KR102575268B1 - Wafer vision inspection apparatus and method with precise autofocus function based on artificial intelligence - Google Patents

Abstract

The present invention provides a vision inspection device for wafer, having a precise autofocus function based on artificial intelligence, and a method thereof. The vision inspection device for a wafer can improve the accuracy of inspection by detecting a fine fault or a foreign substance in a wafer by using an inspection image of high quality. According to an aspect of the present disclosure, an inspection method for a vision inspection device for a wafer includes: a step of obtaining height distribution information about a stage; a step of obtaining an estimated height value about a loaded wafer when the wafer is loaded on the stage; a step of obtaining the inspection image by adjusting a focus position of a camera by each photographing position of the camera of the vision inspection device for a wafer based on the obtained height distribution information and the obtained height distribution information and the obtained estimated height value; and a step of executing the vision inspection of the wafer based on the obtained inspection image. The step of obtaining the estimated height value of the loaded wafer includes: a step of obtaining an image having a predetermined position in the wafer as the center; and a step of inputting the obtained image as a wafer height estimation model based on deep learning and obtaining the estimated height value which is outputted by the wafer height estimation model.

Description

Wafer vision inspection apparatus and method having precision autofocus function based on artificial intelligence

The technical concept of the present disclosure relates to a wafer vision inspection apparatus and method, and particularly to a wafer vision inspection apparatus and method having an artificial intelligence-based precise autofocus function.

A semiconductor manufacturing process can be largely divided into a pre-process of forming a semiconductor integrated circuit (die) on a wafer and a post-process of forming a semiconductor chip by inspecting and packaging the wafer on which the die is formed. Among them, the inspection process may correspond to a process of performing a visual (vision) inspection and an electrical inspection of a wafer on which dies are formed.

As the semiconductor manufacturing process continues to be miniaturized, the influence of fine defects on the wafer or foreign matter such as dust provided from the outside may increase. Accordingly, a device used for wafer vision inspection must be implemented to accurately detect fine defects or foreign materials and perform pass/fail determination on each die on the wafer.

When a wafer on which dies are formed is loaded (and adsorbed) onto a stage, the wafer vision inspection apparatus acquires an image of the wafer using a camera, and performs a visual inspection of each of the dies on the wafer from the obtained image. . At this time, when there is non-uniformity or deformation of the stage surface, the distance (height) between the camera and the stage surface may be different depending on the position. Such a distance difference may cause the camera to be out of focus, thereby degrading the sharpness of the image. In addition, a height (thickness) deviation may exist between wafers, and such height deviation may also cause deterioration of sharpness of an image.

One problem to be solved by the present invention is to provide a wafer inspection apparatus and method capable of obtaining clear images for each position of a wafer on which dies are formed.

In order to achieve the above object, an inspection method of a wafer vision inspection apparatus according to an aspect according to the technical spirit of the present disclosure includes obtaining height distribution information for a stage; obtaining a height estimation value for the loaded wafer when a wafer is loaded on the stage; obtaining an inspection image while adjusting a focus position of the camera for each imaging position of the camera of the wafer vision inspection apparatus, based on the acquired height distribution information and the height estimation value; and performing a vision inspection of the wafer based on the acquired inspection image, wherein obtaining a height estimation value for the loaded wafer includes obtaining an image centered on a preset position in the wafer. ; and inputting the obtained image to a deep learning-based wafer height estimation model, and acquiring the height estimation value output from the wafer height estimation model.

According to an embodiment, the wafer height estimation model may include at least one artificial neural network trained to output a height estimation value of the wafer based on the standard deviation and brightness values of the input image.

According to an embodiment, the wafer height estimation model outputs a height estimate value of a wafer corresponding to the training image based on a training image, standard deviation and brightness values of the training image, and focus position information of the training image, and , It can be learned by comparing the output height estimation value with the correct height value of the learning image, and updating the weight of the node in the at least one artificial neural network based on the comparison result.

According to an embodiment, the obtaining of the height distribution information may include acquiring an image of each of a plurality of points in the stage; inputting each of the obtained plurality of images into a stage surface height estimation model based on deep learning; and obtaining the height distribution information using a height estimation value of each of the plurality of images output from the stage surface height estimation model.

According to an embodiment, the stage surface height estimation model is trained to output an estimated height of the stage surface corresponding to the marker based on standard deviation and brightness values of a partial area including the marker in the input image. It may include one artificial neural network.

According to an embodiment, the obtaining of the height distribution information by using the height estimation value of each of the plurality of images may include: obtaining a surface height value for each of the plurality of points by using the height estimation value; obtaining a surface height value for each of positions other than the plurality of points through interpolation using the surface height values of adjacent points; and obtaining the height distribution information including the obtained surface height values.

According to an embodiment, the obtaining of the inspection image may include obtaining focus adjustment information for each imaging position of the camera based on the height distribution information and the height estimation value; adjusting a focus position for each imaging position of the camera based on the acquired focus adjustment information; and acquiring partial images for each imaging position of the camera, and obtaining an inspection image of the wafer based on the obtained partial images.

A wafer vision inspection apparatus according to an aspect of the present disclosure includes a stage on which a wafer to be inspected is loaded; a camera disposed facing one surface of the stage to acquire an image including a partial area of the stage or the wafer; a focus position adjusting unit adjusting a focus position of the camera; a camera position moving unit that relatively moves the camera in a direction parallel to the surface of the stage; a controller including at least one processor; and a memory, wherein the control unit acquires an image of each of a plurality of points in the stage using the camera before loading the wafer, and estimates the stage surface height based on deep learning for each of the obtained plurality of images. Height distribution information on the surface of the stage is obtained by using a height estimation value of each of the plurality of images that is input to the model and output from the stage surface height estimation model.

According to the technical idea of the present disclosure, the wafer vision inspection apparatus automatically adjusts the focus position of a camera based on height distribution information of a stage on which a wafer is placed and a height estimation value representing the height (thickness) of the wafer, thereby providing a clearer view of the wafer. High-quality inspection images can be acquired. The wafer vision inspection apparatus can improve inspection accuracy by detecting micro-defects or foreign substances on the wafer using the acquired high-quality inspection image.

In addition, since the wafer vision inspection apparatus can acquire the height distribution information and the height estimation value only with an image obtained from a camera using an artificial intelligence (deep learning)-based stage surface height estimation model and a wafer height estimation model, Compared to this, it enables resource saving and speed improvement in the focus search and adjustment process.

Effects that can be obtained by the wafer vision inspection apparatus and method according to the technical idea of the present disclosure are not limited to the above-mentioned effects, and other effects not mentioned are common in the art to which the present invention belongs from the description below. It will be clearly understood by those who have knowledge.

A brief description of each figure is provided in order to more fully understand the figures cited in this disclosure.
1 is a diagram schematically illustrating a process in which a wafer vision inspection apparatus acquires an inspection image of a wafer loaded on a stage.
2 is a block diagram schematically illustrating a control configuration of a wafer vision inspection apparatus according to an exemplary embodiment of the present disclosure.
3 is a flowchart illustrating a wafer vision inspection method according to an exemplary embodiment of the present disclosure.
FIG. 4 is a flowchart for explaining a specific embodiment of step S300 of FIG. 3 .
FIG. 5 is an exemplary diagram for explaining the embodiment shown in FIG. 4 .
6 is a flowchart for explaining a specific embodiment of step S320 of FIG. 3 .
FIG. 7 is an exemplary diagram for explaining the embodiment shown in FIG. 6 .
8 is an exemplary diagram for explaining an operation of obtaining an inspection image and controlling autofocus using height distribution information and a height estimation value obtained according to an embodiment of the present disclosure.

Exemplary embodiments according to the technical idea of the present disclosure are provided to more completely explain the technical idea of the present disclosure to those skilled in the art, and the following embodiments are modified in various forms. It may be, and the scope of the technical idea of the present disclosure is not limited to the following embodiments. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the spirit of the invention to those skilled in the art.

Although terms such as first and second are used in this disclosure to describe various members, regions, layers, regions and/or components, these members, parts, regions, layers, regions and/or components do not refer to these terms. It is self-evident that it should not be limited by These terms do not imply any particular order, top or bottom, or superiority or inferiority, and are used only to distinguish one member, region, region, or component from another member, region, region, or component. Accordingly, a first member, region, region, or component to be described in detail below may refer to a second member, region, region, or component without departing from the teachings of the technical concept of the present disclosure. For example, a first element may be termed a second element, and similarly, the second element may be termed a first element, without departing from the scope of the present disclosure.

Unless defined otherwise, all terms used herein, including technical terms and scientific terms, have the same meaning as commonly understood by those of ordinary skill in the art to which the concepts of the present disclosure belong. In addition, commonly used terms as defined in the dictionary should be interpreted as having a meaning consistent with what they mean in the context of the technology to which they relate, and in an overly formal sense unless explicitly defined herein. will not be interpreted.

When an embodiment is otherwise implementable, a specific process sequence may be performed differently from the described sequence. For example, two processes described in succession may be performed substantially simultaneously, or may be performed in an order reverse to the order described.

In the accompanying drawings, variations of the shapes shown may be expected, eg depending on manufacturing techniques and/or tolerances. Therefore, embodiments according to the technical idea of the present disclosure should not be construed as being limited to the specific shape of the region shown in the present disclosure, and should include, for example, a change in shape resulting from a manufacturing process. The same reference numerals are used for the same components in the drawings, and duplicate descriptions thereof are omitted.

The term 'and/or' as used herein includes each and every combination of one or more of the recited elements.

Hereinafter, embodiments according to the technical idea of the present disclosure will be described in detail with reference to the accompanying drawings.

1 is a diagram schematically illustrating a process in which a wafer vision inspection apparatus acquires an inspection image of a wafer loaded on a stage.

Referring to FIG. 1 , when a wafer 10 is loaded onto a stage 101, the wafer vision inspection apparatus uses a camera 110 to form a die 30 formed on a surface (eg, upper surface) of the loaded wafer 10. ) can be acquired. The wafer vision inspection apparatus may perform a vision inspection on each die 30 using the obtained inspection image.

Meanwhile, since circuit patterns formed on the die 30, micro-defects, and foreign substances on the wafer 10 are very small in size, the camera 110 focuses on the surface of the wafer 10 in order to obtain a clear inspection image. It is necessary to precisely adjust the position.

On the other hand, the surface of the stage 101 on which the wafer 10 is loaded and seated (and adsorbed) has a height from the ground or reference plane due to non-uniformity or deviation occurring during the stage manufacturing process, or deformation due to external influences. can be very different. For example, as illustrated in FIG. 1 , the height H1 of the surface corresponding to the first position of the stage 101 may be different from the height H2 of the surface corresponding to the second position.

As the surface height of the stage 101 is different for each position, the surface height of the wafer 10 loaded on the stage 101 may also vary for each position. In this case, the distance D1 between the die 30a in the first position and the camera 110 may also be different from the distance D2 between the die 30b in the second position and the camera 110 .

When the distance from the camera 110 is different for each die and the focus position of the camera 110 is fixed, the sharpness of some of the inspection images obtained by the camera 110 is degraded as the focus is shifted, and the corresponding inspection image is out of focus. Inspection accuracy using images may deteriorate. Also, height (thickness) deviations may exist between wafers, and sharpness of some of the inspection images may be deteriorated due to such height deviations.

In order to solve this problem, the wafer vision inspection apparatus according to an embodiment of the present disclosure obtains height distribution information for each position of the stage 101 using images obtained from the camera 110 and calculates a height difference between wafers 10. A height estimation value for correction may be obtained. The wafer vision inspection apparatus may obtain a clear inspection image by precisely adjusting the focus of the camera 110 using the obtained height distribution information and the height estimation value.

A wafer vision inspection apparatus and method according to an embodiment of the present disclosure will be described in detail with reference to the following drawings.

2 is a block diagram schematically illustrating a control configuration of a wafer vision inspection apparatus according to an exemplary embodiment of the present disclosure.

Referring to FIG. 2 , the wafer vision inspection apparatus 100 includes a camera 110, a focus position adjusting unit 120, a camera position moving unit 130, a light source 140, an input unit 150, an output unit 160, A memory 170 and a controller 180 may be included. The configuration of the wafer vision inspection apparatus 100 shown in FIG. 2 is an example for convenience of explanation, and the wafer vision inspection apparatus 100 may not have some of the configurations shown in FIG. In addition, additional configurations may be further included.

The camera 110 may be disposed to face one side of the stage 101 and acquire an image including the stage 101 or a predetermined area of the wafer 10 loaded onto the stage 101 . For example, the camera 110 may be disposed above the stage 101 and face the top surface of the stage 101 . The camera 110 may be implemented as a line scan camera or an area scan camera.

The focus position adjusting unit 120 corresponds to a component for adjusting the focus position of the camera 110 . For example, the focus position adjusting unit 120 is a component that adjusts the focus position by adjusting the distance between the camera 110 and the stage 101, and a device (such as a motor) or a stage for adjusting the height of the camera 110 ( 101) may include a device (motor, etc.) for adjusting the height. Depending on the embodiment, the focus position adjusting unit 120 may correspond to a device that adjusts the focus position by adjusting the distance between the camera 110 and a lens (not shown).

The camera position moving unit 130 relatively moves the camera 110 in a direction substantially parallel to the stage 101 or the surface of the wafer 10, and may include at least one motor or a moving mechanism such as a rail. there is. For example, when the surface of the stage 101 or the wafer 10 corresponds to a plane (xy plane) including the x-axis and the y-axis, the camera position moving unit 130 moves the camera 110 in the x-axis and y-axis directions. can be moved The camera 110 acquires images for each region while relatively moving with respect to the stage 101 or the wafer 10 by the camera position moving unit 130, and the controller 180 combines the acquired images for each region to stage (101) or the entire image of the wafer (10) can be obtained.

The light source 140 may be disposed to emit light toward a surface (eg, upper surface) of the stage 101 in order to provide a sufficient amount of light when the camera 110 acquires an image. This light source 140 may include at least one LED.

The input unit 150 is a component for receiving an input related to an inspection operation of the wafer 10 (initiation/end of inspection, confirmation of inspection results, etc.) can include The output unit 160 is a component that outputs various information such as the operating state of the wafer vision inspection apparatus 100, inspection results, or inspection images obtained by the camera 110, and includes a known output means such as a display. can do.

The memory 170 may include at least one volatile/non-volatile memory for storing commands, data, algorithms related to the operation of the wafer vision inspection apparatus 100, the top and bottom images, and inspection result data.

The controller 180 may include one or more processors that control overall operations of elements included in the wafer vision inspection apparatus 100 . The one or more processors may be implemented in hardware such as CPU, MCU, application processor (AP), GPU, FPGA, ASIC, and the like. Meanwhile, the controller 180 may control overall operations of a wafer vision inspection method according to an embodiment of the present disclosure, which will be described below with reference to FIGS. 3 to 8 .

3 is a flowchart illustrating a wafer vision inspection method according to an exemplary embodiment of the present disclosure. FIG. 4 is a flowchart for explaining a specific embodiment of step S300 of FIG. 3 . FIG. 5 is an exemplary diagram for explaining the embodiment shown in FIG. 4 . 6 is a flowchart for explaining a specific embodiment of step S320 of FIG. 3 . FIG. 7 is an exemplary view for explaining the embodiment shown in FIG. 6 . 8 is an exemplary diagram for explaining an operation of acquiring an inspection image and controlling autofocus using height distribution information and a height correction value obtained according to an embodiment of the present disclosure.

Referring to FIG. 3 , the wafer vision inspection apparatus 100 may acquire height distribution information of the stage 101 before the wafer 10 is loaded onto the stage 101 (S300).

As described above, the surface of the stage 101 may have different heights from the reference surface due to non-uniformity or deviation in the manufacturing process, deformation due to external influences, and the like. Accordingly, the wafer vision inspection apparatus 100 may first acquire height distribution information for each position of the stage 101 before the wafer 10 is loaded. For example, the height distribution information may be stored in the form of a table having a height value for each of a plurality of coordinates set for the surface of the stage 101, but is not limited thereto, and various functions such as a function having x coordinates and y coordinates as variables may be stored. It can also be stored in the form

Meanwhile, the wafer vision inspection apparatus 100 obtains height distribution information of the stage 101 using an image acquired through the camera 110 without having a separate additional sensor, thereby enabling an efficient configuration of the apparatus. .

Referring to FIGS. 4 and 5 as specific examples related to this, the wafer vision inspection apparatus 100 may acquire images for each of a plurality of points within the stage 101 (S400).

A plurality of points in the stage 101 may correspond to predefined coordinates spaced apart according to a predetermined interval. The controller 180 controls the camera position moving unit 130 to sequentially move the position of the camera 110 according to each of the plurality of points in order to acquire an image centered on each of the plurality of points. image can be obtained. At this time, the focus position of the camera 110 may be fixed to a preset reference focus position, and surface height information of the stage 101 corresponding to the reference focus position may be stored in the memory 170. .

Depending on the embodiment, a marker 501 may be formed at each of the plurality of points. The marker 501 may be directly formed on the surface of the stage 101 or may be loaded with a separate plate (glass, etc.) on which the marker 501 is formed.

When the marker 501 is directly formed on the surface of the stage 101, the marker 501 needs not be displayed on the inspection image of the wafer 10. For example, the marker 501 may be implemented with a material that expresses color when UV light is irradiated, such as a UV functional ink, and in this case, the light source 140 may include a UV light source.

The plurality of acquired images may have different sharpness depending on whether the focus position matches the surface of the stage. When the sharpness of the image is greater than or equal to the reference sharpness, it may be determined that the focus position coincides with the stage surface, and the height of the stage surface at the corresponding point may be determined to be a height corresponding to the reference focus position.

On the other hand, if the sharpness of the image is less than the standard sharpness, it means that the focus position and the stage surface do not match. Conventional methods of adjusting the focus position based on the image have a problem of not accurately determining in which direction the focus position should be adjusted, and the correct focus position is determined by acquiring the image again while adjusting the focus position. There are inefficient aspects.

Referring to FIGS. 4 to 5 , the wafer vision inspection apparatus 100 may input each of a plurality of obtained images to the stage surface height estimation model 520 (S410). The wafer vision inspection apparatus 100 may obtain height distribution information of the stage 101 using height estimation values output from the stage surface height estimation model (S420).

The stage surface height estimation model 520 may correspond to a deep learning-based model for estimating and providing the stage surface height from an input image. For example, the stage surface height estimation model 520 may include, but is not limited to, a convolutional neural network (CNN), and may include one or more artificial neural networks of various types.

Specifically, the stage surface height estimation model 520 is based on the standard deviation (image contrast) and brightness value (brightness) of the area including the marker 501 in the input image, the position where the marker 501 is displayed. It can be implemented to estimate the stage surface height. The standard deviation is a value related to the sharpness of the image (marker), and may be related to the difference (absolute value) between the height of the stage surface and the focus position, and the brightness value may be related to the distance between the stage surface and the camera 110. can Therefore, in order to estimate the height of the stage surface from the image, not only the standard deviation but also the brightness value should be considered.

To this end, the stage surface height estimation model 520 is used for training images of the above-described marker 501 taken at various focus positions (z-axis position), standard deviation of each of the training images, brightness, and focus position information ( z-axis position) can be used to learn. Learning of the stage surface height estimation model 520 may be performed by the wafer vision inspection device 100 or a separate learning device. Hereinafter, a case in which the learning is performed by the wafer vision inspection apparatus 100 will be described.

For example, when the learning image and focus position information of the learning image are acquired, the controller 180 may apply a known image analysis technique to the learning image to extract a standard deviation and a brightness value. The control unit 180 inputs the extracted standard deviation and brightness values together with the focus position information to the stage surface height estimation model 520, or inputs the extracted standard deviation and brightness values to the stage surface together with the training image and the focus position information. It can be input to the surface height estimation model 520. The stage surface height estimation model 520 may output an estimated height of the stage surface from input data. The control unit 180 compares the output height estimation value with the correct height value (ground truth) corresponding to the training image, and based on the comparison result, the stage surface height estimation model 520 is trained (weight of nodes in the artificial neural network). update) can be performed.

As another example, the stage surface height estimation model 520 includes a first artificial neural network outputting standard deviation and brightness values from the learning image, and the standard deviation and brightness values output from the first artificial neural network, the learning image and focus It may also be configured to include a second artificial neural network that outputs a height estimation value of the stage surface using location information. In this case, the controller 180 performs learning on the first artificial neural network by comparing the standard deviation and brightness values output from the first artificial neural network with the standard deviation and brightness values of the correct answer of the learning image, and performs learning on the second artificial neural network. The second artificial neural network may be trained by comparing the estimated height value output from the correct height value with the correct height value. However, the learning method of the stage surface height estimation model 520 is not limited to the above-described examples, and may be modified within a range that can be easily derived by a person skilled in the art.

The estimated height value corresponds to the z-axis distance (height) between the surface of the stage 101 and a predetermined reference plane, or the adjustment of the focus position adjusting unit 120 for adjusting the focus position to correspond to the height of the surface of the stage 101. value can be displayed.

The control unit 180 sequentially inputs the images obtained for each of the plurality of points and the focus position (z-axis position) of each of the images to the stage surface height estimation model 520 learned as described above, The surface height estimation model 520 may output an estimated height of the stage surface for each of a plurality of points (eg, (Xn, Yn) coordinates (n is a natural number)).

The controller 180 may obtain height distribution information of the stage 101 based on the estimated height values for each of the plurality of points. For example, the controller 180 may obtain a surface height value of the stage 101 for each of the plurality of points by using the estimated height value for each of the plurality of points. Depending on the embodiment, the controller 180 may obtain a surface height value for positions other than the plurality of points by using any one of various interpolation techniques using surface height values of adjacent points. The controller 180 may obtain height distribution information including the surface height values and store the obtained height distribution information in the memory 170 .

The above-described steps S300 or S400 to S420 are performed at predetermined intervals (time period or inspection number period) set in the device 100, performed whenever the power of the device 100 is turned on, or stage ( 101) may be performed according to various criteria, such as being performed whenever replacement is performed or performed according to predetermined conditions (vibration detection, manager input, etc.).

Figure 3 will be described again.

The wafer vision inspection apparatus 100 may load the wafer 10 on the stage 101 (S310) and obtain a height estimation value for the loaded wafer 10 (S320).

Since height distribution information of the stage 101 on which the wafer 10 is not loaded is acquired in step S300, when the focus position of the camera 110 is adjusted based only on the height distribution information, the height (thickness) of the wafer 10 ), a clear image of the wafer surface may not be obtained. Accordingly, the wafer vision inspection apparatus 100 needs to adjust the focus position of the camera 110 by reflecting the height of the loaded wafer.

On the other hand, as described above, there may be a height (thickness) variation between wafers, and reflecting the same height value for each wafer may cause a sharpness variation of an image. Accordingly, the wafer vision inspection apparatus 100 may obtain a height estimation value for the wafer 10 loaded on the stage 101 and adjust the focus position of the camera 110 based on the obtained height estimation value.

Referring to FIGS. 6 and 7 as specific examples related to step S320 , the wafer vision inspection apparatus 100 may acquire an image 720 of a predetermined position within the wafer 700 loaded onto the stage 101 . Yes (S600).

For example, the wafer vision inspection apparatus 100 moves the camera 110 to a predetermined position to measure the height (thickness) of the loaded wafer 700, and acquires an image 720 centered on the moved position. can do.

The predetermined position is set to a position closest to the initial position of the camera 110, so that the time required to move the position of the camera 110 can be minimized. More specifically, the predetermined position corresponds to the center position of a die that is closest to the initial position among the dies 710 formed on the wafer 700, or a pattern such as a circuit formed on the die 710. Among them, it may correspond to the position of a pattern (eg, the position of a bump ball, etc.) corresponding to the main inspection target. The controller 180 may control the camera 110 to obtain an image 720 including the die present at the moved position.

Meanwhile, the focus position of the camera 110 in step S600 may correspond to the reference focus position described above in step S400 of FIG. It may correspond to a focus position.

The wafer vision inspection apparatus 100 may obtain a height estimation value for the loaded wafer 700 by inputting the acquired image 720 to a wafer height estimation model (S610).

Similar to the above description in steps S410 to S420, the wafer height estimation model 730 may correspond to a deep learning-based model for estimating and providing the height (thickness) of the wafer from the input image 720. For example, the wafer height estimation model 730 may include a CNN, but is not limited thereto, and may include one or more artificial neural networks of various types.

An overall implementation method and learning method of the wafer height estimation model 730 may be similar to the stage surface height estimation model 520 described above in steps S410 to S420.

Specifically, the wafer height estimation model 730 may be implemented to estimate the height (thickness) of the wafer 700 based on the standard deviation and brightness values of the input image. As described above, the standard deviation is a value related to the sharpness of the image, and may be related to the difference (absolute value) between the surface height of the wafer 700 and the focus position, and the brightness value is the difference between the wafer surface and the camera 110. It may be related to the distance between them. Therefore, in order to estimate the height of the wafer 700 from the image, not only the standard deviation but also the brightness value should be considered.

To this end, the wafer height estimation model 730 calculates learning images of a die (or a predetermined object) at various focus positions (z-axis position), standard deviation, brightness, and focus position information of each of the training images. can be learned using Learning of the wafer height estimation model 730 may be performed by the wafer vision inspection device 100 or a separate learning device. Hereinafter, a case in which the learning is performed by the wafer vision inspection apparatus 100 will be described.

For example, when the learning image and focus position information of the learning image are acquired, the controller 180 may apply a known image analysis technique to the learning image to extract a standard deviation and a brightness value. The controller 180 inputs the extracted standard deviation and brightness values together with the focus position information into the wafer height estimation model 730, or inputs the extracted standard deviation and brightness values together with the training image and the focus position information to the wafer height estimation model 730. It can be input to the estimation model 730. The wafer height estimation model 730 may output a height estimation value of the wafer 700 from input data. The controller 180 may compare the output height estimation value with a ground truth corresponding to the learning image, and perform learning on the wafer height estimation model 730 based on the comparison result.

As another example, the wafer height estimation model 730 includes a first artificial neural network that outputs standard deviation and brightness values from the training image, and the standard deviation and brightness values output from the first artificial neural network, the training image and the focus position. It may also be configured to include a second artificial neural network that outputs a height estimation value of the wafer 700 using the information. In this case, the controller 180 performs learning on the first artificial neural network by comparing the standard deviation and brightness values output from the first artificial neural network with the standard deviation and brightness values of the correct answer of the learning image, and performs learning on the second artificial neural network. The second artificial neural network may be trained by comparing the estimated height value output from the correct height value with the correct height value. However, the learning method of the wafer height estimation model 730 is not limited to the above-described examples, and may be modified within a range that can be easily derived by a person skilled in the art.

The estimated height value may indicate the height (thickness) of the wafer 700 or may correspond to an adjustment value of the focus position adjusting unit 120 for setting the focus to a position corresponding to the surface of the wafer 700 .

According to an embodiment, the wafer vision inspection apparatus 100 obtains a plurality of height estimation values by performing steps S600 to S610 for each of a plurality of positions of the wafer 700, and based on the obtained height estimation values, the wafer ( 700) height distribution information may be acquired. In this case, when there is a height deviation for each region in the wafer 700 itself, the height deviation may be more accurately reflected.

Referring back to FIG. 3 , the wafer vision inspection apparatus 100 acquires an inspection image while adjusting the focus position of the camera 110 for each imaging position of the camera 110 based on the obtained height distribution information and the height estimation value. It can (S330). The wafer vision inspection apparatus 100 may perform an inspection of the wafer 10 based on the obtained inspection image (S340).

Referring to FIG. 8 together, the control unit 180 controls the camera position moving unit 130 to move the camera 110 (eg, in the x-axis and y-axis directions) to obtain an inspection image of the wafer 800. can do.

At this time, the controller 180 may adjust the focus position for each imaging position of the camera 110 using the height distribution information of the stage obtained in steps S300 and S320 and the estimated height value of the wafer 800 . The control unit 180 obtains focus adjustment information for the imaging position of the camera 110 based on the height distribution information and the height estimation value stored in the memory 170, and adjusts the focus position adjustment unit 120 according to the obtained focus adjustment information. can control. For example, when the height distribution information includes height values from the reference plane for each of a plurality of points on the stage surface and the estimated height value indicates the height (thickness) of the wafer 800, the controller 180 controls the height distribution Focus adjustment information for adjusting a focus position to correspond to a value obtained by adding the estimated height value to each height value included in the information may be obtained. Depending on the embodiment, the estimated height value may indicate a height from the reference surface to the surface of the wafer 800 placed on the stage at a predetermined point. In this case, the control unit 180 calculates the height (thickness) of the wafer 800 itself through the difference between the height value of the predetermined point and the height estimation value among the height distribution information, and the calculated height is the height distribution information. Focus adjustment information may be obtained based on a value added to each height value included in .

The focus adjustment information may be different according to the configuration of the focus position adjusting unit 120. For example, the focus adjustment information moves the camera 110 or the stage 101 in a vertical direction (z-axis) with respect to the surface of the wafer 800. It may include information for moving or information for adjusting a distance between the camera 110 and a lens (not shown).

As the focus position of the camera 110 is adjusted for each imaging position, the sharpness of the partial image 820 obtained by the camera 110 may maintain a predetermined standard or higher. The controller 180 may acquire an inspection image of the wafer 800 by combining the partial images 820 and perform an inspection operation of determining whether the wafer 800 is good or bad by using the acquired inspection image.

According to an embodiment of the present disclosure, the wafer vision inspection apparatus automatically adjusts (autofocusing) a focus position of a camera based on height distribution information of a stage on which a wafer is placed and an estimate of the height of the wafer, thereby providing a clearer, higher quality image of the wafer. An inspection image can be acquired. The wafer vision inspection apparatus can improve inspection accuracy by detecting micro-defects or foreign substances on the wafer using the acquired high-quality inspection image.

In addition, since the wafer vision inspection apparatus can acquire the height distribution information and the height estimation value only with an image obtained from a camera using an artificial intelligence (deep learning)-based stage surface height estimation model and a wafer height estimation model, Compared to this, it enables resource saving and speed improvement in the focus search and adjustment process.

Since the description of the above embodiments is only examples with reference to the drawings for a more thorough understanding of the present disclosure, it should not be construed as limiting the technical spirit of the present disclosure.

In addition, it will be clear to those skilled in the art that various changes and modifications are possible within a range that does not deviate from the basic principles of the present disclosure.

Claims (14)

In the inspection method of the wafer vision inspection device,
obtaining height distribution information of stages of the wafer vision inspection apparatus;
obtaining a height estimation value for the loaded wafer when a wafer is loaded on the stage;
obtaining an inspection image while adjusting a focus position of the camera for each imaging position of the camera of the wafer vision inspection apparatus, based on the acquired height distribution information and the height estimation value; and
Performing a vision inspection of the wafer based on the acquired inspection image;
Obtaining a height estimation value for the loaded wafer,
obtaining an image centered on a predetermined position in the wafer; and
Inputting the obtained image to a deep learning-based wafer height estimation model and obtaining the height estimation value output from the wafer height estimation model,
The wafer height estimation model,
At least one artificial neural network trained to output a height estimation value of the wafer based on the standard deviation (contrast) and brightness value (brightness) of the input image,
outputting an estimated height of a wafer corresponding to the learning image based on the learning image, standard deviation and brightness values of the learning image, and focus position information of the learning image;
Learning by comparing the output height estimation value with the correct height value of the learning image and updating the weight of the node in the at least one artificial neural network based on the comparison result,
method.
delete delete According to claim 1,
Obtaining the height distribution information,
obtaining an image for each of a plurality of points within the stage;
inputting each of the obtained plurality of images into a stage surface height estimation model based on deep learning; and
Obtaining the height distribution information using a height estimation value of each of the plurality of images output from the stage surface height estimation model,
method.
According to claim 4,
The stage surface height estimation model,
At least one artificial neural network trained to output a height estimation value of the stage surface corresponding to the marker based on the standard deviation and brightness value of a partial region including the marker in the input image,
method.
According to claim 4,
Obtaining the height distribution information using a height estimation value for each of the plurality of images,
obtaining a surface height value for each of the plurality of points by using the estimated height value;
obtaining a surface height value for each of positions other than the plurality of points through interpolation using the surface height values of adjacent points; and
Obtaining the height distribution information including the obtained surface height values,
method.
According to claim 1,
Obtaining the inspection image,
obtaining focus adjustment information for each imaging position of the camera based on the height distribution information and the height estimation value;
adjusting a focus position for each imaging position of the camera based on the acquired focus adjustment information; and
Acquiring partial images for each imaging position of the camera, and obtaining an inspection image of the wafer based on the obtained partial images.
method.
In the wafer vision inspection device,
a stage on which a wafer to be inspected is loaded;
a camera disposed facing one surface of the stage to acquire an image including a partial area of the stage or the wafer;
a focus position adjusting unit adjusting a focus position of the camera;
a camera position moving unit that relatively moves the camera in a direction parallel to the surface of the stage;
a controller including at least one processor; and
contains memory;
The control unit,
Before loading the wafer, acquiring an image for each of a plurality of points on the stage using the camera;
Each of the obtained plurality of images is input to a deep learning-based stage surface height estimation model,
Obtaining height distribution information on the surface of the stage using a height estimation value of each of the plurality of images output from the stage surface height estimation model;
Acquiring an image centered on a predetermined position in a wafer loaded on the stage;
Inputting the obtained image to a deep learning-based wafer height estimation model to obtain the height estimation value output from the wafer height estimation model,
The wafer height estimation model,
At least one artificial neural network trained to output a height estimation value of the wafer based on the standard deviation (contrast) and brightness value (brightness) of the input image,
outputting an estimated height of a wafer corresponding to the learning image based on the learning image, standard deviation and brightness values of the learning image, and focus position information of the learning image;
Learning by comparing the output height estimation value with the correct height value of the learning image and updating the weight of the node in the at least one artificial neural network based on the comparison result,
inspection device.
According to claim 8,
The stage surface height estimation model,
Including at least one artificial neural network trained to output a height estimation value of the input image based on the standard deviation and brightness value of the input image,
inspection device.
According to claim 9,
The image obtained for each of the plurality of points includes a marker disposed around the corresponding point,
The marker is formed on the surface of the stage, or included in the image by loading the plate on which the marker is formed on the stage.
inspection device.
According to claim 8,
The control unit,
Controlling the camera to acquire the image while sequentially moving the position of the camera according to each of the plurality of points by controlling the camera position movement unit,
inspection device.
According to claim 8,
The control unit,
Obtaining a surface height value for each of the plurality of points using the height estimation value;
Obtaining a surface height value through interpolation using surface height values of adjacent points for each of the positions other than the plurality of points;
Obtaining the height distribution information including the obtained surface height values,
inspection device.
According to claim 8,
The control unit,
Acquiring an image centered on a predetermined position in a wafer loaded on the stage;
Inputting the acquired image to a deep learning-based wafer height estimation model to obtain a wafer height estimation value output from the wafer height estimation model,
inspection device.
According to claim 13,
The control unit,
Obtaining focus adjustment information for each imaging position of the camera based on the height distribution information and the wafer height estimation value;
Controlling the focus position adjusting unit to adjust the focus position for each imaging position of the camera based on the acquired focus adjustment information;
Obtaining partial images for each imaging position of the camera moving relative to the stage under the control of the camera position moving unit;
Obtaining an inspection image of the wafer based on the obtained partial images,
inspection device.

Images