Mask mark detection method with occlusion, mask position correction method

Abstract

The disclosure provides a kind of shielded mask mark detection method, comprising: S1, the mask mark image collected is converted into gray scale chart;Wherein, the image collected is at least partially shielded;S2, gray scale chart is carried out image segmentation, and non-mark part is filled as white;S3, the gray scale median of gray scale chart is obtained, as initial binary threshold value the gray scale chart is carried out binary processing;Carry out outline detection, and obtain the multiple minimum circumscribed rectangle of image outline;S4, according to the long and wide features of mask mark minimum circumscribed rectangle, determine fuzzy inference rule, obtain the probability that each minimum circumscribed rectangle is actual shielded mark circumscribed rectangle;S5, whether there is probability greater than preset threshold in probability;If not, adaptive adjustment binary threshold value, repeat S3-S5;If yes, determine the minimum circumscribed rectangle of maximum probability, as first detection target;S6, according to first detection target, determine first center coordinates, complete the positioning detection of mask mark.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of optical detection, in particular to a mask mark detection method with shielding and a mask position correction method. BACKGROUND

[0002] During the mask transmission process, position correction is needed at the end. The CCD camera can collect the alignment marks on the mask through the transparent mask chuck to detect the position deviation, and the higher the accuracy of the detection result is, the better the correction effect is. Due to the airway or other structures of the chuck shielding the marks, the collected marks by the CCD camera are incomplete, part of the marks is shielded, and the gray scale in the collected image is unevenly distributed, which will affect the positioning accuracy of the detected marks.

[0003] The binarization method with appropriate threshold can reduce the interference information of the image, that is, under the premise of keeping the integrity of the collected marks, the binarization processing of the collected image can improve the positioning accuracy of the detected marks, but the appropriate binarization threshold is difficult to determine. After the binarization processing with different thresholds under the same image, the detected mark result will have obvious difference, which will further affect the positioning accuracy of the marks. SUMMARY

[0004] (I) Technical problems to be solved

[0005] In view of the above problems, the present disclosure provides a mask mark detection method with shielding and a mask position correction method to solve the technical problems that the traditional method is difficult to obtain an appropriate binarization threshold and improve the positioning accuracy of the detected marks.

[0006] (II) Technical solutions

[0007] In one aspect, the present disclosure provides a mask mark detection method with shielding, comprising: S1, collecting images of mask marks by using a first camera, and converting the collected images into a gray scale image; wherein the collected images are at least partially shielded; S2, performing image segmentation on the gray scale image, and filling the non-mark part as white; S3, obtaining the gray scale median of the gray scale image obtained in S2, and performing binarization processing on the gray scale image by taking the gray scale median as an initial binarization threshold; performing contour detection, and obtaining a plurality of minimum bounding rectangles of the image contour; S4, determining a fuzzy inference rule according to the length-width characteristics of the minimum bounding rectangles of the mask marks, and obtaining the probability that each minimum bounding rectangle obtained in S3 is the actual shielded mark bounding rectangle; S5, judging whether there is a probability greater than a preset threshold in the probability obtained in S4; if not, adaptively adjusting the binarization threshold, and repeating S3-S5; if yes, determining the minimum bounding rectangle with the maximum probability as the first detection target; S6, determining the first center coordinate according to the first detection target, and completing the positioning detection of the mask marks.

[0008] According to an embodiment of the present disclosure, S1 further comprises filtering the grayscale image to suppress noise.

[0009] According to an embodiment of the present disclosure, the image segmentation in S2 comprises selecting the overlapping points of the mask mark and the occluded part as the segmentation points according to the graphical features of the mask mark; and performing image segmentation on the grayscale image with the segmentation points as one end point or edge point of the segmentation rectangle.

[0010] According to an embodiment of the present disclosure, the method of performing contour detection in S3 comprises any one of a contour detection algorithm based on connectivity, a contour detection algorithm based on edge detection, and a contour detection algorithm based on segmentation; and the method of obtaining the plurality of minimum bounding rectangles of the image contour comprises calculating the plurality of minimum bounding rectangles by using a BoundingRect function.

[0011] According to an embodiment of the present disclosure, S4 comprises: S41, estimating the length and width of the bounding rectangle in the collected image; S42, dividing the pixel interval of the length into multiple segments according to the length and determining a length membership function; dividing the pixel interval of the width into multiple segments according to the width and determining a width membership function; S43, obtaining a fuzzy inference result according to the length membership function and the width membership function; S44, de-fuzzifying the fuzzy inference result by using a maximum membership degree method to obtain the probability that each minimum bounding rectangle is an actual occluded mark bounding rectangle.

[0012] According to an embodiment of the present disclosure, the length and width of the bounding rectangle are estimated according to the following formula in S41:

[0013] l 图像 =l 实际 ×n÷Pix 像元

[0014] wherein, l 图像 represents the side length in the image, l 实际 represents the actual side length, n represents the magnification of the camera lens, and Pix 像元 represents the pixel size.

[0015] According to an embodiment of the present disclosure, the length membership function and the width membership function in S42 are one of a triangular membership function, a trapezoidal membership function, and a piecewise linear membership function.

[0016] According to an embodiment of the present disclosure, S5 comprises adaptively adjusting the binary threshold value according to the following formula:

[0017] V self-adaption =V Threshold ±f%×V Threshold ×m

[0018] wherein, VThreshold V is a current binarization threshold self-adaption V is an adaptive binarization threshold, m is the number of adaptive adjustment, f% is the adjustment amplitude, the smaller f is, the more accurate the adaptive binarization threshold is, and the longer the algorithm takes.

[0019] Another aspect of the present disclosure provides a mask position correction method, comprising: obtaining a first center coordinate of a first detection target according to the above-mentioned obscured mask mark detection method; S7, repeating S1-S6 using a second camera to obtain a second center coordinate of a second detection target; S8, obtaining X, Y direction deviation and deflection angle of the mask upper piece according to the first center coordinate, the second center coordinate and a third center coordinate of a calibration pattern; S9, correcting the position of the mask according to the X, Y direction deviation and the deflection angle.

[0020] According to the embodiment of the present disclosure, S8 comprises: calculating the deflection angle θ of the mask upper piece according to the following formula:

[0021]

[0022] Wherein, R is the distance between two mask marks, and L is the difference between the distance of the two detection target centers and the distance of the calibration pattern center.

[0023] (Three) beneficial effects

[0024] The obscured mask mark detection method and the mask position correction method of the present disclosure use a single camera to detect the obscured marks on the mask, determine the fuzzy inference rule according to the length-width features of the detected binary images of the marks, and take the inference result as the termination condition of the threshold optimization process of the adaptive threshold image binarization; the suitable binarization threshold is obtained in a result-oriented manner, which reduces the image interference information and improves the positioning accuracy of the detected marks. On this basis, the distance between the detected mark pattern center coordinate and the calibration pattern center coordinate is calculated by collecting two alignment marks, the X, Y direction deviation and the deflection angle θ of the mask upper piece are calculated, and the accurate data required for the position correction of the mask is obtained. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The flowchart of the obscured mask mark detection method according to the embodiment of the present disclosure is schematically shown;

[0026] Figure 2 The schematic diagram before and after the image segmentation of the gray-scale image according to the embodiment of the present disclosure is schematically shown;

[0027] Figure 3 The result schematic diagram of the binarization processing of the gray-scale image according to the embodiment of the present disclosure is schematically shown;

[0028] Figure 4 Fig. 12 schematically shows a result diagram of contour detection according to an embodiment of the present disclosure;

[0029] Figure 5 Fig. 13 schematically shows a result diagram of a minimum bounding rectangle group of image contours according to an embodiment of the present disclosure;

[0030] Figure 6 Fig. 14 schematically shows a diagram of detecting a label input / output membership function according to an embodiment of the present disclosure;

[0031] Figure 7 Fig. 15 schematically shows a result diagram of a deblurring plane of a label probability according to an embodiment of the present disclosure;

[0032] Figure 8 Fig. 16 schematically shows a diagram of a detection result after fuzzy reasoning according to an embodiment of the present disclosure;

[0033] Figure 9 Fig. 17 schematically shows a diagram of each stage of target detection according to an embodiment of the present disclosure;

[0034] Figure 10 Fig. 18 schematically shows a diagram of a CCD camera image and a detection target image of two labels according to an embodiment of the present disclosure;

[0035] Figure 11 Fig. 19 schematically shows a partial flowchart of a correction method of a mask position according to an embodiment of the present disclosure;

[0036] Figure 12 Fig. 20 schematically shows a diagram of a detection device according to an embodiment of the present disclosure;

[0037] Figure 13 Fig. 21 schematically shows a diagram of a principle of calculating a deflection angle according to an embodiment of the present disclosure;

[0038] Figure 14 Fig. 22 schematically shows a diagram of a data instance of a deflection angle calculation according to an embodiment of the present disclosure;

[0039] Figure 15 Fig. 23 schematically shows a diagram of a real-time label detection result according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0040] In order to make the objects, technical solutions and advantages of the present disclosure clearer, the present disclosure will be further described in detail below with reference to specific embodiments and drawings.

[0041] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the terms "comprises", "comprising", "includes", "including" and the like are specifically intended to be open-ended and to mean that other features, steps, operations, and / or components can be added.

[0042] It should be noted that if the disclosure embodiments involve directionality indication, the directionality indication is only used to explain the relative position relationship, motion condition and the like between the components in a certain specific posture, and if the specific posture changes, the directionality indication also changes accordingly.

[0043] The words such as "first", "second", "third" and the like used in the specification and claims are used to modify the corresponding elements, and do not inherently mean or represent that the elements have any order or manufacturing method, and the use of these ordinal numbers is only used to distinguish the elements with the same name from another element with the same name.

[0044] In the process of determining the position information of the mark, first, the CCD camera is used to detect the blocked mark on the mask. For the blocked part in the collected image, the mark image collected by the CCD camera can be graphically segmented, and then the non-mark part is filled with white color to ensure the success rate of mark position detection. However, there are usually unevenly distributed shadow areas in the image collected by the CCD camera, and small residual areas of non-mark part are easily produced after image segmentation. The method of appropriate threshold binaryzation can reduce the interference information of the image and improve the positioning accuracy of the detected mark, but it is difficult to determine the appropriate binaryzation threshold, and the result of adaptive threshold image binaryzation is different with different termination conditions, which all lead to the fact that the mark positioning cannot be performed through the binaryzation image.

[0045] Based on this, the disclosure provides a method for detecting a blocked mask mark, please see Figure 1, comprising: S1, image acquisition of the mask mark by using the first camera, and converting the acquired image into a gray scale image; wherein the acquired image is at least partially blocked; S2, image segmentation of the gray scale image, and filling the non-mark part as white; S3, obtaining the gray scale median of the gray scale image obtained in S2, and taking the gray scale median as an initial binary threshold to perform binary processing on the gray scale image; performing contour detection, and obtaining a plurality of minimum bounding rectangles of the image contour; S4, determining a fuzzy inference rule according to the length-width characteristics of the minimum bounding rectangle of the mask mark, and obtaining the probability that each minimum bounding rectangle obtained in S3 is the actual bounding rectangle of the blocked mark; S5, judging whether there is a probability greater than a preset threshold in the probability obtained in S4; if not, adaptively adjusting the binary threshold, and repeating S3-S5; if yes, determining the minimum bounding rectangle with the maximum probability as the first detection target; S6, determining the first center coordinate according to the first detection target, and completing the positioning detection of the mask mark.

[0046] The method detects the blocked mark on the mask by using a single camera, determines a fuzzy inference rule according to the length-width characteristics of the detected mark image after binary, and takes the inference result as the termination condition of the threshold optimization process of the adaptive threshold image binary, so that the suitable binary threshold is obtained in a result-oriented manner, the image interference information is reduced, and the positioning accuracy of the detected mark is improved.

[0047] On the basis of the above embodiment, S1 further comprises: filtering the gray scale image to suppress noise.

[0048] The image acquired by the CCD camera is converted into a gray scale image and filtered, so that the image noise is suppressed while the image detail characteristics are preserved as much as possible.

[0049] On the basis of the above embodiment, the image segmentation in S2 comprises: selecting the overlapping points of the mask mark and the blocked part as the segmentation points according to the graphic characteristics of the mask mark; and performing image segmentation on the gray scale image by taking the segmentation points as one end point or edge point of the segmentation rectangle.

[0050] The acquired blocked mark image is segmented, and the non-mark part is filled as white. According to the graphic characteristics of the mark, the selection of the image segmentation points should preserve the mark characteristics as much as possible to ensure the success rate of mark detection, for example, as shown in the drawing, the overlapping points of the mark and the blocked part are selected as one end point or edge point of the segmentation rectangle, which can maximize the preservation of the mark characteristics. Figure 2

[0051] ​On the basis of the above embodiment, the method of profile detection in S3 includes any one of the profile detection algorithm based on connectivity, the profile detection algorithm based on edge detection, and the profile detection algorithm based on segmentation; and the method of obtaining the plurality of minimum circumscribed rectangles of the image profile includes calculating the plurality of minimum circumscribed rectangles by using the BoundingRect function.

[0052] The median value of the gray scale of the gray scale image is obtained by traversing the gray scale, and the median value is the initial value of the binarization threshold. After binarization, the marked image is as shown in Figure 3 The profile detection is then performed, as shown in Figure 4 The method of profile detection is to use the Opencv algorithm. After the profile detection, the BoundingRect method in OpenCv is used to obtain the minimum circumscribed rectangle (Rectangle) of the image profile. These rectangles are the targets obtained by detection, as shown in Figure 5 As can be seen from Figure 5 , due to the non-uniformity of the collected image position and gray scale distribution, and the residual non-marked part after image segmentation, multiple minimum circumscribed rectangles of non-marked parts may be generated.

[0053] On the basis of the above embodiment, S4 includes: S41, estimating the length and width of the circumscribed rectangle in the collected image; S42, dividing the length pixel interval into multiple segments according to the length and determining the length membership function; dividing the width pixel interval into multiple segments according to the width and determining the width membership function; S43, obtaining the fuzzy inference result according to the length membership function and the width membership function; S44, using the maximum membership degree method to de-fuzzify the fuzzy inference result to obtain the probability that each minimum circumscribed rectangle is the actual occluded marked circumscribed rectangle.

[0054] According to the length and width characteristics of the minimum circumscribed rectangle of the marker in the CCD image, the fuzzy inference rule is designed. The classical logic (Boolean logic) insists that all things can be represented by binary items (0 or 1, black or white, yes or no), while the logic of fuzzy inference is to replace the Boolean value with membership degree, that is, the value between 0 and 1 can be used to represent the relationship value between members.

[0055] The fuzzy inference system has two input fuzzy sets {length, width}, the former represents the length of the minimum bounding rectangle, and the latter represents the width. There is an output {Rectangle}, which represents the fuzzy inference output of the labeled minimum bounding rectangle, and its physical meaning represents the similarity between the minimum bounding rectangle detected in the previous two steps and the actual bounding rectangle of the occluded label. The fuzzy subsets of length are {short, normal, long}; the fuzzy subsets of width are {narrow, general, wide}, and the two fuzzy subsets correspond to a membership function, which is the membership degree of each member in the fuzzy subset, such as short, normal, and long, within its value range, which is between 0 and 1. The membership function can take any form, commonly triangular, trapezoidal, piecewise linear, etc., which can be selected according to experience, for example, a trapezoidal membership function is selected as shown in Figure 6 The fuzzy subsets of Rectangle are {small, medium, large}, which represent the similarity between the minimum bounding rectangle detected in the previous two steps and the actual bounding rectangle of the label when the corresponding input is {length, width}.

[0056] For example, with the selected detection label input / output membership function as shown in Figure 6 The fuzzy inference rules are shown in Table 1. The membership functions of the fuzzy subsets of length and width are determined using the same rule, and only the determination method of the length membership function is introduced here.

[0057] Table 1 Fuzzy Inference Rule Table

[0058]

[0059] S41, estimate the side length (unit: pixel) of the bounding rectangle of the label in the image after image segmentation in the CCD imaging. The length and width of the bounding rectangle are estimated according to the following formula:

[0060] l 图像 = l 实际 × n ÷ Pix 像元 (1)

[0061] where l 图像 represents the side length in the image, l 实际 represents the actual side length, n represents the magnification of the camera lens, and Pix 像元 represents the pixel size.

[0062] S42, since the length of the circumscribed rectangle marked in the image imaged by the CCD is about 1000 pixels, the pixel interval of length is divided into five segments (0, 600], [600, 900], [900, 1100], [1100, 1300], [1300, 1900). The vertical coordinate membership degree is used to describe the degree to which an object belongs to a certain definition, such as Figure 6 In (a), when the length of length is less than 600 pixels, it belongs to the degree of short 1, indicating that the length of the smallest circumscribed rectangle in the CCD image is less than 600 pixels, and the length belongs to the degree of short 1. The length of the smallest circumscribed rectangle is too short, and is not the smallest circumscribed rectangle corresponding to the mark; when the length of length is between 900-1100, it belongs to the degree of normal 1; when the length of length is greater than 1300 pixels, it belongs to the degree of long 1. In the pixel interval of [600, 900], [1100, 1300], since there are two membership degrees, the maximum membership degree function is selected, that is, the maximum membership degree value is selected as the membership degree corresponding to the pixel.

[0063] S43-S44, the output of the fuzzy system is the union of the reasoning results of each rule, and the maximum membership degree method is used for defuzzification, which converts the fuzzy reasoning results (small, medium, large) into accurate data. The defuzzification plane is shown in Figure 7 , and the probability P Rectangle that the smallest circumscribed rectangle in the image contour smallest circumscribed rectangle group is the actual occluded mark circumscribed rectangle is obtained.

[0064] A preset threshold of the probability P Rectangle is set, which is generally set according to the accuracy of the image information to be retained. If only the image information completely consistent with the reasoning result needs to be retained, the preset threshold can be set larger; if more image information related to the reasoning result needs to be retained, the preset threshold can be set smaller, for example, the threshold of the selected probability P Rectangle is 0.75. The image contour smallest circumscribed rectangle with the largest probability greater than the threshold is retained, and the loop is ended. Rectangle

[0065] On the basis of the above embodiment, S5 comprises: adaptively adjusting the binary threshold value according to the following formula:

[0066] V self-adaption = V Threshold ± f% × V Threshold × m (2)

[0067] Where V Threshold is the current binary threshold value, V​self-adaption is the adaptively adjusted binarization threshold, m is the number of adaptive adjustments, and f% is the adjustment amplitude. The smaller f is, the more accurate the adaptively adjusted binarization threshold is, and the longer the algorithm takes, for example, f% is 2%.

[0068] If the maximum P of the minimum bounding rectangle of the image contour Rectangle Does not meet the probability P Rectangle The preset threshold value of Figure 8 As shown, for example, the probability of detecting the target as marked in the figure is 0.13, which is less than the set probability P Rectangle The detected marker image is incomplete. This is because the selected binary value is too large or too small, resulting in too many or too few retained image features. Threshold In this case, the binarization threshold is adaptively adjusted according to the formula, and steps S3 to S5 are repeated until a probability P is satisfied. Rectangle For example, find the minimum bounding rectangle of the image contour that satisfies the probability P Rectangle The image of the minimum enclosing rectangle of the image contour is as follows Figure 9 shown.

[0069] The formula selection of "+" or "-" in a certain cycle is based on the fuzzy reasoning result after the binary threshold is selected. For example, if "+" is selected first in the first optimization, if the maximum P Rectanggle Increases, it means that increasing the binarization threshold can increase the maximum P Rectanggle , in the next cycle, you can still use "+"; at the maximum P Rectanggle Greater than probability P Rectanggle Before the threshold, if in a certain cycle, after using "+" to update the binary threshold, the maximum P in the next cycle Rectanggle decreases, it means that increasing the binarization threshold will lead to the maximum P Rectanggle Decrease, then you need to use "-" to update the binarization threshold.

[0070] For example, after multiple cycles, when the fuzzy reasoning obtains the maximum P Rectanggle is 0.856, which is greater than the probability P Rectanggle The preset threshold, the binary image of the occluded mark, the contour detection image, the image contour minimum bounding rectangle group and the occluded mark image contour minimum bounding rectangle (detection target) obtained by fuzzy inference are as follows Figure 9 As shown in the figure, the contour detection results of the two groups of markers are as follows Figure 10 shown.

[0071] The center coordinates of the minimum circumscribed rectangle (detection target) of the occluded marker image outline are calculated according to the following formula.

[0072]

[0073] P x 、P y respectively represent the x coordinate and y coordinate of the center of the minimum circumscribed rectangle of the blocked marker image profile collected by the CCD camera (detection target), x Rectanggle-TopLeft , y Rectanggle-TopLeft respectively represent the x coordinate and y coordinate of the upper left corner of the detection target obtained by the fuzzy inference process, W Rectangle , L Rectangle represent the x and y side lengths of the detection target, thereby achieving accurate positioning of the mask marker.

[0074] On the basis of the above, the disclosure also provides a mask position correction method, please see Figure 11 , comprising: obtaining the first center coordinates of the first detection target according to the above blocked mask marker detection method; S7, repeating S1-S6 using a second camera to obtain the second center coordinates of the second detection target; S8, obtaining the X and Y direction deviations and the deflection angle of the mask upper piece according to the first center coordinates, the second center coordinates and the third center coordinates of the calibration pattern; S9, correcting the position of the mask according to the X and Y direction deviations and the deflection angle.

[0075] The method uses two sets of CCD devices to collect images for marker detection. The detection device includes a mirror group, an adapter mirror, a light source and a CCD camera. The detection device schematic diagram is shown in Figure 12 . The two sets of CCD devices are used for mask upper piece deviation detection and position correction. When performing image processing and positioning of the blocked mask marker, a single detection device can complete the image processing and positioning function of the blocked mask marker.

[0076] Specifically, according to the mask correction accuracy requirement, two sets of alignment marker positions are designed. According to the arctangent function (i.e. the following formula (4)), when the distance between the two mask markers is farther (R is larger), the angle that can be calculated is smaller (θ is smaller), the detection accuracy is higher, and the subsequent effect of using the marker to correct the mask position is better.

[0077]

[0078] Wherein, R is the distance between the two mask markers, L is the difference between the distance of the two detection target centers and the center of the calibration pattern, the y coordinate difference between one detection target center and the center of the calibration pattern is L1, and the y coordinate difference between the other detection target center and the center of the calibration pattern is L2, L is the difference between L2 and L1, as shown in Figure 14 . θ is the deflection angle of the detection image deviating from the calibration image.

[0079] The calculation of the turning angle is shown in the schematic diagram Figure 13The mask is installed according to the parallelism of the X / Y direction, the CCD camera collects images for calibration, OP1 and OP2 are the centers of the two sets of marker calibration patterns, the center refers to the detection contour center of the calibration image, which can be obtained by formula (3) and the like, P1 and P2 are the detection target center coordinates of the marker image collected by the CCD camera after the manipulator transmits the mask to the station x , P y )

[0080] According to formula (5), L1 is the y coordinate difference between P1 and OP1, and L2 is the y coordinate difference between P2 and OP2.

[0081]

[0082] Taking the detection contour center line OP1-OP2 of the calibration image as the reference, first correct the position of the detection target center line P1-P2 in the x direction. For example, in the above figure, OP1 is the correction origin, and the detection target center line P1-P2 is moved in the x direction by -ΔX, and the calculation formula is as follows:

[0083]

[0084] After moving, the x coordinate of P1 is the same as that of OP1, and the moved figure is shown in Figure 14 At this time, the deflection angle θ of the line OP1-OP2 and the line P1-P2 is calculated:

[0085]

[0086] The calculation of θ can be summarized as the following six cases, and the angle calculation data example is shown in Figure 14

[0087] After calculating the deflection angle θ and correcting the deflection angle with P1 as the origin, the line OP1-OP2 and the line P1-P2 are parallel to each other, and then the detection target center line P1-P2 is moved in the y direction by -ΔY, and the mask position correction is completed.

[0088]

[0089] Please note that here is for the convenience of understanding, so OP1 is the correction origin, first move P1-P2 in the x direction, then correct the deflection angle, and finally move in the y direction, and those skilled in the art can also obtain the detection contour center of the calibration image and the detection target center coordinates of the marker image collected by the CCD camera (P x , P y ​), or adjusting the order of correction, and the specific calculation formula can be derived according to the above process, which will not be described here.

[0090] The present disclosure first detects the position coordinates of the blocked mask mark on the mask through a single camera, and then detects the mask position deviation by using the position coordinates of the detected mask mark on the mask through two cameras, so that the deflection angle of the mask on the mask can be obtained through the inverse trigonometric function, thereby improving the accuracy of the mask position correction.

[0091] The present disclosure will be further described through specific embodiments. In the following embodiments, the above-mentioned blocked mask mark detection method and mask position correction method are specifically described. However, the following embodiments are only used to illustrate the present disclosure, and the scope of the present disclosure is not limited thereto.

[0092] The blocked mask mark detection method and mask position correction method of the present disclosure, as shown in Figures 2 to 15 , include the following steps executed in sequence:

[0093] Step 1: The image collected by the single camera is converted into a gray-scale image, and filtering is performed to suppress image noise while retaining image details as much as possible. According to the collected image of the blocked mark, image segmentation is performed, and the non-mark part is filled with white, as shown in Figure 2 , which corresponds to the above steps S1-S2.

[0094] Step 2: Traverse the gray scale of the gray-scale image to obtain the gray scale median for binarization. The gray scale median is the initial value of the binarization threshold. After binarization, the mark image is as shown in Figure 3 . Then, contour detection is performed, and the result is as shown in Figure 4 . The BoundingRect method in OpenCv is used to obtain the minimum circumscribed rectangle (Rectangle) of the image contour. These rectangles are the detected targets, as shown in Figure 5 . This corresponds to the above step S3.

[0095] Step 3: The input / output membership function of the mark detection is as shown in Figure 6 , and the fuzzy inference rule table is as shown in Table 1. In this embodiment, the membership functions of the fuzzy subsets of length and width are determined using the same rule. The output of the fuzzy system is the union of the inference results of each rule. The maximum membership degree method is used for defuzzification to convert the fuzzy inference results (small, medium, large) into precise data. The defuzzification plane is as shown in Figure 7 , and the probability P that a certain minimum circumscribed rectangle in the minimum circumscribed rectangle group of the image contour is the actual circumscribed rectangle of the blocked mark is obtained. Rectangle. Equivalent to the above step S4.

[0096] Step 4: Determine whether there is a probability greater than a preset threshold among the probabilities obtained in S4; if not, adaptively adjust the binarization threshold and repeat S3 to S5; if so, determine the minimum bounding rectangle with the highest probability as the first detection target. In this embodiment, the maximum P of the initial minimum bounding rectangle is Rectangle Does not meet the probability P Rectangle The preset threshold is 0.75, such as Figure 8 As shown in the figure, the probability of detecting the target being marked is 0.13, which is less than the set probability P Rectangle After multiple cycles, fuzzy reasoning obtains the maximum P Rectangle is 0.856, which is greater than the probability P Rectangle The preset threshold is 0.75, the binary image of the occluded mark, the contour detection image, the minimum bounding rectangle group of the image contour and the minimum bounding rectangle of the occluded mark image contour obtained by fuzzy inference (detection target) are as follows Figure 9 This is equivalent to the above step S5.

[0097] Step 5: Determine the first center coordinates based on the first detection target to complete the positioning detection of the mask mark, which is equivalent to the above step S6.

[0098] Step 6: According to the mask correction accuracy requirements, design two sets of alignment mark positions. The minimum bounding rectangle detection results of the two sets of marks are as follows: Figure 10 As shown. Get the center coordinates of the two sets of detection targets. The detection results are as follows Figure 15 As shown, the dashed rectangle represents the detection contour of the calibration pattern, and the solid rectangle represents the detection contour of the real-time image captured by the CCD camera. Assuming the two markers are 120 mm apart and each pixel is approximately 2.2 μm, the pre-alignment algorithm can achieve a detection accuracy of 0.001°. This is equivalent to step S7 above.

[0099] Step 7: Calculate the X and Y deviations and deflection angles of the mask sheet based on the center coordinates of the two sets of detection targets and the center coordinates of the calibration pattern. This is equivalent to the above step S8.

[0100] Step 8: Move the mask according to the X and Y deviations and the deflection angle to complete the correction of the mask position. This is equivalent to the above step S9.

[0101] The present invention uses a single camera to determine the accurate position information of the mark, thereby achieving high-precision detection of the mask mark; further, two cameras are used for detection to calculate the X and Y deviations and deflection angle θ of the mask sheet, and obtain accurate data for the position correction required for the mask.

[0102] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present disclosure, and it should be understood that the above-described specific embodiments are merely specific embodiments of the present disclosure and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for detecting mask marks with occlusion, characterized in that: include: S1, using a first camera to capture an image of the mask mark, and converting the captured image into a grayscale image; wherein the captured image is at least partially obscured; S2, performing image segmentation on the grayscale image and filling the non-marked parts with white; S3, obtaining the grayscale median value of the grayscale image obtained in S2, and binarizing the grayscale image using the grayscale median value as an initial binarization threshold; performing contour detection, and obtaining multiple minimum circumscribed rectangles of the image contour; S4, determining a fuzzy inference rule based on the length and width features of the minimum bounding rectangle of the mask mark, and obtaining a probability that each minimum bounding rectangle obtained in S3 is the actual bounding rectangle of the mask mark; S5, determining whether there is a probability greater than a preset threshold among the probabilities obtained in S4; if not, adaptively adjusting the binarization threshold and repeating S3 to S5; if so, determining the minimum bounding rectangle with the highest probability as the first detection target; S6, determining first center coordinates according to the first detection target, and completing positioning detection of the mask mark; The image segmentation in S2 includes: According to the graphic features of the mask mark, the overlapping point of the mask mark and the blocked part is selected as the segmentation point; Performing image segmentation on the grayscale image using the segmentation point as an end point or an edge point of a segmentation rectangle; The S4 includes: S41, estimating the length and width of the circumscribed rectangle in the acquired image; S42, dividing the pixel interval of the length into multiple segments according to the length, and determining a length membership function; dividing the pixel interval of the width into multiple segments according to the width, and determining a width membership function; S43, obtaining a fuzzy reasoning result according to the length membership function and the width membership function.

2. The method for detecting mask marks with occlusion according to claim 1, wherein: Said S1 further comprises: The grayscale image is filtered to suppress noise.

3. The method for detecting mask marks with occlusion according to claim 1, wherein: The method for performing contour detection in S3 includes any one of a connectivity-based contour detection algorithm, an edge detection-based contour detection algorithm, and a segmentation-based contour detection algorithm; The method for obtaining multiple minimum bounding rectangles of the image outline includes using a BoundingRect function to calculate and obtain multiple minimum bounding rectangles.

4. The method for detecting mask marks with occlusion according to claim 1, wherein: Said S4 further comprises: S44, defuzzifying the fuzzy inference result using a maximum membership method to obtain a probability that each minimum bounding rectangle is the actual bounding rectangle of the obscured mark.

5. The method for detecting mask marks with occlusion according to claim 1, wherein: In S41, the length and width of the circumscribed rectangle are estimated according to the following formula: in, represents the edge length in the image, represents the actual side length, n represents the magnification of the camera lens, Indicates the pixel size.

6. The method for detecting mask marks with occlusion according to claim 1, wherein: The length membership function and the width membership function in S42 are respectively one of a triangle membership function, a trapezoidal membership function, and a piecewise linear membership function.

7. The method for detecting mask marks with occlusion according to claim 1, wherein: The step S5 includes: adaptively adjusting the binarization threshold according to the following formula: in, is the current binarization threshold, is the adaptively adjusted binarization threshold, is the number of adaptive adjustments, To adjust the amplitude.

8. A method for correcting a mask position, characterized in that: include: Obtaining a first center coordinate of a first detection target according to the mask mark detection method with occlusion according to any one of claims 1 to 6; S7, repeating S1 to S6 using a second camera to obtain a second center coordinate of a second detection target; S8, obtaining the X and Y deviations and the deflection angle of the mask sheet according to the first center coordinate, the second center coordinate, and the third center coordinate of the calibration pattern; S9, correcting the position of the mask according to the X and Y deviations and the deflection angle.

9. The method for correcting the mask position according to claim 8, wherein: The step S8 includes: calculating the deflection angle of the mask sheet according to the following formula: : Where R is the distance between the two mask marks, and L is the difference between the distances between the centers of the two detection targets and the center of the calibration pattern.

Images