WO2023185932A1 - 硅片对准标记的检测定位方法、系统、电子设备及介质 - Google Patents

硅片对准标记的检测定位方法、系统、电子设备及介质 Download PDF

Info

Publication number
WO2023185932A1
WO2023185932A1 PCT/CN2023/084727 CN2023084727W WO2023185932A1 WO 2023185932 A1 WO2023185932 A1 WO 2023185932A1 CN 2023084727 W CN2023084727 W CN 2023084727W WO 2023185932 A1 WO2023185932 A1 WO 2023185932A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
detection image
alignment mark
pixel
alignment marks
Prior art date
Application number
PCT/CN2023/084727
Other languages
English (en)
French (fr)
Inventor
易兵
刘涛
周许超
张记晨
鲁阳
高贯义
王敬贤
Original Assignee
上海微电子装备（集团）股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海微电子装备（集团）股份有限公司 filed Critical 上海微电子装备（集团）股份有限公司
Publication of WO2023185932A1 publication Critical patent/WO2023185932A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding

Definitions

  • the present invention relates to the field of semiconductor manufacturing technology, and in particular to a method, system, electronic equipment and medium for detecting and positioning silicon wafer alignment marks.
  • Alignment marks in silicon wafers have low reflectivity and are susceptible to interference from pattern reflections or scattered signals.
  • Existing detection and positioning methods for silicon wafer alignment marks have a high error rate, which affects alignment accuracy and also reduces detection efficiency. lower.
  • the object of the present invention is to provide a method, system, electronic equipment and medium for detecting and positioning silicon wafer alignment marks, so as to solve the problems of high error rate and/or low efficiency of existing methods for detecting and positioning silicon wafer alignment marks.
  • the present invention provides a method for detecting and positioning silicon wafer alignment marks, which includes:
  • a knowledge base has first feature information of the first detection image of several known alignment marks, the first feature information includes the pixel feature statistical information of the first detection image and the Pixel feature information of the area of interest where the known alignment mark is located in the first detection image;
  • the second feature information includes pixel feature statistical information of the second detection image and the second detection Pixel feature information of the area of interest where the alignment mark to be measured is located in the image;
  • the first characteristic information and the second characteristic information are compared item by item for similarity, and the position of the alignment mark to be measured on the silicon wafer is obtained based on several known alignment marks obtained through the comparison. letter interest.
  • the first detection image and the second detection image have the same size.
  • a portion of the captured image with the known alignment mark is intercepted as the first detection image using manual outline.
  • the part of the alignment mark to be measured is used as the second detection image.
  • the image pixel feature processing method includes an image segmentation and positioning method or an image symmetry calculation method.
  • the step of obtaining the first characteristic information includes:
  • the step of obtaining the second characteristic information includes:
  • the image of the area of interest where the alignment mark to be measured is located in the second detection image is obtained based on the position information and size information of the area of interest where the alignment mark to be measured is located in the second detection image. element feature information.
  • the pixel feature statistical information is used to characterize grayscale projection signals in at least two directions, and the image grayscale projection method is used to obtain the pixel feature statistics of the first detection image and/or the second detection image. information.
  • the grayscale projection signal is processed using an image pixel feature processing method to obtain the position information and size information of the region of interest.
  • the image pixel feature processing method includes an image segmentation and positioning method or an image symmetry calculation method.
  • pixel feature statistical information of the first detection image before obtaining the pixel feature statistical information of the first detection image, perform neighborhood pixel feature calculation on each pixel of the first detection image; and/or obtain the pixel features of the second detection image. Before statistical information, neighborhood pixel features are calculated for each pixel of the second detection image.
  • the calculation of neighborhood pixel features includes calculation of pixel mean within the neighborhood, calculation of pixel weighted mean, calculation of pixel gradient or calculation of pixel extreme value range.
  • the first characteristic information and the second characteristic information are compared item by item for similarity, and the alignment mark to be tested is obtained in silicon based on several known alignment marks obtained through the comparison.
  • the steps to generate on-chip location information include:
  • the pixel feature statistical information of the second detection image and the pixel feature statistical information of the first detection image of the known alignment mark are compared one by one for similarity, and the confidence of the similarity is greater than the first set value.
  • the position information of the alignment mark to be measured on the silicon wafer is obtained based on several preferred known alignment marks.
  • the knowledge base divides the known alignment marks into at least two categories based on the pixel feature statistics of the first detection image, and obtains all the known alignments in each category.
  • the first characteristic information and the second characteristic information are compared item by item for similarity, and the position of the alignment mark to be measured on the silicon wafer is obtained based on several known alignment marks obtained through the comparison.
  • Information steps include:
  • the candidates are All the known alignment marks in the category are the known alignment marks as candidates;
  • the position information of the alignment mark to be measured on the silicon wafer is obtained based on several preferred known alignment marks.
  • a phase correlation method based on fast Fourier transform or a template matching search method is used to compare the pixel feature statistical information of the second detection image with the pixel feature statistical information of the first detection image of the known alignment mark. Compare the similarities one by one.
  • the pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image is compared with the location of the known alignment mark in the first detection image of the best candidate.
  • the pixel feature information of the area of interest is compared for similarity to obtain the preferred known alignment mark.
  • Similarity is performed between the pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image and the pixel feature information of the area of interest where all known alignment marks are located in the category of the best candidate. Compare and obtain the preferred known alignment mark.
  • the pixel feature information includes one or more of the overall pixel value of the region of interest, geometric feature information, and grayscale feature information, and the geometric feature information includes gradient information of the region of interest. and/or marginal information.
  • a pixel grayscale template alignment method and/or a geometric template alignment method to compare the pixel feature information of the area of interest where the alignment mark to be measured in the second detection image is with the candidate known
  • the pixel feature information of the area of interest where the known alignment mark is located in the first detection image of the alignment mark is compared one by one for similarity.
  • the pixel grayscale template alignment method includes a standard optical flow method or a reverse sequence synthetic image alignment algorithm.
  • the pixel feature information corresponding to the known alignment marks obtained by comparison uses the pixel feature information corresponding to the known alignment marks obtained by comparison to traverse the second detection image to obtain the second detection image.
  • the most similar position to the pixel feature information corresponding to the known alignment mark obtained by comparison is obtained, thereby obtaining the position information of the alignment mark to be measured on the silicon wafer.
  • the second characteristic information of the alignment mark to be measured is stored in in the knowledge base.
  • the invention also provides a detection and positioning system for silicon wafer alignment marks, including:
  • a storage module configured to store a knowledge base that contains first feature information of the first detection image of several known alignment marks, where the first feature information includes pixel features of the first detection image Statistical information and pixel feature information of the area of interest where the known alignment mark is located in the first detection image;
  • a feature acquisition module configured to provide a second detection image of the alignment mark to be measured, and to obtain second feature information of the second detection image, where the second feature information includes pixel feature statistical information of the second detection image. and the area of interest where the alignment mark to be measured is located in the second detection image. Pixel feature information; and,
  • a similarity comparison module configured to compare the similarity between the first feature information and the second feature information item by item, and obtain the pair to be tested based on several known alignment marks obtained through comparison. Accurate mark position information on the silicon wafer.
  • the present invention also provides an electronic device, including a processor and a memory. Instructions are stored on the memory. When the instructions are executed by the processor, the method for detecting and positioning silicon wafer alignment marks is implemented.
  • the present invention also provides a non-transitory computer-readable storage medium. Instructions are stored on the non-transitory computer-readable storage medium. When the instructions are executed, the detection and positioning of the silicon wafer alignment mark is realized. method.
  • a knowledge base is first established.
  • the knowledge base contains the first characteristic information of the first detection image of several known alignment marks, and then the object to be tested is obtained.
  • the second feature information of the second detection image of the alignment mark is compared item by item for similarity between the first feature information and the second feature information, and is obtained based on several known alignment marks obtained through the comparison.
  • the first feature information includes pixel feature statistical information of the first detection image and pixel feature information of the area of interest where the known alignment mark is located in the first detection image
  • the second feature information including pixel feature statistical information of the second detection image and pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image, after comparing the first feature information and the second feature information
  • the similarity between the first detection image and the second detection image is actually compared as a whole to filter out the number of known alignment marks from a larger number.
  • the present invention also provides a detection and positioning system for silicon wafer alignment marks, electronic equipment and a non-transitory computer-readable storage medium.
  • Figure 1 is a flow chart of a method for detecting and positioning silicon wafer alignment marks provided by Embodiment 1 of the present invention
  • Figure 2 is a schematic diagram of the first detection image intercepted from the first original image provided in Embodiment 1 of the present invention
  • 3a and 3b are schematic diagrams of the first grayscale projection signal and the second grayscale projection signal obtained by grayscale projection of the first detection image along the horizontal direction and the vertical direction respectively according to Embodiment 1 of the present invention;
  • Figure 4 is a schematic diagram of intercepting the first matrix image from the first detection image provided in Embodiment 1 of the present invention
  • 5a and 5b are schematic diagrams of the third grayscale projection signal and the fourth grayscale projection signal obtained by grayscale projection of the second detection image along the horizontal direction and the vertical direction respectively according to Embodiment 1 of the present invention;
  • Figures 6a and 6b are schematic diagrams of the first similarity signal and the second similarity signal provided by Embodiment 1 of the present invention.
  • Figure 7 is a structural block diagram of a silicon wafer alignment mark detection and positioning system provided in Embodiment 1 of the present invention.
  • FIG. 1 is a flow chart of a method for detecting and positioning silicon wafer alignment marks provided in this embodiment. As shown in Figure 1, the detection and positioning method of the silicon wafer alignment mark includes:
  • Step S100 Provide a knowledge base that contains first feature information of the first detection image of several known alignment marks, where the first feature information includes images of the first detection image. Pixel feature statistical information and pixel feature information of the area of interest where the known alignment mark is located in the first detection image;
  • Step S200 Provide a second detection image of the alignment mark to be measured, and obtain second feature information of the second detection image.
  • the second feature information includes pixel feature statistical information of the second detection image and Pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image;
  • Step S300 Compare the first characteristic information and the second characteristic information item by item for similarity, and obtain the alignment mark to be tested on the silicon wafer based on several known alignment marks obtained through the comparison. location information on.
  • step S100 is executed to build the knowledge base.
  • a silicon wafer having the known alignment mark is provided (hereinafter referred to as the silicon wafer having the known alignment mark as the first silicon wafer), and the known alignment mark is on the first silicon wafer.
  • the location information on the chip is certain and known.
  • the first original image needs to have complete known alignment marks , that is, the known alignment mark needs to be within the imaging field of view of the first image acquisition unit.
  • the first image acquisition unit can take photos of the known alignment marks on the first silicon wafer one by one. , thereby obtaining multiple first original images; of course, if the first image acquisition unit simultaneously takes photos of multiple known alignment marks on the first silicon wafer, the photos can be manually The image is processed to obtain a first original image of a single known alignment mark. That is, there is only one known alignment mark in each first original image.
  • silicon wafer generally refers to a substrate formed of semiconductor or non-semiconductor materials.
  • semiconductor or non-semiconductor materials include, but are not limited to, single crystal silicon, gallium nitride, gallium arsenide, indium phosphide, sapphire, and glass.
  • substrates may be commonly found and/or processed in semiconductor fabrication facilities.
  • the size of the first original image collected by the first image acquisition unit is large (the area occupied by the known alignment mark in the first original image is small), in this embodiment, intercept Place The part with the known alignment mark in the first original image is used as the first detection image.
  • the size of the first detection image is smaller and has fewer pixels, which can reduce the time required for subsequent image processing. difficulty and data volume, increasing the speed of image processing.
  • manual outlining can be used to capture the portion of the first detection image with the known alignment mark as the first detection image, thereby ensuring that the first detection image has the known alignment mark.
  • the alignment mark is known, and manual delineation can ensure that the size of the intercepted first detection image is appropriate.
  • Figure 2 is a schematic diagram of the first detection image intercepted from the first original image provided in this embodiment.
  • the size of the first original image 100 is larger, and the size of the intercepted first detection image 101 is smaller, but the first detection image 101 has the complete known pair. Quasi mark 200.
  • the shape of the first detection image 101 is a rectangle, thereby facilitating subsequent image processing; as an optional embodiment, the shape of the first detection image 101 is not limited to a rectangle, and may also be Is other shapes, such as cross, etc.
  • the first original image may not be intercepted.
  • the first original image may be directly used as the first detection image to perform subsequent steps.
  • the first pixel feature statistical information is used to characterize grayscale projection signals in at least two directions of the first detection image, and the first pixel feature statistical information is obtained using an image grayscale projection method.
  • the first pixel feature statistical information is used to characterize the grayscale projection signals in the horizontal and vertical directions of the first detection image
  • the image grayscale projection method is used to obtain the first pixel feature statistical information.
  • grayscale projection is performed on the first detection image along the horizontal direction and the vertical direction, respectively, to obtain a first grayscale projection signal along the horizontal direction and a second grayscale projection signal along the vertical direction.
  • neighborhood pixel features may be calculated for each pixel of the first detection image to remove noise in the first detection image.
  • the calculation of neighborhood pixel features includes calculation of pixel mean value in the neighborhood, calculation of pixel weighted mean value, calculation of pixel gradient or calculation of pixel extreme value range, etc.
  • Pixel feature information (hereinafter referred to as the pixel feature information of the area of interest where the known alignment mark is located in the first detection image is the first pixel feature information).
  • the first pixel feature information may include one or more of the overall pixel value of the area of interest where the known alignment mark is located in the first detection image, geometric feature information, and grayscale feature information.
  • the geometric feature information includes gradient information and/or edge information of the region of interest.
  • the area of interest is a rectangular area.
  • the first pixel feature information may include the overall pixel value of the rectangular area where the known alignment mark is located in the first detection image, One or more of geometric feature information and grayscale feature information, the geometric feature information includes gradient information and/or edge information of the rectangular area.
  • the position information and size information of the area of interest where the known alignment mark is located in the first detection image can be first obtained based on the first pixel feature statistical information. , and then obtain the first pixel feature information based on the position information and size information of the area of interest where the known alignment mark is located in the first detection image.
  • FIG. 3a and FIG. 3b show the first gray-scale projection signal and the second gray-scale projection signal obtained by performing gray-scale projection of the first detection image along the horizontal direction and the vertical direction respectively according to this embodiment.
  • Schematic diagram As shown in Figure 3a and Figure 3b, when obtaining the first pixel feature information, autocorrelation calculations can be performed on the first grayscale projection signal and the second grayscale projection signal respectively to obtain the first detection image.
  • the position information (including position information along the horizontal direction and the vertical direction) of the area of interest where the known alignment mark is located is based on the peaks and troughs of the first gray-scale projection signal and the second gray-scale projection signal.
  • the size information (including size information along the horizontal direction and the vertical direction) of the area of interest where the known alignment mark is located in the first detection image can be obtained.
  • the first area of interest where the known alignment mark is located in the first detection image can be The matrix image is intercepted, and then one or more of the overall pixel value, geometric feature information, and grayscale feature information of the first matrix image are obtained as the first pixel feature information.
  • the image pixel feature processing method can also be used to process the first gray-scale projection signal and the second gray-scale projection signal to obtain the known known value in the first detection image. Position information and size information of the area of interest where the alignment mark is located.
  • Figure 4 shows the first matrix image intercepted from the first detection image provided in this embodiment. schematic diagram. As shown in Figure 4, the size of the first detection image 101 is larger, and the size of the intercepted first matrix image 102 is smaller, but the first matrix image 102 also has the complete known Alignment mark 200.
  • the area of interest is a rectangle, but it should not be limited to this. The area of interest can also be in other shapes, such as a cross shape.
  • the first matrix image may not be intercepted from the first detection image, based on the first detection image and the known alignment mark in the acquired first detection image.
  • the first pixel feature information can be directly calculated from the position information and size information of the area of interest.
  • first pixel feature statistical information and the first pixel feature information together constitute the first feature information of the first detection image. It can be understood that the pixel feature statistical information can characterize the overall pixel features of the first detection image, and the first pixel feature information can characterize the ability of the first detection image to accommodate the known alignment. Label the pixel features of a smaller region of interest.
  • the first feature information generates formatted data according to a standard template, thereby forming the knowledge base.
  • the knowledge base may be stored in a computer memory medium or a non-volatile storage medium in the form of formatted files and/or database form data.
  • the formatted files include but are not limited to formatted binary files, formatted text files, JSON format files, XML format files, YAML format files or CSV format files, etc.
  • the several known alignment marks referred to in this embodiment refer to the known alignment marks with different positions on the first silicon wafer, for example: on the same first silicon wafer.
  • the known alignment marks at different positions on the silicon wafer should be considered as different types of the known alignment marks; the known alignment marks on the different first silicon wafers, even if The positions are corresponding (the first silicon wafers of the same batch are manufactured successively according to the same process), and due to process deviations, they should also be considered as different types of known alignment marks.
  • step S200 is performed to obtain second feature information of the second detection image of the alignment mark to be measured.
  • a silicon wafer having the alignment mark to be measured is provided (hereinafter referred to as the silicon wafer having the alignment mark to be measured as the second silicon wafer), and the alignment mark to be measured is on the second silicon wafer.
  • the on-chip position information is unknown and needs to be detected.
  • a second image acquisition unit is used to take pictures of the alignment mark to be tested on the second silicon wafer to obtain a second original image.
  • the second original image needs to have the complete alignment mark to be tested. , that is, the alignment mark to be measured needs to be within the imaging field of view of the second image acquisition unit.
  • the second image acquisition unit can take photos of the alignment marks to be measured on the second silicon wafer one by one. , thereby obtaining multiple second original images. That is, each second original image has only one alignment mark to be measured.
  • intercept The part of the second original image with the alignment mark to be measured is used as the second detection image.
  • the size of the second detection image is smaller and has fewer pixels, which can reduce subsequent image processing.
  • the difficulty and data volume can increase the speed of image processing.
  • predetermined position information and size information can be used to intercept the portion with the alignment mark to be measured from the second original image as the second detection image; or, an image pixel feature processing method can be used to intercept all the alignment marks.
  • the portion of the second original image with the alignment mark to be measured is used as the second detection image, and the image pixel feature processing method may be an image segmentation and positioning method or an image symmetry calculation method, etc.
  • the second original image may not be intercepted.
  • the second original image may be directly used as the second detection image to perform subsequent steps.
  • the size of the first detection image and the second detection image are the same to facilitate subsequent similarity comparison steps. Therefore, if the predetermined position information and size information are used to intercept the portion with the alignment mark to be measured from the second original image as the second detection image, the location of the alignment mark to be measured can be obtained in advance.
  • the approximate position on the second silicon wafer is used, and the second detection image is intercepted from the second original image using the same size as the first detection image; and the image pixel feature processing method is used to intercept the second detection image.
  • the size of the image to be intercepted is preset according to the size of the first detected image.
  • the second pixel feature statistical information is used to characterize the grayscale projection signals in at least two directions of the second detection image, and the second pixel feature statistical information is obtained using an image grayscale projection method.
  • the second pixel feature statistical information is used to characterize the grayscale projection signals in the horizontal and vertical directions of the second detection image
  • the image grayscale projection method is used to obtain the second pixel feature statistical information.
  • grayscale projection is performed on the second detection image along the horizontal direction and the vertical direction respectively, and a third grayscale projection signal along the horizontal direction and a fourth grayscale projection signal along the vertical direction are respectively obtained.
  • neighborhood pixel features may be calculated for each pixel of the second detection image to remove noise in the second detection image.
  • the calculation of neighborhood pixel features includes calculation of pixel mean value in the neighborhood, calculation of pixel weighted mean value, calculation of pixel gradient or calculation of pixel extreme value range, etc.
  • the information is the second pixel feature information).
  • the second pixel feature information may include one or more of the overall pixel value of the rectangular area where the alignment mark to be measured is located in the second detection image, geometric feature information, and grayscale feature information.
  • the geometric feature information includes gradient information and/or edge information of the rectangular area.
  • the second pixel feature information obtained at this time should contain corresponding feature information.
  • the first pixel feature information only contains geometric feature information, so The second pixel feature information should also include geometric feature information; if the first pixel feature information includes multiple features, the second pixel feature information obtained at this time may include one or more of the corresponding feature information,
  • the first pixel feature information includes geometric feature information and grayscale feature information, and the second pixel feature information may include geometric feature information and/or grayscale feature information.
  • the second pixel feature information may first be obtained based on the second pixel feature information. Obtain the position information and size information of the area of interest where the alignment mark to be measured is located in the second detection image, and then based on the area of interest of the alignment mark to be measured in the second detection image. Position information and size information are used to obtain the second pixel feature information.
  • FIG. 5a and FIG. 5b show the third gray-scale projection signal and the fourth gray-scale projection signal obtained by gray-scale projection of the second detection image along the horizontal direction and the vertical direction respectively provided by this embodiment.
  • Schematic diagram As shown in Figure 5a and Figure 5b, when obtaining the second pixel feature information, autocorrelation calculations can be performed on the third grayscale projection signal and the fourth grayscale projection signal respectively to obtain the second detection image.
  • the size information (including size information along the horizontal direction and the vertical direction) of the area of interest where the alignment mark to be measured is located in the second detection image can be obtained.
  • the second area of interest of the alignment mark to be measured in the second detection image can be The matrix image is intercepted, and then one or more of the overall pixel value, geometric feature information, and grayscale feature information of the second matrix image are obtained as the second pixel feature information.
  • the image pixel feature processing method can also be used to process the third gray-scale projection signal and the fourth gray-scale projection signal to obtain the to-be-measured signal in the second detection image. Position information and size information of the area of interest where the alignment mark is located.
  • the second matrix image may not be intercepted from the second detection image, and the alignment mark to be measured may be determined based on the second detection image and the acquired second detection image.
  • the second pixel feature information can be directly calculated from the position information and size information of the area of interest.
  • the second pixel feature statistical information and the second pixel feature information together constitute the second feature information of the second detection image. It can be understood that the pixel feature statistical information can characterize the overall pixel features of the second detection image, and the second pixel feature information can characterize the ability of the second detection image to accommodate the alignment to be tested. Label the pixel features of a smaller region of interest.
  • step S300 is executed to combine the first feature information and the second feature information item by item. Perform similarity comparison.
  • the second pixel feature statistical information and the first pixel feature statistical information are compared one by one for similarity to obtain a confidence level of similarity (used to evaluate the second pixel feature statistical information and the first pixel feature statistical information).
  • the authenticity of the similarity of the first pixel feature statistical information in the knowledge base is greater than the known alignment mark of several candidates corresponding to the first set value, and the first known alignment mark of the candidate
  • the first feature information of the detected image is called the candidate first feature information (including the candidate first pixel feature statistical information and the candidate first pixel feature information).
  • a phase correlation method based on fast Fourier transform is used to compare the similarity between the second pixel feature statistical information and the first pixel feature statistical information in the knowledge base one by one.
  • both the third grayscale projection signal and the fourth grayscale projection signal in the second pixel feature statistical information can be converted into the frequency domain, and then all the third grayscale projection signals in the first pixel feature statistical information can be converted into the frequency domain.
  • the first gray-scale projection signal and the second gray-scale projection signal are both converted to the frequency domain, and the similarity between the second pixel feature statistical information and each of the first pixel feature statistical information can be obtained based on the phase correlation method.
  • a template matching search method may also be used to compare the similarity between the second pixel feature statistical information and the first pixel feature statistical information one by one. Specifically, the similarity between the second pixel feature statistical information and each of the first pixel feature statistical information is calculated based on a similarity metric in a sliding window form.
  • the similarity measures include 1-norm measure, 2-norm measure, p-norm measure, infinite norm measure, cosine similarity measure, etc.
  • the similarity between the second pixel feature statistical information and the first pixel feature statistical information is obtained using the phase correlation method based on fast Fourier transform or the template matching search method, a first similarity will be generated signal and a second similarity signal, the first similarity signal is used to characterize the third grayscale projection signal of the second pixel feature statistical information and the first grayscale projection signal of all the first pixel feature statistical information.
  • the confidence of the similarity, the second similarity signal is used to characterize the fourth grayscale projection signal of the second pixel feature statistical information and the second grayscale projection signal of all the first pixel feature statistical information. Confidence of similarity, based on the first similarity signal and the second similarity signal, the confidence of the similarity between the second pixel feature statistical information and each of the first pixel feature statistical information can be obtained .
  • the known alignment mark of the candidate is obtained based on the confidence of similarity, rather than simply using similarity to obtain the known alignment mark of the candidate.
  • the known alignment mark of the candidate is obtained.
  • the alignment marks are more likely to match the alignment marks to be measured, thereby improving detection accuracy.
  • Figures 6a and 6b are schematic diagrams of the first similarity signal and the second similarity signal provided in this embodiment.
  • the position of the peak is the position of the first pixel feature statistical information in the knowledge base, and the specific value of the peak can be converted into a measure of the second pixel feature statistical information and the corresponding The confidence level of the similarity of the first pixel feature statistical information.
  • the second pixel feature information is compared with the candidate first pixel feature information one by one to obtain a number of preferred known alignment marks corresponding to a similarity greater than the second set value.
  • the second pixel feature statistical information and the candidate first pixel feature statistical information can be subjected to cross-correlation calculations one by one (specifically It may be that the third gray-scale projection signal is cross-correlated with each of the candidate first gray-scale projection signals, and the fourth gray-scale projection signal is calculated with each of the candidate second gray-scale projection signals. Signal cross-correlation calculation) to obtain the best candidate known alignment mark from the known alignment mark candidates.
  • a best candidate known alignment mark can be selected from the candidate known alignment marks, and the second pixel feature information and the candidate first pixel feature information can be compared one by one. During comparison, only the second pixel feature information and the best candidate first pixel feature information may be compared for similarity, thereby improving comparison efficiency.
  • a pixel grayscale template alignment method and/or a geometric template alignment method may be used to perform similarity comparisons on the second pixel feature information and the candidate first pixel feature information one by one
  • the pixel grayscale template alignment method can include the standard optical flow method or the reverse sequence synthetic image alignment method.
  • the position information of the alignment mark to be measured on the second silicon wafer can be obtained based on several preferred known alignment marks obtained through comparison.
  • the first pixel feature information corresponding to the preferred known alignment mark can be used to traverse the second detection image to obtain the preferred first pixel feature information corresponding to the known alignment mark in the second detection image.
  • the most similar position of the first pixel feature information finally obtains the position information of the alignment mark to be measured on the silicon wafer.
  • all preferred known alignment marks may be The first pixel feature information corresponding to the known alignment mark traverses the second detection image. If there are at least two preferred known alignment marks, one of the most preferred known alignment marks can be selected from the preferred known alignment marks, and the most preferred known alignment mark can be used. The first pixel feature information corresponding to the alignment mark is traversed through the second detection image; alternatively, the first pixel feature information corresponding to the preferred known alignment mark can be averaged, and then the preferred known pair of The average value of the first pixel feature information corresponding to the quasi-mark traverses the second detection image.
  • the second characteristic information of the alignment mark to be measured can be stored in the knowledge base, thereby expanding the Describe the contents of the knowledge base.
  • this embodiment also provides a detection and positioning system for silicon wafer alignment marks.
  • Figure 7 is a structural block diagram of the detection and positioning system for silicon wafer alignment marks provided in this embodiment. As shown in Figure 7, the detection and positioning system for silicon wafer alignment marks includes:
  • Storage module 10 configured to store a knowledge base that contains first feature information of a first detection image of several known alignment marks, where the first feature information includes pixels of the first detection image Feature statistical information and pixel feature information of the area of interest where the known alignment mark is located in the first detection image;
  • Feature acquisition module 20 configured to provide a second detection image of the alignment mark to be measured, and obtain second feature information of the second detection image, where the second feature information includes pixels of the second detection image Feature statistical information and pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image; and,
  • the similarity comparison module 30 is used to compare the similarity between the first feature information and the second feature information one by one, and obtain the test target based on several known alignment marks obtained through comparison. Alignment mark position information on the silicon wafer.
  • This embodiment also provides an electronic device, including a processor and a memory. Instructions are stored in the memory. When the instructions are executed by the processor, the above steps of detecting and positioning the silicon wafer alignment marks are implemented.
  • the processor can perform various actions and processes according to instructions stored in the memory.
  • the processor may be an integrated circuit chip with signal processing capabilities.
  • the above-mentioned processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an on-site Programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA on-site Programmable gate array
  • Various methods, steps and logical block diagrams disclosed in the embodiments of the present invention can be implemented or executed.
  • the general-purpose processor can be a microprocessor or the processor can be any conventional processor, etc., and can be an X86 architecture or an ARM architecture, etc.
  • the memory stores executable instructions, which are executed by the processor for detecting and positioning the silicon wafer alignment mark as described above.
  • the memory may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
  • Non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
  • Volatile memory may be random access memory (RAM), which acts as an external cache.
  • RAM Direct Memory Bus Random Access Memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • SDRAM synchronous dynamic random access memory
  • DDRSDRAM double data rate synchronous dynamic Random Access Memory
  • ESDRAM Enhanced Synchronous Dynamic Random Access Memory
  • SLDRAM Synchronous Linked Dynamic Random Access Memory
  • DR RAM Direct Memory Bus Random Access Memory
  • a non-transitory computer-readable storage medium is proposed. Instructions are stored on the non-transitory computer-readable storage medium. When the instructions are executed, the silicon chip pair described above can be realized. The steps in the method of detecting and positioning quasi-marks.
  • non-transitory computer-readable storage media in embodiments of the present invention may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that computer-readable storage media described herein are intended to include, without limitation, these and any other suitable types of memory.
  • Embodiment 1 The difference from Embodiment 1 is that in this embodiment, the known alignment marks in the knowledge base are distributed in a clustered manner to facilitate similarity comparison.
  • the known alignment marks are divided into At least two categories, for example: using a clustering algorithm to cluster the first pixel feature statistical information, thereby classifying the known alignment marks, and all the known alignment marks belonging to the same category
  • the first feature information of the first detected image will be gathered together instead of being scattered in the knowledge base.
  • obtain the central pixel feature statistical information of all the first feature information in each category for example: perform an average calculation on all the first pixel feature statistical information in each category, thereby obtaining the center pixel feature statistical information of each category.
  • the central pixel feature statistical information of all the first pixel feature statistical information for example: perform an average calculation on all the first pixel feature statistical information in each category, thereby obtaining the center pixel feature statistical information of each category.
  • the comparison efficiency can be improved.
  • the second pixel feature statistical information and the central pixel feature statistical information of the candidate categories can be subjected to cross-correlation calculations one by one to obtain the best result from the candidate categories. Categories for best candidates. In this way, a best candidate category can be selected from the candidate categories.
  • comparing the second pixel feature information and the candidate first pixel feature information one by one only the second pixel can be compared. The feature information is compared for similarity with the first pixel feature information in the category of the best candidate, thereby improving comparison efficiency.
  • the pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image is compared with the location of the known alignment mark in the first detection image of the candidate known alignment mark.
  • the pixel feature information of the area of interest is compared one by one for similarity, and a number of preferred known alignment marks corresponding to a similarity greater than the second set value are obtained; and, all the known alignment marks are obtained based on a number of preferred known alignment marks. Describe the position information of the alignment mark to be measured on the silicon wafer. For specific details, please refer to
  • Embodiment 1 and Embodiment 2 will not be described again in detail.
  • a knowledge base is first established, and the knowledge base contains the first characteristic information of the first detection images of several known alignment marks, Then obtain the second feature information of the second detection image of the alignment mark to be measured, compare the similarity between the first feature information and the second feature information one by one, and based on the comparison, several of the The alignment mark is known to obtain the position information of the alignment mark to be measured on the silicon wafer.
  • the first feature information includes pixel feature statistical information of the first detection image and pixel feature information of the area of interest where the known alignment mark is located in the first detection image
  • the second feature information including pixel feature statistical information of the second detection image and pixel feature information of the area of interest where the alignment mark to be measured is located in the second detection image, after comparing the first feature information and the second feature information
  • the similarity between the first detection image and the second detection image is actually compared as a whole to filter out the number of known alignment marks from a larger number.
  • the present invention also provides a detection and positioning system for silicon wafer alignment marks, electronic equipment and a non-transitory computer-readable storage medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)
  • Container, Conveyance, Adherence, Positioning, Of Wafer (AREA)
  • Testing Or Measuring Of Semiconductors Or The Like (AREA)

Abstract

本发明提供了一种硅片对准标记的检测定位方法，先建立知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，然后获取待测对准标记的第二检测图像的第二特征信息，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息，从而精确地找到所述已知对准标记，提高了待测对准标记的检测定位精度及效率。相应的，本发明还提供了一种硅片对准标记的检测定位系统、电子设备及非暂态计算机可读存储介质。

Description

硅片对准标记的检测定位方法、系统、电子设备及介质 技术领域
本发明涉及半导体制造技术领域，尤其涉及一种硅片对准标记的检测定位方法、系统、电子设备及介质。
背景技术
目前，在ADI(after develop inspection，显影后检测)工艺或CMP(chemical mechanical polishing，化学机械抛光)工艺后的检测对准过程中，均需要获取硅片上的对准标记的位置。
硅片中的对准标记的反射率较低，易受图案反射或散射信号的干扰，现有的硅片对准标记的检测定位方法错误率较高，会影响对准精度，且检测效率也较低。
发明内容
本发明的目的在于提供一种硅片对准标记的检测定位方法、系统、电子设备及介质，以解决现有的硅片对准标记的检测定位方法错误率高和/或效率低等问题。
为了达到上述目的，本发明提供了一种硅片对准标记的检测定位方法，包括：
提供一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
提供待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息；以及，
将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信 息。
可选的，所述第一检测图像与所述第二检测图像的尺寸相同。
可选的，对具有所述已知对准标记的硅片进行拍照，并将拍摄的图像作为所述第一检测图像；或者，对具有所述已知对准标记的硅片进行拍照，并截取拍摄的图像中具有所述已知对准标记的部分作为所述第一检测图像。
可选的，采用人工勾画的方式截取拍摄的图像中具有所述已知对准标记的部分作为所述第一检测图像。
可选的，对具有所述待测对准标记的硅片进行拍照，并将拍摄的图像作为所述第二检测图像；或者，对具有所述待测对准标记的硅片进行拍照，并截取拍摄的图像中具有所述待测对准标记的部分作为所述第二检测图像。
可选的，利用预定的位置信息和尺寸信息截取拍摄的图像中具有所述待测对准标记的部分作为所述第二检测图像；或者，利用图像像素特征处理法截取拍摄的图像中具有所述待测对准标记的部分作为所述第二检测图像。
可选的，所述图像像素特征处理法包括图像分割定位法或图像对称性计算法。
可选的，获取所述第一特征信息的步骤包括：
获取所述第一检测图像的像素特征统计信息；
基于所述第一检测图像的像素特征统计信息获取所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息；以及，
基于所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
和/或，获取所述第二特征信息的步骤包括：
获取所述第二检测图像的像素特征统计信息；
基于所述第二检测图像的像素特征统计信息获取所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息；以及，
基于所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第二检测图像中所述待测对准标记所在感兴趣区域的像 素特征信息。
可选的，所述像素特征统计信息用于表征至少两个方向上的灰度投影信号，采用图像灰度投影法获取所述第一检测图像和/或所述第二检测图像的像素特征统计信息。
可选的，对所述灰度投影信号进行自相关计算，以得到所述感兴趣区域的位置信息，基于所述灰度投影信号的波峰和波谷的位置得到所述感兴趣区域的尺寸信息；或者，利用图像像素特征处理法对所述灰度投影信号进行处理，以得到所述感兴趣区域的位置信息和尺寸信息。
可选的，所述图像像素特征处理法包括图像分割定位法或图像对称性计算法。
可选的，获取所述第一检测图像的像素特征统计信息之前，对所述第一检测图像的每个像素进行邻域像素特征计算；和/或，获取所述第二检测图像的像素特征统计信息之前，对所述第二检测图像的每个像素进行邻域像素特征计算。
可选的，所述邻域像素特征计算包括邻域内像素均值计算、像素加权均值计算、像素梯度计算或像素极值范围计算。
可选的，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息的步骤包括：
将所述第二检测图像的像素特征统计信息与所述已知对准标记的第一检测图像的像素特征统计信息逐一进行相似性比对，得到相似性的置信度大于第一设定值对应的若干候选的所述已知对准标记；
将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记；以及，
基于若干优选的所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
可选的，所述知识库中基于所述第一检测图像的像素特征统计信息将所述已知对准标记分为至少两个类别，并得到每个类别中的所有所述已知对准标记的所述第一检测图像的中心像素特征统计信息；
将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息的步骤包括：
将所述第二检测图像的像素特征统计信息与每个类别的所述中心像素特征统计信息进行相似性比对，得到相似性的置信度大于第三设定值的若干候选的类别，将候选的类别中的所有所述已知对准标记作为候选的所述已知对准标记；
将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记；以及，
基于若干优选的所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
可选的，利用基于快速傅里叶变换的相位相关法或模板匹配搜索法将所述第二检测图像的像素特征统计信息与所述已知对准标记的第一检测图像的像素特征统计信息逐一进行相似性比对。
可选的，获取若干候选的所述已知对准标记之后，将所述第二检测图像的像素特征统计信息与候选的所述已知对准标记的第一检测图像的像素特征统计信息逐一进行互相关计算，以从候选的所述已知对准标记中得到最佳候选的所述已知对准标记；以及，
将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与最佳候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息进行相似性比对，得到优选的所述已知对准标记。
可选的，获取若干候选的类别之后，将所述第二检测图像的像素特征统计信息与候选的类别的所述中心像素特征统计信息逐一进行互相关计算，以 从候选的类别中得到最佳候选的类别；以及，
将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与最佳候选的类别中的所有所述已知对准标记所在感兴趣区域的像素特征信息进行相似性比对，得到优选的所述已知对准标记。
可选的，所述像素特征信息包括所述感兴趣区域的整体像素值、几何特征信息及灰度特征信息中的一种或多种，所述几何特征信息包括所述感兴趣区域的梯度信息和/或边缘信息。
可选的，利用像素灰度模板对准法和/或几何模板对准法将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对。
可选的，所述像素灰度模板对准法包括标准光流法或反序合成图像对准算法。
可选的，根据比对得到若干所述已知对准标记之后，利用比对得到的所述已知对准标记对应的像素特征信息遍历所述第二检测图像，得到所述第二检测图像中与比对得到的所述已知对准标记对应的像素特征信息最相似的位置，从而得到所述待测对准标记在硅片上的位置信息。
可选的，根据比对得到若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息之后，将所述待测对准标记的所述第二特征信息存储至所述知识库中。
本发明还提供了一种硅片对准标记的检测定位系统，包括：
存储模块，用于存储一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
特征获取模块，用于提供待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的 像素特征信息；以及，
相似性比对模块，用于将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
本发明还提供了一种电子设备，包括处理器和存储器，所述存储器上存储有指令，当所述指令被所述处理器执行时，实现所述的硅片对准标记的检测定位方法。
本发明还提供了一种非暂态计算机可读存储介质，所述非暂态计算机可读存储介质上存储有指令，当所述指令被执行时，实现所述硅片对准标记的检测定位方法。
在本发明提供的硅片对准标记的检测定位方法中，先建立知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，然后获取待测对准标记的第二检测图像的第二特征信息，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。由于所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息，所述第二特征信息中包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息，在比对所述第一特征信息和第二特征信息的像素特征统计信息的相似性时，实际上是整体比对所述第一检测图像和所述第二检测图像的相似性，以从数量较多的所述已知对准标记中筛选出数量较少的所述已知对准标记，缩小下一次比对的范围，然后比对所述第一特征信息和第二特征信息的像素特征信息的相似性时，实际上是精确比对第一检测图像中所述已知对准标记所在感兴趣区域与所述第二检测图像中所述待测对准标记所在感兴趣区域的相似性，从而精确地找到所述已知对准标记，提高了待测对准标记的检测定位精度及效率。相应的，本发明还提供了一种硅片对准标记的检测定位系统、电子设备及非暂态计算机可读存储介质。
附图说明
图1为本发明实施例一提供的硅片对准标记的检测定位方法的流程图；
图2为本发明实施例一提供的从所述第一原始图像中截取出所述第一检测图像的示意图；
图3a及图3b为本发明实施例一提供的沿着水平方向和垂直方向分别对所述第一检测图像进行灰度投影得到的第一灰度投影信号和第二灰度投影信号的示意图；
图4为本发明实施例一提供的从所述第一检测图像中截取出所述第一矩阵图像的示意图；
图5a及图5b为本发明实施例一提供的沿着水平方向和垂直方向分别对所述第二检测图像进行灰度投影得到的第三灰度投影信号和第四灰度投影信号的示意图；
图6a及图6b为本发明实施例一提供的第一相似性信号和第二相似性信号的示意图；
图7为本发明实施例一提供的硅片对准标记的检测定位系统的结构框图；
其中，附图标记为：
10-存储模块；20-特征获取模块；30-相似性比对模块。
具体实施方式
下面将结合示意图对本发明的具体实施方式进行更详细的描述。根据下列描述，本发明的优点和特征将更清楚。需说明的是，附图均采用非常简化的形式且均使用非精准的比例，仅用以方便、明晰地辅助说明本发明实施例的目的。
实施例一
图1为本实施例提供的硅片对准标记的检测定位方法的流程图。如图1所示，所述硅片对准标记的检测定位方法包括：
步骤S100：提供一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像的像 素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
步骤S200：提供所述待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息；以及，
步骤S300：将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
首先执行步骤S100，构建所述知识库。
具体而言，提供具有所述已知对准标记的硅片(下称具有所述已知对准标记的硅片为第一硅片)，所述已知对准标记在所述第一硅片上的位置信息是确定且已知的。利用一第一图像采集单元对所述第一硅片上的所述已知对准标记进行拍照，得到第一原始图像，所述第一原始图像中需要具有完整的所述已知对准标记，也即，所述已知对准标记需要在所述第一图像采集单元的成像视场内。
应理解，若所述第一硅片上具有多个所述已知对准标记时，所述第一图像采集单元可以逐个对所述第一硅片上的所述已知对准标记进行拍照，从而得到多个所述第一原始图像；当然，若所述第一图像采集单元同时对所述第一硅片上的多个所述已知对准标记进行了拍照，可以通过人工对拍摄的图像进行处理得到单个所述已知对准标记的第一原始图像。也即，每个所述第一原始图像中仅具有一个所述已知对准标记。
此外，需要说明的是，在本文中，术语“硅片”大体上指由半导体或非半导体材料形成的衬底。此半导体或非半导体材料的实例包含但不限于单晶硅、氮化镓、砷化镓、磷化铟、蓝宝石及玻璃。此类衬底可普遍在半导体制作设施中找到及/或处理。
由于所述第一图像采集单元采集的所述第一原始图像的尺寸较大(所述已知对准标记在所述第一原始图像中所占的区域较小)，本实施例中，截取所 述第一原始图像中具有所述已知对准标记的部分作为所述第一检测图像，如此一来，所述第一检测图像的尺寸较小，像素点较少，可以减少后续图像处理的难度和数据量，增加图像处理的速度。举例而言，可以采用人工勾画的方式截取所述第一检测图像中具有所述已知对准标记的部分作为所述第一检测图像，从而能够保证所述第一检测图像中具有所述已知对准标记，且人工勾画可以保证截取出的所述第一检测图像的尺寸适当。
图2为本实施例提供的从所述第一原始图像中截取出所述第一检测图像的示意图。如图2所示，所述第一原始图像100的尺寸较大，截取出的所述第一检测图像101的尺寸较小，但所述第一检测图像101中具有完整的所述已知对准标记200，本实施例中，所述第一检测图像101的形状为矩形，从而便于后续的图像处理；作为可选实施例，所述第一检测图像101的形状不限于是矩形，还可以是其他形状，如十字形等。
当然，作为可选实施例，也可以不对所述第一原始图像进行截取操作，此时，直接将所述第一原始图像作为所述第一检测图像进行后续步骤。
接下来，获取所述第一检测图像的像素特征统计信息(下称所述第一检测图像的像素特征统计信息为第一像素特征统计信息)。所述第一像素特征统计信息用于表征所述第一检测图像的至少两个方向上的灰度投影信号，采用图像灰度投影法获取所述第一像素特征统计信息。
本实施例中，所述第一像素特征统计信息用于表征所述第一检测图像的水平方向和垂直方向上的灰度投影信号，采用图像灰度投影法获取所述第一像素特征统计信息时，沿着水平方向和垂直方向分别对所述第一检测图像进行灰度投影，分别得到沿着水平方向的第一灰度投影信号和沿垂直方向的第二灰度投影信号。
作为可选实施例，获取所述第一像素特征统计信息之前，可以先对所述第一检测图像的每个像素进行邻域像素特征计算，从而去除所述第一检测图像中的噪声。所述邻域像素特征计算包括邻域内像素均值计算、像素加权均值计算、像素梯度计算或像素极值范围计算等。
接下来，获取所述第一检测图像中所述已知对准标记所在感兴趣区域的 像素特征信息(下称所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息为第一像素特征信息)。所述第一像素特征信息可以包括所述第一检测图像中所述已知对准标记所在感兴趣区域的整体像素值、几何特征信息及灰度特征信息中的一种或多种，所述几何特征信息包括所述感兴趣区域的梯度信息和/或边缘信息。在本申请实施例中，所述感兴趣区域为矩形区域，相应的，所述第一像素特征信息可以包括所述第一检测图像中所述已知对准标记所在矩形区域的整体像素值、几何特征信息及灰度特征信息中的一种或多种，所述几何特征信息包括所述矩形区域的梯度信息和/或边缘信息。
具体而言，获取所述第一像素特征信息时，可以先基于所述第一像素特征统计信息获取所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息，然后基于所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第一像素特征信息。
举例而言，图3a及图3b为本实施例提供的沿着水平方向和垂直方向分别对所述第一检测图像进行灰度投影得到的第一灰度投影信号和第二灰度投影信号的示意图。图3a及图3b所示，获取所述第一像素特征信息时，可以分别对所述第一灰度投影信号和所述第二灰度投影信号进行自相关计算，得到所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息(包括沿水平方向和垂直方向的位置信息)，基于所述第一灰度投影信号和所述第二灰度投影信号的波峰和波谷的位置可以得到所述第一检测图像中所述已知对准标记所在感兴趣区域的尺寸信息(包括沿水平方向和垂直方向的尺寸信息)。获取所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息之后，可以将所述第一检测图像中所述已知对准标记所在感兴趣区域的第一矩阵图像截取下来，然后获取所述第一矩阵图像的整体像素值、几何特征信息及灰度特征信息中的一种或多种作为所述第一像素特征信息。
作为可选实施例，还可以利用所述图像像素特征处理法对所述第一灰度投影信号和所述第二灰度投影信号进行处理，以得到所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息。
图4为本实施例提供的从所述第一检测图像中截取出所述第一矩阵图像 的示意图。如图4所示，所述第一检测图像101的尺寸较大，截取出的所述第一矩阵图像102的尺寸较小，但所述第一矩阵图像102中也具有完整的所述已知对准标记200。本实施例中，所述感兴趣区域为矩形，但不应以此为限，所述感兴趣区域也可以是其他形状，如十字形等。
作为可选实施例，也可以不从所述第一检测图像中截取出所述第一矩阵图像，基于所述第一检测图像及获取的所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息可以直接计算出所述第一像素特征信息。
进一步地，所述第一像素特征统计信息及所述第一像素特征信息共同构成所述第一检测图像的第一特征信息。可以理解的是，所述像素特征统计信息可以表征所述第一检测图像整体的像素特征，而所述第一像素特征信息则可以表征所述第一检测图像中能够容纳所述已知对准标记的一个较小的感兴趣区域的像素特征。
应理解，对于其他种类的所述已知对准标记，可以重复如上步骤，得到若干种所述已知对准标记的所述第一特征信息，然后基于若干种所述已知对准标记的所述第一特征信息按照标准模板生成格式化的数据，从而形成所述知识库。所述知识库可以以格式化的文件和/或数据库表单数据的形式保存在计算机内存介质或者非易失性存储介质中。所述格式化的文件包括但不限于格式化的二进制文件、格式化文本文件、JSON格式文件、XML格式文件、YAML格式文件或CSV格式文件等。
应理解，本实施例中所指的若干种所述已知对准标记，是指在所述第一硅片上位置不同的所述已知对准标记，例如：在同一个所述第一硅片上的不同位置上的所述已知对准标记，应该被认为是不同种类的所述已知对准标记；不同的所述第一硅片上的所述已知对准标记，即使位置是相对应的(同一批所述第一硅片按照同样的工艺先后制造完成)，由于具有工艺偏差，也应该被认为是不同种类的所述已知对准标记。
接下来，执行步骤S200，获取所述待测对准标记的第二检测图像的第二特征信息。
具体而言，提供具有所述待测对准标记的硅片(下称具有所述待测对准标记的硅片为第二硅片)，所述待测对准标记在所述第二硅片上的位置信息是未知的，需要进行检测。利用一第二图像采集单元对所述第二硅片上的所述待测对准标记进行拍照，得到第二原始图像，所述第二原始图像中需要具有完整的所述待测对准标记，也即，所述待测对准标记需要在所述第二图像采集单元的成像视场内。
应理解，若所述第二硅片上具有多个所述待测对准标记时，所述第二图像采集单元可以逐个对所述第二硅片上的所述待测对准标记进行拍照，从而得到多个所述第二原始图像。也即，每个所述第二原始图像中仅具有一个所述待测对准标记。
由于所述第二图像采集单元采集的所述第二原始图像的尺寸较大(所述待测对准标记在所述第二原始图像中所占的区域较小)，本实施例中，截取所述第二原始图像中具有所述待测对准标记的部分作为所述第二检测图像，如此一来，所述第二检测图像的尺寸较小，像素点较少，可以减少后续图像处理的难度和数据量，可以增加图像处理的速度。举例而言，可以利用预定的位置信息和尺寸信息从所述第二原始图像中截取具有所述待测对准标记的部分作为所述第二检测图像；或者，利用图像像素特征处理法截取所述第二原始图像中具有所述待测对准标记的部分作为所述第二检测图像，所述图像像素特征处理法可以是图像分割定位法或图像对称性计算法等。
当然，作为可选实施例，也可以不对所述第二原始图像进行截取操作，此时，直接将所述第二原始图像作为所述第二检测图像进行后续步骤。
本实施例中，所述第一检测图像与所述第二检测图像的尺寸相同，以利于后续的相似性比对步骤。因此，若利用预定的位置信息和尺寸信息从所述第二原始图像中截取具有所述待测对准标记的部分作为所述第二检测图像，可以预先获取所述待测对准标记在所述第二硅片上的大致位置，同时利用与所述第一检测图像尺寸相同的尺寸从所述第二原始图像中截取出所述第二检测图像；而利用图像像素特征处理法截取所述第二原始图像中具有所述待测对准标记的部分作为所述第二检测图像时，也需要预先在所述图像像素特征 处理法中按照所述第一检测图像的尺寸预置好需要截取的图像的尺寸。
接下来，获取所述第二检测图像的像素特征统计信息(下称所述第二检测图像的像素特征统计信息为第二像素特征统计信息)。所述第二像素特征统计信息用于表征所述第二检测图像的至少两个方向上的灰度投影信号，采用图像灰度投影法获取所述第二像素特征统计信息。
本实施例中，所述第二像素特征统计信息用于表征所述第二检测图像的水平方向和垂直方向上的灰度投影信号，采用图像灰度投影法获取所述第二像素特征统计信息时，沿着水平方向和垂直方向分别对所述第二检测图像进行灰度投影，分别得到沿着水平方向的第三灰度投影信号和沿垂直方向的第四灰度投影信号。
作为可选实施例，获取所述第二像素特征统计信息之前，可以先对所述第二检测图像的每个像素进行邻域像素特征计算，从而去除所述第二检测图像中的噪声。所述邻域像素特征计算包括邻域内像素均值计算、像素加权均值计算、像素梯度计算或像素极值范围计算等。
接下来，获取所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息(下称所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息为第二像素特征信息)。所述第二像素特征信息可以包括所述第二检测图像中所述待测对准标记所在矩形区域的整体像素值、几何特征信息及灰度特征信息中的一种或多种，所述几何特征信息包括所述矩形区域的梯度信息和/或边缘信息。
应理解，若所述第一像素特征信息仅包含一种特征，此时获取的所述第二像素特征信息应包含相应的特征信息，如所述第一像素特征信息仅包含几何特征信息，所述第二像素特征信息也应包含几何特征信息；若所述第一像素特征信息包含多种特征，此时获取的所述第二像素特征信息可以包含相应的特征信息的一种或多种，如所述第一像素特征信息包含几何特征信息及灰度特征信息，所述第二像素特征信息可以包含几何特征信息和/或灰度特征信息。
具体而言，获取所述第二像素特征信息时，可以先基于所述第二像素特 征统计信息获取所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息，然后基于所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第二像素特征信息。
举例而言，图5a及图5b为本实施例提供的沿着水平方向和垂直方向分别对所述第二检测图像进行灰度投影得到的第三灰度投影信号和第四灰度投影信号的示意图。图5a及图5b所示，获取所述第二像素特征信息时，可以分别对所述第三灰度投影信号和所述第四灰度投影信号进行自相关计算，得到所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息(包括沿水平方向和垂直方向的位置信息)，基于所述第三灰度投影信号和所述第四灰度投影信号的波峰和波谷的位置可以得到所述第二检测图像中所述待测对准标记所在感兴趣区域的尺寸信息(包括沿水平方向和垂直方向的尺寸信息)。获取所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息之后，可以将所述第二检测图像中所述待测对准标记所在感兴趣区域的第二矩阵图像截取下来，然后获取所述第二矩阵图像的整体像素值、几何特征信息及灰度特征信息中的一种或多种作为所述第二像素特征信息。
作为可选实施例，还可以利用所述图像像素特征处理法对所述第三灰度投影信号和所述第四灰度投影信号进行处理，以得到所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息。
作为可选实施例，也可以不从所述第二检测图像中截取出所述第二矩阵图像，基于所述第二检测图像及获取的所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息可以直接计算出所述第二像素特征信息。
进一步地，所述第二像素特征统计信息及所述第二像素特征信息共同构成所述第二检测图像的第二特征信息。可以理解的是，所述像素特征统计信息可以表征所述第二检测图像整体的像素特征，而所述第二像素特征信息则可以表征所述第二检测图像中能够容纳所述待测对准标记的一个较小的感兴趣区域的像素特征。
接下来，执行步骤S300，将所述第一特征信息与所述第二特征信息逐项 进行相似性比对。
具体而言，将所述第二像素特征统计信息与所述第一像素特征统计信息逐一进行相似性比对，得到相似性的置信度(用于评价所述第二像素特征统计信息与所述知识库中的所述第一像素特征统计信息的相似性的真实性)大于第一设定值对应的若干候选的所述已知对准标记，候选的所述已知对准标记的第一检测图像的第一特征信息称为候选的第一特征信息(包括候选的所述第一像素特征统计信息及候选的所述第一像素特征信息)。
作为可选实施例，利用基于快速傅里叶变换的相位相关法将所述第二像素特征统计信息与所述知识库中的所述第一像素特征统计信息逐一进行相似性比对。具体而言，可以先将所述第二像素特征统计信息中的第三灰度投影信号和第四灰度投影信号均转换到频域，再将所有所述第一像素特征统计信息中的第一灰度投影信号和第二灰度投影信号均转换到频域，基于相位相关法即可获取所述第二像素特征统计信息和每个所述第一像素特征统计信息的相似性。
作为可选实施例，还可以采用模板匹配搜索法将所述第二像素特征统计信息与所述第一像素特征统计信息逐一进行相似性比对。具体而言，以滑窗形式在所述第二像素特征统计信息上基于一相似性度量计算其与每个所述第一像素特征统计信息的相似性。所述相似性度量包括1范数度量、2范数度量、p范数度量、无穷范数度量、余弦相似性度量等。
进一步地，以基于快速傅里叶变换的相位相关法或所述模板匹配搜索法获取所述第二像素特征统计信息和所述第一像素特征统计信息的相似性时，会产生第一相似性信号和第二相似性信号，所述第一相似性信号用于表征所述第二像素特征统计信息的第三灰度投影信号与所有所述第一像素特征统计信息的第一灰度投影信号的相似性的置信度，所述第二相似性信号用于表征所述第二像素特征统计信息的第四灰度投影信号与所有所述第一像素特征统计信息的第二灰度投影信号的相似性的置信度，基于所述第一相似性信号和所述第二相似性信号即可得到所述第二像素特征统计信息与每个所述第一像素特征统计信息的相似性的置信度。
本实施例中，基于相似性的置信度来获取候选的所述已知对准标记，而并非简单的采用相似性来获取候选的所述已知对准标记，得到的候选的所述已知对准标记中与所述待测对准标记匹配的可能性更大，提高了检测精度。
图6a及图6b为本实施例提供的第一相似性信号和第二相似性信号的示意图。在图6a及图6b中，峰值所在的位置即为所述第一像素特征统计信息在所述知识库中的位置，峰值的具体数值即可转换为衡量所述第二像素特征统计信息和相应的所述第一像素特征统计信息的相似性的置信度。
接下来，将所述第二像素特征信息与候选的所述第一像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记。
进一步地，作为可选实施例，获取若干候选的所述已知对准标记之后，可以将所述第二像素特征统计信息与候选的所述第一像素特征统计信息逐一进行互相关计算(具体可以是将所述第三灰度投影信号与候选的每个所述第一灰度投影信号进行互相关计算，将所述第四灰度投影信号与候选的每个所述第二灰度投影信号进行互相关计算)，以从候选的所述已知对准标记中得到最佳候选的所述已知对准标记。如此可以从候选的所述已知对准标记中筛选出一个最佳候选的所述已知对准标记，将所述第二像素特征信息与候选的所述第一像素特征信息逐一进行相似性比对时，可以只将所述第二像素特征信息与最佳候选的所述第一像素特征信息进行相似性比对，从而提高比对效率。
作为可选实施例，可以利用像素灰度模板对准法和/或几何模板对准法将所述第二像素特征信息与候选的所述第一像素特征信息逐一进行相似性比对，所述像素灰度模板对准法可以包括标准光流法或反序合成图像对准法等。
进一步地，根据比对得到的若干优选的所述已知对准标记即可获取所述待测对准标记在所述第二硅片上的位置信息。具体而言，可以利用优选的所述已知对准标记对应的第一像素特征信息遍历所述第二检测图像，得到所述第二检测图像中与优选的所述已知对准标记对应的第一像素特征信息最相似的位置，最终得到所述待测对准标记在硅片上的位置信息。
可以理解的是，若优选的所述已知对准标记仅为一个，可以将优选的所 述已知对准标记对应的第一像素特征信息遍历所述第二检测图像。若优选的所述已知对准标记为至少两个时，可以从优选的所述已知对准标记中选取一个最优选的所述已知对准标记，利用最优选的所述已知对准标记对应的第一像素特征信息遍历所述第二检测图像；或者，可以将优选的所述已知对准标记对应的第一像素特征信息取平均值，然后利用优选的所述已知对准标记对应的第一像素特征信息的平均值遍历所述第二检测图像。
作为可选实施例，获取所述待测对准标记在硅片上的位置信息之后，可以将所述待测对准标记的所述第二特征信息存储至所述知识库中，从而扩充所述知识库的内容。
基于此，本实施例还提供了一种硅片对准标记的检测定位系统。图7为本实施例提供的硅片对准标记的检测定位系统的结构框图，如图7所示，所述硅片对准标记的检测定位系统包括：
存储模块10，用于存储一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
特征获取模块20，用于提供所述待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息；以及，
相似性比对模块30，用于将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
本实施例还提供了一种电子设备，包括处理器和存储器，存储器上存储有指令，当指令被处理器执行时，实现上述硅片对准标记的检测定位的步骤。
其中，处理器可以根据存储在存储器中的指令执行各种动作和处理。具体地，处理器可以是一种集成电路芯片，具有信号的处理能力。上述处理器可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现场 可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明实施例中公开的各种方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等，可以是X86架构或者是ARM架构等。
存储器存储有可执行指令，该指令在被处理器执行上文所述的硅片对准标记的检测定位方法。存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。非易失性存储器可以是只读存储器(ROM)、可编程只读存储器(PROM)、可擦除可编程只读存储器(EPROM)、电可擦除可编程只读存储器(EEPROM)或闪存。易失性存储器可以是随机存取存储器(RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、同步动态随机存取存储器(SDRAM)、双倍数据速率同步动态随机存取存储器(DDRSDRAM)、增强型同步动态随机存取存储器(ESDRAM)、同步连接动态随机存取存储器(SLDRAM)和直接内存总线随机存取存储器(DR RAM)。应注意，本文描述的方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
根据本发明的另一个方面，提出了一种非暂态计算机可读存储介质，非暂态计算机可读存储介质上存储有指令，当指令被执行时，可以实现上文所描述的硅片对准标记的检测定位方法中的步骤。
类似地，本发明实施例中的非暂态计算机可读存储介质可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。应注意，本文描述的计算机可读存储介质旨在包括但不限于这些和任意其它适合类型的存储器。
实施例二
与实施例一的区别在于，本实施例中，所述知识库中的所述已知对准标记以聚类的方式分布，以便于相似性比对。
具体而言，基于所述第一像素特征统计信息将所述已知对准标记分为至 少两个类别，例如：利用聚类算法对所述第一像素特征统计信息进行聚类，以此将所述已知对准标记进行分类，属于同一类别的所述已知对准标记的所述第一检测图像的所述第一特征信息就会聚集在一起，而不是在所述知识库中散乱分布。
接下来，获取每个类别中的所有所述第一特征信息的中心像素特征统计信息，例如：将每个类别中的所有所述第一像素特征统计信息进行均值计算，从而得到每个类别中的所有所述第一像素特征统计信息的中心像素特征统计信息。
将所述第二像素特征统计信息与所述第一像素特征统计信息逐一进行相似性比对时，先将所述第二像素特征统计信息与每个类别的所述中心像素特征统计信息进行相似性比对，得到相似性的置信度大于第三设定值的若干候选的类别，将候选的类别中的所有所述已知对准标记作为候选的所述已知对准标记。
本实施例中，由于不需要将所述第二像素特征统计信息与所述第一像素特征统计信息逐一进行相似性比对，从而可以提高比对效率。
进一步地，作为可选实施例，获取若干候选的类别之后，可以将所述第二像素特征统计信息与候选的类别的中心像素特征统计信息逐一进行互相关计算，以从候选的类别中得到最佳候选的类别。如此可以从候选的类别中筛选出一个最佳候选的类别，将所述第二像素特征信息与候选的所述第一像素特征信息逐一进行相似性比对时，可以只将所述第二像素特征信息与最佳候选的类别中的所述第一像素特征信息进行相似性比对，从而提高比对效率。
接着，将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记；以及，基于若干优选的所述已知对准标记获取所述待测对准标记在硅片上的位置信息。具体的，可相应参考
实施例一，本实施例二对此不再赘述。
综上，在本发明实施例提供的硅片对准标记的检测定位方法中，先建立知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，然后获取待测对准标记的第二检测图像的第二特征信息，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。由于所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息，所述第二特征信息中包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息，在比对所述第一特征信息和第二特征信息的像素特征统计信息的相似性时，实际上是整体比对所述第一检测图像和所述第二检测图像的相似性，以从数量较多的所述已知对准标记中筛选出数量较少的所述已知对准标记，缩小下一次比对的范围，然后比对所述第一特征信息和第二特征信息的像素特征信息的相似性时，实际上是精确比对第一检测图像中所述已知对准标记所在感兴趣区域与所述第二检测图像中所述待测对准标记所在感兴趣区域的相似性，从而精确地找到所述已知对准标记，提高了待测对准标记的检测定位精度及效率。相应的，本发明还提供了一种硅片对准标记的检测定位系统、电子设备及非暂态计算机可读存储介质。
需要说明的是，本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似部分互相参见即可。对于实施例公开的系统而言，由于与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。
还需要说明的是，虽然本发明已以较佳实施例披露如上，然而上述实施例并非用以限定本发明。对于任何熟悉本领域的技术人员而言，在不脱离本发明技术方案范围情况下，都可利用上述揭示的技术内容对本发明技术方案作出许多可能的变动和修饰，或修改为等同变化的等效实施例。因此，凡是未脱离本发明技术方案的内容，依据本发明的技术实质对以上实施例所做的任何简单修改、等同变化及修饰，均仍属于本发明技术方案保护的范围。
还应当理解的是，除非特别说明或者指出，否则说明书中的术语“第一”、“第二”、“第三”等描述仅仅用于区分说明书中的各个组件、元素、步骤等，而不是用于表示各个组件、元素、步骤之间的逻辑关系或者顺序关系等。
此外还应该认识到，此处描述的术语仅仅用来描述特定实施例，而不是用来限制本发明的范围。必须注意的是，此处的以及所附权利要求中使用的单数形式“一个”和“一种”包括复数基准，除非上下文明确表示相反意思。例如，对“一个步骤”或“一个装置”的引述意味着对一个或多个步骤或装置的引述，并且可能包括次级步骤以及次级装置。应该以最广义的含义来理解使用的所有连词。以及，词语“或”应该被理解为具有逻辑“或”的定义，而不是逻辑“异或”的定义，除非上下文明确表示相反意思。此外，本发明实施例中的方法和/或设备的实现可包括手动、自动或组合地执行所选任务。

Claims (26)

  1. 一种硅片对准标记的检测定位方法，其特征在于，包括：
    提供一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
    提供待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息；以及，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
  2. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，所述第一检测图像与所述第二检测图像的尺寸相同。
  3. 如权利要求1或2所述的硅片对准标记的检测定位方法，其特征在于，对具有所述已知对准标记的硅片进行拍照，并将拍摄的图像作为所述第一检测图像；或者，对具有所述已知对准标记的硅片进行拍照，并截取拍摄的图像中具有所述已知对准标记的部分作为所述第一检测图像。
  4. 如权利要求3所述的硅片对准标记的检测定位方法，其特征在于，采用人工勾画的方式截取拍摄的图像中具有所述已知对准标记的部分作为所述第一检测图像。
  5. 如权利要求1或2所述的硅片对准标记的检测定位方法，其特征在于，对具有所述待测对准标记的硅片进行拍照，并将拍摄的图像作为所述第二检测图像；或者，对具有所述待测对准标记的硅片进行拍照，并截取拍摄的图像中具有所述待测对准标记的部分作为所述第二检测图像。
  6. 如权利要求5所述的硅片对准标记的检测定位方法，其特征在于，利用预定的位置信息和尺寸信息截取拍摄的图像中具有所述待测对准标记的部 分作为所述第二检测图像；或者，利用图像像素特征处理法截取拍摄的图像中具有所述待测对准标记的部分作为所述第二检测图像。
  7. 如权利要求6所述的硅片对准标记的检测定位方法，其特征在于，所述图像像素特征处理法包括图像分割定位法或图像对称性计算法。
  8. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，获取所述第一特征信息的步骤包括：
    获取所述第一检测图像的像素特征统计信息；
    基于所述第一检测图像的像素特征统计信息获取所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息；以及，
    基于所述第一检测图像中所述已知对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
    和/或，获取所述第二特征信息的步骤包括：
    获取所述第二检测图像的像素特征统计信息；
    基于所述第二检测图像的像素特征统计信息获取所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息；以及，
    基于所述第二检测图像中所述待测对准标记所在感兴趣区域的位置信息和尺寸信息获取所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息。
  9. 如权利要求8所述的硅片对准标记的检测定位方法，其特征在于，所述像素特征统计信息用于表征至少两个方向上的灰度投影信号，采用图像灰度投影法获取所述第一检测图像和/或所述第二检测图像的像素特征统计信息。
  10. 如权利要求9所述的硅片对准标记的检测定位方法，其特征在于，对所述灰度投影信号进行自相关计算，以得到所述感兴趣区域的位置信息，基于所述灰度投影信号的波峰和波谷的位置得到所述感兴趣区域的尺寸信息；或者，利用图像像素特征处理法对所述灰度投影信号进行处理，以得到所述感兴趣区域的位置信息和尺寸信息。
  11. 如权利要求10所述的硅片对准标记的检测定位方法，其特征在于，所述图像像素特征处理法包括图像分割定位法或图像对称性计算法。
  12. 如权利要求9所述的硅片对准标记的检测定位方法，其特征在于，获取所述第一检测图像的像素特征统计信息之前，对所述第一检测图像的每个像素进行邻域像素特征计算；和/或，获取所述第二检测图像的像素特征统计信息之前，对所述第二检测图像的每个像素进行邻域像素特征计算。
  13. 如权利要求12所述的硅片对准标记的检测定位方法，其特征在于，所述邻域像素特征计算包括邻域内像素均值计算、像素加权均值计算、像素梯度计算或像素极值范围计算。
  14. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息的步骤包括：
    将所述第二检测图像的像素特征统计信息与所述已知对准标记的第一检测图像的像素特征统计信息逐一进行相似性比对，得到相似性的置信度大于第一设定值对应的若干候选的所述已知对准标记；
    将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记；以及，
    基于若干优选的所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
  15. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，所述知识库中基于所述第一检测图像的像素特征统计信息将所述已知对准标记分为至少两个类别，并得到每个类别中的所有所述已知对准标记的所述第一检测图像的中心像素特征统计信息；
    将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息的步骤包括：
    将所述第二检测图像的像素特征统计信息与每个类别的所述中心像素特征统计信息进行相似性比对，得到相似性的置信度大于第三设定值的若干候选的类别，将候选的类别中的所有所述已知对准标记作为候选的所述已知对准标记；
    将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对，得到相似性大于第二设定值对应的若干优选的所述已知对准标记；以及，
    基于若干优选的所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
  16. 如权利要求14或15所述的硅片对准标记的检测定位方法，其特征在于，利用基于快速傅里叶变换的相位相关法或模板匹配搜索法将所述第二检测图像的像素特征统计信息与所述已知对准标记的第一检测图像的像素特征统计信息逐一进行相似性比对。
  17. 如权利要求14所述的硅片对准标记的检测定位方法，其特征在于，获取若干候选的所述已知对准标记之后，将所述第二检测图像的像素特征统计信息与候选的所述已知对准标记的第一检测图像的像素特征统计信息逐一进行互相关计算，以从候选的所述已知对准标记中得到最佳候选的所述已知对准标记；以及，
    将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与最佳候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息进行相似性比对，得到优选的所述已知对准标记。
  18. 如权利要求15所述的硅片对准标记的检测定位方法，其特征在于，获取若干候选的类别之后，将所述第二检测图像的像素特征统计信息与候选 的类别的所述中心像素特征统计信息逐一进行互相关计算，以从候选的类别中得到最佳候选的类别；以及，
    将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与最佳候选的类别中的所有所述已知对准标记所在感兴趣区域的像素特征信息进行相似性比对，得到优选的所述已知对准标记。
  19. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，所述像素特征信息包括所述感兴趣区域的整体像素值、几何特征信息及灰度特征信息中的一种或多种，所述几何特征信息包括所述感兴趣区域的梯度信息和/或边缘信息。
  20. 如权利要求14或15所述的硅片对准标记的检测定位方法，其特征在于，利用像素灰度模板对准法和/或几何模板对准法将所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息与候选的所述已知对准标记的第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息逐一进行相似性比对。
  21. 如权利要求20所述的硅片对准标记的检测定位方法，其特征在于，所述像素灰度模板对准法包括标准光流法或反序合成图像对准算法。
  22. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，根据比对得到若干所述已知对准标记之后，利用比对得到的所述已知对准标记对应的像素特征信息遍历所述第二检测图像，得到所述第二检测图像中与比对得到的所述已知对准标记对应的像素特征信息最相似的位置，从而得到所述待测对准标记在硅片上的位置信息。
  23. 如权利要求1所述的硅片对准标记的检测定位方法，其特征在于，根据比对得到若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息之后，将所述待测对准标记的所述第二特征信息存储至所述知识库中。
  24. 一种硅片对准标记的检测定位系统，其特征在于，包括：
    存储模块，用于存储一知识库，所述知识库中具有若干种已知对准标记的第一检测图像的第一特征信息，所述第一特征信息包括所述第一检测图像 的像素特征统计信息及所述第一检测图像中所述已知对准标记所在感兴趣区域的像素特征信息；
    特征获取模块，用于提供待测对准标记的第二检测图像，并获取所述第二检测图像的第二特征信息，所述第二特征信息包括所述第二检测图像的像素特征统计信息及所述第二检测图像中所述待测对准标记所在感兴趣区域的像素特征信息；以及，
    相似性比对模块，用于将所述第一特征信息与所述第二特征信息逐项进行相似性比对，并根据比对得到的若干所述已知对准标记获取所述待测对准标记在硅片上的位置信息。
  25. 一种电子设备，其特征在于，包括处理器和存储器，所述存储器上存储有指令，当所述指令被所述处理器执行时，实现如权利要求1～23中任一项所述的硅片对准标记的检测定位方法。
  26. 一种非暂态计算机可读存储介质，其特征在于，所述非暂态计算机可读存储介质上存储有指令，当所述指令被执行时，实现如权利要求1～23中任一项所述的硅片对准标记的检测定位方法。
PCT/CN2023/084727 2022-03-31 2023-03-29 硅片对准标记的检测定位方法、系统、电子设备及介质 WO2023185932A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210346590.8A CN116935076A (zh) 2022-03-31 2022-03-31 硅片对准标记的检测定位方法、系统、电子设备及介质
CN202210346590.8 2022-03-31

Publications (1)

Publication Number Publication Date
WO2023185932A1 true WO2023185932A1 (zh) 2023-10-05

Family

ID=88199312

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/084727 WO2023185932A1 (zh) 2022-03-31 2023-03-29 硅片对准标记的检测定位方法、系统、电子设备及介质

Country Status (3)

Country Link
CN (1) CN116935076A (zh)
TW (1) TW202341404A (zh)
WO (1) WO2023185932A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117690846A (zh) * 2024-02-02 2024-03-12 深圳市双翌光电科技有限公司 太阳能硅片视觉检测方法、装置、设备及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002062112A (ja) * 2000-08-21 2002-02-28 Sony Corp 位置決め装置及び位置決め方法
JP2012114181A (ja) * 2010-11-24 2012-06-14 Panasonic Corp 部品実装装置及び部品実装装置における基板マーク認識方法
CN113838137A (zh) * 2021-11-24 2021-12-24 中导光电设备股份有限公司 一种基于tft-lcd获取校正标记位置的方法和系统
KR20220036803A (ko) * 2020-09-16 2022-03-23 주식회사 크래비스 인쇄회로기판 인식마크 검출 장치 및 방법

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002062112A (ja) * 2000-08-21 2002-02-28 Sony Corp 位置決め装置及び位置決め方法
JP2012114181A (ja) * 2010-11-24 2012-06-14 Panasonic Corp 部品実装装置及び部品実装装置における基板マーク認識方法
KR20220036803A (ko) * 2020-09-16 2022-03-23 주식회사 크래비스 인쇄회로기판 인식마크 검출 장치 및 방법
CN113838137A (zh) * 2021-11-24 2021-12-24 中导光电设备股份有限公司 一种基于tft-lcd获取校正标记位置的方法和系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117690846A (zh) * 2024-02-02 2024-03-12 深圳市双翌光电科技有限公司 太阳能硅片视觉检测方法、装置、设备及存储介质
CN117690846B (zh) * 2024-02-02 2024-04-09 深圳市双翌光电科技有限公司 太阳能硅片视觉检测方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN116935076A (zh) 2023-10-24
TW202341404A (zh) 2023-10-16

Similar Documents

Publication Publication Date Title
US9778206B2 (en) Defect inspection device and defect inspection method
US10867116B2 (en) Forecasting wafer defects using frequency domain analysis
TWI750553B (zh) 對樣本的缺陷檢測的方法及其系統
US11244444B2 (en) Method and apparatus for analyzing semiconductor wafer
Caldelli et al. Fast image clustering of unknown source images
WO2023185932A1 (zh) 硅片对准标记的检测定位方法、系统、电子设备及介质
KR102449376B1 (ko) 다이 내부 검사에 있어서의 등록 및 설계 주변부에서 야기된 노이즈의 저감
CN112424826A (zh) 基于机器学习的图案分组方法
US20110038527A1 (en) Image pattern matching systems and methods for wafer alignment
TWI533383B (zh) 使用重覆結構之基於設計之檢測
US20120182442A1 (en) Hardware generation of image descriptors
US7630560B2 (en) Increasing accuracy of discrete curve transform estimates for curve matching in four or more dimensions
US20100111376A1 (en) Assesssing biometric sample quality using wavelets and a boosted classifier
US11880994B2 (en) Method and system for determining plant leaf surface roughness
WO2023035301A1 (en) A camera calibration method
US10074034B2 (en) Image processing including adjoin feature based object detection, and/or bilateral symmetric object segmentation
Ciric et al. Evaluating similarity of spectrogram-like images of DC motor sounds by Pearson correlation coefficient
Han et al. Defect detection on button surfaces with the weighted least-squares model
WO2020125528A1 (zh) 锚定物检测方法及装置、电子设备及存储介质
Ghanbari et al. Radiometric normalization of multitemporal and multisensor remote sensing images based on a Gaussian mixture model and error ellipse
JP7475901B2 (ja) 試験片上の欠陥検出の方法およびそのシステム
CN107679528A (zh) 一种基于AdaBoost‑SVM集成学习算法的行人检测方法
Borole et al. Statistical Feature Based Digital Camera Identification
Mun et al. 2D template matching for automated inspection using generalized hough transform
US11847777B1 (en) Correlation-based overlay key centering system and method thereof

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23778297

Country of ref document: EP

Kind code of ref document: A1