Information processing device, inspection device, information processing method, inspection method, and learning method

Abstract

The present invention provides an information processing device, an inspection device, an information processing method, an inspection method, and a learning method that can achieve high accuracy. [Solution] The information processing device 200 according to this embodiment includes a learning unit 210 for training a model, an image acquisition unit 220 for acquiring an image captured image CI, a reference image generation unit 230 for generating a reference image RI, and an evaluation unit 240 for evaluating the object 300 based on a comparison between the reference image RI and the image captured image CI. The learning unit 210 includes a structure representation data acquisition unit 211 for acquiring a plurality of structure representation data, a class classification unit 212 for classifying the plurality of structure representation data into one of a plurality of classes, a representative position acquisition unit 213 for selecting representative data from the structure representation data belonging to the same class and acquiring a representative position which is the position corresponding to that class, and a training unit 214 for training the model.

Description

Technical Field

[0001] The present disclosure relates to an information processing apparatus, an inspection apparatus, an information processing method, an inspection method, and a learning method.

Background Art

[0002] Patent Document 1 describes an inspection method for inspecting a photomask by comparing a captured image of a photomask manufactured based on design data with a reference image generated from the design data. The inspection method of Patent Document 1 generates a reference image from design data using a machine learning model.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] A machine learning model such as that of Patent Document 1 is preferably configured or customized according to the object to be inspected. Further, the machine learning model is preferably configured in a shorter time as a more highly accurate model. Therefore, it is important to appropriately and quickly select teacher images used for learning the machine learning model. It is desired to improve the accuracy of the machine learning model and thereby improve the inspection accuracy of the object.

[0005] The present disclosure has been made in view of such problems, and provides an information processing apparatus, an inspection apparatus, an information processing method, an inspection method, and a learning method capable of quickly and highly accurately improving a model for generating a reference image.

Means for Solving the Problems

[0006] An information processing device according to one aspect of this embodiment includes a learning unit for training a model, an image acquisition unit for acquiring an image of an object, a reference image generation unit for generating a reference image based on the design data of the object, and an evaluation unit for evaluating the object based on a comparison of the reference image and the image. The learning unit includes a structure representation data acquisition unit for acquiring a plurality of structure representation data corresponding to a plurality of positions of the object, and a feature quantity of the feature vector in the plurality of structure representation data, and based on the feature quantities of the plurality of structure representation data, it evaluates the plurality of structure representation data into one of a plurality of classes. The system includes a classification unit that classifies objects into classes, a representative position acquisition unit that selects representative data from the structural representation data belonging to the same class and acquires a representative position which is the position of the object corresponding to the representative data corresponding to the class, and a training unit that trains the model by including information based on at least one portion of the design data corresponding to the representative position and information based on the captured image of the portion of the object corresponding to that at least one representative position in the training data, and the reference image generation unit generates the reference image based on the design data of the object and the trained model.

[0007] In the above-described information processing device, the structural representation data may be an image generated based on the design data of the object.

[0008] In the above-described information processing device, the structural representation data may also be the captured image of the object.

[0009] In the above-described information processing device, the structural representation data may also be vector data included in the design data of the object.

[0010] In the above-described information processing device, the feature vector may include, as a feature quantity, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more.

[0011] In the above-described information processing device, the representative position acquisition unit may select the structural representation data that is closest to the centroid position in a feature vector space with multiple feature quantities as coordinate axes, from among the multiple structural representation data included in the class, as the representative data.

[0012] In the above-described information processing device, the format and properties of the structural representation data may be the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object.

[0013] In the above-described information processing device, the format and properties of the structural representation data may differ from the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object.

[0014] An inspection apparatus according to one aspect of this embodiment comprises an imaging device for imaging the object to be imaged and the information processing device described above.

[0015] An information processing method according to one aspect of this embodiment includes the steps of training a model, acquiring an image of an object, generating a reference image based on design data of the object, and evaluating the object based on a comparison between the reference image and the image, wherein the step of training the model includes the steps of acquiring a plurality of structural representation data corresponding to a plurality of positions of the object, acquiring feature quantities of feature vectors in the plurality of structural representation data and classifying the plurality of structural representation data into one of a plurality of classes based on the feature quantities of the plurality of structural representation data, selecting representative data to be representative from the structural representation data belonging to the same class and acquiring a representative position which is the position of the representative data corresponding to the class relative to the object, and training the model by including information based on a portion of the design data corresponding to at least one representative position and information based on the image of the portion of the object corresponding to that at least one representative position in the object into the training data, wherein the step of generating the reference image generates the reference image based on the design data of the object and the trained model.

[0016] In the above-described information processing method, in the step of acquiring the structural representation data, the structural representation data may be an image generated based on the design data of the object.

[0017] In the above-described information processing method, in the step of acquiring the structural representation data, the structural representation data may also be the captured image of the object.

[0018] In the above-described information processing method, in the step of acquiring the structural representation data, the structural representation data may be vector data included in the design data of the object.

[0019] In the above-described information processing method, the feature vector may include, as feature quantities, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more.

[0020] In the above information processing method, in the step of obtaining the representative position, the structural representation data that is closest to the centroid position in a feature vector space with multiple features as coordinate axes may be selected from among the multiple structural representation data included in the class as the representative data.

[0021] In the above-described information processing method, the format and properties of the structural representation data may be the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object.

[0022] In the above-described information processing method, the format and properties of the structural representation data may differ from the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object.

[0023] An inspection method according to one aspect of this embodiment comprises the steps of imaging an object and performing information processing using the information processing method described above.

[0024] A learning method according to one aspect of the present embodiment includes: a step of obtaining a plurality of structural representation data corresponding to a plurality of positions of an object; obtaining feature amounts of feature vectors in the plurality of structural representation data, and classifying the plurality of structural representation data into any one of a plurality of classes based on the feature amounts of the plurality of structural representation data; selecting representative data as a representative from the structural representation data belonging to the same class, and obtaining a representative position that is the position corresponding to the object of the representative data corresponding to the class; and training a model by including, in training data, information based on at least a part corresponding to the at least one representative position in the design data of the object and information based on an image of the object captured at the at least one representative position corresponding to the object in the object.

[0025] In the above learning method, in the step of obtaining the structural representation data, the structural representation data may be an image generated based on the design data of the object.

[0026] In the above learning method, in the step of obtaining the structural representation data, the structural representation data may be an image of the object captured.

[0027] In the above learning method, in the step of obtaining the structural representation data, the structural representation data may be vector data included in the design data of the object.

[0028] In the above learning method, the feature vector may include, as the feature amount, at least any one of a differential value of a luminance change in a predetermined direction, a direction in which the luminance changes by a predetermined amount or more, and an interval between pixels indicating the luminance of a predetermined amount or more.

[0029] In the above learning method, in the step of obtaining the representative position, among the plurality of structural representation data included in the class, the structural representation data closest to the centroid position in the feature vector space with a plurality of feature amounts as coordinate axes may be selected as the representative data.

[0030] In the learning method described above, the format and properties of the structural representation data may be the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object.

[0031] In the learning method described above, the format and properties of the structural representation data may differ from the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object. [Effects of the Invention]

[0032] This disclosure provides an information processing device, an inspection device, an information processing method, an inspection method, and a learning method that can achieve high accuracy. [Brief explanation of the drawing]

[0033] [Figure 1] This is a schematic diagram illustrating an inspection device according to Embodiment 1. [Figure 2] This is a schematic diagram illustrating the outline of the automatic calibration point pickup function in the inspection device according to Embodiment 1. [Figure 3] This is a diagram illustrating the configuration of an imaging device in an inspection apparatus according to Embodiment 1. [Figure 4] This is a configuration diagram illustrating another imaging device in the inspection apparatus according to Embodiment 1. [Figure 5] This is a block diagram illustrating the configuration of the information processing device in the inspection device according to Embodiment 1. [Figure 6] This figure illustrates structural representation data generated by the structural representation data acquisition unit in the information processing device according to Embodiment 1. [Figure 7]This figure illustrates the feature vectors of structured representation data distributed on the feature vector space by the classification unit in the information processing device according to Embodiment 1. [Figure 8] This figure illustrates the classification of structural representation data performed by the classification unit in the information processing device according to Embodiment 1. [Figure 9] This figure illustrates the selection of a representative position in the representative position acquisition unit of the information processing device according to Embodiment 1. [Figure 10] This flowchart illustrates an information processing method using the information processing device according to Embodiment 1. [Figure 11] This flowchart illustrates a learning method using a learning unit in the information processing device according to Embodiment 1. [Figure 12] This is a flowchart illustrating an example of the inspection method according to Embodiment 1. [Modes for carrying out the invention]

[0034] Embodiments of the present disclosure will be described below with reference to the drawings. The following description illustrates preferred embodiments of the present disclosure and does not limit the scope of the present disclosure to the following embodiments. In the following description, the same reference numerals indicate substantially the same thing. Some reference numerals may be omitted to avoid cluttering the drawings.

[0035] <Embodiment 1> Embodiment 1 will now be described. First, the <inspection device> will be described, followed by the <imaging device> and <information processing device> within the inspection device. Then, the <information processing method> and <learning method> will be described, followed by the <inspection method>.

[0036] <Inspection equipment> An inspection apparatus according to Embodiment 1 will now be described. Figure 1 is a schematic diagram illustrating an inspection apparatus 1 according to Embodiment 1. As shown in Figure 1, the inspection apparatus 1 according to this embodiment includes an imaging device 100 and an information processing device 200. In Figure 1, the imaging device 100 and the information processing device 200 are shown separately. However, the inspection apparatus 1 may be an integrated unit with the imaging device 100 and the information processing device 200, or the imaging device 100 and the information processing device 200 may function as separate units.

[0037] The inspection device 1 of this embodiment inspects the object 300. For example, the inspection device 1 inspects for defects present in the object 300. The object 300 may be an EUV (Extreme Ultra Violet) photomask used in lithography using EUV light. The EUV photomask is simply referred to as the EUV mask 310. The object 300 may also be a photomask used in lithography using light other than EUV light. Note that the object 300 is not limited to a photomask, but may also be a semiconductor substrate or a semiconductor device, as long as a pattern is formed on it.

[0038] In the following, the object 300 may be described as an EUV mask 310 as an example. In that case, the inspection device 1 is an EUV mask inspection device that inspects the EUV mask 310. The inspection device 1 inspects the EUV mask 310 by capturing an image CI of the EUV mask 310 including the pattern and comparing the captured image CI with a reference image RI. The following describes the outline of the inspection performed by the inspection device 1.

[0039] (1) The inspection device 1 first converts the design data D10 of the EUV mask 310 into a reference image RI. The design data D10 may include design CAD (Computer Aided Design) data. The design data D10 may also include vector data. The process of converting the design data D10 into a reference image RI is called rendering.

[0040] (2) Next, the inspection device 1 acquires an image CI by imaging the EUV mask 310 with the imaging device 100. The inspection device 1 then compares the reference image RI and the acquired image CI and detects the difference. The inspection device 1 detects defects in the EUV mask 310 from the difference obtained by the image comparison.

[0041] In this embodiment, the inspection device 1 generates and trains a rendering model M10 (transformer) that performs the conversion process before carrying out (1). The rendering model M10 may sometimes be simply referred to as the model. The rendering model M10 is generated and trained using machine learning techniques. One of the features of this embodiment is that it automatically selects positions (calibration points) on the EUV mask 310 that serve as training data for machine learning the rendering model M10.

[0042] Figure 2 is a schematic diagram illustrating the overview of the Auto Calibration Point Pickup (ACPP) function in the inspection device 1 according to Embodiment 1. As shown in Figure 2, the inspection device 1 of this embodiment has the ACPP function. When the inspection device 1 receives the design data D10 of the EUV mask 310 and the inspection recipe that defines the inspection target range, it outputs a calibration point CP (Calibration Point) on the EUV mask 310 that is suitable for learning the rendering model M10. This makes it possible to speed up and improve the accuracy of the generation and learning of the rendering model M10.

[0043] By the way, in order to create a rendering model M10 with sufficient rendering accuracy, it is necessary to appropriately select calibration points CP, which are fundamental elements that can reconstruct the pattern shape on the object 300, and to collect data at these calibration points CP. However, performing the selection of these calibration points CP and the collection of data for them by visual inspection and manual work by the user is undesirable in terms of time and cost.

[0044] Therefore, this embodiment automatically selects calibration points CP for the rendering model M10 and automatically collects data to be used for training. For this purpose, this embodiment uses image feature extraction technology and statistical analysis technology to select the minimum necessary calibration points CP so as to cover variations in pattern shapes on the object 300, such as a photomask.

[0045] Specifically, as an example, the inspection device 1 of this embodiment first places multiple positions on the object 300 that are to be used as candidates for calibration point CP on the object 300. Then, at each candidate position, the inspection device 1 collects a pattern image as structural representation data, which is an image (for example, rasterized) of the design data D10 described as vector data.

[0046] Next, the inspection device 1 extracts feature vectors from the pattern images that numerically represent the pattern shapes of the objects 300. Then, the inspection device 1 performs a classification of the extracted set of feature vectors, for example by applying clustering. The inspection device 1 finds a grouping such that pattern images with similar feature vectors are grouped into one class, and patterns with significantly different feature vectors belong to separate classes.

[0047] The inspection device 1 selects representative data from each class thus formed and outputs the position on the object 300 corresponding to this representative data as a calibration point CP. In this way, the inspection device 1 has ACPP functionality. The configurations of the <imaging device> and <information processing device> in the inspection device 1 of this embodiment will be described below.

[0048] <Imaging device> First, the imaging device 100 will be described with reference to the figures. Figure 3 is a configuration diagram illustrating the imaging device 100 in the inspection apparatus 1 according to Embodiment 1. Figure 4 is a configuration diagram illustrating another imaging device 100a in the inspection apparatus 1 according to Embodiment 1. As shown in Figure 3, the imaging device 100 may image the EUV mask 310 using transmitted illumination. Also, as shown in Figure 4, the imaging device 100a may image the EUV mask 310 using reflected illumination. As shown in Figure 3, the imaging device 100 includes an illumination light source 110, an illumination optical system 120, a lens 130, a stage 140, a lens 150, a detection optical system 160, and a detector 170.

[0049] In this explanation, we will use an EUV mask 310 with a pattern 311 as the object 300. However, the object 300 is not limited to an EUV mask 310; it may also be a mask used for lithography other than EUV light with a pattern 311, or a semiconductor substrate or semiconductor device.

[0050] The illumination light source 110 generates illumination light L10 to illuminate the EUV mask 310. The illumination light L10 from the illumination light source 110 is incident on the illumination optical system 120. The illumination optical system 120 is equipped with optical components such as relay lenses and mirrors, and guides the illumination light L10 to the lens 130. The illumination optical system 120 may also be equipped with an optical scanner and an autofocus (AF) function. The illumination light L10 is focused by the lens 130 and incident on the EUV mask 310. The lens 130 focuses the illumination light L10 on the pattern surface of the EUV mask 310 where the pattern 311 is formed. As a result, the EUV mask 310 is illuminated.

[0051] The transmitted light L20 that has passed through the EUV mask 310 passes through the stage 140, which is transparent to the transmitted light L20, and enters the lens 150. The lens 150 is an objective lens and focuses the transmitted light L20 from the EUV mask 310. The transmitted light L20 enters the detection optical system 160 via the lens 150. The detection optical system 160 is equipped with optical components such as an imaging lens and a mirror and guides the transmitted light L20 to the detector 170. The detection optical system 160 forms an image of the EUV mask 310 on the light-receiving surface of the detector 170.

[0052] The detector 170 is a line sensor or a two-dimensional array sensor such as a CCD (Charged Coupled Device) or CMOS camera (Complementary Metal Oxide Semiconductor) containing multiple pixels. A TDI (Time Delay Integration) sensor can also be used as the detector 170. Therefore, the detector 170 images the EUV mask 310 on which the pattern 311 is provided. The reflectance and transmittance to the illumination light L10 differ depending on whether the pattern 311 is present or not. For example, in the case of the EUV mask 310, the transmittance is lower where the pattern 311 is present and higher where it is not. Therefore, the amount of light received changes depending on whether the pattern 311 is present or not. Note that the difference in transmittance depending on the presence or absence of the pattern is just one example, and the opposite may also be true.

[0053] The EUV mask 310 is placed on the stage 140. The stage 140 is an XY stage, and moves the EUV mask 310 in the X-axis and Y-axis directions. The movement coordinates of the stage 140 are input to the information processing device 200. While the stage 140 is moving the EUV mask 310, the detector 170 images the EUV mask 310. In this way, an image CI of the entire EUV mask 310 or a desired region can be obtained. Since the transmittance to the illumination light L10 differs depending on whether the pattern 311 is present or not, the brightness value, i.e., the intensity of the detection signal, differs depending on whether the pattern 311 is present or not.

[0054] The detector 170 outputs a detection signal to the information processing device 200 according to the amount of light received. This inputs the captured image CI to the information processing device 200. Each pixel of the captured image CI has a grayscale value set according to the amount of light received. The information processing device 200 performs image processing on the detection signal. For example, the information processing device 200 is a computer equipped with a processor, memory, etc., as will be described later.

[0055] As shown in Figure 4, the EUV mask 310 may also be imaged using an imaging device 100a that employs reflective illumination. The imaging device 100a includes an illumination light source 110a, an illumination optical system 120a, a mirror 130a, a stage 140, a detection optical system 160a, and a detector 170. When illuminating and imaging the EUV mask 310 with light of a wavelength in the EUV region as illumination light L30, it is preferable to configure the imaging device 100a as a reflective optical system.

[0056] The illumination light source 110a generates illumination light L30 to illuminate the EUV mask 310. The illumination light L30 from the illumination light source 110a is incident on the illumination optical system 120a. The illumination optical system 120a is equipped with optical components such as an elliptical reflector and guides the illumination light L30 to the mirror 130a. The illumination optical system 120a may also be equipped with an optical scanner or AF function. The illumination light L30 is reflected by the mirror 130a and incident on the EUV mask 310. The mirror 130a focuses the illumination light L30 onto the pattern surface of the EUV mask 310 where the pattern 311 is formed. As a result, the EUV mask 310 is illuminated.

[0057] The reflected light L40 reflected by the EUV mask 310 enters the detection optical system 160a. The detection optical system 160a includes optical components such as a mirror and guides the reflected light L40 to the detector 170. The detection optical system 160a forms an image of the EUV mask 310 on the light-receiving surface of the detector 170. The detector 170 outputs a detection signal corresponding to the amount of light received to the information processing device 200. As a result, the captured image CI is input to the information processing device 200.

[0058] <Information Processing Device> Figure 5 is a block diagram illustrating the configuration of the information processing device 200 in the inspection apparatus 1 according to Embodiment 1. As shown in Figure 5, the information processing device 200 comprises a learning unit 210, an image acquisition unit 220, a reference image generation unit 230, an evaluation unit 240, a learning storage unit 250, and a control unit 260. The learning unit 210 includes a structure representation data acquisition unit 211, a classification unit 212, a representative position acquisition unit 213, and a training unit 214. The control unit 260 includes a processor PRC, a memory MMR, a storage device STR, and a user interface UI. The information processing device 200 includes information processing equipment such as a PC (Personal Computer), a server, and a tablet.

[0059] First, the functions of the control unit 260 will be explained. The storage device STR stores programs for the processes to be executed by each component of the information processing device 200. The processor PRC loads the programs from the storage device STR into the memory MMR and executes them. In this way, the processor PRC realizes the functions of each component of the information processing device 200, such as the learning unit 210, the image acquisition unit 220, the reference image generation unit 230, and the evaluation unit 240. The user interface UI may include input devices such as a keyboard, mouse, and imaging equipment, as well as output devices such as a display, printer, and speaker.

[0060] Each component of the information processing device 200 may be implemented with dedicated hardware. Furthermore, some or all of each component may be implemented by general-purpose or dedicated circuits and processors (PRCs), or combinations thereof. These may be implemented by a single chip or by multiple chips connected via a bus. Some or all of each component may be implemented by a combination of the aforementioned circuits and processors (PRCs) and a program. As the processor (PRC), a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field-programmable Gate Array), a quantum processor (quantum computer control chip), etc., can be used.

[0061] Furthermore, if some or all of the components of the information processing device 200 are implemented by multiple information processing devices or circuits, these devices and circuits may be centrally located or distributed. For example, the information processing devices and circuits may be implemented in a form where they are connected via a communication network, such as a client-server system or a cloud computing system. Also, the functions of the information processing device 200 may be provided in SaaS (Software as a Service) format.

[0062] The learning unit 210 trains the rendering model M10. The rendering model M10 may be an image generation model that generates a reference image RI from the design data D10. The learning unit 210 trains the rendering model M10 by operating the structural representation data acquisition unit 211, the classification unit 212, the representative position acquisition unit 213, and the training unit 214. The rendering model M10 includes a transformation function generated by machine learning using information based on the design data D10 in the portion corresponding to the representative position and information based on the captured image CI of the object 300 in the portion corresponding to the representative position as training data. The rendering model M10 includes a transformation function that generates a reference image RI from the design data D10.

[0063] Figure 6 is a diagram illustrating structural representation data acquired by the structural representation data acquisition unit 211 in the information processing device 200 according to Embodiment 1. As shown in Figure 6, the structural representation data acquisition unit 211 acquires multiple structural representation data corresponding to multiple positions P10 in the design data D10 of the object 300. For example, the structural representation data acquisition unit 211 acquires multiple pattern images F10 corresponding to multiple positions P10 of the object 300 as structural representation data based on the design data D10 of the object 300. Specifically, the structural representation data acquisition unit 211 generates and acquires multiple pattern images F10 corresponding to multiple positions P10 by rasterizing the vector data of the parts of the object 300 corresponding to multiple positions P10 in the design data D10. In other words, the structural representation data acquisition unit 211 converts design data D10, such as design CAD data expressed as vector data, into pattern images F10 that can be processed as images. In this way, the structural representation data acquisition unit 211 generates and acquires multiple pattern images F10 as multiple structural representation data based on the design data D10. In the following description, the structural representation data acquisition unit 211 will acquire N pattern images F10 (F11 to F1N) associated with N positions on the object 300 as multiple structural representation data.

[0064] The classification unit 212 classifies multiple structural representation data into one of several classes. The classification procedure will be explained below using clustering as an example. Figure 7 is a diagram illustrating V1 to V3 of the feature vectors V1 to VN of structural representation data distributed on the feature vector space by the classification unit 212 in the information processing device 200 according to Embodiment 1. As shown in Figure 7, first, the classification unit 212 acquires the feature quantities of multiple structural representation data. In other words, the classification unit 212 acquires the feature quantities of the feature vectors V1 to VN in the multiple structural representation data. Thus, if the structural representation data consists of N pattern images F10 (F11 to F1N) generated from multiple positions P10 in the design data D10, the classification unit 212 may acquire the feature quantities of the feature vectors V1 to VN in the multiple pattern images F10 (F11 to F1N).

[0065] Next, the classification unit 212 performs clustering based on the features of multiple structural representation data. Specifically, the classification unit 212 performs clustering based on the features of multiple pattern images F10. For example, the classification unit 212 uses an image feature extraction function to extract the features of the feature vectors V1 to VN of the pattern images F11 to F1N, which are structural representation data. Then, the classification unit 212 maps the extracted feature vectors V1 to VN onto the feature vector space. The classification unit 212 maps the feature vector space using the features C1 to C3 corresponding to each coordinate axis component of the feature vectors V1 to VN in the pattern images F11 to F1N, which are structural representation data. In other words, the classification unit 212 transforms the pattern images F11 to F1N generated by rasterization into a single point on a high-dimensional feature vector space.

[0066] The feature vector space is a space with feature quantities C1, C2, and C3 as coordinate axes. Having three feature quantities, such as C1 to C3, is just one example. The number of feature quantities can be two or fewer, or four or more. The feature vector space is not limited to a three-dimensional space; it has as many dimensions as there are feature quantities. Feature quantities are numerical values ​​obtained by performing predetermined numerical calculations on structural representation data such as captured images or images generated from design data, or on vector data included in structural representation data such as design data. For example, if the structural representation data is an image, the feature quantities of structural representation data may include at least one of the following: the differential value of brightness change in a predetermined direction, the direction in which brightness changes by a predetermined amount or more, and the interval between pixels that show brightness of a predetermined amount or more. That is, feature vectors V1 to VN may include at least one of the following as feature quantities for an image that is structural representation data: the differential value of brightness change in a predetermined direction, the direction in which brightness changes by a predetermined amount or more, and the interval between pixels that show brightness of a predetermined amount or more. Furthermore, for example, if the structural representation data is vector data, the feature quantities of the structural representation data may include at least one of the vector length, the distance between vectors, or the vector density in a given interval. That is, feature vectors V1 to VN may include at least one of the vector length, the distance between vectors, or the vector density in a given interval as feature quantities for the vector data which is the structural representation data. Note that the above feature quantities are just examples, and other elements may be used as feature quantities. Other elements may also be included. The classification unit 212 may perform the classification process using both the feature quantities for the image which is the structural representation data and the feature quantities for the vector data which is the structural representation data.

[0067] Figure 8 illustrates the classification of structural representation data performed by the classification unit 212 in the information processing device 200 according to Embodiment 1. As shown in Figure 8, the classification unit 212 classifies multiple structural representation data into one of the multiple classes G1 to G4. Specifically, for example, the classification unit 212 classifies multiple pattern images F10 having feature vectors similar to feature vector V1 in the feature vector space as one class G1. The classification unit 212 also classifies multiple pattern images F10 having feature vectors similar to feature vector V2 in the feature vector space as one class G2. Furthermore, similarly, the classification unit 212 classifies multiple pattern images F10 having feature vectors similar to feature vectors V3 and V4 in the feature vector space as classes G3 and G4, respectively. The above example of four classes G1 to G4 is just one example. The number of classes may be three or fewer, or five or more.

[0068] Thus, the classification unit 212 classifies the set of feature vectors V1 to VN in the feature vector space so that similar vectors (those close together in the feature vector space) are grouped into the same class. The classification unit 212 also groups the set of feature vectors V1 to VN in the feature vector space so that different vectors (those far apart in the feature vector space) are grouped into different classes. Note that the classes to be classified may include classes that classify only unique structural representation data, such as outliers, that do not belong to other specific classes. Also, there may be groups that have only one element. These are also included in the process of classifying into one of the classes.

[0069] In this way, the classification unit 212 classifies the multiple pattern images F10 into one of the multiple classes G1 to G4.

[0070] The range of a single class in the feature vector space may be set by predetermined conditions. For example, the range of each class in the feature vector space may be set in advance, or, as mentioned above, a threshold may be set for the distance between feature vectors, and the data may be classified based on whether it is closer or further than the threshold. The classification unit 212 only needs to be able to classify the structural representation data based on the features of multiple structural representation data, and the classification unit 212 may classify the structural representation data by other processing. For example, the classification unit 212 may classify the structural representation data by applying a rule-based processing to the structural representation data that determines whether the structural representation data has a predetermined amount of a predetermined feature. Alternatively, the classification unit 212 may classify the structural representation data by applying a trained classifier for classifying the structural representation data to the structural representation data. Or, the classification unit 212 may classify the structural representation data by a combination of these processes.

[0071] Figure 9 illustrates the selection of representative data by the representative position acquisition unit 213 in the information processing device 200 according to Embodiment 1. As shown in Figure 9, the representative position acquisition unit 213 selects representative data from a plurality of structural representation data belonging to the same class. The representative position acquisition unit 213 performs this selection of representative data for a plurality of classes G1 to G4. The representative position acquisition unit 213 then selects a plurality of representative data corresponding to each of the plurality of classes G1 to G4. In other words, the representative position acquisition unit 213 selects representative data from a plurality of structural representation data belonging to the same class and acquires a representative position which is the position corresponding to the object 300 of the representative data corresponding to that class.

[0072] Specifically, the representative position acquisition unit 213 selects a representative image that shows a representative feature vector from multiple pattern images F10 belonging to the same class as representative data. The representative position acquisition unit 213 performs this selection of representative images as representative data for multiple classes G1 to G4. Then, the representative position acquisition unit 213 selects multiple representative images corresponding to each of the multiple classes G1 to G4 as representative data.

[0073] The representative position acquisition unit 213 may select representative data for each class under predetermined conditions. For example, the representative position acquisition unit 213 may select as representative data the structural representation data closest to the centroid position in a feature vector space with multiple features as coordinate axes, from among the multiple structural representation data included in the class. Alternatively, the representative position acquisition unit 213 may select as representative data the structural representation data closest to the geometric center position of the region enclosed by the structural representation data located at the edge of the class, from among the multiple structural representation data included in the class. Furthermore, the representative position acquisition unit 213 may select as representative data the structural representation data with a large number of structural representation data located at the same coordinates, from among the multiple structural representation data included in the class.

[0074] Next, the representative position acquisition unit 213 acquires representative positions for each of the representative data of multiple classes G1 to G4 by associating the pattern image F10 corresponding to the representative data selected from each class G1 to G4 with a position P10 on the object 300. The position P10 associated with the representative data is called the representative position, and this becomes the calibration point CP. For example, the pattern image F11 selected as the representative data for class G1 is associated with position P101. Therefore, position P101 is the representative position and becomes the calibration point CP1. Similarly, the positions P102 to P104 associated with the pattern images F12 to F14 selected as the representative data for groups G2 to G4 are representative positions and become the calibration points CP2 to CP4. In this way, the representative position acquisition unit 213 acquires the positions (P101 to P104) on the object 300 associated with the selected representative data as representative positions. The representative position acquisition unit 213 may acquire multiple representative positions corresponding to multiple classes G1 to G4, or it may acquire only representative positions for a desired subset of classes G1 to G4.

[0075] The training unit 214 trains the rendering model M10 using as training data information based on the portion of the design data D10 corresponding to the representative positions (P101 to P104) for multiple representative positions (P101 to P104), and information based on the captured image C1 of the portion of the object 300 corresponding to the representative positions (P101 to P104).

[0076] In other words, the training unit 214 trains the rendering model M10 using as training data information the portion of the design data D10 corresponding to the calibration points CP(CP1~CP4) for multiple calibration points CP(CP1~CP4), and the captured image C1 of the portion of the object 300 corresponding to the calibration points CP(CP1~CP4).

[0077] Furthermore, the training unit 214 trains the rendering model M10 by including information based on a portion corresponding to at least one representative position in the design data and information based on an image C1 of the portion of the object 300 corresponding to that at least one representative position in the training data. Here, the training unit 214 may, for example, depending on the training status of the rendering model M10, appropriately include information for only some of the representative positions among the multiple representative positions in the training data to train the rendering model M10. Alternatively, the training unit 214 may also train the rendering model M10 by including information for other positions that do not belong to the representative positions acquired by the representative position acquisition unit 213 in the training data.

[0078] The information based on the portion corresponding to the representative positions (P101~P104) in design data D10, or the information based on the portion corresponding to the calibration points CP (CP1~CP4), is, for example, pattern images F11~F14 generated based on the vector data in the portion corresponding to the representative positions (P101~P104) in design data D10, or pattern images F11~F14 generated based on the vector data in the portion corresponding to the calibration points CP (CP1~CP4). Alternatively, the information based on the portion corresponding to the representative positions (P101~P104) in design data D10, or the information based on the portion corresponding to the calibration points CP (CP1~CP4), may be images obtained by correcting or normalizing the pattern images F11~F14 generated based on the vector data in the portion corresponding to the representative positions (P101~P104) in design data D10, or the pattern images F11~F14 generated based on the vector data in the portion corresponding to the calibration points CP (CP1~CP4), using a predetermined method. Furthermore, the information based on the captured image C1 of the portion corresponding to the representative positions (P101~P104) of the object 300, or the information based on the captured image C1 of the portion corresponding to the calibration points CP (CP1~CP4), is, for example, the captured image C1 of the portion corresponding to the representative positions (P101~P104) of the object 300, which has a pattern formed based on the design data D10, or the captured image C1 of the portion corresponding to the calibration points CP (CP1~CP4). Alternatively, the information based on the captured image C1 of the portion corresponding to the representative positions (P101~P104) of the object 300, or the information based on the captured image C1 of the portion corresponding to the calibration points CP (CP1~CP4), may be an image obtained by correcting or normalizing the captured image C1 of the portion corresponding to the representative positions (P101~P104) of the object 300, which has a pattern formed based on the design data D10, or the captured image C1 of the portion corresponding to the calibration points CP (CP1~CP4), in a predetermined manner.

[0079] The image acquisition unit 220 acquires an image CI from the imaging device 100. The image acquisition unit 220 acquires the image CI based on the detection signal from the detector 170 of the imaging device 100. The image acquisition unit 220 acquires a two-dimensional image of the EUV mask 310 by associating the coordinates of the stage 140 with the intensity of the detection signal. The image CI is an image acquired by imaging the object 300. The image acquisition unit 220 may also acquire an image CI that has been previously stored in a storage medium such as a storage device STR from the storage device STR.

[0080] The reference image generation unit 230 generates a reference image RI based on the design data D10 of the object 300, such as the EUV mask 310. The reference image generation unit 230 may also generate the reference image RI based on the design data D10 of the object 300 and a trained rendering model M10. Specifically, the reference image generation unit 230 generates the reference image RI from the design data D10 using the rendering model M10 trained in the learning unit 210. In other words, the reference image generation unit 230 generates the reference image RI by applying the rendering model M10, which is a converter that performs conversion processing and is trained in the learning unit 210, to the design data D10.

[0081] The evaluation unit 240 evaluates the object 300, such as the EUV mask 310, based on a comparison between the reference image RI and the captured image CI.

[0082] The learning memory unit 250 may store training data used for learning in the learning unit 210. The learning memory unit 250 may also store coefficients of the rendering model M10 learned in the learning unit 210.

[0083] <Information Processing Methods> Next, an information processing method using the information processing device 200 of this embodiment will be described. Figure 10 is a flowchart illustrating an information processing method using the information processing device 200 according to Embodiment 1. As shown in Figure 10, the information processing method of this embodiment includes a step S10 for training a model, a step S20 for acquiring an image CI of the object 300, a step S30 for generating a reference image RI based on the design data D10 of the object 300, and a step S40 for evaluating the object 300 based on a comparison between the reference image RI and the image CI.

[0084] In step S10, the learning unit 210 trains the rendering model M10. Specifically, the learning unit 210 classifies multiple structural representation data corresponding to multiple positions P10 of the object 300. The learning unit 210 selects representative data from each class of the multiple structural representation data classified into multiple classes G1 to G4, and obtains a representative position, which is the position of the object 300 associated with the representative data. Then, the learning unit 210 trains the rendering model M10 using information based on design data D10 and information based on captured image C1 for the part corresponding to the representative position as training data.

[0085] In step S20, the image acquisition unit 220 acquires, for example, an image CI of the object 300 captured by the imaging device 100. The image acquisition unit 220 may also acquire an image CI stored in a storage medium such as a storage device STR.

[0086] In step S30, the reference image generation unit 230 generates a reference image RI based on the design data D10 of the object 300. Specifically, the reference image generation unit 230 generates a reference image RI based on the design data D1 of the object 300 and the trained rendering model M10. Step S30 may be performed before step S20 or in parallel with step S20.

[0087] In step S40, the evaluation unit 240 compares the reference image RI and the captured image CI, and evaluates defects and other issues contained in the object 300 based on the difference between the two.

[0088] <Learning Method> Next, a learning method using the learning unit 210 of this embodiment will be described. Figure 11 is a flowchart illustrating a learning method using the learning unit 210 in the information processing device 200 according to Embodiment 1. As shown in Figure 11, the learning method of this embodiment includes a step S11 of acquiring multiple structural representation data corresponding to multiple positions P10 of the object 300, a step S12 of classifying the multiple structural representation data into one of multiple classes, a step S13 of acquiring representative positions, and a step S14 of training a model. Specifically, step S13 is a step of selecting multiple representative data corresponding to each of the multiple classes and acquiring representative positions, which are the positions in the object 300 associated with the representative data. Specifically, step S14 is a step of training a model using information based on the portion of the design data D10 corresponding to the representative positions and information based on the captured image C1 of the portion of the object 300 corresponding to the representative positions as training data.

[0089] In step S11, the structural representation data acquisition unit 211 acquires multiple structural representation data corresponding to multiple locations P10 of the object 300. Specifically, the structural representation data acquisition unit 211 generates and acquires multiple pattern images F10 corresponding to multiple locations P10 of the object 300 as structural representation data, based on the design data D10 of the object 300.

[0090] In step S12, the classification unit 212 obtains feature quantities from the feature vectors of multiple structural representation data and performs clustering based on the feature quantities of the multiple structural representation data. Then, the classification unit 212 classifies the multiple structural representation data into one of the multiple classes G1 to G4. Specifically, the classification unit 212 obtains feature quantities from the feature vectors of multiple pattern images F10 and performs clustering based on the feature quantities of the multiple pattern images F10. Then, the classification unit 212 classifies the multiple pattern images F10 into one of the multiple classes G1 to G4. Note that the classification unit 212 may perform classification by applying rule-based processing or a pre-trained classifier instead of, or in conjunction with, the clustering process. As mentioned above, the classification unit 212 only needs to be able to classify the structural representation data based on the feature quantities of the multiple structural representation data.

[0091] In step S13, the representative position acquisition unit 213 selects representative data from multiple structural representation data belonging to the same class. The representative position acquisition unit 213 performs this selection of representative data for multiple classes. That is, the representative position acquisition unit 213 selects representative data from multiple structural representation data belonging to the same class and acquires a representative position which is the position corresponding to the object 300 of the representative data corresponding to that class. In this way, the representative position acquisition unit 213 selects multiple representative data corresponding to each of the multiple classes. Specifically, the representative position acquisition unit 213 selects a representative image from multiple pattern images F10 belonging to the same class for multiple classes G1 to G4. Then, the representative position acquisition unit 213 selects multiple representative images corresponding to each of the multiple classes G1 to G4. Then, the representative position acquisition unit 213 acquires a representative position which is the position on the object 300 associated with the multiple representative data (representative images) corresponding to each of the multiple classes G1 to G4.

[0092] The representative position acquisition unit 213 may also select the structural representation data closest to the centroid position in a feature vector space with multiple features as coordinate axes from among the multiple structural representation data included in the class as the representative data.

[0093] In step S14, the training unit 214 trains the rendering model M10 using information based on the portion of the design data D10 corresponding to the representative position for multiple representative positions, and information based on the captured image C1 of the portion of the object 300 corresponding to the representative position, as training data. Specifically, the training unit 214 trains the rendering model M10 using a pattern image F10 generated based on vector data in the portion of the design data D10 corresponding to the representative position, and a captured image C1 of the portion of the object 300 corresponding to the representative position, as training data. The training unit 214 also trains the rendering model M10 by including information based on the portion of the design data D10 corresponding to at least one representative position and information based on the captured image CI of the portion of the object 300 corresponding to that at least one representative position as training data.

[0094] <Testing Method> Next, the inspection method according to Embodiment 1 will be described. Figure 12 is a flowchart illustrating the inspection method according to Embodiment 1. As shown in Figure 12, the inspection method of this embodiment comprises a step S100 of imaging the object 300 and a step S200 of performing information processing using the information processing method described above.

[0095] Next, the effects of this embodiment will be described. The inspection device 1 of this embodiment inspects the object 300 using a reference image RI generated based on the design data D10 of the object 300 and a trained rendering model M10. At that time, the inspection device 1 trains the rendering model M10 using as training data information based on the portion of the design data D10 corresponding to the representative position for multiple representative positions, and information based on the captured image C1 of the portion of the object 300 corresponding to the representative position. Here, the representative position is the position of a pattern image F10 selected as representative from among those classified based on the features of multiple pattern images F10, and represents a calibration point CP that can cover variations in the structure on the object 300. Therefore, since the trained rendering model M10 is configured or customized to match the object 300, the rendering model M10 can be made highly accurate. As a result, the inspection device 1 can make the inspection accuracy more accurate. Furthermore, the inspection device 1 can provide an information processing method that makes the inspection accuracy more accurate.

[0096] <Example 1> Next, an information processing device 200 according to a modified example 1 of Embodiment 1 will be described. In Embodiment 1 described above, the structural representation data acquisition unit 211 acquires a plurality of pattern images F10 corresponding to a plurality of positions P10 of the object 300 as structural representation data, based on the design data D10 of the object 300.

[0097] On the other hand, in this modified example 1, the structural representation data acquisition unit 211 acquires multiple captured images C1 corresponding to multiple positions P10 in the object 300 on which a pattern has been formed based on the design data D10, as structural representation data. In other words, the structural representation data acquisition unit 211 may use each image obtained by extracting multiple parts from the captured image CI as structural representation data. The rest of the configuration is the same as in Embodiment 1.

[0098] Next, the information processing method for Modification 1 of this embodiment will be described. In this embodiment, in step S11 in Figure 11, the structural representation data acquisition unit 211 acquires multiple captured images CI corresponding to multiple positions P10 of the object 300 as structural representation data. The other steps are the same as in this embodiment 1.

[0099] According to this modified example 1, the structural representation data acquisition unit 211 uses images of multiple positions P10 in the captured image CI of the object 300 as structural representation data. Therefore, since a representative position (calibration point CP) is selected based on the actually manufactured object 300, it can be adapted to the actual conditions of the object 300. Other configurations and effects are included in the description of Embodiment 1.

[0100] <Modification 2> Next, an information processing device 200 according to a modified example 2 of Embodiment 1 will be described. In Embodiment 1 described above, the structural representation data acquisition unit 211 acquires a plurality of pattern images F10 corresponding to a plurality of positions P10 of the object 300 as structural representation data, based on the design data D10 of the object 300.

[0101] On the other hand, in this modified example 2, the structural representation data acquisition unit 211 acquires multiple vector data corresponding to multiple positions P10 included in the design data D10 of the object 300 as structural representation data. In other words, as shown in Figure 6, the vector data may be used as structural representation data without rasterization. The other configurations are the same as in Embodiment 1.

[0102] Next, the information processing method for a modified example 2 of this embodiment will be described. In this modified example 2, in step S11 in Figure 11, the structural representation data acquisition unit 211 acquires multiple vector data corresponding to multiple positions P10 included in the design data D10 as structural representation data. The other steps are the same as in the first embodiment.

[0103] According to this modified example 2, the structural representation data acquisition unit 211 acquires vector data as structural representation data. Therefore, since a representative position (calibration point CP) is selected based on the result of classifying multiple vector data, the representative position can be selected with a simpler process than image processing. Other configurations and effects are described in Embodiments 1 and 2.

[0104] While embodiments of the present disclosure have been described above, the disclosure includes appropriate modifications that do not impair its purpose and advantages, and is not limited by the embodiments described above.

[0105] The structural representation data acquired by the structural representation data acquisition unit 211 and the information based on the portion corresponding to the representative position in the design data D10 used as training data by the training unit 214 may be information of the same format and nature. For example, the structural representation data may be a pattern image F10, and the information based on the portion corresponding to the representative position in the design data D10, which is the training data, may also be a pattern image F10. This is an example of the embodiment 1 described above.

[0106] The structural representation data acquired by the structural representation data acquisition unit 211 and the information based on the captured image C1 of the portion of the object 300 corresponding to a representative position, which the training unit 214 uses as training data, may be information of the same format and nature. For example, the structural representation data may be the captured image C1 of the object 300, and the information based on the captured image C1 of the portion of the object 300 corresponding to a representative position, which is the training data, may be the captured image C1. This is an example of modification 1 of the above-described embodiment 1.

[0107] Thus, the structural representation data acquired by the structural representation data acquisition unit 211 may be information with the same format and properties as either the information based on the portion corresponding to the representative position in the design data D10 used as training data by the training unit 214, or the information based on the captured image C1 of the portion corresponding to the representative position in the object 300. In other words, the format and properties of the structural representation data may be the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data D10, which is the training data, or the information based on the captured image C1 of the portion corresponding to the representative position in the object 300. This allows the structural representation data to be used as training data as well, simplifying the processing.

[0108] The structural representation data acquired by the structural representation data acquisition unit 211 may be in a different format and nature from the information based on the portion corresponding to the representative position in the design data D10, which the training unit 214 uses as training data, and the information based on the captured image CI of the portion corresponding to the representative position in the object 300. For example, the structural representation data may be vector data included in the design data D10, and the training data may be the pattern image F10 and the captured image CI. This is an example of a modified example 2 of the embodiment 1 described above. Alternatively, the structural representation data acquired by the structural representation data acquisition unit 211 may be the captured image CI, and the training data may be the pattern image F10 and the captured image CI, in which case the captured image CI of the structural representation data may be in a lower resolution format than the captured image CI of the training data. That is, the resolution of the captured image CI as structural representation data only needs to be sufficient to allow the feature quantities of the captured image CI to be grasped and classified with a predetermined accuracy, and may be lower in resolution than the captured image CI as training data. This allows for the identification and acquisition of representative locations to be performed using lower-resolution images with less computational load, while the training of the rendering model M10 can be performed using higher-resolution images to enhance learning effectiveness.

[0109] Thus, the structural representation data acquired by the structural representation data acquisition unit 211 may be information with a format and properties different from both the information based on the portion corresponding to the representative position in the design data D10, which is used as training data by the training unit 214, and the information based on the captured image C1 of the portion corresponding to the representative position in the object 300. In other words, the format and properties of the structural representation data may be different from both the information based on the portion corresponding to the representative position in the design data D10, which is the training data, and the information based on the captured image C1 of the portion corresponding to the representative position in the object 300. This makes it possible to use the structural representation data as information suitable for identifying and acquiring representative positions, and the training data as information suitable for training the rendering model M10, by using them separately.

[0110] Furthermore, combinations of the configurations of Embodiment 1, Modification 1, and Modification 2 are also within the scope of the technical concept of this disclosure. In addition, the following learning program for causing a computer to execute the learning method of the embodiment is also within the scope of the technical concept of this disclosure. (Note 1) The steps include obtaining multiple structural representation data corresponding to multiple locations of the object, The steps include obtaining the feature quantities of the feature vectors in the plurality of structured representation data, and classifying the plurality of structured representation data into one of the plurality of classes based on the feature quantities of the plurality of structured representation data, A step of selecting representative data from the structural representation data belonging to the same class, and obtaining a representative position which is the position corresponding to the object of the representative data corresponding to the class, A step of training a model by including information based on a portion of the design data of the object corresponding to at least one representative position and information based on an image of the object of the portion corresponding to that at least one representative position in the object as training data, A learning program that instructs a computer to execute a command. (Note 2) In the step of obtaining the aforementioned structural representation data, The structural representation data is an image generated based on the design data of the object. The learning program described in Appendix 1. (Note 3) In the step of obtaining the aforementioned structural representation data, The structural representation data is the captured image of the object. The learning program described in Appendix 1. (Note 4) In the step of obtaining the aforementioned structural representation data, The structural representation data is vector data included in the design data of the object. The learning program described in Appendix 1. (Note 5) The feature vector includes, as a feature quantity, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more. The learning program described in any one of the following appendices 1-4. (Note 6) In the step of obtaining the representative position, From among the multiple structural representation data included in the class, the structural representation data closest to the centroid position in a feature vector space with multiple features as coordinate axes is selected as the representative data. A learning program described in any one of the appendices 1 to 4 that causes a computer to perform the following actions. (Note 7) The format and properties of the structural representation data are the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object. The learning program described in any one of the following appendices 1-4. (Note 8) The format and properties of the aforementioned structural representation data are different from both the format and properties of the information based on the portion corresponding to the representative position in the design data which is the training data and the information based on the captured image of the portion corresponding to the representative position in the object. The learning program described in any one of the following appendices 1-4.

[0111] Furthermore, the learning program described above includes a set of instructions (or software code) for causing the computer to perform one or more of the functions described in the embodiments when loaded into a computer. The learning program may be stored in a non-temporary computer-readable medium or a physical storage medium. Examples, but not limited to, include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technologies, CD-ROM, digital versatile disc (DVD), Blu-ray® disc or other optical disc storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices. The learning program may be transmitted over a temporary computer-readable medium or a communication medium. Examples, but not limited to, include temporary computer-readable medium or a communication medium that includes electrical, optical, acoustic or other forms of propagating signals. [Explanation of symbols]

[0112] 1. Inspection device 100, 100a Imaging device 110, 110a illumination light source 120, 120a illumination optics 130 lens 130a Mirror 140 stages 150 lenses 160, 160a detection optics 170 detectors 200 Information Processing Devices 210 Learning Department 211 Structural representation data acquisition unit 212 Classification Section 213 Representative position acquisition section 214 Training Department 220 Image acquisition unit 230 Reference Image Generation Unit 240 Evaluation Department 250 Learning memory unit 260 Control Unit 300 objects 310 EUV Mask 311 patterns CI acquired images D10 Design Data F10, F11, F12, F13, F14 pattern images G1, G2, G3 classes L10 illumination light L20 transmitted light L30 illumination light L40 reflected light M10 Rendering Model MMR memory P10, P101, P102, P103, P104 position PRC Processor RI Reference Image STR storage UI User Interface

Claims

1. A learning unit that trains the model, An image acquisition unit that acquires an image of the target object, A reference image generation unit generates a reference image based on the design data of the aforementioned object, An evaluation unit that evaluates the object based on a comparison between the reference image and the captured image, Equipped with, The aforementioned learning unit, A structural representation data acquisition unit that acquires multiple structural representation data corresponding to multiple locations of the object, A classification unit that obtains feature quantities from the feature vectors in the plurality of structural representation data and classifies the plurality of structural representation data into one of the classes based on the feature quantities of the plurality of structural representation data, A representative position acquisition unit selects representative data from the structural representation data belonging to the same class and acquires a representative position which is the position corresponding to the object of the representative data corresponding to the class, A training unit that trains the model by including information based on a portion of the design data corresponding to at least one representative position and information based on an image of the portion of the object corresponding to that at least one representative position in the object as training data, It has, The reference image generation unit generates the reference image based on the design data of the object and the trained model. Information processing device.

2. The structural representation data is an image generated based on the design data of the object. The information processing apparatus according to claim 1.

3. The structural representation data is the captured image of the object. The information processing apparatus according to claim 1.

4. The structural representation data is vector data included in the design data of the object. The information processing apparatus according to claim 1.

5. The feature vector includes, as a feature quantity, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more. The information processing apparatus according to any one of claims 1 to 4.

6. The representative position acquisition unit selects, from among the multiple structural representation data included in the class, the structural representation data that is closest to the centroid position in a feature vector space with multiple features as coordinate axes, as the representative data. The information processing apparatus according to any one of claims 1 to 4.

7. The format and properties of the structural representation data are the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object. The information processing apparatus according to any one of claims 1 to 4.

8. The format and properties of the aforementioned structural representation data are different from both the format and properties of the information based on the portion corresponding to the representative position in the design data which is the training data and the information based on the captured image of the portion corresponding to the representative position in the object. The information processing apparatus according to any one of claims 1 to 4.

9. An imaging device for imaging the aforementioned object, An information processing device according to any one of claims 1 to 4, An inspection device equipped with the following features.

10. Steps to train the model, The steps include: acquiring an image of the target object, The steps include generating a reference image based on the design data of the aforementioned object, A step of evaluating the object based on a comparison between the reference image and the captured image, Equipped with, The step of training the aforementioned model is: The steps include: obtaining multiple structural representation data corresponding to multiple locations of the object; The steps include obtaining the feature quantities of the feature vectors in the plurality of structured representation data, and classifying the plurality of structured representation data into one of the plurality of classes based on the feature quantities of the plurality of structured representation data, A step of selecting representative data from the structural representation data belonging to the same class, and obtaining a representative position which is the position of the representative data corresponding to the class with respect to the object, A step of training the model by including information based on a portion of the design data corresponding to at least one representative position and information based on an image of the portion of the object corresponding to that at least one representative position in the object as training data. It has, In the step of generating the reference image, the reference image is generated based on the design data of the object and the trained model. Information processing methods.

11. In the step of obtaining the aforementioned structural representation data, The structural representation data is an image generated based on the design data of the object. The information processing method according to claim 10.

12. In the step of obtaining the aforementioned structural representation data, The structural representation data is the captured image of the object. The information processing method according to claim 10.

13. In the step of obtaining the aforementioned structural representation data, The structural representation data is vector data included in the design data of the object. The information processing method according to claim 10.

14. The feature vector includes, as a feature quantity, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more. The information processing method according to any one of claims 10 to 13.

15. In the step of obtaining the representative position, From among the multiple structural representation data included in the class, the structural representation data closest to the centroid position in a feature vector space with multiple features as coordinate axes is selected as the representative data. The information processing method according to any one of claims 10 to 13.

16. The format and properties of the structural representation data are the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object. The information processing method according to any one of claims 10 to 13.

17. The format and properties of the aforementioned structural representation data are different from both the format and properties of the information based on the portion corresponding to the representative position in the design data which is the training data and the information based on the captured image of the portion corresponding to the representative position in the object. The information processing method according to any one of claims 10 to 13.

18. The steps include: imaging the target object, A step of performing information processing using the information processing method described in any one of claims 10 to 13, A testing method that includes [a specific feature / feature].

19. The steps include obtaining multiple structural representation data corresponding to multiple locations of the object, The steps include obtaining the feature quantities of the feature vectors in the plurality of structured representation data, and classifying the plurality of structured representation data into one of the plurality of classes based on the feature quantities of the plurality of structured representation data, A step of selecting representative data from the structural representation data belonging to the same class, and obtaining a representative position which is the position corresponding to the object of the representative data corresponding to the class, A step of training a model by including information based on a portion of the design data of the object corresponding to at least one representative position and information based on an image of the object of the portion corresponding to that at least one representative position in the object as training data, A learning method that includes [something].

20. In the step of obtaining the aforementioned structural representation data, The structural representation data is an image generated based on the design data of the object. The learning method according to claim 19.

21. In the step of obtaining the aforementioned structural representation data, The structural representation data is the captured image of the object. The learning method according to claim 19.

22. In the step of obtaining the aforementioned structural representation data, The structural representation data is vector data included in the design data of the object. The learning method according to claim 19.

23. The feature vector includes, as a feature quantity, at least one of the following: the differential value of the brightness change in a predetermined direction, the direction in which the brightness changes by a predetermined amount or more, and the interval between pixels exhibiting the brightness of a predetermined amount or more. The learning method according to any one of claims 19 to 22.

24. In the step of obtaining the representative position, From among the multiple structural representation data included in the class, the structural representation data closest to the centroid position in a feature vector space with multiple features as coordinate axes is selected as the representative data. The learning method according to any one of claims 19 to 22.

25. The format and properties of the structural representation data are the same as the format and properties of either the information based on the portion corresponding to the representative position in the design data which is the training data, or the information based on the captured image of the portion corresponding to the representative position in the object. The learning method according to any one of claims 19 to 22.

26. The format and properties of the aforementioned structural representation data are different from both the format and properties of the information based on the portion corresponding to the representative position in the design data which is the training data and the information based on the captured image of the portion corresponding to the representative position in the object. The learning method according to any one of claims 19 to 22.

Images