Graph correction method, storage medium and storage terminal

Abstract

The invention discloses a graph correction method, a storage medium and a storage terminal, and the method comprises the steps: providing a test layout which comprises a plurality of test graphs; based on the test layout, a convolutional neural network model is obtained, the convolutional neural network model comprises first environment distribution matrixes of a plurality of first line segments around the test layout and first offsets of the plurality of first line segments after optical proximity effect correction of the test layout, the plurality of first environment distribution matrixes are in one-to-one correspondence with the plurality of first offsets, and the plurality of first offsets are in one-to-one correspondence with the plurality of first line segments; the plurality of first line segments are obtained by segmenting the contour of the test pattern; providing a to-be-corrected layout which comprises a plurality of to-be-corrected graphs; cutting the contour of the to-be-corrected graph into a plurality of second line segments; based on the to-be-corrected layout, obtaining a second environment distribution matrix around each second line segment; taking the second environment distribution matrix as an input value, and obtaining a second offset of a second line segment according to the convolutional neural network model; and obtaining a corrected layout based on the second offset of each second line segment. The process of obtaining the corrected layout is simplified, and the accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a graphic correction method, a storage medium and a storage terminal. Background Art

[0002] The pattern on the mask is projected onto the photoresist through the exposure system. Due to the diffraction limitation of the optical system, the pattern on the photoresist will be inconsistent with the pattern on the mask. Optical Proximity Correction (OPC) is to correct the pattern on the mask through calculation so that the final photoresist pattern meets the design requirements.

[0003] Model-based optical proximity correction is computationally expensive, and the required runtime increases rapidly as technology nodes shrink. This is because model-based optical proximity correction requires more iterations to meet smaller edge placement errors, while lithography simulations take longer to simulate more complex light fields.

[0004] Therefore, it is necessary to optimize the model-based optical proximity effect correction process. Summary of the invention

[0005] The technical problem solved by the present invention is to provide a graphic correction method, a storage medium and a storage terminal to optimize the optical proximity effect correction process based on a model.

[0006] In order to solve the above technical problems, the technical solution of the present invention provides a method for correcting graphics, comprising: providing a test layout, the test layout comprising a plurality of test graphics; based on the test layout, obtaining a convolutional neural network model, the convolutional neural network model comprising: a first environment distribution matrix around a plurality of first line segments on the test layout, and a first offset of a plurality of first line segments of the test layout after optical proximity effect correction, the plurality of first environment distribution matrices corresponding to the plurality of first offsets one by one, and the plurality of first line segments being obtained by segmenting the contour of the test graphics; providing a layout to be corrected, the layout to be corrected comprising a plurality of graphics to be corrected; segmenting the contour of the graphics to be corrected so that the contour of the graphics to be corrected is segmented into a plurality of second line segments; based on the layout to be corrected, obtaining a second environment distribution matrix around each second line segment; taking the second environment distribution matrix as an input value, according to the convolutional neural network model, obtaining a second offset of the second line segment; based on the second offset of each second line segment, obtaining a correction layout, the correction layout comprising a plurality of correction graphics, the correction graphics corresponding to the graphics to be corrected one by one.

[0007] Optionally, based on the test layout, a convolutional neural network model is obtained, including: segmenting the outline of the test pattern into a plurality of first line segments; based on the test layout, obtaining a first environmental distribution matrix around each first line segment; performing optical proximity effect correction on the test layout to obtain a first offset of each first line segment, wherein the first environmental distribution matrix around each first line segment corresponds one-to-one to the first offset of each first line segment; providing an initial convolutional neural network model; and training the initial convolutional neural network model using the first environmental distribution matrix of each first line segment as an input value and the first offset of each first line segment as an output value to obtain the convolutional neural network model.

[0008] Optionally, based on the test layout, a first environment distribution matrix around each first line segment is obtained, including: generating a rectangular area with a preset side length on the test layout with the midpoint of the first line segment as the center; setting a number of sampling points distributed in an array within the rectangular area; marking the sampling points according to whether there are graphics at the locations of the sampling points; and generating the first environment distribution matrix according to the markings of the sampling points.

[0009] Optionally, a plurality of sampling points distributed in an array are set in the rectangular area, including: generating sub-rectangular areas with different preset side lengths in the rectangular area with the midpoint of the first line segment as the center; generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas; connecting the sampling points of the sub-rectangular areas horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extension lines and vertical extension lines; generating side sampling points at positions where the horizontal and vertical extension lines intersect with the side lengths of the rectangular area and the sub-rectangular area, and the plurality of vertex sampling points and the plurality of side sampling points are distributed in an array.

[0010] Optionally, the shape of the rectangular area includes a square; the shape of the sub-rectangular area includes a square.

[0011] Optionally, the sampling point is marked according to whether there is a graphic at the location of the sampling point, including: if there is a graphic at the location of the sampling point, it is marked as "1"; if there is no graphic at the location of the sampling point, it is marked as "0".

[0012] Optionally, the distribution of the sampling points in the rectangular area includes: the closer to the midpoint of the first line segment, the greater the density of the sampling points; the farther from the midpoint of the first line segment, the smaller the density of the sampling points.

[0013] Optionally, the test layout is corrected for an optical proximity effect to obtain a first offset of each first line segment, including: performing simulated exposure on the test layout to obtain a first simulated exposure layout; obtaining an edge placement error between the first simulated exposure layout and a target layout; moving each of the first line segments according to the edge placement error to obtain a second test layout; performing simulated exposure on the second test layout to obtain a second simulated exposure layout; obtaining an edge placement error between the second simulated exposure layout and the target layout; if the edge placement error satisfies a preset range, obtaining a corrected test layout to obtain a first offset of each first line segment in the corrected test layout compared to the first line segment in the test layout; if the edge placement error does not satisfy the preset range, continuing to perform simulated exposure on the corrected test layout until the edge placement error satisfies the preset range.

[0014] Optionally, the convolutional neural network model includes a convolutional attention module, and the convolutional attention module includes a spatial attention module and a channel attention module.

[0015] Optionally, based on the layout to be corrected, a second environment distribution matrix around each second line segment is obtained, including: generating a rectangular area with a preset side length on the layout to be corrected with the midpoint of the second line segment as the center; setting a number of sampling points distributed in an array within the rectangular area; marking the sampling points according to whether there are graphics at the locations of the sampling points; and generating the second environment distribution matrix according to the markings of the sampling points.

[0016] Optionally, a plurality of sampling points distributed in an array are set in the rectangular area, including: taking the midpoint of the second line segment as the center, generating sub-rectangular areas with different preset side lengths in the rectangular area; generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas; connecting the sampling points of the sub-rectangular areas horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extension lines and vertical extension lines; generating side sampling points at positions where the horizontal extension lines and vertical extension lines intersect with the side lengths of the rectangular area and the sub-rectangular area, and the plurality of vertex sampling points and the plurality of side sampling points are distributed in an array

[0017] Optionally, the shape of the rectangular area includes a square; the shape of the sub-rectangular area includes a square.

[0018] Optionally, the sampling point is marked according to whether there is a graphic at the location of the sampling point, including: if there is a graphic at the location of the sampling point, it is marked as "1"; if there is no graphic at the location of the sampling point, it is marked as "0".

[0019] Optionally, the distribution of the sampling points in the rectangular area includes: the closer to the midpoint of the second line segment, the greater the density of the sampling points; the farther from the midpoint of the second line segment, the smaller the density of the sampling points.

[0020] Optionally, based on the second offset of each second line segment, a revised layout is obtained, including: according to the second offsets of the second line segments, several second line segments of several graphics to be corrected are moved to obtain the revised layout, wherein the revised layout includes several revised graphics, and the revised graphics correspond one-to-one to the graphics to be corrected.

[0021] Correspondingly, the technical solution of the present invention also provides a storage medium on which computer instructions are stored, and the steps of the above method are executed when the computer instructions are executed.

[0022] Correspondingly, the technical solution of the present invention also provides a storage terminal, including a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, and the processor executes the steps of the above method when executing the computer instructions.

[0023] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0024] The graphic correction method of the present invention obtains the second offset of the second line segment based on the second environment distribution matrix of the second line segment according to the convolutional neural network model, and then moves the second line segments of the graphic to be corrected according to the second offset to obtain the correction map. The process can simplify the process of obtaining the second offset, can meet the requirements of smaller edge prevention errors, reduce calculation costs, improve accuracy, and improve the efficiency of optical proximity effect correction.

[0025] Furthermore, the convolutional neural network model includes a convolutional attention module, which can enhance the accuracy of the convolutional neural network model's output values. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figures 1 to 7 is a schematic flow chart of a layout correction method in an embodiment of the present invention;

[0027] Figures 8 to 17 is a structural schematic diagram of a graphics correction process in an embodiment of the present invention;

[0028] Figures 18 to 20 It is a schematic diagram of the structure of the convolutional neural network attention module in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned objects, features and beneficial effects of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0030] Figures 1 to 7 is a schematic flow chart of a layout correction method in an embodiment of the present invention; Figures 8 to 17 is a structural schematic diagram of a graphics correction process in an embodiment of the present invention; Figures 18 to 20 It is a schematic diagram of the structure of the convolutional neural network attention module in an embodiment of the present invention.

[0031] Please refer to Figure 1 , the graphics correction method comprises:

[0032] Step S1: providing a test layout, wherein the test layout includes a plurality of test patterns;

[0033] Step S2: Based on the test layout, a convolutional neural network model is obtained, wherein the convolutional neural network model includes: a first environment distribution matrix around a plurality of first line segments on the test layout, and first offsets of the plurality of first line segments of the test layout after correction of the optical proximity effect, wherein the plurality of first environment distribution matrices correspond to the plurality of first offsets in a one-to-one manner, and the plurality of first line segments are obtained by segmenting the contour of the test pattern;

[0034] Step S3: providing a layout to be corrected, wherein the layout to be corrected includes a plurality of graphics to be corrected;

[0035] Step S4: segmenting the contour of the to-be-corrected figure into a plurality of second line segments;

[0036] Step S5: based on the layout to be corrected, obtaining a second environment distribution matrix around each second line segment;

[0037] Step S6: using the second environment distribution matrix as an input value and obtaining a second offset of the second line segment according to the convolutional neural network model;

[0038] Step S7: based on the second offset of each of the second line segments, a revised layout is obtained, wherein the revised layout includes a plurality of revised graphics, and the revised graphics correspond one-to-one to the graphics to be revised.

[0039] The graphic correction method obtains the second offset of the second line segment based on the second environment distribution matrix of the second line segment according to the convolutional neural network model, and then moves the second line segments of the graphic to be corrected according to the second offset to obtain the correction layout. The process can simplify the process of obtaining the second offset, can meet the requirements of smaller edge prevention errors, reduce calculation costs, improve accuracy, and improve the efficiency of optical proximity effect correction.

[0040] Next, combine Figures 8 to 17 Each step is analyzed and explained. Figures 8 to 17 It is a structural schematic diagram of the graphic correction process in an embodiment of the present invention.

[0041] Please combine Figure 8 Continue to refer Figure 1 , executing step S1: providing a test layout 100 , wherein the test layout 100 includes a plurality of test patterns 101 .

[0042] The test layout 100 is used to obtain input values ​​and output values ​​for training a convolutional neural network model. The test layout 100 has the same minimum graphic size as a subsequent layout to be corrected, so that the convolutional neural network model trained based on the input values ​​and output values ​​obtained from the test layout 100 can be applicable to the layout to be corrected.

[0043] Next, combine Figures 2 to 4 Execute step S2: Based on the test layout 100, obtain a convolutional neural network model, wherein the convolutional neural network model includes: a first environmental distribution matrix around a number of first line segments on the test layout 100, and first offsets of the number of first line segments of the test layout 100 after correction of the optical proximity effect, wherein the first environmental distribution matrices correspond one-to-one to the first offsets, and the first line segments are obtained by segmenting the contour of the test pattern 101.

[0044] Please refer to Figure 2 In this embodiment, based on the test layout 100, a convolutional neural network model is obtained, including:

[0045] Step S21: segmenting the outline of the test pattern 101 into a plurality of first line segments;

[0046] Step S22: based on the test layout 100, obtaining a first environment distribution matrix around each first line segment;

[0047] Step S23: performing optical proximity effect correction on the test layout 100 to obtain a first offset of each first line segment, wherein a first environment distribution matrix around each first line segment corresponds to the first offset of each first line segment one by one;

[0048] Step S24: providing an initial convolutional neural network model;

[0049] Step S25: Taking the first environment distribution matrix of each first line segment as an input value and the first offset of each first line segment as an output value, the initial convolutional neural network model is trained to obtain the convolutional neural network model.

[0050] Please continue to combine Figure 8 refer to Figure 2, executing step S21: segmenting the outline of the test pattern 101 so that the outline of the test pattern 101 is segmented into a plurality of first line segments.

[0051] The contour of the test pattern 101 is segmented, that is, according to a set segmentation rule, the contour of the test pattern 101 is segmented into a plurality of first line segments of the same length, or into a plurality of first line segments of different lengths.

[0052] Please continue to refer to Figure 2 , execute step S22: based on the test layout 100, obtain a first environment distribution matrix around each first line segment.

[0053] Please refer to Figure 3 In this embodiment, based on the test layout 100, obtaining a first environment distribution matrix around each first line segment includes:

[0054] Step S221: generating a rectangular area with a preset side length on the test layout with the midpoint of the first line segment as the center;

[0055] Step S222: setting a plurality of sampling points distributed in an array in the rectangular area;

[0056] Step S223: marking the sampling point according to whether there is a graphic at the location of the sampling point;

[0057] Step S224: generating the first environment distribution matrix according to the marks of the sampling points.

[0058] Please combine Fig. 9 refer to Figure 3 , execute step S221: generate a rectangular area M with a preset side length on the test layout 100 with the midpoint O of the first line segment L as the center.

[0059] Fig. 9 One of the first line segments L is schematically shown in FIG.

[0060] In this embodiment, the rectangle includes a square, and the side length of the rectangle is set according to requirements.

[0061] Please combine Fig.10 refer to Figure 3 , execute step S222: set a plurality of sampling points A distributed in an array in the rectangular area M.

[0062] Please combine Figures 11 to 15 refer to Figure 4 In this embodiment, a plurality of sampling points A distributed in an array are set in the rectangular area M, including:

[0063] Step S2221: taking the midpoint O of the first line segment L as the center, generating sub-rectangular areas with different preset side lengths in the rectangular area M;

[0064] Step S2222: generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas;

[0065] Step S2223: connecting the sampling points of the sub-rectangular area horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extended lines and vertical extended lines;

[0066] Step S2224: generating side sampling points at the positions where the horizontally extending connecting line and the vertically extending connecting line intersect the sides of the rectangular area and the sub-rectangular area, and the plurality of the vertex sampling points and the plurality of the side sampling points are distributed in an array.

[0067] Please refer to Fig.11 , Fig.11 Schematically, a sub-rectangular area M1 with a first preset side length is generated with the midpoint O of the first line segment L as the center; and a vertex sampling point At1 is generated at the vertex of the sub-rectangular area M1.

[0068] Please refer to Fig.12 , Fig.12 It is illustrated in the figure that a sub-rectangular area M2 with a second preset side length is generated with the midpoint O of the first line segment L as the center, and the second preset side length is greater than the first preset side length; and a vertex sampling point At2 is generated at the vertex of the sub-rectangular area M2.

[0069] Please refer to Fig.13 , Fig.13 It is shown that the vertex sampling point At1 of the sub-rectangular area M1 is connected horizontally and vertically, and the horizontal and vertical lines are extended to obtain horizontal extension lines and vertical extension lines; and the side sampling point Am1 is generated at the position where the horizontal extension line and the vertical extension line intersect the side length of the sub-rectangular area M2.

[0070] Please refer to Fig.14 , Fig.14 It is illustrated in FIG. 1 that a rectangular area M with a third preset side length is generated with the midpoint O of the first line segment L as the center, and the third preset side length is greater than the second preset side length; and a vertex sampling point At3 is generated at the vertex of the rectangular area M.

[0071] Please refer to Fig.15 , Fig.15It is shown that the vertex sampling points At1, At2, At3, and Am1 are connected horizontally and vertically, and the horizontal and vertical lines are extended to obtain horizontal extended lines and vertical extended lines; and a side sampling point Am2 is generated at the position where the horizontal extended lines and the vertical extended lines intersect the side length of the rectangular area M.

[0072] The plurality of sampling points A distributed in an array include: vertex sampling points At1, At2, At3 and side sampling points Am1 and Am2. The plurality of sampling points At1, At2, At3, Am1 and Am2 are distributed in an array.

[0073] In this embodiment, the sub-rectangular area M1 and the sub-rectangular area M2 are in the shape of a square.

[0074] In this embodiment, the distribution of the sampling points A in the rectangular area M includes: the closer to the midpoint O of the first line segment L, the greater the density of the sampling points A; the farther from the midpoint O of the first line segment L, the smaller the density of the sampling points A.

[0075] That is, the first preset side length, the second preset side length and the third preset side length do not increase in equal proportion, and the first single-side spacing between the sub-rectangular area M1 and the sub-rectangular area M2 is not equal to the second single-side spacing between the sub-rectangular area M2 and the rectangular area M.

[0076] In other embodiments, the first preset side length, the second preset side length and the third preset side length can increase proportionally, and the first single-side spacing between the sub-rectangular area M1 and the sub-rectangular area M2 and the second single-side spacing between the sub-rectangular area M2 and the rectangular area M can be equal.

[0077] Please combine Fig.10 refer to Figure 3 , execute step S223: mark the sampling point A according to whether there is a graphic at the location of the sampling point A.

[0078] The sampling point A is marked according to whether there is a graphic at the location of the sampling point A, including: if there is a graphic at the location of the sampling point A, it is marked as "1"; if there is no graphic at the location of the sampling point A, it is marked as "0".

[0079] Please combine Fig.16 refer to Figure 3 , execute step S224: generate the first environment distribution matrix according to the mark of the sampling point A.

[0080] Fig.16 Indicated in Fig.10 Schematic diagram of the marking results of several sampling points A in the rectangular area M. Fig.16 The marks of “0” and “1” in the first environment distribution matrix correspond one-to-one to the positions of several sampling points A in the rectangular area M.

[0081] Please continue to refer to Figure 2 , executing step S23: performing optical proximity effect correction on the test layout 100, obtaining a first offset of each first line segment, and a first environment distribution matrix around each first line segment corresponds to the first offset of each first line segment one by one.

[0082] Please combine Fig.17 refer to Figure 5 In this embodiment, optical proximity effect correction is performed on the test layout to obtain a first offset of each first line segment, including:

[0083] Step S231: performing simulated exposure on the test layout to obtain a first simulated exposure layout;

[0084] Step S232: obtaining an edge placement error between the first simulated exposure layout and the target layout;

[0085] Step S233: moving each of the first line segments according to the edge placement error to obtain a second test layout;

[0086] Step S234: performing simulated exposure on the second test layout to obtain a second simulated exposure layout;

[0087] Step S235: obtaining an edge placement error between the second simulated exposure layout and the target layout;

[0088] Step S236: if the edge placement error satisfies a preset range, obtaining the corrected test layout 200, and obtaining a first offset of each first line segment in the corrected test layout 200 compared to the first line segment in the test layout 100;

[0089] Step S237: If the edge placement error does not meet the preset range, continue to simulate the exposure of the corrected test layout until the edge placement error meets the preset range.

[0090] The target layout is the layout that needs to be made on the mask to transfer the pattern to the wafer after the optical proximity effect correction.

[0091] Fig.17 The modified test layout 200 is shown schematically, and the modified test layout 200 includes a plurality of modified test patterns 201 , and the plurality of modified test patterns 201 correspond to the plurality of test patterns 101 one by one.

[0092] Please combine Figures 18 to 20 Continue to refer Figure 2, execute step S24: provide an initial convolutional neural network model.

[0093] The initial convolutional neural network model includes a convolutional attention module, and the convolutional attention module includes a spatial attention module and a channel attention module.

[0094] The initial convolutional neural network model includes a convolutional attention module, which can enhance the accuracy of the initial convolutional neural network model in predicting output values.

[0095] Please refer to Figures 18 to 20 , Fig.18 The structure of the convolutional attention module is shown in Figure 2. Fig.19 This shows the structure of the spatial attention module. Fig. 20 Illustration of the structure of the channel attention module.

[0096] Fig.19 The spatial attention module mechanism focuses on specific parts of the input image. By calculating the importance of each pixel and adjusting the processing weights, the model can focus more on important areas. Fig. 20 The channel attention module mechanism focuses on each channel in the feature map and dynamically adjusts the weight of each channel, allowing the network to more effectively utilize key feature information. Both the spatial attention module mechanism and the channel attention module mechanism significantly improve the performance of the model in various computer vision tasks by optimizing the feature extraction process and enhancing the accuracy of the initial convolutional neural network model in predicting the output value.

[0097] Please continue to refer to Figure 2 , execute step S25: take the first environment distribution matrix of each first line segment as an input value, take the first offset of each first line segment as an output value, train the initial convolutional neural network model, and obtain the convolutional neural network model.

[0098] The first environmental distribution matrix and the first offset corresponding to each first line segment correspond one to one. The first environmental distribution matrix of each first line segment is used as an input value, and the first offset of each first line segment is used as an output value. The initial convolutional neural network model is trained until the model converges to obtain the convolutional neural network model.

[0099] Please continue to refer to Figure 1 , execute step S3: provide a layout to be corrected, wherein the layout to be corrected includes a plurality of graphics to be corrected.

[0100] The layout to be corrected is a mask design pattern that has not been corrected for optical proximity effect, and the pattern of the layout to be corrected after optical proximity effect correction is a pattern that needs to be transferred to the mask.

[0101] The minimum graphic size of the to-be-corrected layout is the same as the minimum size of the test layout 100 , that is, the resolution of the to-be-corrected layout is the same as that of the test layout 100 .

[0102] Please continue to refer to Figure 1 , executing step S4: segmenting the contour of the to-be-corrected figure so as to segment the contour of the to-be-corrected figure into a plurality of second line segments.

[0103] For the process of segmenting the contour of the to-be-corrected figure into a plurality of second line segments, please refer to Figure 2 The process of segmenting the outline of the test pattern 101 is not described in detail here.

[0104] Please continue to refer to Figure 1 , execute step S5: based on the layout to be corrected, obtain a second environment distribution matrix around each second line segment.

[0105] Please refer to Figure 6 In this embodiment, based on the layout to be corrected, obtaining a second environment distribution matrix around each second line segment includes:

[0106] Step S51: generating a rectangular area with a preset side length on the layout to be corrected, taking the midpoint of the second line segment as the center;

[0107] Step S52: setting a plurality of sampling points distributed in an array in the rectangular area;

[0108] Step S53: marking the sampling point according to whether there is a graphic at the location of the sampling point;

[0109] Step S54: Generate the second environment distribution matrix according to the marks of the sampling points.

[0110] Please refer to Figure 7 In this embodiment, a plurality of sampling points distributed in an array are set in the rectangular area, including:

[0111] Step S521: generating sub-rectangular areas with different preset side lengths within the rectangular area with the midpoint of the second line segment as the center;

[0112] Step S522: generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas;

[0113] Step S523: connecting the sampling points of the sub-rectangular area horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extended lines and vertical extended lines;

[0114] Step S524: generating side sampling points at the positions where the horizontally extending connecting line and the vertically extending connecting line intersect the sides of the rectangular area and the sub-rectangular area, and the plurality of the vertex sampling points and the plurality of the side sampling points are distributed in an array.

[0115] In this embodiment, the shape of the rectangular area includes a square; the shape of the sub-rectangular area includes a square.

[0116] The sampling point is marked according to whether there is a graphic at the location of the sampling point, including: if there is a graphic at the location of the sampling point, it is marked as "1"; if there is no graphic at the location of the sampling point, it is marked as "0".

[0117] The distribution of the sampling points in the rectangular area includes: the closer to the midpoint of the second line segment, the greater the density of the sampling points; the farther from the midpoint of the second line segment, the smaller the density of the sampling points.

[0118] For the process of obtaining the second environment distribution matrix around each second line segment based on the to-be-corrected layout, please refer to Figure 3 and Figure 4 , Figures 9 to 16 The contents described will not be repeated here.

[0119] Please continue to refer to Figure 1 , execute step S6: take the second environment distribution matrix as input value, and obtain the second offset of the second line segment according to the convolutional neural network model.

[0120] The second offset of the second line segment corresponds one-to-one to the second environment distribution matrix. Therefore, by inputting the second environment distribution matrix into the convolutional neural network model, the second offset of the second line segment can be obtained from the convolutional neural network model as an output value.

[0121] Since the convolutional neural network model is obtained after multiple trainings based on several groups of input values ​​and output values ​​obtained from the test layout, the convolutional neural network model can be used continuously for offset prediction of the layout to be corrected, eliminating multiple iterations, improving the accuracy of the second offset obtained, and improving the accuracy of the correction model obtained.

[0122] The convolutional neural network model can obtain the second offset of the second line segment according to the second environmental distribution matrix of the second line segment. The process can simplify the process of obtaining the second offset, meet the requirements of smaller edge prevention errors, reduce calculation costs, and improve the efficiency of optical proximity effect correction.

[0123] Please continue to refer to Figure 1, executing step S7: based on the second offset of each of the second line segments, obtaining a revised layout, wherein the revised layout includes a plurality of revised graphics, and the revised graphics correspond one-to-one to the graphics to be revised.

[0124] Based on the second offset of each second line segment, a revised layout is obtained, including: according to the second offsets of the second line segments, several second line segments of several graphics to be corrected are moved to obtain the revised layout, wherein the revised layout includes several revised graphics, and the revised graphics correspond one-to-one to the graphics to be corrected.

[0125] The revised layout is a pattern that needs to be transferred to the mask plate later.

[0126] The correction layout is obtained according to the convolutional neural network model, based on the second environment distribution matrix of the second line segment, to obtain the second offset of the second line segment, and then the second line segments of the to-be-corrected pattern are moved according to the second offset to obtain the correction layout. The process can simplify the process of obtaining the second offset, can meet the requirements of smaller edge prevention errors, reduce the calculation cost, improve the accuracy, and improve the efficiency of optical proximity effect correction.

[0127] Accordingly, an embodiment of the present invention further provides a storage medium on which computer instructions are stored, and the computer instructions are executed as follows when they are executed: Figures 1 to 7 The steps of the method.

[0128] Accordingly, an embodiment of the present invention further provides a storage terminal, including a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, and when the processor executes the computer instructions, the following is performed: Figures 1 to 7 The steps of the method.

[0129] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A graphics correction method, characterized in that: include: Providing a test layout, wherein the test layout includes a plurality of test patterns; Based on the test layout, a convolutional neural network model is obtained, wherein the convolutional neural network model includes: a first environment distribution matrix around a plurality of first line segments on the test layout, and first offsets of the plurality of first line segments of the test layout after correction of the optical proximity effect, wherein the plurality of first environment distribution matrices correspond to the plurality of first offsets in a one-to-one manner, and the plurality of first line segments are obtained by segmenting the contour of the test pattern; Providing a layout to be revised, wherein the layout to be revised includes a plurality of graphics to be revised; Segmenting the contour of the to-be-corrected figure so that the contour of the to-be-corrected figure is divided into a plurality of second line segments; Based on the to-be-corrected layout, obtaining a second environment distribution matrix around each second line segment; Taking the second environment distribution matrix as an input value, and according to the convolutional neural network model, obtaining a second offset of the second line segment; Based on the second offset of each of the second line segments, a revised layout is obtained, wherein the revised layout includes a plurality of revised graphics, and the revised graphics correspond to the graphics to be revised one by one.

2. The image correction method according to claim 1, characterized in that: Based on the test layout, a convolutional neural network model is obtained, including: segmenting the outline of the test pattern so that the outline of the test pattern is divided into a plurality of first line segments; based on the test layout, a first environmental distribution matrix around each first line segment is obtained; performing optical proximity effect correction on the test layout to obtain a first offset of each first line segment, wherein the first environmental distribution matrix around each first line segment corresponds one-to-one to the first offset of each first line segment; providing an initial convolutional neural network model; taking the first environmental distribution matrix of each first line segment as an input value and the first offset of each first line segment as an output value, the initial convolutional neural network model is trained to obtain the convolutional neural network model.

3. The image correction method according to claim 2, characterized in that: Based on the test layout, a first environment distribution matrix around each first line segment is obtained, including: generating a rectangular area with a preset side length on the test layout with the midpoint of the first line segment as the center; setting a plurality of sampling points distributed in an array within the rectangular area; marking the sampling points according to whether there are graphics at the locations of the sampling points; and generating the first environment distribution matrix according to the markings of the sampling points.

4. The image correction method according to claim 3, characterized in that: A plurality of sampling points distributed in an array are set in the rectangular area, including: taking the midpoint of the first line segment as the center, generating sub-rectangular areas with different preset side lengths in the rectangular area; generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas; connecting the sampling points of the sub-rectangular areas horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extension lines and vertical extension lines; generating side sampling points at positions where the horizontal and vertical extension lines intersect with the side lengths of the rectangular area and the sub-rectangular area, and a plurality of the vertex sampling points and a plurality of the side sampling points are distributed in an array.

5. The image correction method according to claim 4, characterized in that: The shape of the rectangular area includes a square; the shape of the sub-rectangular area includes a square.

6. The image correction method according to claim 3, characterized in that: The sampling point is marked according to whether there is a graphic at the location of the sampling point, including: if there is a graphic at the location of the sampling point, it is marked as "1"; if there is no graphic at the location of the sampling point, it is marked as "0".

7. The image correction method according to claim 3, characterized in that: The distribution of the sampling points in the rectangular area includes: the closer the sampling points are to the midpoint of the first line segment, the greater the density of the sampling points; the farther the sampling points are from the midpoint of the first line segment, the smaller the density of the sampling points.

8. The image correction method according to claim 2, characterized in that: The optical proximity effect correction is performed on the test layout to obtain the first offset of each first line segment, including: performing simulated exposure on the test layout to obtain a first simulated exposure layout; obtaining an edge placement error between the first simulated exposure layout and a target layout; moving each first line segment according to the edge placement error to obtain a second test layout; performing simulated exposure on the second test layout to obtain a second simulated exposure layout; obtaining an edge placement error between the second simulated exposure layout and the target layout; if the edge placement error meets a preset range, obtaining a corrected test layout, and obtaining the first offset of each first line segment in the corrected test layout compared to the first line segment in the test layout; if the edge placement error does not meet the preset range, continuing to simulate the exposure of the corrected test layout until the edge placement error meets the preset range.

9. The image correction method according to claim 1, characterized in that: The convolutional neural network model includes a convolutional attention module, and the convolutional attention module includes a spatial attention module and a channel attention module.

10. The image correction method according to claim 1, characterized in that: Based on the layout to be corrected, a second environment distribution matrix around each second line segment is obtained, including: generating a rectangular area with a preset side length on the layout to be corrected with the midpoint of the second line segment as the center; setting a number of sampling points distributed in an array within the rectangular area; marking the sampling points according to whether there are graphics at the locations of the sampling points; and generating the second environment distribution matrix according to the markings of the sampling points.

11. The image correction method according to claim 10, characterized in that: A plurality of sampling points distributed in an array are set in the rectangular area, including: taking the midpoint of the second line segment as the center, generating sub-rectangular areas with different preset side lengths in the rectangular area; generating vertex sampling points at the vertices of the rectangular area and at the vertices of a plurality of sub-rectangular areas; connecting the sampling points of the sub-rectangular areas horizontally and vertically, and extending the horizontal and vertical lines to obtain horizontal extension lines and vertical extension lines; generating side sampling points at positions where the horizontal and vertical extension lines intersect with the side lengths of the rectangular area and the sub-rectangular area, and a plurality of the vertex sampling points and a plurality of the side sampling points are distributed in an array.

12. The image correction method according to claim 11, characterized in that: The shape of the rectangular area includes a square; the shape of the sub-rectangular area includes a square.

13. The image correction method according to claim 10, characterized in that: The sampling point is marked according to whether there is a graphic at the location of the sampling point, including: if there is a graphic at the location of the sampling point, it is marked as "1"; if there is no graphic at the location of the sampling point, it is marked as "0".

14. The image correction method according to claim 10, characterized in that: The distribution of the sampling points in the rectangular area includes: the closer to the midpoint of the second line segment, the greater the density of the sampling points; the farther from the midpoint of the second line segment, the smaller the density of the sampling points.

15. The image correction method according to claim 1, characterized in that: Based on the second offset of each second line segment, a revised layout is obtained, including: according to the second offsets of the second line segments, several second line segments of several graphics to be corrected are moved to obtain the revised layout, wherein the revised layout includes several revised graphics, and the revised graphics correspond one-to-one to the graphics to be corrected.

16. A storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed, the steps of the method according to any one of claims 1 to 15 are executed.

17. A storage terminal, comprising a memory and a processor, wherein the memory stores computer instructions that can be run on the processor, characterized in that: When the processor runs the computer instructions, the steps of the method according to any one of claims 1 to 15 are performed.

Images