Length measurement system, model creation system, and length measurement method

Abstract

A measurement object captured image in which a sample is captured is converted into a measurement object length measurement image by a length measurement image conversion model, and the size of a measurement pattern image included in the measurement object length measurement image is measured. Here, the length measurement image conversion model learns using training data, which is a combination of a learning captured image in which a sample has been captured and a learning length measurement image corresponding to the learning captured image. The contrast between a measurement pattern image in the learning length measurement image and a non-measurement pattern image existing in proximity to the measurement pattern image is higher than the contrast between the measurement pattern image and the non-measurement pattern image in the learning captured image, or the non-measurement pattern image in the learning captured image is deleted from the learning length measurement image.

Description

Technical Field

[0001] The present disclosure relates to a length measurement system, a model making system and a length measurement method. Background Art

[0002] Patent Document 1 discloses an image processing method for optimizing the brightness and grayscale of a specific object area in an image. Therefore, a specific area (e.g., a person's face) is detected, a tone curve for image enhancement is calculated using a histogram of the entire image or a histogram of the specific area, and contrast correction is performed using the calculated tone curve.

[0003] Patent document 2 discloses a learning model that can selectively transform a specific pattern contained in an image, or other partial images such as the edges of structures, with high precision. By changing the degree of learning according to the importance of each part of the image, it is possible to generate high-quality images for important parts, and on the other hand, reduce the processing required for learning for unimportant parts, thereby achieving more efficient learning.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2009-200743

[0007] Patent Document 2: International Publication No. 2021 / 140620 Summary of the invention

[0008] Problems to be solved by the invention

[0009] use Figure 1A The subject of the present disclosure is described. A captured image (schematic diagram) is shown in the left column. The captured image 100 is, for example, a SEM image of a pattern formed on a semiconductor wafer. On the semiconductor wafer, a second pattern that is not a measurement object is provided close to the first pattern of the measurement object. The right column shows a brightness profile 111 along a line 103 on the captured image 100. In addition, together with the brightness profile 111, a brightness profile 112 obtained in the absence of the second pattern is also shown.

[0010] If the first pattern is an isolated pattern, the contrast with the surroundings is also large as shown in the brightness distribution 112, and it can be expected that the edge extraction from the first pattern image (measurement pattern image) 101 for measurement can also be performed with high accuracy. On the other hand, in the case of the captured image 100 in which the non-measurement pattern exists close, due to the influence of the second pattern image (non-measurement pattern image) 102, the brightness in the boundary area between the first pattern and the second pattern does not decrease as in the case of an isolated pattern, and the brightness representing the first pattern image 101 is connected to the brightness representing the second pattern image 102 without completely decreasing, resulting in a brightness distribution like the brightness distribution 111. In this case, the accuracy of edge extraction from the first pattern image 101 for measurement is reduced, and as a result, the measurement accuracy is also reduced.

[0011] for Figure 1A If the technology disclosed in Patent Document 1 is applied to the captured image shown, the second pattern image 102 will also be made more prominent by making the first pattern image 101 more prominent, and the above-mentioned problem cannot be solved. In addition, according to Patent Document 2, even if the accuracy is changed to perform transformation in important / unimportant parts, the difference in contrast is not increased, and thus the above-mentioned problem cannot be solved.

[0012] In the present disclosure, when measuring the length of a pattern based on a captured image, the captured image 100 is transformed into Figure 1B 1 is a length measurement image 120 for dimension measurement as a schematic diagram in FIG. The length measurement image 120 is an image in which the contrast between the first pattern image 121 as a measurement pattern image and the second pattern image 122 as a non-measurement pattern image is increased relative to the original captured image 100. For example, the length measurement image 120 may be an image in which the second pattern image 122 as a non-measurement pattern image is deleted. Alternatively, the length measurement image 120 may be an image in which the contrast between the second pattern image 122 and the background portion of the length measurement image 120 (referring to the area other than the first pattern image 121 and the second pattern image 122 in the length measurement image 120) is reduced.

[0013] Means for solving problems

[0014] A length measurement system as one embodiment of the present disclosure is a length measurement system for measuring the size of a pattern formed on a sample, and includes a computer system having a length measurement image conversion unit and a length measurement unit, wherein the length measurement image conversion unit converts a measurement object captured image obtained by capturing the sample into a measurement object length measurement image by using a length measurement image conversion model, and the length measurement unit measures the size of a measurement pattern image included in the measurement object length measurement image, and learns the length measurement image conversion model using training data of a combination of a learning captured image obtained by capturing the sample and a learning length measurement image corresponding to the learning captured image, wherein a contrast between a measurement pattern image in the learning length measurement image and a non-measurement pattern image existing close to the measurement pattern image is higher than a contrast between a measurement pattern image and a non-measurement pattern image in the learning captured image, or, in the learning length measurement image, the non-measurement pattern image in the learning captured image is deleted.

[0015] Effects of the Invention

[0016] Other objects and novel features will become apparent from the description of this specification and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1A It is a figure for explaining the subject of this disclosure.

[0018] Figure 1B It is a figure for explaining the subject of this disclosure.

[0019] Figure 2 This is an example of the structure of a length measurement system.

[0020] Figure 3 This is an example of the structure of a model making system.

[0021] Figure 4A This is a functional block diagram of the image generation model making unit.

[0022] Figure 4B This is an example of the generated model confirmation screen.

[0023] Figure 5 This is a functional block diagram of the region segmentation unit.

[0024] Fig. 6A This is a functional block diagram of the length measurement edge detection unit.

[0025] Figure 6B is a diagram showing a situation where interactive machine learning is applied to detect edges.

[0026] Figure 7 This is a functional block diagram of the image transformation model creation unit.

[0027] Figure 8This is a diagram for explaining the area specifying method.

[0028] Fig.9A is an example of a label image.

[0029] Fig. 9B is based on Fig.9A An example of an image for length measurement generated from a labeled image.

[0030] Fig. 10A This is an example of the area specification screen.

[0031] Fig. 10B This is an example of manually specifying a screen.

[0032] Fig. 10C This is an example of automatically specifying a screen.

[0033] Fig.11 This is an example of the conversion model confirmation screen.

[0034] Fig.12 This is a functional block diagram of the length measurement image conversion unit. DETAILED DESCRIPTION

[0035] Figure 2 This is a diagram showing a structural example of a length measuring system. Figure 2 The length measurement system is a system that uses a charged particle beam device such as a scanning electron microscope (SEM) to measure the length of a pattern formed on a semiconductor wafer. The length measurement system obtains an image of a fine pattern formed on a semiconductor wafer and performs dimension measurement. The length measurement system includes: SEM11, a control device 12 that controls SEM11, a computer system 1 that performs the length measurement process, an input device 13 for inputting necessary information to the computer system 1, and a display 14 for displaying a screen for information input or measurement results.

[0036] The computer system 1 includes a calculation processing unit 2 including one or more CPUs (Central Processing Units) and a storage device 3. In the present disclosure, the storage device 3 stores programs related to length measurement processing, other programs related to the control of the SEM, data used in processing by these programs, such as the length measurement image conversion model 31 and the length measurement process 32 described later, and processing results of these programs. In addition, the computer system 1 may also be configured integrally with the control device 12.

[0037] The calculation processing unit 2 performs the length measurement processing by executing the program related to the length measurement processing. That is, the computer system 1 functions as a length measurement processing device. In other words, the program causes the computer system to function as a length measurement processing device. The program executed by the computer system 1 and the like and its function are referred to as "function", "unit", etc.

[0038] The calculation processing unit 2 functions as a captured image input unit 21 which is an interface for inputting captured images of the SEM 11, a length measurement image conversion unit 22 which converts the captured images input by the captured image input unit 21 into length measurement images, and a length measurement unit 23 which measures the dimensions of the length measurement images produced by the length measurement image conversion unit 22 according to a length measurement process 32. When the length measurement image conversion unit 22 converts the captured images into the length measurement images, it uses a length measurement image conversion model 31.

[0039] Figure 3 This is a diagram showing a configuration example of a model making system for making an image transformation model 31 for length measurement used in a length measurement system.

[0040] The computer system 5 functions as a captured image input unit 61 which is an interface for inputting captured images, an image generation model creation unit 62 which uses the captured images input by the captured image input unit 61 as training data to create an image generation model 30, and an image transformation model creation unit 63 which uses the captured images input by the captured image input unit 61 as training data and uses the image generation model 30 to create an image transformation model 31 for length measurement. The captured images of the semiconductor wafer as the measurement object captured by the SEM 11 of the length measurement system are stored in the image storage medium 15. The hardware structure of the computer system 5 is the same as that of the computer system 1, and repeated descriptions are omitted. The computer system 5 is connected to an input device 16 and a display 17.

[0041] In addition, if there is sufficient computing resources, the processing of the computer system 5 can also be performed by the computer system 1 in the length measurement system. Figure 3 In FIG. 1 , an example of storing captured images for training data in the image storage medium 15 is shown, but the image storage medium 15 may also be connected to the computer system 1 to perform a length measurement system for measuring the dimensions of the captured images stored in the image storage medium 15.

[0042] First, the model making system is used to make a model. Figure 1A The captured image shown is transformed into Figure 1B Therefore, training data is created which is a combination of a captured image and a length measurement image.

[0043] Figure 4AThe functional block diagram of the image generation model production unit 62 is shown. The image generation model production unit 62 is a functional block for producing the image generation model 30. The image generation model 30 is a model that generates an image (hereinafter sometimes referred to as a simulated image) having the shape shown in the label image and regarded as an actual image based on the label image representing the shape of the pattern area. As described later, the image generation model 30 is used in the process of producing the length measurement image for training data. The image generation model production unit 62 includes a region segmentation unit 71, an image enhancement processing unit 72, and a first learning unit 73.

[0044] The region dividing unit 71 divides the captured image into a measurement pattern region and other background regions. Figure 1A In the example, the captured image is divided into a measurement pattern area corresponding to the first pattern image (measurement pattern image) 101 and other background areas. Such area segmentation can be performed manually or by image processing. In addition, any method in which the user verifies and corrects the combination thereof, that is, the area segmentation performed by image processing, can be adopted. As a method for performing area segmentation by image processing, methods using image features (clustering of brightness based on the k-means method), methods using machine learning (trained and untrained segmentation), etc. are known, and they can be applied. Here, it is sufficient to extract only the measurement pattern area, so the area segmentation processing using the length measurement process for measuring the size of the measurement pattern image is described. The length measurement process 32 is data representing the length measurement conditions of the measurement pattern image, and the shape of the measurement pattern (for example, a line, a hole, an ellipse, a circle, etc.), the length measurement cursor (for example, the coordinates, width, length, number, etc. of the upper left vertex), the measurement algorithm, the length measurement parameters, etc. are registered.

[0045] Figure 5 The functional block diagram of the region segmentation unit 71 that performs the region segmentation processing using the length measurement process 32 is shown. The region segmentation unit 71 includes a length measurement edge detection unit 74, a region determination unit 75, and a region extraction unit 76. Schematic diagrams 74a, 75a, and 76a are schematic diagrams showing the processing results in these functional blocks. The length measurement edge detection unit 74 detects the edge 74c of the captured image using the length measurement process 32. In the processing result 74a, the edge 74c is represented by a white circle. The edge 74c can be extracted as a change point of the brightness distribution along the line 74b, for example. However, as Figure 1AAs shown, the pattern image sometimes becomes unclear due to the existence of close patterns and other reasons. Therefore, the edge 74c can also be detected with high precision using a deep learning method in the length measurement edge detection unit 74. The details will be described later. In the area determination unit 75, the pattern boundary 75b is determined by interpolating and connecting the close edges 74c, as shown in the processing result 75a. The area extraction unit 76 determines the measurement pattern area based on the pattern boundary 75b determined by the area determination unit 75. In the processing result 76a, the area 76b corresponding to the area surrounded by the pattern boundary 75b is the measurement pattern area. Thus, a label image representing the measurement pattern area corresponding to the measurement pattern image contained in the captured image is obtained. In addition, when there are multiple measurement pattern areas in the captured image, the measurement pattern area can be extracted for each measurement pattern area.

[0046] Fig. 6A The functional block diagram of the length measurement edge detection unit 74 for detecting edges from a captured image is shown. The length measurement edge detection unit 74 includes a length measurement edge learning unit 77 and a length measurement edge inference unit 78, and can detect edges with high accuracy by an interactive machine learning method. The length measurement edge learning unit 77 uses training data to learn a learning model for inferring edges, and the length measurement edge inference unit 78 uses the learning model learned by the length measurement edge learning unit 77 to infer edges.

[0047] Figure 6B The figure shows how the edge is detected by applying interactive machine learning in the length measurement edge detection unit 74. In addition, the white frame 79 in the figure is a length measurement cursor, and the change point of the brightness distribution in this area is detected as an edge. The details of the detection process are defined in the length measurement process 32. In addition, here, the white circle is the edge used as training data, and the white triangle represents the edge inferred using the learning model.

[0048] The first training data 77a is provided by the user. The result inferred by the learning model that has learned the first training data 77a is the inferred edge 78a. The inferred edge 78a includes the inferred result located on the pattern boundary and the inferred result that deviates from the pattern boundary. Therefore, the user deletes the inferred result that deviates from the pattern boundary, and sets the inferred result located on the pattern boundary and the training data initially provided as new training data 77b. Similarly, as the result of inference by the learning model that has learned the training data 77b, the inferred edge 78b is obtained. This operation is performed repeatedly, and finally, when the result inferred by the learning model converges on the pattern boundary (here is the third inferred edge 78c), the learning of the learning model is terminated.

[0049] As described above, the area dividing unit 71 can generate a label image indicating an area corresponding to the measurement pattern image based on the captured image.

[0050] return Figure 4A In the first learning unit 73, the combination of the label image representing the measurement pattern area generated by the area segmentation unit 71 and the measurement pattern image of the captured image is used as training data to learn the image generation model 30. If the error between the simulated measurement pattern image generated by the image generation model 30 and the measurement pattern image of the original captured image converges to a certain value or less, the learning of the image generation model 30 is terminated. In order to increase the amount of training data to improve the accuracy of the model, not only the number of captured images can be increased, but also a known data expansion method for increasing the amount of training data can be applied.

[0051] In this case, the shape of the simulated measurement pattern image generated by the image generation model 30 follows the shape of the pattern area of ​​the label image, and on the other hand, as an image, an image equivalent to the original captured image is output. In contrast, for example, an image in which the brightness of the generated simulated measurement pattern image is improved compared to the original captured image can be output. In this case, as training data used in the first learning unit 73, an image in which the captured image is processed by the image enhancement processing unit 72 to enhance the measurement pattern area can be used. In addition, in the first learning unit 73, a learning model for a label image representing an area corresponding to the measurement pattern image output based on the captured image is learned, and the image generation model 30 outputs a simulated image based on the label output by the learning model using the output label image and the brightness of the pixels of the part corresponding to the label output in the captured image. In addition, in the first learning unit 73, multiple types of measurement pattern images can also be learned by one image generation model based on the number of image generation models 30 collected and stored in the storage device 7. For example, in Figure 1A In the example of FIG. 1 , the first pattern image 101 and the second pattern image 102 may be collectively learned by one image generation model 30 . The image generation model 30 after learning is stored in the storage device 7 .

[0052] Furthermore, the image generation model creation unit 62 may include a GUI (Graphical User Interface) for confirming the performance of the created image generation model 30 . Figure 4BAn example of the generated model confirmation screen 130 is shown. When the user presses the label image / reference image loading button 131 of the generated model confirmation screen 130, the captured image and the label image produced by segmenting the captured image are read and displayed as a reference image 135 and a label image 134, respectively. In this example, there are two types of measurement pattern images in the reference image 135. When the user presses the AI ​​model loading button 132, the corresponding image generation model is loaded. For example, the first image generation model for the first measurement pattern area 137 and the second image generation model for the second measurement pattern area 138 are loaded. Thereafter, when the user presses the execution button 133, the first image generation model is used to generate a simulated measurement pattern image 137b based on the label image representing the first measurement pattern area 137, and the second image generation model is used to generate a simulated measurement pattern image 138b based on the label image representing the second measurement pattern area 138. The generated simulated image 136 is displayed side by side with the reference image 135. The user can check that an appropriate simulated measurement pattern image is generated by comparing the measurement pattern image of the reference image 135 with the simulated measurement pattern image of the generated simulated image 136 .

[0053] Figure 7 1 is a functional block diagram showing the image transformation model creation unit 63. The image transformation model creation unit 63 is a functional block that generates the length measurement image transformation model 31. The image transformation model creation unit 63 includes a region specifying unit 81, a region dividing unit 82, an image generating unit 83, and a second learning unit 84. The region specifying unit 81, the region dividing unit 82, and the image generating unit 83 generate a length measurement image that serves as learning data for the length measurement image transformation model 31.

[0054] The area designation unit 81 designates an area including the measurement pattern image from the captured image. The method of area designation is arbitrary. The user can also Figure 8 The area designation can be performed manually by using the designation box 81a shown. The area designation can also be performed by using the coordinate information of the length measurement cursor defined in the length measurement process. Alternatively, the area designation can also be performed by using the layout design drawing of the device and expanding the layout of the corresponding pattern.

[0055] Fig. 10A An example of a GUI for performing area designation is shown. A captured image for area designation processing is selected from the image load button 91 of the area designation screen 90 and displayed as a captured image 94. A designation frame 95 indicating the designated area is displayed on the captured image 94, and the coordinates and size of the designation frame 95 are determined. When the user performs manual designation, the user presses the automatic designation button 93 when the manual designation button 92 is automatically designated by the length measurement process or other information. By pressing the manual designation button 92, the image is displayed as shown in FIG. Fig. 10BThe manual designation screen 96 shown in FIG. 1 is displayed. In the manual designation screen 96, the shape, coordinates, and size of the designation frame 95 can be designated. In addition, the size of the designation frame 95 displayed on the captured image 94 can be adjusted. Furthermore, the visibility of the designated area can be increased or decreased. On the other hand, by pressing the automatic designation button 93, the display shown in FIG. 1 is displayed. Fig. 10C The automatic designation screen 97 is shown. For example, when a plurality of automatic designation methods are available, the user can select an automatic designation method to be applied.

[0056] In addition, the area designation in the area designation screen 90 is not limited to the rectangular shape shown as the designation frame 95, and other shapes such as an ellipse may be used. In addition, other designation methods such as the center coordinates of the area may be used. In such a case, the area designation screen 90 has a GUI corresponding to the area designation method. In addition, the label image 134 (see Figure 4B ) and displayed to the user, and the user specifies, for example, the first measurement pattern area 137 in the label image 134 by operating the GUI.

[0057] return Figure 7 The area segmentation unit 82 segments the area in the captured image in this manner into the measurement pattern area and other areas. The processing of the area segmentation unit 82 is similar to that in Figure 4A The processing of the region segmentation unit 71 described in the above is the same as that of the region segmentation unit 71 described in the above, so repeated description is omitted. In addition, as the region segmentation processing, when the region segmentation processing using the length measurement process is performed, the region designation unit 81 may not be used to designate the region, but the captured image may be directly region segmented. Through the processing of the region segmentation unit 82, Fig.9A The label image 85 shown. The label image 85 is obtained by Figure 8 The captured image 100 shown is an image obtained by performing region segmentation processing, and includes two measurement pattern regions 85 a .

[0058] The image generation unit 83 generates a simulated captured image corresponding to the length measurement image using the label image 85 generated by the area division unit 82 and the image generation model 30 . Fig. 9BAn example of a length measurement image 86 corresponding to the captured image 100 generated based on the label image 85 is shown. The measurement pattern image 86a is a simulated measurement pattern image generated by the image generation model 30 for the measurement pattern area 85a, and for example, the background 86b is set to a uniform monochrome image. Not limited to this, the brightness of the area other than the measurement pattern area 85a of the original captured image 100 may be reduced and synthesized with the simulated measurement pattern image. That is, the contrast of the simulated measurement pattern image in the length measurement image 86 may be higher than the contrast of the measurement pattern image 101 in the original captured image 100.

[0059] For example, the image generator 83 may adjust the brightness of the area other than the measurement pattern area 85a of the original captured image 100 in the length measurement image 86 to be lowered and the brightness in the measurement pattern area 85a to be increased, and the two images may be synthesized. In this case, the contrast of the measurement pattern image 86a in the length measurement image 86 is greatly improved compared to the contrast of the measurement pattern image 101 in the original captured image 100. Regarding the adjustment to increase the brightness in the measurement pattern area 85a, the user can use the GUI such as the area designation screen 90 to perform operations such as how much the brightness is increased.

[0060] In the second learning unit 84, the combination of the captured image and the length measurement image generated by the image generating unit 83 is used as training data to perform learning of the length measurement image conversion model 31. If the error between the length measurement image generated by the length measurement image conversion model 31 and the length measurement image generated by the image generating unit 83 converges to a certain value or less, the learning of the length measurement image conversion model 31 is terminated. In order to increase the number of training data to improve the accuracy of the model, not only the number of captured images can be increased, but also a known data expansion method for increasing the number of training data can be applied.

[0061] In addition, the image conversion model creation unit 63 may include a GUI for confirming the performance of the created length measurement image conversion model 31 . Fig.11 An example of the transformation model confirmation screen 140 is shown. When the user presses the captured image load button 141 on the transformation model confirmation screen 140, the captured image is read and displayed as an input image 144. In this example, the input image 144 includes a measurement pattern image 146 and a non-measurement pattern image 147 close thereto. When the user presses the AI ​​model load button 142, the length measurement image transformation model is loaded. Thereafter, when the user presses the execute button 143, the length measurement image 145 obtained by transforming the input image 144 by the length measurement image transformation model is displayed. The user can confirm that an appropriate length measurement image has been generated by comparing the input image 144 with the transformed length measurement image 145.

[0062] Next, the length measurement process performed by the length measurement system will be described. The storage device 3 of the computer system 1 of the length measurement system stores the image transformation model 31 for length measurement and the length measurement process 32 (see Figure 2 The control device 12 uses the information of the length measurement process 32 read from the computer system 1 to obtain an image of, for example, a semiconductor wafer through the SEM 11. The image captured by the SEM 11 is taken from the control device 12 to the computer system 1 through the captured image input unit 21.

[0063] Fig.12 FIG. 2 is a functional block diagram of the length measurement image conversion unit 22. The length measurement image conversion unit 22 includes an image conversion unit 150. The image conversion unit 150 uses the length measurement image conversion model 31 to convert the image into Figure 1A The captured image shown is transformed into Figure 1B The length measurement image shown. Then, the length measurement unit 23 measures the dimensions of the length measurement image transformed by the length measurement image transformation unit 22 according to the content defined in the length measurement process 32. Thus, even if there is a non-measurement pattern image that hinders the length measurement near the measurement pattern image, the contrast between the measurement pattern image and the non-measurement pattern image is improved in the length measurement image compared with the original captured image, thereby suppressing erroneous measurement. In addition, the length measurement image is also measured using a length measurement image transformation model that has been subjected to machine learning, thereby enabling high-speed and automatic image transformation.

[0064] In addition, the present disclosure is not limited to the above-mentioned embodiments, and includes various modified examples. For example, the above-mentioned embodiments are examples described in detail in order to easily explain the present disclosure, and are not necessarily limited to having all the structures described. In addition, a part of the structure of a certain embodiment can be replaced with the structure of other embodiments, and in addition, the structure of other embodiments can be added to the structure of a certain embodiment. In addition, with respect to a part of the structure of each embodiment, other structures can be added, deleted, or replaced.

[0065] Explanation of symbols

[0066] 1.5: Computer system

[0067] 2, 6: Calculation processing unit

[0068] 3.7: Storage device

[0069] 11: SEM

[0070] 12: Control device

[0071] 13, 16: Input device

[0072] 14, 17: Display

[0073] 15: Image storage medium

[0074] 21: Shooting image input unit

[0075] 22: Image conversion unit for length measurement

[0076] 23: Length measurement department

[0077] 30: Image Generation Model

[0078] 31: Image transformation model for length measurement

[0079] 32: Length measurement process

[0080] 61: Photographing image input unit

[0081] 62: Image Generation Model Production Department

[0082] 63: Image Transformation Model Production Department

[0083] 71: Area segmentation

[0084] 72: Image enhancement processing unit

[0085] 73: First Learning Department

[0086] 74: Length measurement edge detection unit

[0087] 75: Area Determination Department

[0088] 76: Region Extraction Department

[0089] 74a, 75a, 76a: Processing results

[0090] 74b: Line

[0091] 74c: Edge

[0092] 75b: Pattern border

[0093] 76b: Measuring pattern area

[0094] 77: Length Measurement Edge Learning Department

[0095] 78: Length measurement edge inference unit

[0096] 77a, 77b, 77c: Training data

[0097] 78a, 78b, 78c: Inferred edge

[0098] 81: Regional Designation Department

[0099] 82: Area segmentation

[0100] 83: Image generation unit

[0101] 84: Second Learning Department

[0102] 81a: Designated box

[0103] 85: Label Image

[0104] 85a: Measuring pattern area

[0105] 86: Image for length measurement

[0106] 86a: Measuring pattern image

[0107] 86b: Background

[0108] 90: Area designation screen

[0109] 91: Image loading button

[0110] 92: Manually assign buttons

[0111] 93: Automatically assign buttons

[0112] 94: Capture image

[0113] 95: Specify the frame

[0114] 96: Manually specify the screen

[0115] 97: Automatically specify the screen

[0116] 100: Take an image

[0117] 101: First pattern image (measurement pattern image)

[0118] 102: Second pattern image (non-measurement pattern image)

[0119] 103: Line

[0120] 111, 112: Brightness distribution

[0121] 120: Image for length measurement

[0122] 121: The first pattern image

[0123] 122: Second pattern image

[0124] 130: Generate model confirmation screen

[0125] 131: Label image / reference image loading button

[0126] 132: AI model loading button

[0127] 133: Execute button

[0128] 134: Label Image

[0129] 135: Reference image

[0130] 136: Simulated Image

[0131] 137: First measurement pattern area

[0132] 137b: First simulation measurement pattern image

[0133] 138: Second measurement pattern area

[0134] 138b: Second simulated measurement pattern image

[0135] 140: Confirmation screen for model conversion

[0136] 141: Shooting image loading button

[0137] 142: AI model loading button

[0138] 143: Execute button

[0139] 144: Input image

[0140] 145: Image for length measurement

[0141] 146: Measuring pattern image

[0142] 147: Non-measurement pattern image

[0143] 150: Image conversion unit.

Claims

1. A length measuring system for measuring the size of a pattern formed on a sample, characterized in that: The length measurement system comprises a computer system, wherein the computer system comprises a length measurement image conversion unit and a length measurement unit. The length measurement image conversion unit converts the measurement object captured image obtained by capturing the sample into the measurement object length measurement image using the length measurement image conversion model. The length measuring unit measures the size of the measurement pattern image included in the measurement object length measurement image. The length measurement image transformation model is learned using training data, the training data being a combination of a learning image obtained by photographing the sample and a learning length measurement image corresponding to the learning image. The contrast between the measurement pattern image and the non-measurement pattern image existing close to the measurement pattern image in the learning length measurement image is higher than the contrast between the measurement pattern image and the non-measurement pattern image in the learning captured image, or, in the learning length measurement image, the non-measurement pattern image in the learning captured image is deleted.

2. The length measurement system according to claim 1, characterized in that: The length measuring system includes a charged particle beam device for photographing the sample. The computer system includes a captured image input unit that takes in a captured image captured by the charged particle beam device.

3. A model making system for making an image conversion model for length measurement, wherein the image conversion model for length measurement converts an image obtained by photographing a sample having a pattern formed thereon into an image for length measurement, wherein: The model making system comprises a computer system, wherein the computer system comprises an image transformation model making unit. The image transformation model making unit comprises a region segmentation unit, an image generation unit and a learning unit. The region dividing unit of the image conversion model creating unit generates a label image divided into a measurement pattern region as a region of the measurement pattern image included in the captured image and a background region other than the measurement pattern region. The image generation unit of the image conversion model generation unit generates a simulated captured image in which a simulated measurement pattern image is generated in the measurement pattern region of the label image. The learning unit of the image transformation model creation unit performs learning of the image transformation model for length measurement using training data which is a combination of the captured image and the simulated captured image. The contrast between the simulated measurement pattern image and the non-measurement pattern image existing in the background area and close to the simulated measurement pattern image in the simulated captured image is higher than the contrast between the measurement pattern image and the non-measurement pattern image in the captured image, or, in the simulated captured image, the non-measurement pattern image in the captured image is deleted.

4. The model making system according to claim 3, characterized in that: The region segmentation unit of the image transformation model creation unit includes a length measurement edge detection unit, a region identification unit, and a region extraction unit. The length measurement edge detection unit detects the edge of the measurement pattern image. The area determination unit interpolates the detected edge to determine a pattern boundary of the measurement pattern image. The region extraction unit extracts a region surrounded by the determined pattern boundary as the measurement pattern region.

5. The model making system according to claim 4, characterized in that: The computer system stores a length measurement recipe, wherein the length measurement recipe is data indicating a length measurement condition of the measurement pattern image. The length measurement edge detection unit includes a length measurement edge learning unit and a length measurement edge inference unit, and detects the edge of the measurement pattern image within the length measurement cursor defined in the length measurement process through interactive machine learning. The length measurement edge learning unit learns a learning model using training data representing the edge of the measurement pattern image. The length measurement edge estimation unit estimates the edge of the measurement pattern image within the length measurement cursor using the learning model. The edges of the measurement pattern image deviating from the boundary of the measurement pattern image among the edges of the measurement pattern image inferred by the learning model are deleted and used as new training data, and the learning model is learned by the length-measuring edge learning unit.

6. The model making system according to claim 3, characterized in that: The image conversion model creation unit further includes a region specifying unit that specifies a region in which the measurement pattern image included in the captured image exists.

7. The model making system according to claim 3, characterized in that: The computer system further comprises an image generation model making unit. The image generation unit of the image conversion model generation unit generates the simulated measurement pattern image in the measurement pattern region of the label image using the image generation model generated by the image generation model generation unit.

8. The model making system according to claim 7, characterized in that: The image generation model making unit includes a region segmentation unit and a learning unit. The region dividing unit of the image generation model creating unit generates a label image divided into a measurement pattern region as a region of the measurement pattern image included in the captured image and a background region other than the measurement pattern region. The learning unit of the image generation model preparation unit learns the image generation model using training data that is a combination of the label image and the captured image or an image obtained by performing image processing on the captured image to enhance the measurement pattern image.

9. The model making system according to claim 8, characterized in that: The region segmentation unit of the image generation model creation unit includes a length measurement edge detection unit, a region identification unit, and a region extraction unit. The length measurement edge detection unit detects the edge of the measurement pattern image. The area determination unit interpolates the detected edge to determine a pattern boundary of the measurement pattern image. The region extraction unit extracts a region surrounded by the determined pattern boundary as the measurement pattern region.

10. The model making system according to claim 9, characterized in that: The computer system stores a length measurement recipe, which is data indicating the length measurement conditions of the measurement pattern image. The length measurement edge detection unit includes a length measurement edge learning unit and a length measurement edge inference unit, and detects the edge of the measurement pattern image within the length measurement cursor defined in the length measurement process through interactive machine learning. The length measurement edge learning unit learns a learning model using training data representing the edge of the measurement pattern image. The length measurement edge estimation unit estimates the edge of the measurement pattern image within the length measurement cursor using the learning model. The edges of the measurement pattern image deviating from the boundaries of the measurement pattern image among the edges of the measurement pattern image inferred by the learning model are deleted and used as new training data, and the learning model is learned by the length-measuring edge learning unit.

11. A length measurement method, comprising: measuring the size of a pattern formed on a sample by a length measurement system having a computer system, wherein: The computer system comprises a length measurement image conversion unit and a length measurement unit. The length measurement image conversion unit converts the measurement object captured image obtained by capturing the sample into the measurement object length measurement image using the length measurement image conversion model. The length measuring unit measures the size of the measurement pattern image included in the measurement object length measurement image. The length measurement image transformation model is learned using training data, the training data being a combination of a learning image obtained by photographing the sample and a learning length measurement image corresponding to the learning image. The contrast between the measurement pattern image and the non-measurement pattern image existing near the measurement pattern image in the learning length measurement image is higher than the contrast between the measurement pattern image and the non-measurement pattern image in the learning captured image, or, in the learning length measurement image, the non-measurement pattern image in the learning captured image is deleted.

12. The length measuring method according to claim 11, characterized in that: The computer system further comprises an image transformation model creating unit that creates the image transformation model for length measurement. The image transformation model preparation unit includes a region segmentation unit, an image generation unit, and a learning unit. The region segmentation unit of the image conversion model creation unit generates a label image segmented into a measurement pattern region as a region of the measurement pattern image included in the captured image for learning and a background region other than the measurement pattern region. The image generation unit of the image conversion model generation unit generates a simulated captured image in which a simulated measurement pattern image is generated in the measurement pattern region of the label image as the learning length measurement image. The learning unit of the image transformation model creation unit learns the length measurement image transformation model using training data, the training data being a combination of the learning captured image and the simulated captured image.

13. The length measuring method according to claim 12, characterized in that: The region segmentation unit of the image transformation model creation unit includes a length measurement edge detection unit, a region identification unit, and a region extraction unit. The length measurement edge detection unit detects the edge of the measurement pattern image. The area determination unit interpolates the detected edge to determine a pattern boundary of the measurement pattern image. The region extraction unit extracts a region surrounded by the determined pattern boundary as the measurement pattern region.

14. The length measuring method according to claim 13, characterized in that: The computer system stores a length measurement recipe, which is data indicating the length measurement conditions of the measurement pattern image. The length measurement edge detection unit includes a length measurement edge learning unit and a length measurement edge inference unit, and detects the edge of the measurement pattern image within the length measurement cursor defined in the length measurement process through interactive machine learning. The length measurement edge learning unit learns a learning model using training data representing the edge of the measurement pattern image. The length measurement edge estimation unit estimates the edge of the measurement pattern image within the length measurement cursor using the learning model. The edges of the measurement pattern image deviating from the boundaries of the measurement pattern image among the edges of the measurement pattern image inferred by the learning model are deleted and used as new training data, and the learning model is learned by the length-measuring edge learning unit.

15. The length measuring method according to claim 12, characterized in that: The image conversion model creation unit further includes a region specifying unit that specifies a region in which the measurement pattern image included in the learning captured image exists.

16. The length measuring method according to claim 12, characterized in that: The computer system further comprises an image generation model making unit. The image generation unit of the image conversion model generation unit generates a simulated measurement pattern image in the measurement pattern region of the label image using the image generation model generated by the image generation model generation unit.

17. The length measuring method according to claim 16, characterized in that: The image generation model making unit includes a region segmentation unit and a learning unit. The region dividing unit of the image generation model creating unit generates a label image divided into a measurement pattern region as a region of the measurement pattern image included in the captured image for learning and a background region other than the measurement pattern region. The learning unit of the image generation model preparation unit learns the image generation model using training data that is a combination of the label image and the captured image for learning or an image obtained by performing image processing on the captured image for learning to enhance the measurement pattern image.

18. The length measuring method according to claim 17, characterized in that: The region segmentation unit of the image generation model creation unit includes a length measurement edge detection unit, a region identification unit, and a region extraction unit. The length measurement edge detection unit detects the edge of the measurement pattern image. The area determination unit interpolates the detected edge to determine a pattern boundary of the measurement pattern image. The region extraction unit extracts a region surrounded by the determined pattern boundary as the measurement pattern region.

19. The length measuring method according to claim 18, characterized in that: The computer system stores a length measurement recipe, which is data indicating the length measurement conditions of the measurement pattern image. The length measurement edge detection unit includes a length measurement edge learning unit and a length measurement edge inference unit, and detects the edge of the measurement pattern image within the length measurement cursor defined in the length measurement process through interactive machine learning. The length measurement edge learning unit learns a learning model using training data representing the edge of the measurement pattern image. The length measurement edge estimation unit estimates the edge of the measurement pattern image within the length measurement cursor using the learning model. The edges of the measurement pattern image deviating from the boundary of the measurement pattern image among the edges of the measurement pattern image inferred by the learning model are deleted and used as new training data, and the learning model is learned by the length-measuring edge learning unit.

Images