inspection provided in this embodiment. To further illustrate the above-mentioned effect of this embodiment, a test case is set up. Specifically, for the same silicon wafer, a planar array camera is used to perform a rotating shooting method consistent with this embodiment to acquire all images on the silicon wafer. All images are then stitched together to obtain a complete stitched image. Figure 10 This is the test result image of the complete stitched image for this test case.
[0054] Combination Figure 9 and Figure 10 As can be seen, the stitching seam of the complete stitched image in the test example is more obvious than the stitching trace of the complete stitched image obtained in this embodiment. The image processing method provided in this embodiment can ensure the integrity and continuity of the image and improve the quality of the complete stitched image.
[0055] Implementation Example 2 The main difference between this embodiment and Embodiment 1 lies in the seam elimination method in step S3. This embodiment provides an image processing method for silicon wafer inspection, including the following steps: S1. Obtain all images of the silicon wafer and the coordinates of the center points of each image by rotating and scanning; S2. Group the images according to the coordinates of the center points of each image, and perform flat field correction on each image in the same group to obtain the corrected image of each image; S3. Perform linear orientation stitching on the corrected images, extract the stitching images of adjacent corrected images, eliminate the stitching seams of adjacent corrected images to obtain a linear stitched image, and perform inverse polar coordinate transformation on the linear stitched image to obtain the ring stitched image of each group; S4. Stitch adjacent ring stitched images, extract the stitching images of adjacent ring stitched images, eliminate the ring stitching seams of adjacent ring stitched images to obtain a complete stitched image.
[0056] In step S3, the corrected images are linearly oriented and stitched together. The stitching area between adjacent corrected images is extracted, and the stitching seams between adjacent corrected images are eliminated by grayscale fusion to obtain a linearly stitched image. The linearly stitched image is then transformed in reverse polar coordinates to obtain the ring-shaped stitched image of each group.
[0057] Implementation Example 3 The main difference between this implementation example and implementation example 1 lies in the seam elimination method in step S4.
[0058] This implementation example provides an image processing method for silicon wafer inspection, including the following steps: S1. Obtain all images of the silicon wafer and the coordinates of the center points of each image by rotating and scanning; S2. Group the images according to the coordinates of the center points of each image, and perform flat field correction on each image in the same group to obtain the corrected image of each image; S3. Perform linear orientation stitching on the corrected images, extract the stitching images of adjacent corrected images, eliminate the stitching seams of adjacent corrected images to obtain a linear stitched image, and perform inverse polar coordinate transformation on the linear stitched image to obtain the ring stitched image of each group; S4. Stitch adjacent ring stitched images, extract the stitching images of adjacent ring stitched images, eliminate the ring stitching seams of adjacent ring stitched images to obtain a complete stitched image.
[0059] In step S4, S41, adjacent ring-shaped stitched images are stitched together to obtain a preliminary combined image; that is, two adjacent ring-shaped stitched images are combined. S42, the preliminary combined image is transformed into a linear combined image through polar coordinate transformation; that is, the stitched rings are replaced with straight lines; this can be horizontal or vertical. S43, the stitching area images of adjacent ring-shaped stitched images are extracted based on the linear combined image. Preferably, a mask image is created for the stitching area image, and average weighted fusion is used to eliminate the ring-shaped stitching seams of adjacent ring-shaped stitched images. S44, a complete stitched image is obtained.
[0060] It should be noted that, taking into account the mirroring effect of scanning and capturing images, in order to obtain the true coordinates of the complete image, in a preferred embodiment, an image processing method for silicon wafer inspection further includes, after obtaining the complete stitched image, mirroring the complete stitched image to obtain the final stitched image with the true coordinates. Of course, it can be understood that, in practice, after step S1, all images of the acquired silicon wafer can be mirrored first, the true coordinates of each image can be matched, and then the images can be stitched together.
[0061] Additionally, it should be noted that in other embodiments of the present invention, step S1 may also be selected as achieving full rotational scanning and imaging of the silicon wafer by precisely controlling the rotation of the silicon wafer using a fixed area array camera.
[0062] In other embodiments of the present invention, in step S3, the corrected image is linearly oriented and stitched, which may be selected as stitching along the horizontal direction.
[0063] In another embodiment of the present invention, a silicon wafer inspection device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-described image processing method for silicon wafer inspection.
[0064] More preferably, the silicon wafer inspection equipment includes an area scan camera connected to the memory and a control device that drives the area scan camera to rotate according to a preset trajectory.
[0065] The silicon wafer inspection equipment provided by this invention can realize coordinate grouping correction and staged seam elimination, fundamentally solving the problems of uneven grayscale and obvious stitching marks in images obtained by rotational shooting. It not only adapts to the high-efficiency imaging characteristics of rotational shooting, but also significantly improves the reliability and practicality of silicon wafer inspection.
[0066] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0067] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0068] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The aforementioned units can be implemented in hardware or software.
[0069] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. An image processing method for silicon wafer inspection, characterized in that, Includes the following steps: S1, obtain all images of the silicon wafer and the coordinates of the center points of each image by rotating and scanning the silicon wafer; S2, group the images according to the coordinates of the center points of each image, perform flat field correction on each image in the same group, and obtain the corrected image of each image; S3, perform linear orientation stitching on the corrected images, extract the stitching images of adjacent corrected images, eliminate the stitching seams of adjacent corrected images to obtain linear stitched images, and perform inverse polar coordinate transformation on the linear stitched images to obtain the annular stitched images of each group. S4, stitch the adjacent ring-shaped stitched images together, extract the stitching area of the adjacent ring-shaped stitched images, eliminate the ring-shaped stitching seams of the adjacent ring-shaped stitched images, and obtain a complete stitched image.
2. The image processing method for silicon wafer inspection according to claim 1, characterized in that, Step S4 includes: S41, the adjacent ring-shaped stitched images are stitched together to obtain a preliminary combined image; S42, the preliminary combined image is transformed using polar coordinates to obtain a linear combined image; S43, based on the linear combination image, extract the splicing image of adjacent ring-shaped spliced images, and eliminate the ring-shaped splicing seam of adjacent ring-shaped spliced images; S44, to obtain the complete stitched image.
3. An image processing method for silicon wafer inspection according to claim 1 or 2, characterized in that, The elimination of the ring seam between adjacent ring-stitched images includes: The images at the stitching point are subjected to grayscale fusion processing to eliminate the ring-shaped stitching seams between adjacent ring-shaped stitched images.
4. The image processing method for silicon wafer inspection according to claim 2, characterized in that, In step S43, extracting the stitching image at the junction of adjacent ring-shaped stitched images based on the linearly combined image includes: Based on the linear combination image, multiple regions to be processed are divided according to the width of a preset pixel, and the splicing image of the adjacent ring spliced image is extracted for each region to be processed.
5. The image processing method for silicon wafer inspection according to claim 1, characterized in that, In step S3, extracting the stitching area image of adjacent corrected images and eliminating the stitching seam between adjacent corrected images includes: The edge images of two adjacent corrected images are extracted, a mask image is created based on the edge images, and average weighted fusion is used to eliminate the stitching seams between adjacent corrected images.
6. The image processing method for silicon wafer inspection according to claim 1, characterized in that, In step S3, the linear orientation stitching of the corrected image includes: The corrected images are stitched together vertically in the scanning order.
7. The image processing method for silicon wafer inspection according to claim 1, characterized in that, In step S2, grouping the images according to the coordinates of their center points includes: Based on the coordinates of the center points of each image, the distance between the coordinates of each center point and the coordinates of the center point of the silicon wafer is determined, and the images at the same distance are grouped together.
8. The image processing method for silicon wafer inspection according to claim 1, characterized in that, In step S2, performing flat-field correction on each image in the same group to obtain corrected images for each image includes: Each image in the same group is split into three channels, R, G, and B, according to color. The pixel mean of all images in a single channel is calculated, and a correction template image corresponding to a single channel is generated. The correction template image is used to perform flat field correction on each image, and the images of each channel are fused to obtain the corrected image of each image.
9. The image processing method for silicon wafer inspection according to claim 1, characterized in that, In step S2, before performing flat field correction on each image in the same group, the field of view of the gripper area of the silicon wafer is filtered out.
10. A silicon wafer inspection device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executed, implements the image processing method for silicon wafer inspection as described in any one of claims 1-9.