WO2024101186A1

WO2024101186A1 - Substrate inspection method, substrate inspection device, and substrate inspection program

Info

Publication number: WO2024101186A1
Application number: PCT/JP2023/038877
Authority: WO
Inventors: 健太郎本田; 修児岩永; 達弥徳丸
Original assignee: 東京エレクトロン株式会社
Priority date: 2022-11-10
Filing date: 2023-10-27
Publication date: 2024-05-16

Abstract

A substrate inspection method according to the present invention involves using an inspection image obtained by capturing an image of a substrate to be inspected to inspect the substrate. The method involves: generating a plurality of component images in a color space from a plurality of normal images and a plurality of normal difference images, and generating respective preprocessed images for the plurality of normal images on the basis of distribution information about the distribution of respective component values for the pixels in the plurality of component images; generating an inspection difference image that is the difference between the inspection image and an average image, generating a plurality of component images in a color space from the inspection image and the inspection difference image, and generating a preprocessed image for the inspection image on the basis of distribution information about the distribution of respective component values for the pixels in the plurality of component images; and inspecting the substrate for defects on the basis of the preprocessed images for the normal images and the preprocessed image for the inspection image.

Description

Substrate inspection method, substrate inspection device, and substrate inspection program

This disclosure relates to a substrate inspection method, a substrate inspection device, and a substrate inspection program.

Patent Document 1 discloses a device that classifies defects occurring on a substrate based on an image of the substrate that is the subject of inspection.

JP 2019-124591 A

This disclosure provides technology that is useful for accurately detecting abnormalities on a substrate surface.

A substrate inspection method according to one aspect of the present disclosure is a substrate inspection method for inspecting a substrate to be inspected using an inspection image obtained by imaging the substrate to be inspected, the method comprising: generating a plurality of normal difference images which are the differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images; generating a plurality of component images in color space from the plurality of normal images and the plurality of normal difference images; generating a preprocessed image for each of the plurality of normal images based on variance information related to the variance of the component values of each pixel included in the plurality of component images; generating an inspection difference image which is the difference between the inspection image and the average image; generating a plurality of component images in color space from the inspection image and the inspection difference image; generating a preprocessed image for the inspection image based on variance information related to the variance of the component values of each pixel included in the plurality of component images; and inspecting the substrate to be inspected for defects based on the preprocessed image for the normal image and the preprocessed image for the inspection image.

FIG. 1 is a perspective view illustrating a substrate processing system according to a first embodiment. FIG. 2 is a side view illustrating an example of a coating and developing apparatus. FIG. 3 is a schematic diagram showing an example of an inspection unit. FIG. 4 is a block diagram illustrating an example of a functional configuration of the control device. FIG. 5 is a block diagram illustrating an example of a hardware configuration of the control device. FIG. 6 is a flowchart showing an example of a substrate inspection method. FIG. 7 is a schematic diagram showing an example of a method for creating a model map. FIG. 8 is a schematic diagram showing an example of an image processing method for a normal image. 9A, 9B, and 9C are schematic diagrams showing an example of an image processing method for a target image. 10A, 10B, 10C, and 10D are schematic diagrams showing an example of an image processing method for a target image. FIG. 11 is a schematic diagram showing an example of a method for creating a preprocessed image from an inspection image. FIG. 12 is a schematic diagram showing an example of a defect detection method.

Various exemplary forms are described below.

A substrate inspection method according to an exemplary embodiment of the present disclosure is a substrate inspection method for inspecting a substrate to be inspected using an inspection image obtained by imaging the substrate to be inspected, and includes: generating a plurality of normal difference images which are the differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images; generating a plurality of component images in color space from the plurality of normal images and the plurality of normal difference images; generating a preprocessed image for each of the plurality of normal images based on variance information related to the variance of the component values of each pixel included in the plurality of component images; generating an inspection difference image which is the difference between the inspection image and the average image; generating a plurality of component images in color space from the inspection image and the inspection difference image; generating a preprocessed image for the inspection image based on variance information related to the variance of the component values of each pixel included in the plurality of component images; and inspecting the substrate to be inspected for defects based on the preprocessed image for the normal image and the preprocessed image for the inspection image.

　According to the above-mentioned substrate inspection method, a preprocessed image for each of the multiple normal images is prepared based on variance information related to the variance of the component values of each pixel contained in the multiple component images in the color space generated from the multiple normal images. Also, a preprocessed image for the inspection image is prepared based on variance information related to the variance of the component values of each pixel contained in the multiple component images in the color space generated from the inspection image of the substrate to be inspected. Then, based on these images, the substrate to be inspected is inspected for defects. In this way, a preprocessed image is generated using variance information related to the variance of the component values of multiple components in the color space in the normal image and the inspection image, and the substrate is inspected for defects using this, so that the inspection can be performed using an image that takes into account the variance of the component values in the color space. Therefore, it is useful for accurately detecting abnormalities on the substrate surface.

In this case, in generating a preprocessed image for the normal image, a model map is generated based on variance information of component values of each pixel included in the component images for the normal images, and a preprocessed image for the normal image is created by calculating a difference between the model map and variance information of component values of each pixel included in the component images obtained from the normal image; and in generating a preprocessed image for the inspection image, a preprocessed image for the inspection image is created by calculating a difference between the model map and variance information of component values of each pixel included in the component images obtained from the inspection image.

In the above configuration, a model map created based on the variance information of the component values of each pixel contained in multiple component images related to multiple normal images can be said to be a compilation of the variance of the component values of each pixel in multiple normal images. By configuring preprocessing images related to normal images and inspection images to be created using such a model map, it is possible to create preprocessing images that take into account the variance of component values that may occur in normal images, making it possible to more accurately detect abnormalities on the substrate surface.

The variance information may be obtained by calculating the Mahalanobis distance for the component values of each pixel included in the multiple component images.

By configuring in this way, variance information can be calculated more accurately by calculating the Mahalanobis distance, making it possible to create a preprocessed image that is more useful when detecting abnormalities on the substrate surface.

The color space may be at least one of the RGB color space, the HSV color space, the XYZ color space, and the Lab color space.

As described above, by creating component images using at least one of the RGB color space, the HSV color space, the XYZ color space, and the Lab color space, a pre-processed image can be created that is more useful in detecting anomalies on the substrate surface.

The normal difference image and the test difference image may include images that have been subjected to enhancement processing to enhance the difference from the average image.

The normal difference image and the inspection difference image may include images after unevenness removal processing has been performed to remove unevenness in the image.

The normal difference image and the test difference image may include an image obtained after averaging the component values of pixels in a specified area and then performing a process to determine the difference from the averaged value.

The normal difference image and the inspection difference image may include images after noise removal processing has been performed to remove noise from the images.

Inspecting the substrate to be inspected for defects may include creating a normal image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the normal image, and an inspection image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the inspection image, and identifying an area in which a defect is estimated to exist from the difference between the normal image abnormality degree map and the inspection image abnormality degree map.

With the above configuration, the area where a defect is estimated to exist is identified from the difference between the normal image abnormality map created from the preprocessed image related to the normal image and the inspection image abnormality map created from the preprocessed image related to the inspection image. The abnormality map is a map that indicates the degree of abnormality for each pixel, and the area where a defect exists is estimated from the difference in the degree of abnormality for each pixel between the normal image and the inspection image, so that a more accurate estimation is performed based on the information of each pixel.

In creating the normal image abnormality map and the test image abnormality map, an algorithm is used that generates an abnormality map by inputting a preprocessed image, and the algorithm may be created by machine learning using as training data at least a portion of the normal images, the normal difference images, and the component images obtained from these.

With the above configuration, an algorithm for generating an anomaly map is created by machine learning using multiple normal images, multiple normal difference images, and at least some of the multiple component images obtained from these as training data. With this configuration, the number of normal images prepared for creating the algorithm can be reduced, making it easier to create a highly accurate algorithm.

A substrate inspection device according to an exemplary embodiment of the present disclosure is a substrate inspection device that inspects a substrate to be inspected using an inspection image obtained by imaging the substrate to be inspected, and includes: a normal preprocessing image creation unit that generates a plurality of normal difference images that are the differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images, generates a plurality of component images in color space from the plurality of normal images and the plurality of normal difference images, and generates a preprocessing image for the normal images based on variance information related to the variance of the component values of each pixel included in the plurality of component images; an inspection preprocessing image creation unit that generates an inspection difference image that is the difference between the inspection image and the average image, generates a plurality of component images in color space from the inspection image and the inspection difference image, and generates a preprocessing image for the inspection image based on variance information related to the variance of the component values of each pixel included in the plurality of component images; and a defect detection unit that inspects the substrate to be inspected for defects based on the preprocessing image for the normal image and the preprocessing image for the inspection image.

　According to the above-mentioned substrate inspection device, a preprocessed image for each of the multiple normal images is prepared based on variance information related to the variance of the component values of each pixel contained in the multiple component images in the color space generated from the multiple normal images. Also, a preprocessed image for the inspection image is prepared based on variance information related to the variance of the component values of each pixel contained in the multiple component images in the color space generated from the inspection image of the substrate to be inspected. Then, based on these images, the substrate to be inspected is inspected for defects. In this way, a preprocessed image is generated using variance information related to the variance of the component values of multiple components in the color space in the normal image and the inspection image, and the substrate is inspected for defects using this, so that the inspection can be performed using an image that takes into account the variance of the component values in the color space. Therefore, it is useful for accurately detecting abnormalities on the substrate surface.

The normal preprocessed image creation unit may generate a model map created based on variance information of component values of each pixel included in the component images related to the normal images, and create a preprocessed image related to the normal image by determining the difference between the model map and the variance information of component values of each pixel included in the component images obtained from the normal image, and the inspection preprocessed image creation unit may create a preprocessed image related to the inspection image by determining the difference between the model map and the variance information of component values of each pixel included in the component images obtained from the inspection image.

The defect detection unit may be configured to create a normal image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the normal image, and an inspection image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the inspection image, and identify an area where a defect is estimated to exist from the difference between the normal image abnormality degree map and the inspection image abnormality degree map.

When creating the normal image anomaly map and the inspection image anomaly map, the defect detection unit may use an algorithm that generates an anomaly map by inputting a preprocessed image, and the algorithm may be created by machine learning using as training data at least a portion of the normal images, the normal difference images, and the component images obtained from these.

A substrate inspection program according to an exemplary embodiment of the present disclosure is a substrate inspection program that causes a computer to perform substrate inspection using an inspection image obtained by imaging a substrate to be inspected, and causes the computer to perform the following: generate a normal difference image that is the difference between an average image generated from multiple normal images obtained by imaging multiple normal substrates and each of the multiple normal images; generate multiple component images in color space from the multiple normal images and the multiple normal difference images; generate a preprocessed image for each of the multiple normal images based on variance information related to the variance of the component values of each pixel included in the multiple component images; generate an inspection difference image that is the difference between the inspection image and the average image; generate multiple component images in color space from the inspection image and the inspection difference image; generate a preprocessed image for the inspection image based on variance information related to the variance of the component values of each pixel included in the multiple component images; and inspect the substrate to be inspected for defects based on the preprocessed image for the normal image and the preprocessed image for the inspection image.

The above-mentioned circuit board inspection program achieves the same effect as the above-mentioned circuit board inspection method.

Exemplary embodiments
Various exemplary embodiments will be described in detail below with reference to the drawings. Note that the same or corresponding parts in each drawing are denoted by the same reference numerals. Some drawings show an orthogonal coordinate system defined by an X-axis, a Y-axis, and a Z-axis. In the following description, the Z-axis corresponds to the up-down direction, and the X-axis and the Y-axis correspond to the horizontal direction.

[Substrate Processing System]
The substrate processing system 1 shown in FIG. 1 is a system that forms a photosensitive film on a workpiece W, exposes the photosensitive film, and develops the photosensitive film. The workpiece W to be processed is, for example, a substrate, or a substrate on which a film or circuit or the like has been formed by performing a predetermined process. The substrate included in the workpiece W is, for example, a wafer containing silicon. The workpiece W (substrate) may be formed in a circular shape. The workpiece W to be processed may be a glass substrate, a mask substrate, or an FPD (Flat Panel Display), or may be an intermediate body obtained by performing a predetermined process on such a substrate. The photosensitive film is, for example, a resist film.

The substrate processing system 1 comprises a coating and developing apparatus 2 and an exposure apparatus 3. The exposure apparatus 3 performs an exposure process on a resist film (photosensitive coating) formed on a workpiece W (substrate). Specifically, the exposure apparatus 3 irradiates an energy beam onto an exposure target portion of the resist film using a method such as immersion exposure. The coating and developing apparatus 2 performs a process of forming a resist film on the surface of the workpiece W before the exposure process by the exposure apparatus 3, and performs a development process of the resist film after the exposure process.

(Substrate Processing Apparatus)
1 and 2 , the coating and developing apparatus 2 includes a carrier block 4, a processing block 5, an interface block 6, and a control device 100.

The carrier block 4 introduces the workpiece W into the coating and developing apparatus 2 and removes the workpiece W from the coating and developing apparatus 2. For example, the carrier block 4 can support multiple carriers C (containers) for the workpieces W, and has a built-in transport device A1 including a transfer arm. The carrier C stores, for example, multiple circular workpieces W. The transport device A1 removes the workpiece W from the carrier C and passes it to the processing block 5, and receives the workpiece W from the processing block 5 and returns it to the carrier C. The processing block 5 has

multiple processing modules

11, 12, 13, and 14.

The processing module 11 incorporates a liquid processing unit U1, a heat processing unit U2, an inspection unit U3, and a transport device A3 that transports the workpiece W to these units. The processing module 11 forms an underlayer film on the surface of the workpiece W using the liquid processing unit U1 and the heat processing unit U2. The liquid processing unit U1 of the processing module 11 applies a processing liquid for forming the underlayer film onto the workpiece W. The heat processing unit U2 of the processing module 11 performs various heat treatments associated with the formation of the underlayer film. The inspection unit U3 performs processing to inspect the condition of the surface of the workpiece W before the formation of the underlayer film, after the formation of the underlayer film, or before the processing liquid for forming the underlayer film is applied and heat treatment is performed.

The processing module 12 incorporates a liquid processing unit U1, a heat processing unit U2, an inspection unit U3, and a transport device A3 that transports the workpiece W to these units. The processing module 12 forms a resist film on the underlayer film using the liquid processing unit U1 and the heat processing unit U2. The liquid processing unit U1 of the processing module 12 applies a processing liquid (resist) for forming a resist film onto the underlayer film. The heat processing unit U2 of the processing module 12 performs various heat treatments associated with the formation of the resist film. The inspection unit U3 performs processing to inspect the condition of the surface of the workpiece W before the resist film is formed, after the resist film is formed, or before the resist is applied and heat treatment is performed.

The processing module 13 incorporates a liquid processing unit U1, a heat processing unit U2, an inspection unit U3, and a transport device A3 that transports the workpiece W to these units. The processing module 13 forms an upper layer film on the resist film using the liquid processing unit U1 and the heat processing unit U2. The liquid processing unit U1 of the processing module 13 applies a processing liquid for forming the upper layer film onto the resist film. The heat processing unit U2 of the processing module 13 performs various heat treatments associated with the formation of the upper layer film. The inspection unit U3 performs processing to inspect the condition of the surface of the workpiece W before the upper layer film is formed, after the upper layer film is formed, or before the processing liquid for forming the upper layer film is applied and heat treatment is performed.

The processing module 14 incorporates a liquid processing unit U1, a heat processing unit U2, an inspection unit U3, and a transport device A3 that transports the workpiece W to these units. The processing module 14 uses the liquid processing unit U1 and the heat processing unit U2 to perform a development process on the resist film after exposure. The liquid processing unit U1 of the processing module 14 performs a development process on the resist film, for example, by supplying a developer onto the surface of the exposed workpiece W and then rinsing it off with a rinse liquid.

The heat treatment unit U2 of the processing module 14 performs various heat treatments associated with the development process. Specific examples of heat treatments include heating before the development process (PEB: Post Exposure Bake) and heating after the development process (PB: Post Bake). The inspection unit U3 performs processing to inspect the condition of the surface of the workpiece W before the development process and PEB are performed, after the development process and PB are performed, or before a developer is supplied and PB is performed.

A shelf unit U10 is provided on the carrier block 4 side within the processing block 5. The shelf unit U10 is divided into multiple cells arranged in the vertical direction. A transport device A7 including a lifting arm is provided near the shelf unit U10. The transport device A7 raises and lowers the workpiece W between the cells of the shelf unit U10.

A shelf unit U11 is provided on the interface block 6 side of the processing block 5. The shelf unit U11 is divided into multiple cells arranged vertically.

The interface block 6 transfers the workpiece W to and from the exposure device 3. For example, the interface block 6 has a built-in transport device A8 that includes a transfer arm, and is connected to the exposure device 3. The transport device A8 transfers the workpiece W placed on the shelf unit U11 to the exposure device 3, receives the workpiece W from the exposure device 3, and returns it to the shelf unit U11.

The control device 100 controls each device included in the coating and developing apparatus 2 to perform a coating and developing process (substrate processing), for example, in the following procedure. First, the control device 100 controls the transport device A1 to transport the workpiece W in the carrier C to the shelf unit U10, and then controls the transport device A7 to place the workpiece W in the cell for the processing module 11.

Then, the control device 100 controls the transport device A3 to transport the workpiece W from the shelf unit U10 to the liquid processing unit U1 in the processing module 11. The control device 100 controls the liquid processing unit U1 to form a film of the processing liquid for forming the underlayer film on the surface of the workpiece W. The control device 100 controls the heat processing unit U2 to heat the workpiece W with the film of the processing liquid for forming the underlayer film formed thereon to form the underlayer film. The control device 100 then controls the transport device A3 to return the workpiece W with the underlayer film formed thereon to the shelf unit U10, and controls the transport device A7 to place the workpiece W in a cell for the processing module 12. The control device 100 may also control the inspection unit U3 to inspect the surface of the workpiece W at any timing during processing in the processing module 11.

Then, the control device 100 controls the transport device A3 to transport the workpiece W from the shelf unit U10 to the liquid processing unit U1 in the processing module 12. The control device 100 controls the liquid processing unit U1 to form a film of the processing liquid for forming a resist film on the surface of the workpiece W. The control device 100 controls the heat processing unit U2 to heat the workpiece W with the film of the processing liquid for forming a resist film formed thereon to form a resist film. The control device 100 then controls the transport device A3 to return the workpiece W to the shelf unit U10, and controls the transport device A7 to place the workpiece W in a cell for the processing module 13. The control device 100 may also control the inspection unit U3 to inspect the surface of the workpiece W at any timing during processing in the processing module 12.

The control device 100 then controls the transport device A3 to transport the workpiece W from the shelf unit U10 to the liquid processing unit U1 in the processing module 13. The control device 100 also controls the liquid processing unit U1 to form a film of the processing liquid for forming an upper layer film on the resist film of the workpiece W. The control device 100 controls the heat processing unit U2 to heat the workpiece W with the film of the processing liquid for forming an upper layer film formed thereon to form an upper layer film. The control device 100 then controls the transport device A3 to transport the workpiece W to the shelf unit U11. The control device 100 may also control the inspection unit U3 to inspect the surface of the workpiece W at any timing during processing in the processing module 13.

The control device 100 then controls the transport device A8 to send the workpiece W from the shelf unit U11 to the exposure device 3. The control device 100 then controls the transport device A8 to receive the workpiece W that has been subjected to the exposure process from the exposure device 3 and place it in a cell for the processing module 14 in the shelf unit U11.

The control device 100 then controls the transport device A3 to transport the workpiece W from the shelf unit U11 to each unit in the processing module 14, and controls the liquid processing unit U1 and the heat processing unit U2 to perform a developing process on the resist film of the workpiece W. The control device 100 then controls the transport device A3 to return the workpiece W to the shelf unit U10, and controls the transport device A7 and the transport device A1 to return the workpiece W into the carrier C. The control device 100 may also control the inspection unit U3 to inspect the surface of the workpiece W at any timing during the processing in the processing module 14. This completes the coating and developing process for one workpiece W. The control device 100 controls each device of the coating and developing apparatus 2 to perform the coating and developing process for each of the subsequent multiple workpieces W in the same manner as described above.

The specific configuration of the substrate processing apparatus is not limited to the configuration of the coating and developing apparatus 2 exemplified above. The substrate processing apparatus may be any type that includes a unit that inspects the surface of the workpiece W that is to be subjected to a specified process, and a control device that controls this unit.

(Inspection unit)
Next, the inspection unit U3 included in the processing modules 11 to 14 will be described. The inspection unit U3 has a function of capturing an image of the surface of the workpiece W (hereinafter referred to as "surface Wa") to acquire image data. The inspection unit U3 may capture an image of the entire surface Wa of the workpiece W to acquire image data of the entire surface Wa. As shown in FIG. 3, the inspection unit U3 includes, for example, a housing 30, a holding unit 31, a linear driving unit 32, an imaging unit 33, and a light projecting/reflecting unit 34.

The holding unit 31 holds the workpiece W horizontally with the surface Wa facing upward. The linear drive unit 32 includes a power source such as an electric motor, and moves the holding unit 31 along a horizontal linear path. The imaging unit 33 has a camera 35 such as a CCD camera. The camera 35 is provided near one end of the inspection unit U3 in the moving direction of the holding unit 31, and is directed toward the other end in the moving direction. The light projecting/reflecting unit 34 projects light into the imaging range and guides the reflected light from the imaging range to the camera 35. For example, the light projecting/reflecting unit 34 has a half mirror 36 and a light source 37. The half mirror 36 is provided in the middle of the moving range of the linear drive unit 32 at a position higher than the holding unit 31, and reflects light from below to the camera 35. The light source 37 is provided above the half mirror 36, and irradiates illumination light downward through the half mirror 36.

The inspection unit U3 operates as follows to acquire image data of the surface Wa of the workpiece W. First, the linear drive unit 32 moves the holder 31. This causes the workpiece W to pass under the half mirror 36. During this passage, reflected light from each part of the surface Wa of the workpiece W is sent sequentially to the camera 35. The camera 35 forms an image of the reflected light from each part of the surface Wa of the workpiece W, and acquires image data of the surface Wa of the workpiece W (the entire surface Wa). The captured image obtained by capturing the surface Wa of the workpiece W changes depending on the condition of the surface Wa of the workpiece W. Therefore, in the inspection unit U3, an image (captured image data) of the surface Wa of the workpiece W is acquired as information indicating the condition of the surface Wa of the workpiece W, and this is used to evaluate the condition of the surface Wa, in particular the presence or absence of defects.

The captured image data acquired by the camera 35 is sent to the control device 100. The control device 100 can inspect the condition of the surface Wa of the workpiece W based on the captured image data of the surface Wa. For example, the presence or absence of defects on the surface Wa of the workpiece W can be inspected. In this disclosure, image data in which pixel values for each pixel are defined may be simply referred to as an "image."

[Control device (circuit board inspection device)]
4, the control device 100 has a process control unit 102 and an inspection control unit 110 as functional components (hereinafter referred to as "functional modules"). The processes executed by the process control unit 102 and the inspection control unit 110 correspond to the processes executed by the control device 100. The process control unit 102 controls the liquid treatment unit U1 and the heat treatment unit U2 so as to perform the liquid treatment and heat treatment in the above-mentioned coating and developing process on the workpiece W.

The inspection control unit 110 (substrate inspection device) inspects the workpiece W based on image data obtained from the inspection unit U3 at any stage when performing the coating and developing process. The inspection of the workpiece W includes determining whether there are any abnormalities (defects) on the surface Wa of the workpiece W. Defects on the surface Wa include, for example, scratches, adhesion of foreign matter, uneven application of the processing liquid, and non-application of the processing liquid.

Before performing the inspection, the inspection control unit 110 prepares normal substrate data to be used in the inspection, using a normal workpiece W (normal substrate) with no defects found on the surface Wa. The inspection control unit 110 performs an inspection of the workpiece W (substrate to be inspected) to be inspected based on the normal substrate data. The normal workpiece W and the workpiece W to be inspected are the same type of workpiece (substrate). The normal workpiece W and the workpiece W to be inspected are subjected to a coating and development process under the same processing conditions, and the preparation of normal substrate data and the inspection of the workpiece W are performed at the same timing in the coating and development process (for example, after the application of resist and before heat treatment).

The inspection control unit 110 has, as its functional modules, a model map creation unit 120, a preprocessing unit 130, a defect detection unit 112, and a result output unit 114. Furthermore, the model map creation unit 120 has an image acquisition unit 122, a component image generation unit 124, and a model map generation unit 126. And the preprocessing unit 130 has an image acquisition unit 132, a component image generation unit 134, and a preprocessing image generation unit 136. The model map creation unit 120 and the preprocessing unit 130 of the inspection control unit 110 function as a normal preprocessing image creation unit that creates a preprocessing image related to a normal image. Furthermore, the model map creation unit 120 and the preprocessing unit 130 of the inspection control unit 110 also function as an inspection preprocessing image creation unit that creates a preprocessing image related to an inspection image. The processing performed by each functional module of the inspection control unit 110 corresponds to the processing performed by the inspection control unit 110 (control device 100).

The model map creation unit 120 has a function of creating a model map to be used for inspection from an image of a normal workpiece W. Note that as image data relating to the normal workpiece W, multiple pieces of image data (e.g., about 3 to 10 pieces) of image data of the surface Wa of the normal workpiece W are prepared.

The image acquisition unit 122 has the function of acquiring image data of the surface Wa of a normal workpiece W from the inspection unit U3.

The component image generating unit 124 has a function of generating a component image used to create a model map from the image acquired by the image acquiring unit 122. The procedure for generating a component image will be described later.

The model map generating unit 126 has a function of creating a model map using the component images generated by the component image generating unit 124. A model map is a map of a shape corresponding to the workpiece W, in which the degree of color variation on the surface Wa in image data relating to a normal workpiece W is quantified for each pixel contained in the image data. One model map is created from multiple image data relating to normal workpieces W. The detailed procedure will be described later, but the color variation for each pixel is quantified using covariance and Mahalanobis distance. The model map created by the model map creating unit 120 is used for pre-processing of image data used in defect detection.

The pre-processing unit 130 has the function of creating pre-processed images to be used for inspection from images of a normal workpiece W and images of the workpiece W to be inspected.

The image acquisition unit 132 has a function of acquiring image data of the surface Wa of the workpiece W to be inspected from the inspection unit U3. Note that image data of the surface Wa of a normal workpiece W is acquired by the image acquisition unit 122, so this image data is used.

The component image generating unit 134 has a function of generating a component image to be used for creating a preprocessed image from the image acquired by the image acquiring unit 132. The procedure for generating the component image will be described later.

The preprocessed image generating unit 136 has a function of generating a preprocessed image using the component image generated by the component image generating unit 134. A preprocessed image is an image in which the degree of color variation on the surface Wa in the image data of a normal workpiece W and the image data of the workpiece W to be inspected is quantified for each pixel contained in the image data. The detailed procedure will be described later, but in generating a preprocessed image, the color variation for each pixel is quantified using the covariance used to create the model map and the Mahalanobis distance. One preprocessed image is created for each image data to be processed. A preprocessed image is also created for the image data of a normal workpiece W. The preprocessed image created by the preprocessing unit 130 is used in defect detection.

The defect detection unit 112 uses a preprocessed image created from image data of a normal workpiece W and image data of the workpiece W to be inspected to check for the presence or absence of defects on the surface Wa of the workpiece W to be inspected. An example of the processing procedure in the defect detection unit 112 will be described later.

The result output unit 114 has a function of outputting the detection result in the defect detection unit 112. As an example, when the defect detection unit 112 determines that there is an abnormality on the surface Wa of the workpiece W, it may output an abnormality signal indicating that the workpiece W being inspected is abnormal. When the result output unit 114 outputs an abnormality signal, it may output the abnormality signal to the process control unit 102, may output it to a higher-level controller, or may output it to an output device such as a monitor for notifying an operator, etc. of information.

The control device 100 is composed of one or more computers. The control device 100 has, for example, a circuit 150 shown in FIG. 5. The circuit 150 has one or more processors 152, a memory 154, a storage 156, and an input/output port 158. The storage 156 has a computer-readable storage medium, such as a hard disk. The storage medium stores a program (substrate inspection program) for causing the control device 100 to execute the substrate inspection method described below. The storage medium may be a removable medium, such as a non-volatile semiconductor memory, a magnetic disk, or an optical disk.

Memory 154 temporarily stores the programs loaded from the storage medium of storage 156 and the results of calculations performed by processor 152. Processor 152 configures each of the functional modules described above by executing the above programs in cooperation with memory 154. Input/output port 158 inputs and outputs electrical signals between liquid processing unit U1, heat processing unit U2, inspection unit U3, etc., according to instructions from processor 152.

The hardware configuration of the control device 100 is not necessarily limited to configuring each functional module by a program. For example, each functional module of the control device 100 may be configured by a dedicated logic circuit or an ASIC (Application Specific Integrated Circuit) that integrates these. When the control device 100 is configured by multiple computers (multiple circuits), some of the above functional modules may be realized by one computer (circuit), and the remaining parts of the above functional modules may be realized by another computer (circuit).

[Substrate inspection method]
As an example of a substrate inspection method, a series of processes executed by the control device 100 (inspection control unit 110) will be described. Fig. 6 is a flow diagram illustrating the series of processes executed by the control device 100. Figs. 7 to 12 are diagrams illustrating the details of the processes performed in each step and an example of image data generated in each step.

The control device 100 executes steps S01 to S03, for example, as shown in FIG. 6. In step S01, the image acquisition unit 122 of the model map creation unit 120 acquires image data (normal image) relating to a normal workpiece W from the inspection unit U3. Next, in step S02, the component image generation unit 124 of the model map creation unit 120 creates a component image from the normal image acquired by the image acquisition unit 122. Furthermore, in step S03, the model map generation unit 126 of the model map creation unit 120 generates a model map from the component image generated in the component image generation unit 124.

FIG. 7 is a diagram explaining the series of steps S01 to S03 described above. In step S01, it is assumed that there are three normal images G1 acquired by the image acquisition unit 122 and used in the procedure described below. For example, image data of 2048 x 2048 pixels is used as the normal images G1. In addition, both normal images G1 (and the inspection images) are color images. In the case of an RGB color image, the pixel value for each pixel includes, for example, information on the pixel values related to each of the RGB colors. In this way, one normal image G1 includes information on the pixel values of 2048 x 2048 pixels.

FIG. 7 shows an example in which image generation process P1 is executed on three normal images G1. Image generation process P1 shown in FIG. 7 corresponds to step S02 in FIG. 6, i.e., the process of creating component images.

In FIG. 7, the image generation process P1 includes enhancement processing, difference processing, component decomposition processing, and filter processing. The procedure for creating these component images will be described with reference to FIG. 8 to FIG. 10. First, an average image G2 is created by averaging the pixel values of each pixel in multiple normal images G1. Next, an enhancement process is performed on the average image G2 to create an average enhancement image G3. As an example, an enhancement process using a LUT (lookup table) based on a tone conversion function is performed on a color image. A tone conversion function is a function that associates pixel values in an input image and an output image, and a function is set for each RGB according to the characteristic part to be emphasized. Also, a LUT is a table that shows the correspondence of pixel values before and after conversion. In other words, a tone conversion function that focuses on the part to be emphasized is set in advance, and a LUT corresponding to this tone conversion function is prepared, and a process is performed to convert the RGB values of each pixel of the color image based on the LUT. As a result, for example, an average enhancement image G3 is obtained in which a specific pixel value area is more emphasized.

Furthermore, an unevenness removal process is performed on this average emphasized image G3. The unevenness removal process refers to a process for removing concentric unevenness, for example. Specifically, in the average emphasized image G3, concentric circles are set with the pixel that captures the center of the workpiece W as the center. Then, by averaging the pixel values of each of the RGB pixels on the concentric circles, a filter image can be prepared in which the pixel values of each of the RGB pixels on the concentric circles are the same value (average value). Then, an average unevenness-removed image G4 is obtained by taking the difference between the pixel values of each pixel included in the filter image from the pixel values of each pixel in the average emphasized image G3. By performing such a process, it is possible to remove concentric unevenness caused by different distances from the center. The average emphasized image G3 and the average unevenness-removed image G4 are obtained by the above procedure.

Next, the image from which the component images are to be generated is set as the target image I1, and the target image is subjected to the same processing as in generating the above-mentioned average emphasized image G3 to generate the target emphasized image I2. Furthermore, the target emphasized image I2 is subjected to the same processing as in generating the above-mentioned average unevenness removed image G4 to generate the target unevenness removed image I3. As a result, as shown in FIG. 9(a), the target emphasized image I2 and the target unevenness removed image I3 are generated. Note that the target image I1 corresponds to one of the normal images G1 used to generate the above-mentioned average image G2.

Furthermore, as shown in FIG. 9(b), a difference emphasized image I4 is obtained by taking the difference between the target emphasized image I2 and the average emphasized image G3. Then, as shown in FIG. 9(c), a difference unevenness removed image I5 is obtained by taking the difference between the target unevenness removed image I3 and the average unevenness removed image G4. The difference emphasized image I4 and the difference unevenness removed image I5 are images from which the components of the average emphasized image G3 and the average unevenness removed image G4 have been removed, respectively. Therefore, the process of obtaining the difference emphasized image I4 and the difference unevenness removed image I5 corresponds to the process of removing the fluctuation components contained in the average emphasized image G3 and the average unevenness removed image G4 generated from the image obtained by averaging three normal images G1. In other words, when the normal image G1 is the target image I1, both the difference emphasized image I4 and the difference unevenness removed image I5 correspond to normal difference images.

Furthermore, for each of the object emphasis image I2, object unevenness removed image I3, difference emphasis image I4, and difference unevenness removed image I5 obtained by the above processes, averaging, difference, and noise removal processes are performed as shown in Figs. 10(a) to 10(d). The averaging process is, for example, a process in which the pixel value of one pixel is set to the average value of the pixel values of the pixels surrounding that pixel, for example, a process in which the pixel value of one pixel is set to the average value of the pixel values of the surrounding 101 x 101 pixels using a mask of 101 x 101 pixels. Next, the difference process is a process in which a large area of color change that is not due to defects is removed by taking the difference between the image data after the averaging process and the image before the process. Furthermore, the noise removal process is a process in which noise that is thought to be unrelated to defects is removed, for example, a process in which a mask of 15 x 15 pixels is used to remove components that are due to noise. By carrying out these processes, four processed images I11, I12, I13, and I14 are created from the object-emphasized image I2, the object-unevenness-removed image I3, the difference-emphasized image I4, and the difference-unevenness-removed image I5, as shown in Figures 10(a) to 10(d).

The above procedure is performed on an RGB color image. Therefore, object-enhanced image I2, object unevenness-removed image I3, and the four processed images I11, I12, I13, and I14 each contain information on pixel values of the R, G, and B components. If only each component is extracted from these to generate image data, for example, an object-enhanced image I2 for the R component, an object-enhanced image I2 for the G component, and an object-enhanced image I2 for the B component can be obtained from object-enhanced image I2 for the RGB color image. Similarly, images relating to the R, G, and B components can be generated for the other images I3, I11 to I14.

In addition, instead of expressing a color image using the R, G, and B color components, it is also possible to define it in another color space. Examples include the HSV color space, the XYZ color space, and the Lab color space. By expressing the color of a pixel using components of different color spaces, it is possible to capture the color characteristics of that pixel from a different perspective. Also, a color defined in a specific color space can be expressed using components of another color space using a predefined conversion formula.

Therefore, one color image is expressed in the above three components in the HSV color space (H component, S component, V component), three components in the XYZ color space (X component, Y component, Z component), and three components in the Lab color space (aa component, ba component, La component), and then images related to each of these components are created: an object-emphasized image I2, an object-unevenness-removed image I3, and four processed images I11, I12, I13, and I14. As a result, six types of images (object-emphasized image I2, object-unevenness-removed image I3, and four processed images I11, I12, I13, and I14) are obtained by preprocessing the 12 color components (R, G, B, H, S, V, X, Y, Z, aa, ba, La) from one color image. In other words, 72 component images are obtained based on 12 components x 6 types of processed images for one color image. In Figure 7, the component image for the three normal images G1 is shown as component image G10.

FIG. 7 shows that a component image G10 is obtained as a result of executing image generation process P1 on three normal images G1. Furthermore, FIG. 7 shows that a model map G20 is created from the component image G10. It also shows that the Mahalanobis distance is calculated to create the model map G20. The procedure for creating the model map G20, including the calculation of the Mahalanobis distance, corresponds to step S03 in FIG. 6, i.e., the process for creating the model map.

The 72 component images G10 relating to one normal image G1 are considered to contain different features. In other words, the 72 component images G10 can be considered to contain 72 types of features for each pixel. In this case, it is considered that 72 features are obtained for each pixel contained in the 2048 x 2048 pixels, and the average and covariance matrix of each pixel are calculated, and the Mahalanobis distance is calculated for each pixel. The average values of the three images are used for the average and covariance matrix. The Mahalanobis distance at pixel M can be calculated, for example, using the following formula (1). In formula (1), μ is the average, and Σ is the covariance matrix.

By calculating the Mahalanobis distance, it is possible to calculate how far the data for that pixel is from the data group, i.e., the degree of abnormality of the data. Since the 72 feature values obtained from one normal image G1 are different for each pixel, it is possible to use the Mahalanobis distance to calculate the degree to which the data for that pixel is abnormal in a pixel group of 2048 x 2048 pixels.

As described above, the 72 component images G10 obtained from one normal image G1 can be used to calculate the Mahalanobis distance for each pixel in the normal image G1. By performing this process on each of the three normal images G1, three Mahalanobis distances are calculated from the three normal images G1 for pixels in the same position. For each pixel, the maximum value of these three Mahalanobis distances is adopted and arranged to correspond to the 2048 x 2048 pixel arrangement. This is how the model map G20 can be created. In other words, the model map G20 can be said to be a mapping of the degree of abnormality of the data for each pixel contained in the three normal images G1.

After creating the model map G20 using the above procedure, the control device 100 executes steps S04 to S07. In step S04, the model map G20 is used to prepare a preprocessed image related to the normal image G1. In addition, in steps S05 to S07, the model map G20 is used to prepare a preprocessed image related to the image to be inspected.

More specifically, in step S05, the image acquisition unit 132 of the pre-processing unit 130 acquires the inspection target image. In step S06, the component image generation unit 134 of the pre-processing unit 130 generates a component image of the inspection target image. Then, in step S07, the pre-processed image generation unit 136 of the pre-processing unit 130 creates a pre-processed image for the normal image G1 using the component image and model map G20 created in step S06. Meanwhile, in step S04, the pre-processed image generation unit 136 of the pre-processing unit 130 creates a pre-processed image for the normal image G1 using the component image G10 and model map G20 created in step S02.

The above procedure will be explained with reference to FIG. 11. FIG. 11 shows the procedure for creating a preprocessed image T30 from an inspection image T1, which is an image of the inspection target. First, an image generation process P1 is executed on the inspection image T1. The image generation process P1 shown in FIG. 11 is the same as the image generation process P1 shown in FIG. 7. Moreover, the image generation process P1 corresponds to step S06 in FIG. 6, i.e., the process of creating a component image.

11, as in FIG. 7, the image generation process P1 includes an emphasis process, a difference process, a component decomposition process, and a filter process. The procedure for creating these component images is the same as that described in the above embodiment. However, the average emphasized image G3 and the average unevenness removed image G4 are created from the normal image G1. Based on these images, the target emphasized image I2 and the target unevenness removed image I3 are generated by the procedure shown in FIG. 9(a) using the inspection image T1 as the target image I1. Next, as shown in FIG. 9(b), the difference emphasized image I4 is generated by taking the difference between the target emphasized image I2 and the average emphasized image G3. Furthermore, as shown in FIG. 9(c), the difference unevenness removed image I5 is generated by taking the difference between the target unevenness removed image I3 and the average unevenness removed image G4. In this way, the difference emphasized image I4 and the difference unevenness removed image I5 created when the inspection image T1 is used as the target image I1 both correspond to the inspection difference image.

Then, as shown in Figs. 10(a) to 10(d), four processed images I11, I12, I13, and I14 are created by performing averaging, subtraction, and noise removal processes on the object-emphasized image I2, object-unevenness-removed image I3, difference-emphasized image I4, and difference-unevenness-removed image I5 obtained in the process so far. By performing this process on the 12 components described above, six types of images (object-emphasized image I2, object-unevenness-removed image I3, and four processed images I11, I12, I13, and I14) are generated from one color image by performing pre-processing on the 12 color components (R, G, B, H, S, V, X, Y, Z, aa, ba, and La). As a result, 72 component images T10 are obtained from one inspection image T1 based on the 12 components x 6 types of processed images, as shown in Fig. 11.

Then, as shown in FIG. 11, a map image T20 is created from the component image T10. When creating the map image T20, the Mahalanobis distance is calculated. The above-mentioned formula (1) is used to calculate the Mahalanobis distance when creating the map image T20. However, the mean and covariance matrix of each pixel in the normal image G1 (the average value of the three images) are used. In other words, the procedure differs from that when creating the model map G20 in that the mean and covariance matrix in the normal image G1 are used instead of the mean and covariance matrix derived from the test image T1.

By calculating the Mahalanobis distance, it is possible to calculate how far the data for that pixel is from the data group, i.e., the degree of abnormality of the data. Since the 72 feature values obtained from one inspection image T1 are different for each pixel, it is possible to use the Mahalanobis distance to calculate the degree to which the data for that pixel is abnormal in a pixel group of 2048 x 2048 pixels.

As described above, the 72 component images T10 obtained from one test image T1 can be used to calculate the Mahalanobis distance of each pixel in the test image T1. The Mahalanobis distance of each pixel is arranged so that it corresponds to the 2048 x 2048 pixel arrangement. In this way, the map image T20 can be created. In other words, like the model map G20, the map image T20 is a mapping of the degree of abnormality of the data of each pixel contained in the test image T1.

Then, as shown in FIG. 11, a preprocessed image T30 of the inspection image T1 is created by calculating the difference between the map image T20 and the model map G20. By performing difference processing using the model map G20, information relating to the degree of abnormality that appears in the model map G20 created from the normal image G1 is removed. The image obtained after processing is performed on the map image T20 to find the difference with the model map G20 becomes the preprocessed image T30. Note that the creation of the map image T20 and the preprocessed image T30 after preparing the component image T10 corresponds to the generation of the preprocessed image in step S07.

By carrying out the above series of processes, a pre-processed image T30 is created from a single inspection image T1.

In step S04, in which a preprocessed image corresponding to the normal image G1 is created, the preprocessed image generating unit 136 of the preprocessing unit 130 creates a preprocessed image for each of the three normal images G1 using the component image G10 and model map G20 created in step S02. At this time, in the procedure for creating the model map G20, the component image G10 for the three normal images G1 is prepared. Therefore, in the procedure shown in FIG. 11, the steps up to the preparation of the component image G10 can be omitted. Therefore, for the normal image G1, the creation of the map image (corresponding to the map image T20 in FIG. 11) and the creation of the preprocessed image (corresponding to the preprocessed image T30 in FIG. 11) after the component image G10 is prepared corresponds to the generation of the preprocessed image in step S07. In addition, by executing step S07, a preprocessed image G30 (see FIG. 12) can be created for each of the three normal images G1.

After creating a preprocessed image G30 corresponding to normal image G1 and a preprocessed image T30 corresponding to inspection image T1 in the above procedure, the control device 100 executes steps S11 to S14. In steps S11 and S12, the control device 100 extracts features from the preprocessed image G30 of normal image G1 and calculates the degree of abnormality (step S11), thereby creating an abnormality map to be used for defect detection (step S12). In steps S13 and S14, the control device 100 extracts features and calculates the degree of abnormality (step S13), thereby creating an abnormality map to be used for defect detection (step S14). Steps S11 to S14 are performed by the defect detection unit 112 of the control device 100.

The extraction of features performed in steps S11 and S13 is a process of extracting features related to defects from the preprocessed images G30 and T30, and can be performed, for example, using Vision Transformer (ViT), which is one of the image classification algorithms. The calculation of anomaly level performed in steps S12 and S14 is a process of calculating the degree of anomaly related to defects for each pixel from the preprocessed images G30 and T30 based on the features, and can be performed, for example, using Graph Mixture Density Networks (GMDN), which is one of the methods for obtaining a probability distribution using a neural network. The same algorithm is used in steps S11 and S13 and steps S12 and S14.

In this way, an algorithm generated by machine learning using training data can be used to extract features and calculate the degree of abnormality. In the learning stage of the algorithm used to extract features and calculate the degree of abnormality, the preprocessed image created from multiple normal images prepared in advance is adjusted so that when features are extracted and the degree of abnormality is calculated, the calculated numerical value of the degree of abnormality is lowered. By applying the algorithm (model) whose parameters have been adjusted through such preprocessing to the preprocessed image G30 created from the normal image G1 and the preprocessed image T30 created from the inspection image T1, the degree of abnormality for each pixel can be calculated.

The defect detection unit 112 of the control device 100 maps the information on the degree of abnormality for each pixel calculated by the above procedure in correspondence with the arrangement of the pixels. As a result, one abnormality map G40 (normal image abnormality map) is created from the three normal images G1, and one abnormality map T40 (inspection image abnormality map) is created from the one inspection image T1.

Next, the control device 100 executes step S15. In step S15, the defect detection unit 112 of the control device 100 uses the above-mentioned anomaly maps G40 and T40 to determine whether or not there is a defect in the inspection image T1. In addition, the result output unit 114 of the control device 100 outputs the determination result of whether or not there is a defect.

FIG. 12 shows an example of a procedure for determining the presence or absence of a defect. First, a difference image T41 is obtained by calculating the difference between the abnormality map G40 created from three normal images G1 as described above and the abnormality map T40 created from one inspection image T1. Furthermore, a binarized image T42 in which the value of each pixel is binarized using a predetermined threshold value for this difference image T41, and a binarized image T43 in which the value of each pixel is binarized using a predetermined threshold value for the pre-processed image T30 of the inspection image T1 created by the above procedure are prepared. Then, the two binarized images T42 and T43 are superimposed to create a defect detection image T45 in which pixels converted to white in both images are identified. In the defect detection image T45, a portion displayed in white is determined to have a defect. Note that the above procedure is merely an example, and the defect detection method is not particularly limited.

By carrying out the above processing, areas that are estimated to have defects are identified based on the inspection image T1. The result output unit 114 of the control device 100 may output the defect detection image T45 as the judgment result, or may output only the judgment result of the presence or absence of a defect based on whether or not the defect detection image T45 contains an area that is judged to have a defect.

Note that, in the above procedure, an algorithm for machine learning is used, but the learning method can be changed as appropriate. For example, as described above, in the learning stage of the algorithm used for extracting the feature amount and calculating the degree of abnormality, when the feature amount is extracted from the preprocessed image created from a plurality of normal images prepared in advance and the degree of abnormality is calculated, the calculated value of the degree of abnormality is adjusted to be low. In the above example, the preprocessed image created from the normal image is used for learning, but instead of using only the preprocessed image for learning, the component images used to create the preprocessed image may be used as training data for machine learning. As described above, the preprocessed image generated from the normal image is created using 72 component images created from one normal image. These component images are images that contain information related to the characteristic parts of the normal image, so they can be used as training data for the algorithm for extracting the feature amount and calculating the degree of abnormality. Therefore, the algorithm for extracting the feature amount and calculating the degree of abnormality may be performed with at least a part of the component images used to generate the preprocessed image included in the training data. In this configuration, not only one preprocessed image obtained from one normal image but also component images are used as training data, so that the number of training data can be increased while reducing the number of normal images prepared in advance. Therefore, it is easier to prepare training data that would be necessary for sufficient training.

[effect]
According to the above-mentioned substrate inspection device (inspection control unit 110) and substrate inspection method, a preprocessed image G30 for each of the plurality of normal images G1 is prepared based on variance information related to the variance of component values of each pixel included in a plurality of component images in a color space generated from a plurality of normal images. Also, a preprocessed image T30 related to the inspection image is prepared based on variance information related to the variance of component values of each pixel included in a plurality of component images in a color space generated from an inspection image T1 obtained by capturing an inspection target substrate. Then, based on these images, an inspection of defects of the substrate to be inspected is performed. In this way, preprocessed images G30 and T30 using variance information related to the variance of component values of a plurality of components in a color space in the normal image G1 and the inspection image T1 are generated, and the substrate is inspected for defects using these. This allows the inspection to be performed using an image that takes into account the variance of component values in a color space. Therefore, it is useful for detecting abnormalities on the substrate surface with high accuracy.

The use of component images in a color space has long been known as a method for inspecting substrates for defects using images. However, in conventional methods, information including differences in color that may be contained in each image is used for defect inspection, leaving room for improvement in terms of inspection accuracy. In contrast, according to the method described in the above embodiment, preprocessed images G30, T30 used for inspection are created using information related to the variance of component values obtained from the component images. With this configuration, preprocessed images G30, T30 are obtained that reflect information related to the variance of component values contained in the component images. With a configuration in which such preprocessed images are used for defect inspection, the accuracy of defect detection can be improved.

In the above embodiment, a model map G20 is created based on the variance information of the component values of each pixel included in the multiple component images. A preprocessed image G30 for one normal image is created by calculating the difference between the variance information of the component values of each pixel included in the multiple component images obtained from one normal image and the model map G20. A preprocessed image T30 for the inspection image is created by calculating the difference between the variance information of the component values of each pixel included in the multiple component images obtained from the inspection image and the model map G20. A model map can be said to be a compilation of the variances of the component values of each pixel in multiple normal images G1. By using such a model map to create preprocessed images for the normal image G1 and the inspection image T1, it is possible to create preprocessed images that take into account the variance of the component values that may occur in the normal image. Therefore, abnormalities on the substrate surface can be detected more accurately.

The variance information may also be obtained by calculating the Mahalanobis distance related to the component values of each pixel contained in the multiple component images. With this configuration, the variance information can be obtained more accurately by calculating the Mahalanobis distance, making it possible to create a preprocessed image that is more useful when detecting abnormalities on the substrate surface.

At least one of the RGB color space, HSV color space, XYZ color space, and Lab color space may be used as the color space. As described above, by creating a component image using at least one of the RGB color space, HSV color space, XYZ color space, and Lab color space, a preprocessed image that is more useful for detecting abnormalities on the substrate surface can be created. As described in the above embodiment, component images in multiple color spaces among the RGB color space, HSV color space, XYZ color space, and Lab color space may be used. In this case, a component image when one image is defined in another color space is obtained. By creating a preprocessed image using such a component image, a component image obtained by decomposing information of one image into multiple different components is used, and the features contained in the image can be captured from multiple perspectives. Therefore, a preprocessed image useful for improving the accuracy of defect detection can be created.

The normal difference image and the test difference image may include images that have been subjected to enhancement processing to enhance the difference from the average image. With this configuration, it is possible to obtain images that further enhance the characteristic components of each image, and these images can be used to create pre-processed images.

The normal difference image and the inspection difference image may include an image after unevenness removal processing to remove unevenness in the image. With this configuration, an image can be obtained in which unevenness that occurs in the image has been appropriately removed, and can be used to create a preprocessed image.

The normal difference image and the test difference image may include an image obtained after averaging the component values of pixels in a specified area and then performing a process to determine the difference from the averaged value. By configuring in this way, an image can be obtained in which abnormal values that occur accidentally in single pixels, etc. have been removed, and this can be used to create a preprocessed image.

The normal difference image and the test difference image may include an image after noise removal processing to remove noise from the image. By configuring in this way, an image can be obtained from which noise that may be contained in the image has been removed, and the image can be used to create a preprocessed image.

Inspecting the substrate for defects may include creating a normal image abnormality map G40 showing the degree of abnormality for each pixel in the preprocessed image G30 related to the normal image G1, and an inspection image abnormality map T40 showing the degree of abnormality for each pixel in the preprocessed image T30 related to the inspection image T1, and identifying an area where a defect is estimated to exist from the difference between the normal image abnormality map G40 and the inspection image abnormality map T40. With the above configuration, an area where a defect is estimated to exist is identified from the difference between the normal image abnormality map G40 created from the preprocessed image G30 related to the normal image G1 and the inspection image abnormality map T40 created from the preprocessed image T30 related to the inspection image T1. The abnormality maps G40 and T40 are maps showing the degree of abnormality for each pixel, and the area where a defect exists is estimated from the difference in the degree of abnormality for each pixel between the normal image G1 and the inspection image T1, so that a more accurate estimation is performed based on the information of each pixel.

In creating the normal image abnormality map G40 and the test image abnormality map T40, an algorithm may be used that generates an abnormality map by inputting a preprocessed image. In this case, the algorithm may be created by machine learning using the multiple normal images G1, the multiple normal difference images, and at least a portion of the multiple component images obtained therefrom as training data. According to the above configuration, the algorithm that generates the abnormality map is created by machine learning using the multiple normal images, the multiple normal difference images, and at least a portion of the multiple component images obtained therefrom as training data. With this configuration, the number of normal images prepared for creating the algorithm can be reduced, making it easier to create a highly accurate algorithm.

[Modification]
Although various exemplary embodiments have been described above, the present invention is not limited to the exemplary embodiments described above, and various omissions, substitutions, and modifications may be made. In addition, elements in different embodiments can be combined to form other embodiments.

For example, in the above embodiment, a case has been described in which six types of images (target emphasis image I2, target unevenness removal image I3, and four processed images I11, I12, I13, I14) are created from one color image by preprocessing each of the 12 color components (R, G, B, H, S, V, X, Y, Z, aa, ba, La) using the RGB color space, HSV color space, XYZ color space, and Lab color space. However, this configuration can be modified as appropriate, and for example, only a portion of the above-mentioned color spaces may be used. Also, component images may be created using a color space different from the above-mentioned color spaces.

In addition, after creating the preprocessed image, the procedure for creating the anomaly map and detecting defects can be changed as appropriate. The procedure described in the above embodiment is one example, and can be changed to various methods that can be used in defect detection using images. In addition, further image processing, etc. can be performed on the preprocessed image depending on the method that can be used in defect detection.

The above-described series of processes can be modified as appropriate. In the above-described series of processes, the control device 100 (inspection control unit 110) may execute one step and the next step in parallel, or may execute each step in an order different from the example described above. The control device 100 (inspection control unit 110) may omit any step, or may execute a process in any step that is different from the example described above.

The computer constituting the inspection control unit 110 may be provided outside the coating and developing apparatus 2. In this case, the control device 100 and the inspection control unit 110 may be communicatively connected by wired or wireless communication. The control device 100 may acquire an image from the inspection unit U3 and then transmit the image to the inspection control unit 110. The inspection control unit 110 may transmit information indicating the determination result of whether or not there is an abnormality to the control device 100.

From the foregoing, it will be understood that the various embodiments of the present disclosure have been described herein for purposes of illustration, and that various modifications may be made without departing from the scope and spirit of the present disclosure. Accordingly, the various embodiments disclosed herein are not intended to be limiting, with the true scope and spirit being indicated by the appended claims.

1...substrate processing system, 2...coating and developing apparatus, 100...control device, 102...processing control unit, 110...inspection control unit, 112...defect detection unit, 114...result output unit, 120...model map creation unit, 122...image acquisition unit, 124...component image generation unit, 126...model map generation unit, 130...pre-processing unit, 132...image acquisition unit, 134...component image generation unit, 136...pre-processing image generation unit.

Claims

A substrate inspection method for inspecting a substrate to be inspected by using an inspection image obtained by imaging the substrate, comprising:
generating a plurality of normal difference images which are differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images; generating a plurality of component images in a color space from the plurality of normal images and the plurality of normal difference images; and generating a preprocessed image for each of the plurality of normal images based on variance information relating to the variance of component values of each pixel included in the plurality of component images;
generating an inspection difference image which is a difference between the inspection image and the average image, generating a plurality of component images in a color space from the inspection image and the inspection difference image, and generating a preprocessed image for the inspection image based on variance information relating to the variance of component values of each pixel included in the plurality of component images;
inspecting the substrate to be inspected for defects based on a preprocessed image related to the normal image and a preprocessed image related to the inspection image;
A substrate inspection method comprising:
In generating a preprocessed image for the normal image, a model map is generated based on variance information of component values of each pixel included in the component images for the normal images, and a difference between the model map and variance information of component values of each pixel included in the component images obtained from one of the normal images is calculated to generate a preprocessed image for one of the normal images;
The substrate inspection method of claim 1, wherein in generating a preprocessed image for the inspection image, the preprocessed image for the inspection image is created by calculating the difference between variance information of component values of each pixel contained in the multiple component images obtained from the inspection image and the model map.
The substrate inspection method according to claim 1, wherein the variance information is obtained by calculating the Mahalanobis distance related to the component values of each pixel included in the plurality of component images.
The substrate inspection method according to claim 1, wherein at least one of the RGB color space, the HSV color space, the XYZ color space, and the Lab color space is used as the color space.
The substrate inspection method according to claim 1, wherein the normal difference image and the inspection difference image include images after enhancement processing has been performed to enhance the difference from the average image.
The substrate inspection method according to claim 1, wherein the normal difference image and the inspection difference image include images after irregularity removal processing for removing irregularities in the images.
The substrate inspection method according to claim 1, wherein the normal difference image and the inspection difference image include an image obtained after averaging component values of pixels in a specified area and then performing a process to determine the difference from the averaged value.
The substrate inspection method according to claim 1, wherein the normal difference image and the inspection difference image include images after noise removal processing for removing noise from the images.
Inspecting the substrate to be inspected for defects includes:
creating a normal image abnormality degree map showing an abnormality degree for each pixel in a preprocessed image related to the normal image, and an inspection image abnormality degree map showing an abnormality degree for each pixel in a preprocessed image related to the inspection image;
Identifying an area in which a defect is estimated to exist based on a difference between the normal image anomaly degree map and the inspection image anomaly degree map;
The method of claim 1 , further comprising:
In creating the normal image abnormality degree map and the inspection image abnormality degree map, an algorithm is used that generates an abnormality degree map by inputting a preprocessed image;
The substrate inspection method according to claim 9 , wherein the algorithm is created by machine learning using the plurality of normal images, the plurality of normal differential images, and at least a portion of the plurality of component images obtained therefrom as training data.
A substrate inspection apparatus for inspecting a substrate to be inspected by using an inspection image obtained by imaging the substrate, comprising:
a normal preprocessed image creation unit that generates a plurality of normal difference images which are differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images, generates a plurality of component images in a color space from the plurality of normal images and the plurality of normal difference images, and generates a preprocessed image for the normal images based on variance information related to the variance of component values of each pixel included in the plurality of component images;
a pre-inspection image creation unit that generates an inspection difference image which is the difference between the inspection image and the average image, generates a plurality of component images in a color space from the inspection image and the inspection difference image, and generates a pre-processed image for the inspection image based on variance information relating to the variance of component values of each pixel included in the plurality of component images;
a defect detection unit that inspects the substrate to be inspected for defects based on a preprocessed image related to the normal image and a preprocessed image related to the inspection image;
A substrate inspection apparatus comprising:
the normal preprocessed image creation unit generates one model map created based on variance information of component values of each pixel included in the plurality of component images related to the plurality of normal images, and creates a preprocessed image related to one of the normal images by calculating a difference between the model map and variance information of component values of each pixel included in the plurality of component images obtained from one of the normal images;
The substrate inspection apparatus of claim 11, wherein the pre-inspection image creation unit creates a pre-processed image related to the inspection image by calculating the difference between variance information of component values of each pixel contained in the multiple component images obtained from the inspection image and the model map.
The substrate inspection device according to claim 11, wherein the variance information is obtained by calculating the Mahalanobis distance related to the component values of each pixel included in the plurality of component images.
The defect detection unit includes:
creating a normal image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the normal image, and an inspection image abnormality degree map showing the degree of abnormality for each pixel in a preprocessed image related to the inspection image;
The substrate inspection device according to claim 11 , wherein an area in which a defect is estimated to exist is identified from a difference between the normal image anomaly degree map and the inspection image anomaly degree map.
the defect detection unit uses an algorithm for generating an anomaly degree map by inputting a preprocessed image when creating the normal image anomaly degree map and the inspection image anomaly degree map;
The substrate inspection device according to claim 14 , wherein the algorithm is created by machine learning using the plurality of normal images, the plurality of normal differential images, and at least a portion of the plurality of component images obtained therefrom as training data.
A substrate inspection program for causing a computer to execute a substrate inspection for inspecting a substrate to be inspected by using an inspection image obtained by imaging the substrate, the substrate inspection program comprising:
generating a plurality of normal difference images which are differences between an average image generated from a plurality of normal images obtained by imaging a plurality of normal substrates and each of the plurality of normal images; generating a plurality of component images in a color space from the plurality of normal images and the plurality of normal difference images; and generating a preprocessed image for each of the plurality of normal images based on variance information relating to the variance of component values of each pixel included in the plurality of component images;
generating an inspection difference image which is a difference between the inspection image and the average image, generating a plurality of component images in a color space from the inspection image and the inspection difference image, and generating a preprocessed image for the inspection image based on variance information relating to the variance of component values of each pixel included in the plurality of component images;
inspecting the substrate to be inspected for defects based on a preprocessed image related to the normal image and a preprocessed image related to the inspection image;
A circuit board inspection program that causes a computer to execute the above.