A high-precision nanoimprint wafer alignment method based on multi-scale image recognition

Abstract

The present invention discloses a high-precision alignment method for nanoimprint wafers based on multi-scale image recognition, which belongs to the field of image processing technology. After graying the image, the present invention constructs images of three scales, extracts contours from the images of the three scales respectively, and refines the contours to make the contour lines clear, and then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines in the images of the three scales, combines the images of the three scales, improves the accuracy of finding the coordinates of the marking points, maps the coordinates of the marking points to the spatial coordinates, obtains the deviation position according to the spatial coordinates, and aligns the wafer based on the deviation position. The present invention extracts the contour center points of the images of the three scales, constructs the refined images of the three scales, makes the contour structure clearer, improves the accuracy of determining the marking points, and then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines under the three scales, further improves the accuracy of determining the marking points, thereby improving the alignment accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a nanoimprint wafer high-precision alignment method based on multi-scale image recognition. Background Art

[0002] Nanoimprint lithography is an emerging nanomanufacturing technology with broad application prospects, especially in the fields of semiconductors, optoelectronics and biomedicine, which can achieve high-resolution and low-cost replication of micro-nano structures. This technology transfers patterns to substrates quickly and efficiently through imprint molds, and is suitable for creating nanoscale structures and patterns. Therefore, it plays an important role in promoting the miniaturization and high performance of electronic products.

[0003] However, in the nanoimprint process, precise alignment between the wafer and the template is a key step to ensure processing quality and final product performance. Poor alignment may lead to pattern distortion, poor repeatability, and even production scrap. Ensuring high-precision alignment is essential to achieve controllable processing and stable product output, which is a major challenge faced by nanoimprint technology in practical applications. Traditional alignment methods mainly rely on optical or mechanical positioning, but as the process size continues to shrink, there is a problem of low alignment accuracy. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a high-precision nanoimprint wafer alignment method based on multi-scale image recognition to solve the problem of low alignment accuracy in the prior art.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a nanoimprint wafer high-precision alignment method based on multi-scale image recognition, comprising the following steps:

[0006] S1, collecting an image of the wafer after placement, graying the image to obtain a grayscale image;

[0007] S2, constructing an enlarged image, a reduced image and an original image for the grayscale image;

[0008] S3, extracting contours of the enlarged image, the reduced image and the original image respectively to obtain an enlarged contour map, a reduced contour map and an original contour map;

[0009] S4, respectively extracting contour center points of the enlarged contour map, the reduced contour map and the original contour map, and constructing an enlarged refinement map, a reduced refinement map and the original refinement map;

[0010] S5, extracting the bending characteristic values ​​of the contour lines from the enlarged refinement map, the reduced refinement map and the original refinement map, and finding the coordinates of the marking points;

[0011] S6. Map the coordinates of the marking points to the spatial coordinates to obtain the deviation position, and align the wafer according to the deviation position.

[0012] Further, S2 includes the following sub-steps:

[0013] S21, inserting a new pixel point between every two adjacent pixels in the grayscale image, taking the grayscale value of the new pixel point as the average of the grayscale values ​​of the two adjacent pixels points, to obtain an enlarged image;

[0014] S22, taking the grayscale image as the original image;

[0015] S23, divide the grayscale image into The weighted value of each gray value in the small block is taken as the gray value of the new pixel to obtain the reduced image.

[0016] Further, S4 includes the following sub-steps:

[0017] S41, setting the sliding window to slide on each contour line on the enlarged contour map, the reduced contour map and the original contour map respectively, and each time the sliding is performed, the forward step length is 1 pixel length;

[0018] S42, calculating the center point of the contour under the sliding window at each sliding;

[0019] S43, using each contour center point on the enlarged contour image as a new contour point to form an enlarged refinement image;

[0020] S44, using the center points of each contour on the reduced contour image as new contour points to form a reduced refinement image;

[0021] S45, taking the center points of each contour on the original contour map as new contour points to form the original refined map.

[0022] Furthermore, the formula for calculating the contour center point in S42 is: , , where x o is the horizontal coordinate of the center point of the contour, y o is the ordinate of the center point of the contour, x i is the horizontal coordinate of the i-th pixel point under the sliding window, y i is the ordinate of the i-th pixel under the sliding window, N is the number of pixels under the sliding window, and i is a positive integer.

[0023] Further, S5 includes the following sub-steps:

[0024] S51, respectively calculating the curvature of each contour point on the enlarged refinement map, the reduced refinement map and the original refinement map;

[0025] S52, calculating the curvature characteristic value of each contour line according to the curvature of each contour point;

[0026] S53, calculating the matching degree according to the bending characteristic values ​​of the same contour line on the enlarged refinement map, the reduced refinement map and the original refinement map;

[0027] S54, when the matching degree is greater than the matching threshold, marking the contour line on the original thinned image as a marker contour;

[0028] S55. Taking the geometric coordinate center of the mark outline as the coordinates of the marking point.

[0029] Furthermore, the formula for calculating the curvature in S51 is: , where s n is the curvature of the nth contour point, is the first-order derivative in the x direction of the nth contour point, is the first-order derivative in the y direction of the nth contour point, is the second-order derivative in the x direction of the nth contour point, is the second-order derivative in the y direction of the nth contour point, where n is a positive integer.

[0030] Furthermore, the formula for calculating the bending characteristic value of each contour line in S52 is: , where γ is the curvature eigenvalue of the contour line, s n is the curvature of the nth contour point, max is the maximum value, min is the minimum value, n is a positive integer, and M is the number of contour points on the contour line.

[0031] Furthermore, the formula for calculating the matching degree in S53 is: , where θ is the matching degree, γ A is the bending eigenvalue of the contour line on the enlarged refinement map, γ N To reduce the curvature eigenvalue of the contour line on the refinement map, γ o is the bending eigenvalue of the contour line on the original refinement map, Stores values ​​for the curvature characteristics of contour lines on the amplified refinement map. Stores values ​​for reducing the curvature of contours on the refinement map. Stores values ​​for the curvature characteristics of contours on the original refinement map.

[0032] Further, S6 includes the following sub-steps:

[0033] S61, mapping the coordinates of the marking point to the spatial coordinates;

[0034] S62, subtracting the mapped spatial coordinates from the set coordinates to obtain a deviation position;

[0035] S63, aligning the wafer according to the deviation position.

[0036] Furthermore, the mapping formula in S61 is: , , where X is the horizontal coordinate in space, Y is the vertical coordinate in space, and x c is the horizontal coordinate of the marking point, y c is the ordinate of the marking point, a x is the scaling factor of the horizontal axis, a y is the scaling factor of the ordinate, b x is the translation of the horizontal axis, b y is the vertical axis translation.

[0037] The beneficial effects of the present invention are as follows: after graying the image, the present invention constructs images of three scales, extracts contours from the images of the three scales respectively, and refines the contours to make the contour lines clear, then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines in the images of the three scales, combines the images of the three scales to improve the accuracy of finding the coordinates of the marking points, maps the coordinates of the marking points to the spatial coordinates, obtains the deviation position according to the spatial coordinates, and aligns the wafer based on the deviation position. The present invention extracts the contour center points from the images of the three scales, constructs the refined images of the three scales, makes the contour structure clearer, improves the accuracy of determining the marking points, then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines under the three scales, further improves the accuracy of determining the marking points, and thus improves the alignment accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of a high-precision nanoimprint wafer alignment method based on multi-scale image recognition. DETAILED DESCRIPTION

[0039] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0040] like Figure 1 As shown, a nanoimprint wafer high-precision alignment method based on multi-scale image recognition includes the following steps:

[0041] S1, collecting an image of the wafer after placement, graying the image to obtain a grayscale image;

[0042] S2, constructing an enlarged image, a reduced image and an original image for the grayscale image;

[0043] S3, extracting contours of the enlarged image, the reduced image and the original image respectively to obtain an enlarged contour map, a reduced contour map and an original contour map;

[0044] S4, respectively extracting contour center points of the enlarged contour map, the reduced contour map and the original contour map, and constructing an enlarged refinement map, a reduced refinement map and the original refinement map;

[0045] S5, extracting the bending characteristic values ​​of the contour lines from the enlarged refinement map, the reduced refinement map and the original refinement map, and finding the coordinates of the marking points;

[0046] S6. Map the coordinates of the marking points to the spatial coordinates to obtain the deviation position, and align the wafer according to the deviation position.

[0047] In this embodiment, S2 includes the following sub-steps:

[0048] S21, inserting a new pixel point between every two adjacent pixels in the grayscale image, taking the grayscale value of the new pixel point as the average of the grayscale values ​​of the two adjacent pixels points, to obtain an enlarged image;

[0049] S22, taking the grayscale image as the original image;

[0050] S23, divide the grayscale image into The weighted value of each gray value in the small block is taken as the gray value of the new pixel to obtain the reduced image.

[0051] In this embodiment, step S23 specifically includes: dividing the grayscale image into The image is then divided into small blocks of different sizes for subsequent processing. Next, the grayscale value of each pixel in each small block is extracted, and these values ​​represent the brightness information of different areas in the image. Then, the weighted sum of the grayscale values ​​of all pixels in each small block is calculated to obtain the weighted value. Next, the weighted value of each small block is used as the pixel value of the corresponding position in the new image, thereby generating a reduced image.

[0052] The present invention enlarges the image by inserting new pixels, thereby enlarging the contour structure, divides the grayscale image into multiple small blocks, uses the weighted values ​​in the small blocks as the grayscale values ​​of the new pixels, shrinks the image, thereby shrinking the contour structure, uses the grayscale image as the original image, retains the original structure, and realizes the search for marker point coordinates from three scales.

[0053] The size of the original image is When , the size of the enlarged image is When the image is reduced to , K is the length or width of the original image. When obtaining the reduced image, if K is an odd number, the last row and the last column are discarded.

[0054] In this embodiment, in step S23, each gray value is weighted according to the weight of each gray value in the small block to obtain a weighted value.

[0055] In this embodiment, the method for extracting contours from the image in step S3 includes: threshold segmentation method, region growing method, etc.

[0056] In this embodiment, S4 includes the following sub-steps:

[0057] S41, setting the sliding window to slide on each contour line on the enlarged contour map, the reduced contour map and the original contour map respectively, and each time the sliding is performed, the forward step length is 1 pixel length;

[0058] In this embodiment, the size of the sliding window is , L is a positive integer, ranging from 3, 4 or 5;

[0059] S42, calculating the center point of the contour under the sliding window at each sliding;

[0060] S43, using each contour center point on the enlarged contour image as a new contour point to form an enlarged refinement image;

[0061] S44, using the center points of each contour on the reduced contour image as new contour points to form a reduced refinement image;

[0062] S45, taking the center points of each contour on the original contour map as new contour points to form the original refined map.

[0063] In this embodiment, the formula for calculating the contour center point in S42 is: , , where x o is the horizontal coordinate of the center point of the contour, y o is the ordinate of the center point of the contour, x i is the horizontal coordinate of the i-th pixel point under the sliding window, y i is the ordinate of the i-th pixel under the sliding window, N is the number of pixels under the sliding window, and i is a positive integer.

[0064] In the present invention, the length of each advance of the sliding window is 1 pixel. After sliding, a contour center point is calculated as a new contour point for this sliding. One contour point is obtained after one sliding, thereby changing the contour line to a contour line with a width of 1, making the contour line clear.

[0065] In this embodiment, S5 includes the following sub-steps:

[0066] S51, respectively calculating the curvature of each contour point on the enlarged refinement map, the reduced refinement map and the original refinement map;

[0067] S52, calculating the curvature characteristic value of each contour line according to the curvature of each contour point;

[0068] S53, calculating the matching degree according to the bending characteristic values ​​of the same contour line on the enlarged refinement map, the reduced refinement map and the original refinement map;

[0069] S54, when the matching degree is greater than the matching threshold, marking the contour line on the original thinned image as a marker contour;

[0070] S55. Taking the geometric coordinate center of the mark outline as the coordinates of the marking point.

[0071] In this embodiment, the matching threshold is set based on experiments or experience.

[0072] In this embodiment, the formula for calculating the curvature in S51 is: , where s n is the curvature of the nth contour point, is the first-order derivative in the x direction of the nth contour point, is the first-order derivative in the y direction of the nth contour point, is the second-order derivative in the x direction of the nth contour point, is the second-order derivative in the y direction of the nth contour point, where n is a positive integer.

[0073] In this embodiment, , , , , x n is the horizontal coordinate of the nth contour point, y n is the ordinate of the nth contour point, x n+1 is the horizontal coordinate of the n+1th contour point, y n+1 is the ordinate of the n+1th contour point, x n-1 is the horizontal coordinate of the n-1th contour point, y n-1 is the ordinate of the n-1th contour point.

[0074] The present invention calculates the curvature of each contour point, and calculates the curvature characteristic value of the contour line according to the curvature of each contour point on the contour line, reflecting the morphological characteristics of the contour line. The present invention performs scaling processing on the original image, so for a contour line, there are three scales: magnified, reduced and original. The matching degree is calculated according to the curvature characteristic value of the same contour line at the three scales, and when the matching degree is greater than the matching threshold, the contour line is marked as a marker contour. The present invention realizes the evaluation of whether the contour line is a marker contour from three scales, and improves the determination accuracy of the marker contour.

[0075] In this embodiment, the formula for calculating the bending characteristic value of each contour line in S52 is: , where γ is the curvature characteristic value of the contour line, s n is the curvature of the nth contour point, max is the maximum value, min is the minimum value, n is a positive integer, and M is the number of contour points on the contour line.

[0076] The present invention can comprehensively reflect the curvature characteristics of the entire contour line by accumulating the curvature of all points on the contour line, and uses the difference between the maximum value and the minimum value as a weight, thereby highlighting the part of the contour line with a large change in curvature, which helps to identify key feature points.

[0077] In this embodiment, the formula for calculating the matching degree in S53 is: , where θ is the matching degree, γ A is the bending eigenvalue of the contour line on the enlarged refinement map, γ N To reduce the curvature eigenvalue of the contour line on the refinement map, γ o is the bending eigenvalue of the contour line on the original refinement map, Stores values ​​for the curvature characteristics of contour lines on the amplified refinement map. Stores values ​​for reducing the curvature of contours on the refinement map. Stores values ​​for the curvature characteristics of contours on the original refinement map.

[0078] In this embodiment, the stored values ​​of the bending characteristics of the contour line on the enlarged and refined image, the stored values ​​of the bending characteristics of the contour line on the reduced and refined image, and the stored values ​​of the bending characteristics of the contour line on the original and refined image are the bending characteristics values ​​obtained based on the known marker contour in the image, and are stored in the system, that is, three pre-stored comparison values. The three stored values ​​are calculated based on the known marker contour in the image and serve as pre-stored comparison values, representing the bending characteristics of the known marker contour at different scales (enlarged, reduced, original). These pre-stored values ​​will be used for subsequent matching degree calculations to evaluate the similarity between the newly recognized contour and the known marker contour.

[0079] In this embodiment, S6 includes the following sub-steps:

[0080] S61, mapping the coordinates of the marking point to the spatial coordinates;

[0081] S62, subtracting the mapped spatial coordinates from the set coordinates to obtain a deviation position;

[0082] S63, aligning the wafer according to the deviation position.

[0083] In this embodiment, the mapping formula in S61 is: , , where X is the horizontal coordinate in space, Y is the vertical coordinate in space, and x c is the horizontal coordinate of the marking point, y c is the ordinate of the marking point, a x is the scaling factor of the horizontal axis, a y is the scaling factor of the ordinate, b x is the translation of the horizontal axis, b y is the vertical axis translation.

[0084] After graying the image, the present invention constructs images of three scales, extracts contours from the images of the three scales respectively, and refines the contours to make the contour lines clear, then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines in the images of the three scales, combines the images of the three scales to improve the accuracy of finding the coordinates of the marking points, maps the coordinates of the marking points to the spatial coordinates, obtains the deviation position according to the spatial coordinates, and aligns the wafer based on the deviation position. The present invention extracts the contour center points from the images of the three scales, constructs the refined images of the three scales, makes the contour structure clearer, improves the accuracy of determining the marking points, then finds the coordinates of the marking points according to the bending characteristic values ​​of the contour lines under the three scales, further improves the accuracy of determining the marking points, and thus improves the alignment accuracy.

[0085] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A high-precision nanoimprint wafer alignment method based on multi-scale image recognition, characterized in that: The following steps are involved: S1, collecting an image of the wafer after placement, graying the image to obtain a grayscale image; S2, constructing an enlarged image, a reduced image and an original image for the grayscale image; S3, extracting contours of the enlarged image, the reduced image and the original image respectively to obtain an enlarged contour map, a reduced contour map and an original contour map; S4, respectively extracting contour center points of the enlarged contour map, the reduced contour map and the original contour map, and constructing an enlarged refinement map, a reduced refinement map and the original refinement map; S5, extracting the bending characteristic values ​​of the contour lines from the enlarged refinement map, the reduced refinement map and the original refinement map, and finding the coordinates of the marking points; S6, mapping the coordinates of the marking point to the spatial coordinates to obtain the deviation position, and aligning the wafer according to the deviation position; The S5 comprises the following sub-steps: S51, respectively calculating the curvature of each contour point on the enlarged refinement map, the reduced refinement map and the original refinement map; S52, calculating the curvature characteristic value of each contour line according to the curvature of each contour point; S53, calculating the matching degree according to the bending characteristic values ​​of the same contour line on the enlarged refinement map, the reduced refinement map and the original refinement map; S54, when the matching degree is greater than the matching threshold, marking the contour line on the original thinned image as a marker contour; S55, taking the geometric coordinate center of the mark outline as the coordinate of the marking point; The formula for calculating the matching degree in S53 is: , where θ is the matching degree, γ A is the bending eigenvalue of the contour line on the enlarged refinement map, γ N To reduce the curvature eigenvalue of the contour line on the refinement map, γ o is the bending eigenvalue of the contour line on the original refinement map, Stores values ​​for the curvature characteristics of contour lines on the amplified refinement map. Stores values ​​for reducing the curvature of contours on the refinement map. Stores values ​​for the curvature characteristics of contours on the original refinement map.

2. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21, inserting a new pixel point between every two adjacent pixels in the grayscale image, taking the grayscale value of the new pixel point as the average of the grayscale values ​​of the two adjacent pixels points, to obtain an enlarged image; S22, taking the grayscale image as the original image; S23, divide the grayscale image into The weighted value of each gray value in the small block is taken as the gray value of the new pixel to obtain the reduced image.

3. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41, setting the sliding window to slide on each contour line on the enlarged contour map, the reduced contour map and the original contour map respectively, and each time the sliding is performed, the forward step length is 1 pixel length; S42, calculating the center point of the contour under the sliding window at each sliding; S43, using each contour center point on the enlarged contour image as a new contour point to form an enlarged refinement image; S44, using the center points of each contour on the reduced contour image as new contour points to form a reduced refinement image; S45, taking the center points of each contour on the original contour map as new contour points to form the original refined map.

4. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 3 is characterized in that: The formula for calculating the contour center point in S42 is: , , where x o is the horizontal coordinate of the center point of the contour, y o is the ordinate of the center point of the contour, x i is the horizontal coordinate of the i-th pixel point under the sliding window, y i is the ordinate of the i-th pixel under the sliding window, N is the number of pixels under the sliding window, and i is a positive integer.

5. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 1, characterized in that: The formula for calculating the curvature in S51 is: , where s n is the curvature of the nth contour point, is the first-order derivative in the x direction of the nth contour point, is the first-order derivative in the y direction of the nth contour point, is the second-order derivative in the x direction of the nth contour point, is the second-order derivative in the y direction of the nth contour point, where n is a positive integer.

6. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 1, characterized in that: The formula for calculating the bending characteristic value of each contour line in S52 is: , where γ is the curvature eigenvalue of the contour line, s n is the curvature of the nth contour point, max is the maximum value, min is the minimum value, n is a positive integer, and M is the number of contour points on the contour line.

7. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 1, characterized in that: The S6 comprises the following sub-steps: S61, mapping the coordinates of the marking point to the spatial coordinates; S62, subtracting the mapped spatial coordinates from the set coordinates to obtain a deviation position; S63, aligning the wafer according to the deviation position.

8. The high-precision nanoimprint wafer alignment method based on multi-scale image recognition according to claim 7, characterized in that: The mapping formula in S61 is: , , where X is the horizontal coordinate in space, Y is the vertical coordinate in space, and x c is the horizontal coordinate of the marking point, y c is the ordinate of the marking point, a x is the scaling factor of the horizontal axis, a y is the scaling factor of the ordinate, b x is the translation of the horizontal axis, b y is the vertical axis translation.

Images