A real-time detection method for nanoimprint wafer demoulding defects

Abstract

The present invention discloses a real-time detection method for nanoimprint wafer demoulding defects, which belongs to the field of image processing technology. The present invention performs difference processing on the initial image and the standard image, enhances the contrast between the defect and the background, and makes the potential defect more obvious. The present invention decomposes the difference image into R, G, and B channel images, retains all channel information, extracts the contour of each channel image, performs channel contour fusion, obtains RG contour fusion image, GB contour fusion image and RB contour fusion image, and then performs pixel value enhancement processing on the three contour fusion images, further improves the contrast of the defect, makes the defect more prominent, finds the defect position according to the abnormality of the local area on the RG contour enhancement image, the GB contour enhancement image and the RB contour enhancement image, realizes capturing the defect characteristics from multiple dimensions, effectively reduces the false alarm rate, and improves the accuracy of defect detection.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a real-time detection method for nanoimprint wafer demoulding defects. Background Art

[0002] As an emerging nanomanufacturing method, nanoimprinting technology has shown great application potential in the fields of semiconductors, optoelectronics, and biomedicine. However, various defects such as residues, tearing, and deformation often occur during the wafer demolding process, which seriously affect the yield and performance of the product. At present, the defect detection methods commonly used in the industry mainly rely on manual observation or offline detection. Although the manual observation method is intuitive, it is inefficient and easily affected by the subjective factors of the operator, making it difficult to ensure the consistency and accuracy of the detection. Although offline detection can provide higher accuracy, it cannot meet the urgent needs of modern production lines for real-time quality control.

[0003] When detecting nanoimprint wafer demoulding defects, image processing methods are often used to extract the image contour after wafer demoulding and compare the extracted contour with the stored contour. However, comprehensive contour comparison of images is a computationally intensive task and is difficult to meet the real-time detection requirements on high-speed production lines. In addition, the contour image extracted in the prior art is a grayscale image, which loses a lot of information compared to a color image, resulting in low accuracy in defect detection. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a real-time detection method for nanoimprint wafer demoulding defects, which solves the problems of high computational complexity and low defect detection 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 real-time detection method for nanoimprint wafer demoulding defects, comprising the following steps:

[0006] S1, collecting images of the wafer after demolding to obtain an initial image;

[0007] S2, subtracting the pixel values ​​of the initial image from the standard image to obtain a difference image;

[0008] S3, decomposing the difference image into R, G, and B channel images, extracting contours of the R, G, and B channel images, and performing channel contour fusion to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image;

[0009] S4, respectively performing pixel value enhancement processing on the RG contour fusion image, the GB contour fusion image, and the RB contour fusion image to obtain an RG contour enhanced image, a GB contour enhanced image, and a RB contour enhanced image;

[0010] S5. Find the defect location according to the abnormality of the local area on the RG contour enhancement image, the GB contour enhancement image, and the RB contour enhancement image.

[0011] Furthermore, S2 specifically includes: subtracting pixel values ​​of the initial image and the standard image at the same pixel position to obtain a pixel difference, and using the pixel difference as a new pixel value at the pixel position to obtain a difference image.

[0012] Furthermore, S3 includes the following sub-steps:

[0013] S31, decomposing the difference image into an R channel image, a G channel image and a B channel image;

[0014] S32, taking each pixel point as the center on the R channel image, calculating the complexity of the R channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the R channel contour image;

[0015] S33, taking each pixel point as the center on the G channel image, calculating the complexity of the G channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point at the center to obtain a G channel contour image;

[0016] S34, taking each pixel point as the center on the B channel image, calculating the complexity of the B channel value within the neighborhood range of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the B channel contour image;

[0017] S35 , fusing the R channel contour image, the G channel contour image, and the B channel contour image to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image.

[0018] Furthermore, the formula for calculating the complexity in S32, S33 and S34 is: , where γ is the complexity, x i The i-th R channel value, G channel value or B channel value within the neighborhood of the center, max{x i} is the maximum channel value in the neighborhood of the center, min{x i} is the minimum channel value in the neighborhood of the center, The channel mean in the neighborhood centered at .

[0019] Furthermore, the S35 comprises the following sub-steps:

[0020] S351, according to the R channel contour image and the G channel contour image, weighted processing is performed on the R channel value and the G channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RG contour fusion image;

[0021] S352, according to the G channel contour image and the B channel contour image, weighted processing is performed on the G channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain a GB contour fusion image;

[0022] S353, according to the R channel contour image and the B channel contour image, weighted processing is performed on the R channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RB contour fusion image.

[0023] Furthermore, the pixel value enhancement formula in S4 is: ,in, is the pixel value of the nth pixel in the RG contour enhanced image, GB contour enhanced image or RB contour enhanced image after enhancement, P n is the pixel value of the nth pixel in the RG contour fusion image, GB contour fusion image or RB contour fusion image, where n is a positive integer.

[0024] Furthermore, the S5 comprises the following sub-steps:

[0025] S51, cutting the RG contour enhanced image, the GB contour enhanced image and the RB contour enhanced image into a plurality of image regions respectively, wherein the cutting method for the three images is the same;

[0026] S52, calculating the abnormality of the image area on the RG contour enhanced image;

[0027] S53, calculating the abnormality of the image area on the GB contour enhanced image;

[0028] S54, calculating the abnormality of the image area on the RB contour enhanced image;

[0029] S55. At the same image area position, when the abnormality of the image area on the RG contour enhanced image, the abnormality of the image area on the GB contour enhanced image, and the abnormality of the image area on the RB contour enhanced image are all greater than the abnormality threshold, the image area position is a defect position.

[0030] Furthermore, S52, S53, and S54 all include the following sub-steps:

[0031] A1. On the contour enhanced image, count the frequency of each pixel value appearing in the image area and the frequency of each pixel value appearing in the contour enhanced image, and calculate the frequency anomaly coefficient;

[0032] A2. On the contour enhanced image, the pixel distribution anomaly coefficient is calculated according to the distribution of pixel values ​​in each image region;

[0033] A3, adding the frequency abnormality coefficient and the pixel distribution abnormality coefficient to obtain the abnormality degree of the image area on the contour enhanced image, wherein the contour enhanced image in A1, A2 and A3 is the RG contour enhanced image, the GB contour enhanced image or the RB contour enhanced image.

[0034] Furthermore, the formula for calculating the frequency anomaly coefficient in A1 is: , where θ f is the frequency anomaly coefficient, f local,m is the frequency of the mth pixel value appearing in the image area, f whole,m is the frequency of the mth pixel value in the image area appearing in the edge enhanced image, m is a positive integer, and M is the number of types of pixel values ​​in the image area.

[0035] Furthermore, the formula for calculating the pixel distribution abnormality coefficient in A2 is: , where θ p is the pixel distribution anomaly coefficient, P k is the pixel value of the kth pixel in the image area, K is the number of pixels in the image area, E is the area of ​​the image area, and k is a positive integer.

[0036] The beneficial effects of the present invention are as follows: the present invention performs difference processing on the initial image and the standard image, thereby enhancing the contrast between the defect and the background, making the potential defect more obvious. The present invention decomposes the difference image into R, G, and B channel images, retains all channel information, extracts the contour of each channel image, performs channel contour fusion, obtains RG contour fusion image, GB contour fusion image and RB contour fusion image, and then performs pixel value enhancement processing on the three contour fusion images, further improving the contrast of the defect, making the defect more prominent, and finding the defect position according to the abnormality of the local area on the RG contour enhancement image, GB contour enhancement image and RB contour enhancement image, realizing the capture of defect features from multiple dimensions, effectively reducing the false alarm rate, and improving the accuracy of defect detection. At the same time, the present invention does not need to perform comprehensive contour comparison on the image, thereby reducing the computational complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flow chart of a method for real-time detection of nanoimprint wafer demoulding defects. DETAILED DESCRIPTION

[0038] 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.

[0039] like Figure 1 As shown, a real-time detection method for nanoimprint wafer demoulding defects comprises the following steps:

[0040] S1, collecting images of the wafer after demolding to obtain an initial image;

[0041] S2, subtracting the pixel values ​​of the initial image from the standard image to obtain a difference image;

[0042] S3, decomposing the difference image into R, G, and B channel images, extracting contours of the R, G, and B channel images, and performing channel contour fusion to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image;

[0043] S4, respectively performing pixel value enhancement processing on the RG contour fusion image, the GB contour fusion image, and the RB contour fusion image to obtain an RG contour enhanced image, a GB contour enhanced image, and a RB contour enhanced image;

[0044] S5. Find the defect location according to the abnormality of the local area on the RG contour enhancement image, the GB contour enhancement image, and the RB contour enhancement image.

[0045] In this embodiment, the standard image is defined as: an image of a defect-free wafer after demolding.

[0046] In this embodiment, the angle, position and parameters of shooting the initial image and the standard image are the same.

[0047] The S2 specifically includes: subtracting pixel values ​​of the initial image and the standard image at the same pixel position to obtain a pixel difference, and using the pixel difference as a new pixel value at the pixel position to obtain a difference image.

[0048] The S3 comprises the following sub-steps:

[0049] S31, decomposing the difference image into an R channel image, a G channel image and a B channel image;

[0050] S32, taking each pixel point as the center on the R channel image, calculating the complexity of the R channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the R channel contour image;

[0051] S33, taking each pixel point as the center on the G channel image, calculating the complexity of the G channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point at the center to obtain a G channel contour image;

[0052] S34, taking each pixel point as the center on the B channel image, calculating the complexity of the B channel value within the neighborhood range of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the B channel contour image;

[0053] S35 , fusing the R channel contour image, the G channel contour image, and the B channel contour image to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image.

[0054] The present invention processes the three channels of R, G, and B separately to retain more image detail information and improve the accuracy of defect detection. Secondly, by calculating the complexity of the area around each pixel and setting a threshold to generate a contour image, the edge and texture features in the image can be effectively highlighted, which helps to identify small or inconspicuous defects. Finally, the contour images of the three channels are fused to comprehensively utilize the information of different color channels to further enhance the visibility of defect features.

[0055] In this embodiment, the complexity threshold is specifically set according to experiments or requirements.

[0056] The formula for calculating the complexity in S32, S33 and S34 is: , where γ is the complexity, x i The i-th R channel value, G channel value or B channel value within the neighborhood of the center, max{x i} is the maximum channel value in the neighborhood of the center, min{x i} is the minimum channel value in the neighborhood of the center, The channel mean in the neighborhood centered at .

[0057] In this embodiment, the size of the neighborhood range is .

[0058] The present invention uses the square of the difference between the maximum value and the minimum value and the average value, so that the method is particularly sensitive to extreme value changes in a local area, which is helpful for detecting small or inconspicuous defects.

[0059] The S35 comprises the following sub-steps:

[0060] S351, according to the R channel contour image and the G channel contour image, weighted processing is performed on the R channel value and the G channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RG contour fusion image;

[0061] S352, according to the G channel contour image and the B channel contour image, weighted processing is performed on the G channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain a GB contour fusion image;

[0062] S353, according to the R channel contour image and the B channel contour image, weighted processing is performed on the R channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RB contour fusion image.

[0063] In this embodiment, during the weighted processing in steps S351, S352, and S353, the weighted coefficients for the two channel values ​​may both be 0.5.

[0064] The present invention makes full use of the information of the three channels of RGB, and can capture the features and details in the image more comprehensively through the fusion of the three combinations of RG, GB and RB. By fusing the contour information of multiple channels, the noise or misjudgment that may be caused by a single channel can be effectively reduced, and the accuracy and reliability of detection can be improved.

[0065] The pixel value enhancement formula in S4 is: ,in, is the pixel value of the nth pixel in the RG contour enhanced image, GB contour enhanced image or RB contour enhanced image after enhancement, P n is the pixel value of the nth pixel in the RG contour fusion image, GB contour fusion image or RB contour fusion image, where n is a positive integer.

[0066] The present invention squares the pixel value of each pixel point, so that the pixel point with a large pixel value has a more significant pixel value, thereby highlighting the defective area.

[0067] The S5 comprises the following sub-steps:

[0068] S51, cutting the RG contour enhanced image, the GB contour enhanced image and the RB contour enhanced image into a plurality of image regions respectively, wherein the cutting method for the three images is the same;

[0069] In this embodiment, in step S51, each image may be cut horizontally L times and vertically L times to obtain image area;

[0070] S52, calculating the abnormality of the image area on the RG contour enhanced image;

[0071] S53, calculating the abnormality of the image area on the GB contour enhanced image;

[0072] S54, calculating the abnormality of the image area on the RB contour enhanced image;

[0073] S55. At the same image area position, when the abnormality of the image area on the RG contour enhanced image, the abnormality of the image area on the GB contour enhanced image, and the abnormality of the image area on the RB contour enhanced image are all greater than the abnormality threshold, the image area position is a defect position.

[0074] The present invention can achieve more detailed local analysis and improve the spatial resolution of detection by cutting the three contour enhancement images of RG, GB and RB into multiple regions. Secondly, the abnormality is calculated separately for the image area of ​​each channel combination, which makes full use of the information of different color channels and enhances the sensitivity to various defects. Finally, by comparing the abnormality of the three channel combinations at the same position and setting the threshold judgment, the false alarm rate is greatly reduced and the accuracy and reliability of detection are improved.

[0075] In this embodiment, the abnormality threshold is specifically set based on experiments or experience.

[0076] The S52, S53 and S54 all include the following sub-steps:

[0077] A1. On the contour enhanced image, count the frequency of each pixel value appearing in the image area and the frequency of each pixel value appearing in the contour enhanced image, and calculate the frequency anomaly coefficient;

[0078] A2. On the contour enhanced image, the pixel distribution anomaly coefficient is calculated according to the distribution of pixel values ​​in each image region;

[0079] A3, adding the frequency abnormality coefficient and the pixel distribution abnormality coefficient to obtain the abnormality degree of the image area on the contour enhanced image, wherein the contour enhanced image in A1, A2 and A3 is the RG contour enhanced image, the GB contour enhanced image or the RB contour enhanced image.

[0080] By calculating the frequency anomaly coefficient (step A1), the present invention can effectively capture the difference in pixel value distribution between the local area and the whole image, and identify potential defects from the perspective of the frequency of pixel value occurrence. Secondly, the calculation of the pixel distribution anomaly coefficient (step A2) focuses on the statistical distribution characteristics of the pixel value, which helps to discover deviations in the overall structure. The comprehensive anomaly degree obtained by adding these two anomaly coefficients (step A3) improves the comprehensiveness and accuracy of the detection. The present invention is applied to three contour-enhanced images, RG, GB and RB, respectively, making full use of the information of different color channel combinations, further improving the sensitivity and reliability of the detection.

[0081] The formula for calculating the frequency anomaly coefficient in A1 is: , where θ f is the frequency anomaly coefficient, f local,m is the frequency of the mth pixel value appearing in the image area, f whole,m is the frequency of the mth pixel value in the image area appearing in the edge enhanced image, m is a positive integer, and M is the number of types of pixel values ​​in the image area.

[0082] The present invention counts the frequency of each pixel value appearing in the image area, and the frequency of each pixel value appearing in the image area on the edge enhancement image. For the same pixel value, f local,m =f whole,m When this pixel value only appears in the image area, the frequency abnormality coefficient is large. whole,m Greater than f local,m When , the greater the frequency of this pixel value appearing in the contour enhanced image, the smaller the frequency anomaly coefficient.

[0083] In the present invention, the same kind of pixel value refers to a type of pixel values ​​with equal pixel values.

[0084] The formula for calculating the pixel distribution anomaly coefficient in A2 is: , where θ p is the pixel distribution anomaly coefficient, P k is the pixel value of the kth pixel in the image area, K is the number of pixels in the image area, E is the area of ​​the image area, and k is a positive integer.

[0085] In step S2, the pixel values ​​of the initial image and the standard image are subtracted to obtain a difference image. Under ideal conditions, the pixel value of each pixel in the difference image is 0. However, due to different environmental factors, there may be slight vibrations during shooting. Therefore, there is an imaging difference between the initial image and the standard image. The pixel distribution anomaly coefficient is calculated in A2 to count the number of pixel values ​​and pixel points in each image area to reflect the abnormality of the area. The larger the pixel distribution anomaly coefficient, the more pixel values ​​and pixel points there are, the greater the difference between the area and the standard image, and the higher the degree of abnormality.

[0086] The present invention performs difference processing on the initial image and the standard image, enhances the contrast between the defect and the background, and makes the potential defect more obvious. The present invention decomposes the difference image into R, G, and B channel images, retains all channel information, extracts the contour of each channel image, performs channel contour fusion, obtains RG contour fusion image, GB contour fusion image and RB contour fusion image, and then performs pixel value enhancement processing on the three contour fusion images, further improves the contrast of the defect, makes the defect more prominent, finds the defect position according to the abnormality of the local area on the RG contour enhancement image, GB contour enhancement image and RB contour enhancement image, realizes capturing the defect features from multiple dimensions, effectively reduces the false alarm rate, and improves the accuracy of defect detection. At the same time, the present invention does not need to perform comprehensive contour comparison on the image, which reduces the computational complexity.

[0087] 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 real-time detection method for nanoimprint wafer demoulding defects, characterized in that: The following steps are involved: S1, collecting images of the wafer after demolding to obtain an initial image; S2, subtracting the pixel values ​​of the initial image from the standard image to obtain a difference image; S3, decomposing the difference image into R, G, and B channel images, extracting contours of the R, G, and B channel images, and performing channel contour fusion to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image; S4, respectively performing pixel value enhancement processing on the RG contour fusion image, the GB contour fusion image and the RB contour fusion image to obtain an RG contour enhanced image, a GB contour enhanced image and a RB contour enhanced image; S5. Find the defect location according to the abnormality of the local area on the RG contour enhancement image, the GB contour enhancement image, and the RB contour enhancement image; The S5 comprises the following sub-steps: S51, cutting the RG contour enhanced image, the GB contour enhanced image and the RB contour enhanced image into a plurality of image regions respectively, wherein the cutting method for the three images is the same; S52, calculating the abnormality of the image area on the RG contour enhanced image; S53, calculating the abnormality of the image area on the GB contour enhanced image; S54, calculating the abnormality of the image area on the RB contour enhanced image; S55, at the same image region position, when the abnormality of the image region on the RG contour enhanced image, the abnormality of the image region on the GB contour enhanced image, and the abnormality of the image region on the RB contour enhanced image are all greater than the abnormality threshold, the image region position is a defect position; The S52, S53 and S54 all include the following sub-steps: A1. On the contour enhanced image, count the frequency of each pixel value appearing in the image area and the frequency of each pixel value appearing in the contour enhanced image, and calculate the frequency anomaly coefficient; A2. On the contour enhanced image, the pixel distribution anomaly coefficient is calculated according to the distribution of pixel values ​​in each image region; A3, adding the frequency abnormality coefficient and the pixel distribution abnormality coefficient to obtain the abnormality degree of the image area on the contour enhanced image, wherein the contour enhanced image in A1, A2 and A3 is the RG contour enhanced image, the GB contour enhanced image or the RB contour enhanced image; The formula for calculating the frequency anomaly coefficient in A1 is: , where θ f is the frequency anomaly coefficient, f local,m is the frequency of the mth pixel value appearing in the image area, f whole,m is the frequency of the mth pixel value in the image region appearing on the edge enhanced image, m is a positive integer, and M is the number of pixel value types in the image region; The formula for calculating the pixel distribution anomaly coefficient in A2 is: , where θ p is the pixel distribution anomaly coefficient, P k is the pixel value of the kth pixel in the image area, K is the number of pixels in the image area, E is the area of ​​the image area, and k is a positive integer.

2. The method for real-time detection of nanoimprint wafer demoulding defects according to claim 1, characterized in that: The S2 is specifically as follows: subtracting the pixel values ​​of the initial image and the standard image at the same pixel position to obtain a pixel difference, and using the pixel difference as a new pixel value at the pixel position to obtain a difference image.

3. The method for real-time detection of nanoimprint wafer demoulding defects according to claim 1, characterized in that: The S3 comprises the following sub-steps: S31, decomposing the difference image into an R channel image, a G channel image and a B channel image; S32, taking each pixel point as the center on the R channel image, calculating the complexity of the R channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the R channel contour image; S33, taking each pixel point as the center on the G channel image, calculating the complexity of the G channel value within the neighborhood of the center, and when the complexity is less than the complexity threshold, removing the pixel point at the center to obtain a G channel contour image; S34, taking each pixel point as the center on the B channel image, calculating the complexity of the B channel value within the neighborhood range of the center, and when the complexity is less than the complexity threshold, removing the pixel point of the center to obtain the B channel contour image; S35 , fusing the R channel contour image, the G channel contour image, and the B channel contour image to obtain an RG contour fusion image, a GB contour fusion image, and a RB contour fusion image.

4. The method for real-time detection of nanoimprint wafer demoulding defects according to claim 3, characterized in that: The formula for calculating the complexity in S32, S33 and S34 is: , where γ is the complexity, x i The i-th R channel value, G channel value or B channel value within the neighborhood of the center, max{x i } is the maximum channel value in the neighborhood of the center, min{x i } is the minimum channel value in the neighborhood of the center, The channel mean in the neighborhood centered at .

5. The method for real-time detection of nanoimprint wafer demoulding defects according to claim 3, characterized in that: The S35 comprises the following sub-steps: S351, according to the R channel contour image and the G channel contour image, weighted processing is performed on the R channel value and the G channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RG contour fusion image; S352, according to the G channel contour image and the B channel contour image, weighted processing is performed on the G channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain a GB contour fusion image; S353, according to the R channel contour image and the B channel contour image, weighted processing is performed on the R channel value and the B channel value at the same pixel point position, and the weighted value is used as the new pixel value of the pixel point to obtain the RB contour fusion image.

6. The method for real-time detection of nanoimprint wafer demoulding defects according to claim 1, characterized in that: The pixel value enhancement formula in S4 is: ,in, is the pixel value of the nth pixel in the RG contour enhanced image, GB contour enhanced image or RB contour enhanced image after enhancement, P n is the pixel value of the nth pixel in the RG contour fusion image, GB contour fusion image or RB contour fusion image, where n is a positive integer.

Images