JP2001165618A - Image recognition device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To perform image recognition easily, at low cost and with high accuracy. SOLUTION: An arbitrary point P on a semiconductor wafer placed on a suction plate 19 is translated by an inspection stage 14 in an image captured by an optical unit 21. Then, by the image recognition computer 60,
3 points the coordinates of the point P, and acquired by the coordinate P _M in the mechanical coordinate system, respectively, the coordinates P _A in the camera coordinate system. Then, by the image recognition computer 60,
Seeking linear transformation matrix A of the movement amount of the point P from the mechanical coordinate system to the camera coordinate system, equation P _M = A ^-1 when the origin coordinates of the camera coordinate system and O _A in mechanical coordinate system
Performing image recognition by associating the mechanical coordinate system and the camera coordinate system using the P _A + O _A.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an image recognition apparatus used for inspecting a semiconductor wafer on which a predetermined device pattern has been formed.

[0002]

2. Description of the Related Art A semiconductor device is manufactured by forming a fine device pattern on a semiconductor wafer. When such a device pattern is formed, dust or the like may adhere to the semiconductor wafer or may be damaged, resulting in defects. A semiconductor device having such a defect becomes a defective device and reduces the yield.

[0003] Therefore, in order to stabilize the yield of a production line at a high level, defects generated by dust and scratches are found at an early stage, the causes thereof are identified, and effective measures are taken for production facilities and production processes. It is preferable to take it.

[0004] Therefore, when a defect is found, an image recognizing device is used to check what the defect is and to classify the defect to identify the equipment or process that caused the defect. ing. Here, the image recognition device for examining what the defect is is like an optical microscope, and identifies the defect by looking at the defect in an enlarged manner.

[0005]

By the way, in recent years, semiconductor devices have been increasingly miniaturized with high integration, and the line width has been reduced to about 0.18 μm or less. For this reason, in a semiconductor device, normal operation cannot be performed even with extremely fine dust or the like which has not been regarded as a problem in the past.

Therefore, in order to perform a defect inspection of a semiconductor device formed with such a fine device pattern, the image recognition apparatus captures an image of the semiconductor wafer with a lens having a higher magnification, and furthermore, images with a higher precision. It is necessary to perform recognition. At this time, it is increasingly important for the image recognition apparatus to make the position of the defect identified in the captured image correspond to the position on the actual semiconductor wafer with high accuracy. Specifically, it is necessary to identify the position of the defect on the semiconductor wafer with an accuracy of, for example, about 1 μm or less.

As described above, in order to identify a position of a defect with high accuracy, when imaging a predetermined position on a semiconductor wafer, it is necessary to move a stage fixedly supporting the semiconductor wafer to which position. It is necessary to make a correspondence with high accuracy with high accuracy. However, in an image recognition device, the mounting position accuracy of an optical system component such as a camera or a lens is not sufficient, and a component mounting error unique to each device is measured to determine a correspondence between a stage and a camera field of view. It is necessary to perform correction with high accuracy (calibration). However, in the conventional image recognition device, using a jig whose absolute accuracy is guaranteed,
Such calibration is performed. However, in an image recognition apparatus, it is extremely difficult to manufacture a jig with a positional accuracy of about 1 μm or less and dispose it on the apparatus.

Further, the image recognition apparatus includes, for example, a low-magnification lens having a wide field of view and a low-magnification lens in order to efficiently perform a defect inspection of a semiconductor device formed with a fine device pattern.
It is desirable to use a high-power lens with a narrow field of view. However, in a conventional image recognition apparatus, the accuracy of placing a semiconductor wafer to be inspected on a stage that fixedly supports the semiconductor wafer is about 100 μm. Therefore, in the image recognition device, every time a semiconductor wafer is placed on the stage, it is necessary to perform correction according to the semiconductor wafer. In addition, as described above, when capturing images by switching a plurality of lenses, or when changing the lens magnification, it is necessary to keep the subject from losing sight. It is necessary that the positional accuracy be about 1 μm or less.
However, in the image recognition apparatus, it is extremely costly and extremely difficult to achieve high-accuracy mounting positions among a plurality of lenses.

Accordingly, the present invention has been proposed in view of the above-mentioned conventional situation, and has as its object to provide an image recognition apparatus capable of performing image recognition easily, at low cost, and with high accuracy. Aim.

[0010]

An image recognition apparatus according to the present invention is provided on a base and the base, fixedly supports an object to be inspected, and holds the object within a predetermined plane. Fixed support means that is movable in parallel, imaging means that is provided on the base, and that captures the object from outside a plane in which the fixed support is free to move the object; Image recognizing means for recognizing the image by arithmetically processing the image taken by the means. The image recognizing means performs image recognition by associating a mechanical coordinate system that is a coordinate system of the base with a camera coordinate system that is a coordinate system in an image captured by the imaging device. An arbitrary point P on the object to be inspected is translated by the fixed support means in an image picked up by the image pickup means.
Points the coordinates 3, and the coordinates P _M in the mechanical coordinate system, respectively, obtained at the coordinate P _A in the camera coordinate system, the coordinates of three points in the mechanical coordinate system _{P M1, P M2, P M3} ,
When the coordinates of three points in the camera coordinate system are P _A1 , P _A2 , and P _A3 ,

[0011]

(Equation 10)

[0012] As, determined by Equation 1 below linear transformation matrix A movement amount of the point P from the mechanical coordinate system to the camera coordinate system, the relationship when the origin coordinates of the camera coordinate system and O _A in mechanical coordinate system performing image recognition by associating the mechanical coordinate system and the camera coordinate system using the equation ^{_{P M = a -1 P a +}} O a.

[0013]

[Equation 11]

The image recognition apparatus configured as described above
The mechanical coordinate system and the camera coordinate system can be associated with each other based on the image captured by the imaging unit without using a special jig. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

An image recognition apparatus according to the present invention is provided on a base and the base, fixedly supports an object to be inspected, and is capable of moving the object in parallel within a predetermined plane. And fixed support means rotatable about an axis perpendicular to this surface and provided with an index indicating the center of rotation, and fixed support means provided on the base and the fixed support means provided on the base. An image pickup means for picking up an image of the object to be inspected from outside the surface in which the object is movable, and an image recognizing means for performing image processing on the image picked up by the image pickup means for image recognition. The image recognizing means calculates the coordinates of an arbitrary point P on the inspection object in the mechanical coordinate system by P
_M , when the coordinates of the point P in the camera coordinate system are P _A , and the primary transformation matrix of the amount of movement of the point P from the mechanical coordinate system to the camera coordinate system is A, the above-mentioned imaging means captures the above-mentioned index. by acquires coordinates O _S of the rotation center of said fixed support means in a mechanical coordinate system, determined by equation 2 below the origin coordinates O _a of the camera coordinate system in the mechanical coordinate system, equation P _M = a ^-1 P performing image recognition by associating the mechanical coordinate system and the camera coordinate system by using the _a + O _a.

[0016]

(Equation 12)

The image recognition device configured as described above
The mechanical coordinate system and the camera coordinate system can be associated with each other based on the image captured by the imaging unit without using a special jig. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

Further, the image recognition apparatus according to the present invention is provided on the base and the base, fixedly supports the inspection object, and is capable of moving the inspection object in parallel within a predetermined plane. And fixed support means rotatable about an axis perpendicular to this plane, and the fixed support means provided on the base, wherein the fixed support means makes the object to be inspected movable from outside the plane. An image pickup means for picking up an object to be inspected, and an image recognizing means for recognizing an image picked up by the image pickup means by performing arithmetic processing. Then, the image recognition means, the coordinates of an arbitrary point P P _M on the object to be inspected in a mechanical coordinate system, the coordinates in the camera coordinate system of the point P P _A, to the camera coordinate system from the mechanical coordinate system the linear transformation matrix of the movement amount of the point P a, the coordinates of the center of rotation of the fixed support means when the O _S in the mechanical coordinate system, rotating the coordinates of the point P within the image which the imaging means for imaging the moved acquired two points the camera coordinate system, the acquired coordinate is taken as P _A1, P _A2, determined by equation 3 below O _a the origin coordinates of the camera coordinate system in the mechanical coordinate system, equation P Image recognition is performed by associating the mechanical coordinate system with the camera coordinate system using _M = A ^-1 P _A + O _A.

[0019]

(Equation 13)

The image recognition apparatus configured as described above
The mechanical coordinate system and the camera coordinate system can be associated with each other based on the image captured by the imaging unit without using a special jig. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

An image recognition apparatus according to the present invention is provided on a base and fixedly supports the object to be inspected, and moves the object in parallel within a predetermined plane. Fixed support means that is freely movable and rotatable about an axis perpendicular to this surface, and provided on the base, and wherein the fixed support means makes the object to be inspected movable from outside the surface. An image pickup means for picking up an image of the object to be inspected, and an image recognizing means for recognizing the image picked up by the image pickup means by performing arithmetic processing. Then, the image recognition means, the coordinates of an arbitrary point P P _M on the object to be inspected in a mechanical coordinate system, the coordinates in the camera coordinate system of the point P P _A, to the camera coordinate system from the mechanical coordinate system When the primary transformation matrix of the movement amount of the point P is A, the coordinates P _A1 of the point P in the camera coordinate system and the coordinates O _S1 of the rotation center of the fixed support means in the mechanical coordinate system are acquired.
After rotating the position of the point P out of the range of the image captured by the imaging unit, the fixed support unit is moved in parallel until the point P falls within the image, and the coordinate P of the point P in the camera coordinate system is obtained. obtains the coordinates O _S2 of the rotation center of said fixed support means in _A2 and the mechanical coordinate system, determined by equation 4 below O _a the origin coordinates of the camera coordinate system in the mechanical coordinate system, equation P _M = a ^- Image recognition is performed by associating the mechanical coordinate system with the camera coordinate system using ¹ P _A + O _A.

[0022]

[Equation 14]

The image recognition device configured as described above
The mechanical coordinate system and the camera coordinate system can be associated with each other based on the image captured by the imaging unit without using a special jig. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

Further, an image recognition apparatus according to the present invention is provided on a base and the base, fixedly supports an object to be inspected, and is capable of moving the object in parallel within a predetermined plane. A fixed support means and a light receiving lens provided on the base, the fixed support means taking an image of the object to be inspected from outside the surface in which the object to be inspected is movable, and a light receiving lens for receiving imaging light. An image recognizing unit includes at least two image capturing units and an image recognizing unit that performs arithmetic processing on the image captured by the image capturing unit to recognize the image. Then, the image recognition means, the coordinates of an arbitrary point P P _M on the object to be inspected in a mechanical coordinate system, the origin of the camera coordinate system the coordinates in the camera coordinate system of the point P P _A, in the mechanical coordinate system coordinates and O _a, between two light-receiving lenses of the imaging means, a ₁ a linear transformation matrix point movement amount P from the mechanical coordinate system to the camera coordinate system when using the first light receiving lens, the The primary transformation matrix when the second light receiving lens is used is A ₂ , and the imaging means translates the coordinates of the point P within the range of the image to be imaged using the second light receiving lens. two points P of coordinates, each obtained in the mechanical coordinate system and the camera coordinate system, P _M1 coordinates of two points in the mechanical coordinate system, P _M2, the coordinates of two points in the camera coordinate system and P _A1, P _A2 When you do

[0025]

(Equation 15)

The primary transformation matrix A ₂ when the second light receiving lens is used is obtained by the following equation 5, and the mechanical coordinate system and the camera coordinate system are calculated using the relational expression P _M = A ⁻¹ P _A + O _A. Image recognition is performed by associating with the system.

[0027]

(Equation 16)

The image recognition device configured as described above
Among the light receiving lenses that receive the imaging light, the primary conversion matrix A ₂ of another light receiving lens can be obtained based on the primary conversion matrix A ₁ of the predetermined light receiving lens. for that reason,
Based on the image captured by the imaging means, the mechanical coordinate system and the camera coordinate system can be associated without using a special jig, and the mechanical coordinate system can be used regardless of the mechanical and optical position accuracy of the device. And the camera coordinate system can be made to correspond with high accuracy and easily.

An image recognition apparatus according to the present invention is provided on a base and the base, fixedly supports the inspection object, and is capable of moving the inspection object in a predetermined plane in parallel. A fixed support means and a light receiving lens provided on the base, the fixed support means taking an image of the object to be inspected from outside the surface in which the object to be inspected is movable, and a light receiving lens for receiving imaging light. Imaging means provided with at least two,
Image recognition means for performing image processing on the image picked up by the image pickup means for image recognition. Then, the image recognition means, the coordinates of an arbitrary point P P _M on the object to be inspected, the coordinates in the camera coordinate system of the point P and P _A in the mechanical coordinate system, first received by the image pickup means When a lens is used, the coordinates P _A1 of the point P in the camera coordinate system, the linear transformation matrix A ₁ of the movement amount of the point P from the mechanical coordinate system to the camera coordinate system, and the camera coordinate system in the mechanical coordinate system The origin coordinate O _A1 is obtained, and the coordinate P _A2 of the point P in the camera coordinate system when the second light receiving lens is used by the above-mentioned imaging means, and the point P from the mechanical coordinate system to the camera coordinate system.
And a primary transformation matrix A ₂ of the moving amount of the camera, and the origin coordinate O _A2 of the camera coordinate system in the mechanical coordinate system when the second light receiving lens is used between the two light receiving lenses in the imaging means is as follows: It is obtained based on the origin coordinate O _A1 of the first light receiving lens by Expression 6, and the relational expression P _M = A ⁻¹ P _A +
Use O _A.

[0030]

[Equation 17]

The image recognition device configured as described above
Based on the origin coordinate O _A1 of the camera coordinate system in the mechanical coordinate system when using the predetermined light receiving lens among the light receiving lenses that receive the imaging light, the origin coordinate in the camera coordinate system when using another light receiving lens O _A2 can be determined. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond to each other without using a special jig based on the image captured by the image capturing means. The coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily.

Further, the image recognition device according to the present invention is provided on the base and fixedly supports the inspection object, and can move the inspection object in parallel in a predetermined plane. A fixed support means and a light receiving lens provided on the base, the fixed support means taking an image of the object to be inspected from outside the surface in which the object to be inspected is movable, and a light receiving lens for receiving imaging light. An image recognizing unit includes at least two image capturing units and an image recognizing unit that performs arithmetic processing on the image captured by the image capturing unit to recognize the image. Then, the image recognition means, an arbitrary point coordinates P on the mechanical coordinate system the inspection object P _M, the coordinates in the camera coordinate system of the point P and P _A, the first light receiving lens by the image pickup means Is used, the coordinates P _M1 of the point P in the mechanical coordinate system, the coordinates P _A1 of the point P in the camera coordinate system, and the linear transformation matrix of the movement amount of the point P from the mechanical coordinate system to the camera coordinate system A
_1, obtains the origin coordinate O _A1 of the camera coordinate system in the mechanical coordinate system, when using the second light receiving lens by the imaging means, the range of the image outside the position of the point P is imaged by the imaging means , The fixed support means is moved in parallel until the point P falls within the image, and the coordinate P _M2 of the point P in the mechanical coordinate system and the coordinate P _A2 of the point P in the camera coordinate system are obtained. , The point P from the mechanical coordinate system to the camera coordinate system
And a primary transformation matrix A ₂ of the moving amount of the camera, and the origin coordinate O _A2 of the camera coordinate system in the mechanical coordinate system when the second light receiving lens is used between the two light receiving lenses in the imaging means is as follows: It is obtained based on the origin coordinate O _A1 of the first light receiving lens by Expression 7, and the relational expression P _M = A ⁻¹ P _A +
Performing image recognition by associating the mechanical coordinate system and the camera coordinate system using O _A.

[0033]

(Equation 18)

The image recognition apparatus configured as described above
Based on the origin coordinate O _A1 of the camera coordinate system in the mechanical coordinate system when using the predetermined light receiving lens among the light receiving lenses that receive the imaging light, the origin coordinate in the camera coordinate system when using another light receiving lens O _A2 can be determined. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond to each other without using a special jig based on the image captured by the image capturing means. The coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily.

[0035]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows the appearance of an image recognition apparatus to which the present invention is applied. The image recognition device 1 is for inspecting a semiconductor wafer on which a predetermined device pattern is formed. If a defect is found in a device pattern formed on the semiconductor wafer, what is the defect? Is checked and classified.

As shown in FIG. 1, this image recognition device 1
Has a clean unit 2 for keeping the environment for inspecting semiconductor wafers clean. The clean unit 2 includes a clean box 3 formed by bending a stainless steel plate or the like to form a hollow box, and a clean air unit 4 integrally provided above the clean box 3.

The clean box 3 is provided with a window 3a at a predetermined location so that an inspector can visually recognize the inside of the clean box 3 from the window 3a.

The clean air unit 4 is for supplying clean air into the clean box 3, and has two blowers 5a and 5b respectively arranged at different positions on the upper part of the clean box 3, and these blowers 5a, 5
b and an air filter (not shown) disposed between the clean box 3. The air filter is, for example, a HEPA filter (High Efficiency Particulate).
Air Filter) and ULPA Filter (Ultra Low Penetrat)
ion Air Filter). The clean air unit 4 is provided with blowers 5a, 5b
The dust and the like in the air blown by the above are removed by a high-performance air filter and supplied as clean air into the clean box 3.

The clean box 3 is supported on a floor plate by supporting legs 6, and has a structure in which the lower end is opened. The air supplied from the clean air unit 4 into the clean box 3 is mainly discharged from the lower end of the clean box 3 to the outside of the clean box 3. As will be described in detail later, an opening area is provided at a predetermined location on a side surface of the clean box 3, and air supplied from the clean air unit 4 into the clean box 3 is used for cleaning the clean box 3. The air is also discharged to the outside from an opening region provided on the side surface.

As described above, the clean unit 2 always supplies the clean air from the clean air unit 4 to the clean box 3, and circulates the air circulated in the clean box 3 as an air flow outside the clean box 3. To be discharged. As a result, dust and the like generated in the clean box 3 are discharged to the outside of the clean box 3 together with the air, so that the internal environment of the clean box 3 is maintained at a very high degree of cleanness, for example, about class 1. .

The internal pressure of the clean box 3 is always maintained at a positive pressure in order to prevent air containing dust and the like from entering from the outside.

As shown in FIG. 2, in the image recognition apparatus 1, the apparatus main body 10 is housed inside the clean box 3, and the apparatus main body 1 is stored in the clean box 3.
0 indicates that a semiconductor wafer on which a predetermined device pattern is formed is inspected. Here, the semiconductor wafer to be inspected is transported in a predetermined hermetically sealed container 7 and transferred into the clean box 3 via the container 7. FIG. 2 shows the inside of the clean box 3 viewed from the direction of arrow A1 in FIG.

The container 7 has a bottom 7a, a cassette 7b fixed to the bottom 7a, and a cover 7c detachably engaged with the bottom 7a to cover the cassette 7b. The semiconductor wafers to be inspected are mounted on the cassette 7b such that a plurality of semiconductor wafers are stacked at a predetermined interval, and are sealed by the bottom 7a and the cover 7c.

When inspecting a semiconductor wafer,
First, the container 7 containing the semiconductor wafer is installed in a container installation space 8 provided at a predetermined position of the clean box 3. In the container installation space 8, an upper surface of an elevator 22a of an elevator 22 to be described later is
The container 7 is installed in the container installation space 8 such that the bottom 7a is located on the elevator 22a of the elevator 22.

When the container 7 is set in the container setting space 8, the engagement between the bottom 7a of the container 7 and the cover 7c is released. When the elevator 20a of the elevator 20 is lowered in the direction of arrow B in FIG. 2, the bottom 7a of the container 7 and the cassette 7b are separated from the cover 7c and moved into the clean box 3. Thus, the semiconductor wafer to be inspected is transferred into the clean box 3 without being exposed to the outside air.

When the semiconductor wafer is transferred into the clean box 3, the semiconductor robot to be inspected is taken out of the cassette 7b and inspected by the transfer robot 23 described later.

As shown in FIG. 1, the image recognition apparatus 1 includes an external unit 50 on which a computer for operating the apparatus main body 10 is arranged. This external unit 50 is installed outside the clean box 3. The external unit 50 includes a display device 51 for displaying an image or the like obtained by imaging a semiconductor wafer, a display device 52 for displaying various conditions at the time of inspection, and a device main body 1.
An input device 53 for inputting an instruction to 0 and the like is also provided. And the inspector who inspects the semiconductor wafer,
While viewing the display devices 51 and 52 provided in the external unit 50, necessary instructions are input from the input device 53 provided in the external unit 50 to inspect the semiconductor wafer.

Next, the apparatus main body 10 disposed inside the clean box 3 will be described in detail.

As shown in FIG. 2, the apparatus main body 10 has a support 11. The support base 11 is used to
Is a table for supporting each of the mechanisms. A support leg 12 is attached to the bottom of the support base 11.
And each mechanism provided on the support base 11 supports the support leg 1
2 has a structure supported on a floor plate independently of the clean box 3.

On the support table 11, via a vibration isolation table 13,
An inspection stage 14 on which a semiconductor wafer to be inspected is placed is provided.

The anti-vibration table 13 is for suppressing vibrations from the floor and vibrations generated when the inspection stage 14 is moved, and is provided with a stone surface plate 13a on which the inspection stage 14 is installed. And a plurality of movable legs 13b for supporting the stone surface plate 13a. And this vibration isolation table 13
When the vibration occurs, the vibration is detected and the movable leg 13b is detected.
Is driven to quickly cancel the vibration of the stone base 13a and the inspection stage 14 installed on the stone base 13a.

In this image recognition apparatus 1, since a semiconductor wafer on which a fine device pattern is formed is inspected,
Even a slight vibration may interfere with the inspection. In particular, in the image recognition device 1, since the inspection is performed at a high resolution using ultraviolet light, the influence of vibration is likely to be large. Therefore, in the image recognition apparatus 1, by installing the inspection stage 14 on the vibration isolation table 13, the inspection stage 1
Even if slight vibrations occur in 4, the vibrations are quickly canceled out, the influence of the vibrations is suppressed, and the inspection capability at the time of performing inspection with high resolution using ultraviolet light is improved. I have.

The inspection stage 14 is placed on the vibration isolation table 13.
In order to stably install the vibration isolator, it is desirable that the center of gravity of the vibration isolation table 13 is located at a somewhat lower position. Therefore, in the image recognition device 1, a notch 13c is provided at the lower end of the stone surface plate 13a, and the movable leg 13b is connected to the notch 13c.
In order to support the stone surface plate 13a, the center of gravity of the vibration isolation table 13 is lowered.

It should be noted that vibrations and the like generated when the inspection stage 14 is moved can be predicted to some extent in advance. If the vibration isolation table 13 is operated by predicting such vibrations in advance, it is possible to prevent the vibrations occurring on the inspection stage 14 from occurring. Therefore, it is desirable that the image recognition apparatus 1 operates the vibration damping table 13 by predicting in advance vibrations and the like generated when the inspection stage 14 is moved and operated.

The inspection stage 14 is a stage for supporting a semiconductor wafer to be inspected. The inspection stage 14 has a function of supporting a semiconductor wafer to be inspected and a function of moving the semiconductor wafer to a predetermined inspection target position.

Specifically, the inspection stage 14 includes an X stage 15 installed on the vibration isolation table 13 and an X stage 1
5, a Y stage 16 installed on the Y stage 16, a θ stage 17 installed on the Y stage 16, a Z stage 18 installed on the θ stage 17, and a suction plate 19 installed on the Z stage 18. Have.

The X stage 15 and the Y stage 16 are stages that move in the horizontal direction.
The stage 16 moves semiconductor wafers to be inspected in directions orthogonal to each other, and guides a device pattern to be inspected to a predetermined inspection position.

The .theta. Stage 17 is a so-called rotary stage for rotating a semiconductor wafer.
When inspecting the semiconductor wafer, the semiconductor wafer is rotated by the θ stage 17 so that, for example, the device pattern on the semiconductor wafer is horizontal or vertical to the screen.

The Z stage 18 is a stage that moves in the vertical direction and adjusts the height of the stage. During the inspection of the semiconductor wafer, the Z stage 18 is used so that the inspection surface of the semiconductor wafer has an appropriate height.
Adjust the height of the stage.

The suction plate 19 is for sucking and fixing the semiconductor wafer to be inspected. When inspecting the semiconductor wafer, the semiconductor wafer to be inspected is placed on the suction plate 19 and is sucked by the suction plate 18 to suppress unnecessary movement.

An optical unit 21 supported by a support member 20 is disposed on the vibration isolation table 12 so as to be positioned on the inspection stage 14. The optical unit 21 is for taking an image of a semiconductor wafer when inspecting the semiconductor wafer. The optical unit 21 has a function of capturing an image of a semiconductor wafer to be inspected at low resolution using visible light, and a function of capturing an image of a semiconductor wafer to be inspected at high resolution using ultraviolet light. It also has the function to perform.

Also, as shown in FIGS. 2 and 3, an elevator 22 for taking out a cassette 7b on which a semiconductor wafer to be inspected is mounted from the container 7 and moving the cassette 7b into the clean box 3, as shown in FIGS. Is provided. Further, as shown in FIG. 3, a transfer robot 23 for transferring the semiconductor wafer, and centering and phase setting of the semiconductor wafer before placing the semiconductor wafer on the inspection stage 14 are provided on the support base 11, as shown in FIG. Is provided. FIG. 3 is a plan view schematically showing the apparatus main body 10 viewed from above.

The elevator 22 has an elevating platform 22a that can be raised and lowered. The container 7 is installed in the container installation space 8 of the clean box 3 and the bottom 7 of the container 7
a when the engagement between the cover a and the cover 7c is released.
The lower portion 2a is operated to lower the bottom portion 7a of the container 7.
Then, the cassette 7b fixed thereto is moved into the clean box 3.

The transfer robot 23 has an operation arm 23b provided with a suction mechanism 23a at the distal end thereof. The operation arm 23b is moved to operate the semiconductor arm by the suction mechanism 23a provided at the distal end thereof. The wafer is sucked, and the semiconductor wafer is transported in the clean box 3.

The pre-aligner 24 carries out phase and centering of the semiconductor wafer based on an orientation flat and a notch formed in advance on the semiconductor wafer. The image recognizing apparatus 1 is configured to improve the efficiency of the inspection by performing the phase alignment or the like by the pre-aligner 24 before placing the semiconductor wafer on the inspection stage 14.

When the semiconductor wafer is placed on the inspection stage 14, first, the bottom 7 a of the container 7 and the cassette 7 b are moved into the clean box 3 by the elevator 22. Then, a semiconductor wafer to be inspected is selected from the plurality of semiconductor wafers mounted on the cassette 7b, and the selected semiconductor wafer is taken out of the cassette 7b by the transfer robot 23.

The semiconductor wafer taken out of the cassette 7b is transferred to the pre-aligner 24 by the transfer robot 23. The semiconductor wafer conveyed to the prealigner 24 is subjected to phase and centering by the prealigner 24. Then, the semiconductor wafer on which the phase setting and the center setting are performed is transferred to the inspection stage 14 by the transfer robot 23, and is placed on the suction plate 19 to perform the inspection.

When the semiconductor wafer to be inspected is transported to the inspection stage 14, the semiconductor wafer to be inspected next is taken out of the cassette 7 b by the transport robot 23 and transported to the pre-aligner 24. Then, while the inspection of the semiconductor wafer previously conveyed to the inspection stage 14 is being performed, the phase determination and the centering of the semiconductor wafer to be subsequently inspected are performed. When the inspection of the semiconductor wafer previously transported to the inspection stage 14 is completed, the next semiconductor wafer to be inspected is immediately transported to the inspection stage 14.

As described above, before the semiconductor wafer to be inspected is conveyed to the inspection stage 14, the image recognition apparatus 1 performs the phase adjustment and the center adjustment by the pre-aligner 24, thereby enabling the inspection stage 14 to be inspected. Can reduce the time required for positioning the semiconductor wafer. Further, the image recognition device 1 takes out the semiconductor wafer to be inspected next from the cassette 7 b using the time during which the semiconductor wafer previously transported to the inspection stage 14 is being inspected, By performing centering and centering, the overall time can be reduced, and inspection can be performed efficiently.

By the way, in this image recognition device 1,
As shown in FIG. 3, the elevator 22, the transfer robot 23, and the pre-aligner 24 are installed on the support base 11 so as to be aligned in a straight line. Then, a distance L1 between the elevator 22 and the transport robot 23,
Distance L between transfer robot 23 and pre-aligner 24
The respective installation positions are determined so that the distance 2 is substantially equal to the distance. Further, the arrangement is such that the inspection stage 14 is positioned in a direction substantially orthogonal to the direction in which the elevators 22 and the pre-aligners 24 are arranged when viewed from the transfer robot 23.

The image recognition apparatus 1 can quickly and accurately transfer a semiconductor wafer as an object to be inspected because the mechanisms are arranged as described above.

That is, in the image recognition apparatus 1, the distance L1 between the elevator 22 and the transport robot 23 is
Distance L between transfer robot 23 and pre-aligner 24
Since the distance is substantially equal to 2, the semiconductor wafer taken out of the cassette 7b can be transferred to the pre-aligner 24 without changing the length of the arm 23b of the transfer robot 23. Therefore, this image recognition device 1
In this case, since an error or the like generated when the length of the arm 23b of the transfer robot 23 is changed does not matter, the operation of transferring the semiconductor wafer to the pre-aligner 24 can be performed accurately. In addition, since the elevator 22, the transfer robot 23, and the pre-aligner 24 are arranged in a straight line,
The transfer robot 23 moves the semiconductor wafers taken out of the cassette 7b by the pre-aligner 24 only by linear movement.
Can be transported. Therefore, in the image recognition device 1, the operation of transporting the semiconductor wafer to the pre-aligner 24 can be performed extremely accurately and quickly.

Further, in the image recognition apparatus 1, when viewed from the transfer robot 23, the inspection stage 1 is arranged in a direction substantially orthogonal to the direction in which the elevators 22 and the pre-aligners 24 are arranged.
The semiconductor wafer can be transferred to the inspection stage 14 by the linear movement of the transfer robot 23 because the transfer robot 4 is located. Therefore, in the image recognition device 1, the operation of transporting the semiconductor wafer to the inspection stage 14 can be performed extremely accurately and quickly. In particular, in the image recognition apparatus 1, since the semiconductor wafer on which the fine device pattern is formed is inspected, it is necessary to transport and position the semiconductor wafer to be inspected extremely accurately. The arrangement is very effective.

Next, the image recognition apparatus 1 will be described with reference to FIG.
This will be described in more detail with reference to the block diagram of FIG.

As shown in FIG. 4, the external unit 50 of the image recognition device 1 includes an image processing computer 60 to which a display device 51 and an input device 53a are connected, and a display device 5
2 and a control computer 61 to which the input device 53b is connected. In FIG. 1 described above, the input device 53a connected to the image processing computer 60 is used.
And an input device 53 connected to the control computer 61
b are collectively shown as an input device 53.

When inspecting a semiconductor wafer, the image processing computer 60 operates to charge the CCD (charge-coupled device) camera 30 installed inside the optical unit 21.
31 is a computer that takes in and processes an image of a semiconductor wafer captured by the computer 31. That is, the image recognition device 1 converts the image of the semiconductor wafer captured by the CCD cameras 30 and 31 installed inside the optical unit 21
The semiconductor wafer is inspected by processing and analyzing by the image processing computer 60.

The input device 53a connected to the image processing computer 60 is for inputting, to the image processing computer 30, instructions necessary for analyzing images captured from the CCD cameras 30 and 31. And comprises, for example, a pointing device such as a mouse, a keyboard, and the like. The display device 51 connected to the image processing computer 60 is for displaying analysis results of images taken from the CCD cameras 30 and 31, and is composed of, for example, a CRT display or a liquid crystal display.

When inspecting a semiconductor wafer, the control computer 61 controls the inspection stage 14, the elevator 2
2. A computer for controlling the transfer robot 23, the pre-aligner 24, and each device inside the optical unit 12. That is, this image recognition device 1
When inspecting a semiconductor wafer, an image of the semiconductor wafer to be inspected is stored in a CC provided inside the optical unit 21.
The control computer 61 controls the inspection stage 14, the elevator 22, the transport robot 23 and the pre-aligner 24, and each device inside the optical unit 21 so as to be imaged by the D cameras 30 and 31.

The control computer 61 has a function of controlling the blowers 5a and 5b of the clean air unit 4. That is, in the image recognition device 1, the control computer 61 controls the blowers 5 a and 5 b of the clean air unit 4, so that when inspecting a semiconductor wafer, clean air is always supplied into the clean box 3. In addition, the air flow in the clean box 3 can be controlled.

The input device 53 b connected to the control computer 61 includes the inspection stage 14, the elevator 22, the transfer robot 23 and the pre-aligner 24, each device inside the optical unit 21, and the blower of the clean air unit 4. This is for inputting an instruction necessary for controlling 5a, 5b and the like to the control computer 61, and includes, for example, a pointing device such as a mouse or a keyboard. The control computer 6
1 is for displaying various conditions and the like at the time of inspection of a semiconductor wafer.
It consists of an RT display, a liquid crystal display and the like.

The image processing computer 60 and the control computer 61 can exchange data with each other by a memory link mechanism. That is,
Image processing computer 60 and control computer 61
Are connected to each other via memory link interfaces 60a and 61a provided respectively, so that the image processing computer 60 and the control computer 51 can exchange data with each other.

On the other hand, inside the clean box 3 of the image recognition apparatus 1, semiconductor wafers which have been transported in a sealed container 7 are taken out from the cassette 7b of the container 7 and set on the inspection stage 14. As described above, the elevator 22, the transfer robot 23, and the pre-aligner 24 are arranged as described above. These are connected to a control computer 61 arranged in the external unit 50 via a robot control interface 61b. Then, control signals are sent from the control computer 61 to the elevator 22, the transfer robot 23, and the pre-aligner 24 via the robot control interface 61b.

That is, the semiconductor wafers conveyed in the sealed container 7 are transferred to the cassette 7 of the container 7.
b, when it is taken out and installed on the inspection stage 14,
Control signals are sent from the control computer 61 to the elevator 22, the transfer robot 23, and the pre-aligner 24 via the robot control interface 61b. Then, the elevator 22, the transfer robot 23, and the pre-aligner 24 operate based on the control signal, and as described above, the semiconductor wafers that have been transported in the closed container 7 are transferred to the cassette 7b of the container 7. , And the phase and center are set by the pre-aligner 25 and set on the inspection stage 14.

Further, an anti-vibration table 13 is disposed inside the clean box 3 of the image recognition apparatus 1.
As described above, the inspection stage 14 provided with the X stage 15, the Y stage 16, the θ stage 17, the Z stage 18, and the suction plate 19 is provided on 3.

Here, the X stage 15 and the Y stage 1
6, the θ stage 17, the Z stage 18, and the suction plate 19 are connected to a control computer 61 arranged in the external unit 50 via a stage control interface 61c. Then, control signals are sent from the control computer 61 to the X stage 15, the Y stage 16, the θ stage 17, the Z stage 18, and the suction plate 19 via the stage control interface 61c.

That is, when inspecting a semiconductor wafer, control signals are sent from the control computer 61 to the X stage 15, the Y stage 16, the θ stage 17, the Z stage 18, and the suction plate 19 via the stage control interface 61c. Sent out. And X stage 1
5, Y stage 16, θ stage 17, Z stage 18
The suction plate 19 operates based on the control signal, and the suction plate 19 sucks and fixes the semiconductor wafer to be inspected, and the X stage 15, the Y stage 16, the θ stage 17 and the Z stage 18 The wafer is moved to a predetermined position, angle and height.

The optical unit 21 is also provided on the vibration isolation table 12 as described above. The optical unit 21 is for capturing an image of the semiconductor wafer during the inspection of the semiconductor wafer, and as described above, a function of capturing an image of the semiconductor wafer to be inspected at a low resolution using visible light. And a function of capturing an image of a semiconductor wafer to be inspected with high resolution using ultraviolet light.

Inside the optical unit 21, a mechanism for capturing an image of a semiconductor wafer with visible light is provided.
A CCD camera 30 for visible light, a halogen lamp 32,
An optical system 33 for visible light, an objective lens 34 for visible light, and an autofocus controller 35 for visible light are arranged.

When the image of the semiconductor wafer is taken with visible light, the halogen lamp 32 is turned on. Here, the driving source of the halogen lamp 32 is the external unit 50.
Is connected via a light source control interface 61d to a control computer 61 arranged in the computer. The driving source of the halogen lamp 32 includes the control computer 6.
1 transmits a control signal via the light source control interface 61d. The turning on / off of the halogen lamp 32 is performed based on this control signal.

When taking an image of a semiconductor wafer with visible light, the halogen lamp 32 is turned on, and the visible light from the halogen lamp 32 is transmitted to the visible light optical system 33.
Then, the semiconductor wafer is illuminated by being applied to the semiconductor wafer via the visible light objective lens 34. Then, the image of the semiconductor wafer illuminated by the visible light is converted into a visible light objective lens 34.
And enlarges the enlarged image with the visible light CCD camera 30.
To capture an image.

Here, the visible light CCD camera 30 is connected to an image processing computer 60 provided in the external unit 50.
Are connected via an image capture interface 60b. The image of the semiconductor wafer captured by the visible light CCD camera 30 is captured by the image processing computer 60 via the image capturing interface 60b.

When an image of a semiconductor wafer is captured with visible light as described above, the visible light autofocus controller 35 performs automatic focus position adjustment. That is, the visible light autofocus controller 35 detects whether or not the distance between the visible light objective lens 34 and the semiconductor wafer matches the focal length of the visible light objective lens 34. Is a visible light objective lens 34
Alternatively, the Z stage 18 is moved so that the inspection target surface of the semiconductor wafer coincides with the focal plane of the objective lens 34 for visible light.

Here, the visible light auto-focus control unit 35 is connected to an auto-focus control interface 61 by a control computer 61 arranged in the external unit 50.
e. Then, a control signal is sent from the control computer 61 to the visible light autofocus control unit 35 via the autofocus control interface 61e. Automatic focus positioning of the visible light objective lens 34 by the visible light autofocus controller 35 is performed based on this control signal.

The optical unit 21 includes a CCD camera 31 for ultraviolet light and an ultraviolet laser light source 3 as a mechanism for capturing an image of a semiconductor wafer with ultraviolet light.
6, an optical system 37 for ultraviolet light, and an objective lens 38 for ultraviolet light
And an ultraviolet light autofocus control section 39.

When an image of the semiconductor wafer is taken with ultraviolet light, the ultraviolet laser light source 36 is turned on. Here, the drive source of the ultraviolet laser light source 36 is connected to a control computer 61 arranged in the external unit 50 via a light source control interface 61d. A control signal is sent from the control computer 61 to the drive source of the ultraviolet laser light source 36 via the light source control interface 61d. Turning on / off the ultraviolet laser light source 36
The light is turned off based on this control signal.

It is preferable that the ultraviolet laser light source 36 emits an ultraviolet laser having a wavelength of about 266 nm. An ultraviolet light laser having a wavelength of about 266 nm is obtained as the fourth harmonic of the YAG laser. Also,
A laser light source having an oscillation wavelength of about 166 nm has also been developed, and such a laser light source may be used as the ultraviolet laser light source 36.

When an image of a semiconductor wafer is picked up by ultraviolet light, the ultraviolet laser light source 36 is turned on, and the ultraviolet light from the ultraviolet laser light source 36 is transmitted to the ultraviolet optical system 37 and the objective lens for ultraviolet light. The semiconductor wafer is illuminated by being applied to the semiconductor wafer via 38. Then, the image of the semiconductor wafer illuminated by the ultraviolet light is enlarged by the ultraviolet light objective lens 38, and the enlarged image is captured by the ultraviolet light CCD camera 31.

Here, the ultraviolet light CCD camera 31 is connected to an image processing computer 60 provided in the external unit 50.
Are connected via an image capture interface 60c. Then, the image of the semiconductor wafer taken by the ultraviolet CCD camera 31 is taken into the image processing computer 60 via the image taking interface 60c.

When the image of the semiconductor wafer is picked up by the ultraviolet light as described above, the automatic focus control is performed by the ultraviolet light autofocus control section 39. In other words, the autofocus controller 39 for ultraviolet light detects whether or not the distance between the ultraviolet light objective lens 38 and the semiconductor wafer matches the focal length of the ultraviolet light objective lens 38. Is the objective lens 38 for ultraviolet light.
Alternatively, the Z stage 18 is moved so that the inspection target surface of the semiconductor wafer coincides with the focal plane of the ultraviolet light objective lens 38.

Here, the ultraviolet light autofocus control section 39 sends an autofocus control interface 61 to a control computer 61 arranged in the external unit 50.
e. Then, a control signal is sent from the control computer 61 to the ultraviolet light autofocus control section 39 via the autofocus control interface 61e. Automatic focusing of the ultraviolet light objective lens 38 by the ultraviolet light autofocus control unit 39 is performed based on this control signal.

Further, the clean air unit 4 is provided with two blowers 5a and 5b as described above. These blowers 5a and 5b are connected to a control computer 61 arranged in the external unit 50 via an air volume control interface 61f. A control signal is sent from the control computer 61 to the blowers 5a and 5b of the clean air unit 4 via the air volume control interface 61f. Control of the number of rotations of the blowers 5a and 5b, switching on / off, and the like are performed based on this control signal.

Next, the optical system of the optical unit 21 of the image recognition device 1 will be described in more detail with reference to FIG. Here, the auto focus control unit 35,
The description of 39 will be omitted, and an optical system for illuminating the semiconductor wafer to be inspected and an optical system for imaging the semiconductor wafer to be inspected will be described.

As shown in FIG. 5, the optical unit 21
As an optical system for capturing an image of a semiconductor wafer with visible light, a halogen lamp 32 and an optical system 33 for visible light are used.
And a visible light objective lens 34.

The visible light from the halogen lamp 32 is guided to the visible light optical system 33 by the optical fiber 40.
The visible light guided to the visible light optical system 33 first passes through the two lenses 41 and 42 and enters the half mirror 43. Then, the visible light incident on the half mirror 43 is
The light is reflected by the half mirror 43 toward the visible light objective lens 34 and enters the semiconductor wafer via the visible light objective lens 34. Thereby, the semiconductor wafer is illuminated by the visible light.

Then, the image of the semiconductor wafer illuminated by the visible light is enlarged by the visible light objective lens 34,
The light passes through the half mirror 43 and the imaging lens 44 and is imaged by the visible light CCD camera 30. That is,
The reflected light from the semiconductor wafer illuminated by visible light
The light enters the visible light CCD camera 30 via the visible light objective lens 34, the half mirror 43, and the imaging lens 44.
The image is taken by the D camera 30. And C for visible light
An image of the semiconductor wafer (hereinafter, referred to as a visible image) captured by the CD camera 30 is sent to the image processing computer 60.

The optical unit 21 includes an ultraviolet laser light source 36, an ultraviolet light optical system 37, and an ultraviolet light objective lens 38 as an optical system for capturing an image of a semiconductor wafer with ultraviolet light. Have.

The ultraviolet light from the ultraviolet laser light source 36 is guided by an optical fiber 45 to an optical system 37 for ultraviolet light.
The ultraviolet light guided to the ultraviolet light optical system 37 first passes through the two lenses 46 and 47 and enters the half mirror 48. Then, the visible light incident on the half mirror 48 is
The light is reflected by the half mirror 48 toward the ultraviolet light objective lens 38 and is incident on the semiconductor wafer via the ultraviolet light objective lens 38. Thereby, the semiconductor wafer is illuminated by the ultraviolet light.

Then, the image of the semiconductor wafer illuminated by the ultraviolet light is enlarged by the ultraviolet light objective lens 38,
The light passes through the half mirror 48 and the imaging lens 49 and is imaged by the ultraviolet light CCD camera 31. That is,
The reflected light from the semiconductor wafer illuminated by ultraviolet light,
The light enters the ultraviolet CCD camera 31 via the ultraviolet objective lens 38, the half mirror 48, and the imaging lens 49.
The image is captured by the D camera 31. And UV light C
An image of the semiconductor wafer (hereinafter, referred to as an ultraviolet image) captured by the CD camera 31 is sent to the image processing computer 60.

In the image recognition apparatus 1 as described above, since an image of a semiconductor wafer can be imaged and inspected by ultraviolet light having a wavelength shorter than that of visible light, defect detection can be performed using visible light. It is possible to detect and classify finer defects as compared with the case of performing classification and classification.

In addition, the image recognition apparatus 1 has both an optical system for visible light and an optical system for ultraviolet light, so that a semiconductor wafer can be inspected at a low resolution using visible light and an ultraviolet light can be used. Inspection of a semiconductor wafer at a high resolution. Therefore, the image recognition device 1 detects and classifies large defects by inspecting the semiconductor wafer at low resolution using visible light, and performs inspection of the semiconductor wafer at high resolution using ultraviolet light. It is also possible to detect and classify small defects.

In the image recognition apparatus 1, the numerical aperture NA of the ultraviolet light objective lens 40 is preferably large, for example, 0.9 or more. As described above, by using a lens having a large numerical aperture NA as the ultraviolet light objective lens 40, it is possible to detect a finer defect.

When a defect in a semiconductor wafer consists of only irregularities without color information, such as a scratch, it is almost impossible to see the defect with light having no coherence. On the other hand, in the case of using light having excellent coherence such as laser light, even if a defect has no color information and consists only of irregularities, the light interferes near the steps of the irregularities. The defect is clearly visible. The image recognition apparatus 1 uses an ultraviolet laser light source 36 that emits ultraviolet laser light as a light source of ultraviolet light. Therefore, the image recognition device 1 can clearly detect a defect including only irregularities without color information. That is, in the image recognition device 1, the phase information that is difficult to detect with the visible light (incoherent light) from the halogen lamp 32 can be easily used by using the ultraviolet laser (coherent light) from the ultraviolet laser light source 36. Can be detected.

Next, an example of a procedure for inspecting a semiconductor wafer by the image recognition apparatus 1 will be described with reference to a flowchart of FIG. Note that the flowchart of FIG. 6 shows a procedure of a process after a state where the semiconductor wafer to be inspected is set on the inspection stage 14. In addition, the flowchart shown in FIG. 6 shows an example of a procedure for inspecting and classifying a defect on the semiconductor wafer by the image recognition device 1 when the position of the defect on the semiconductor wafer is known in advance. Here, it is assumed that a number of similar device patterns are formed on a semiconductor wafer, and the detection and classification of defects are performed by using an image of a defective area (a defect image) and an image of another area (a reference image). ) And image,
It shall be performed by comparing them.

First, as shown in step S1-1, a defect position coordinate file is read into the control computer 61. Here, the defect position coordinate file is a file in which information relating to the position of the defect on the semiconductor wafer is described, and is created by measuring the position of the defect on the semiconductor wafer in advance by a defect detection device or the like. Then, here, the defect position coordinate file is read into the control computer 61.

Next, in step S1-2, the X stage 15 and the Y stage 16 are driven by the control computer 61, and the semiconductor wafer is moved to the defect position coordinates indicated by the defect position coordinate file. In the field of view of the objective lens 34 for visible light.

Next, in step S1-3, the visible light autofocus controller 35 is driven by the control computer 61 to perform automatic focus positioning of the visible light objective lens.

Next, in step S1-4, an image of the semiconductor wafer is captured by the visible light CCD camera 30, and the captured visible image is sent to the image processing computer 60. The visible image captured here is an image at the defect position coordinates indicated by the defect position coordinate file, that is, an image of a region where a defect is present (hereinafter, referred to as a defect image).

Next, in step S1-5, the X stage 15 and the Y stage 16 are driven by the control computer 61 to move the semiconductor wafer to the reference position coordinates. 3
4 so that it is within the field of view. Here, the reference region is a region other than the inspection target region of the semiconductor wafer, and is a region where a device pattern similar to the device pattern in the inspection target region of the semiconductor wafer is formed.

Next, in step S1-6, the visible light autofocus controller 35 is driven by the control computer 61 to perform automatic focus positioning of the visible light objective lens.

Next, in step S1-7, an image of the semiconductor wafer is captured by the visible light CCD camera 30, and the captured visible image is sent to the image processing computer 60. The visible image captured here is an image of a region where a device pattern similar to the device pattern in the inspection target region of the semiconductor wafer is formed (hereinafter, referred to as a reference image).

Next, in step S1-8, the image processing computer 60 compares the defect image captured in step S1-4 with the reference image captured in step S1-7, and detects a defect from the defect image. I do. If a defect is detected, the process proceeds to step S1-9. If a defect is not detected, the process proceeds to step S1-9.
Proceed to 11.

In step S1-9, the image processing computer 60 checks what the detected defect is and classifies it. If the defect can be classified, the process proceeds to step S1-10. If the defect cannot be classified, the process proceeds to step S1-11.

In step S1-10, the defect classification result is stored. Here, the defect classification result is stored in, for example, a storage device connected to the image processing computer 60 or the control computer 61. Note that the defect classification result may be transferred to another computer connected to the image processing computer 60 or the control computer 61 via a network, and may be stored.

When the processing in step S1-10 is completed, the classification of the defects on the semiconductor wafer has been completed, and the processing is completed. However, if there are a plurality of defects on the semiconductor wafer, the process may return to step S1-2 to detect and classify other defects.

On the other hand, if the defect cannot be detected in step S1-8, or if the defect cannot be classified in step S1-9, the process proceeds to step S1-11 and the subsequent steps. Defect detection and classification are performed by imaging at a resolution.

In this case, first, in step S1-11, the X stage 15 is controlled by the control computer 61.
Then, the Y stage 16 is driven to move the semiconductor wafer to the defect position coordinates indicated by the defect position coordinate file so that the inspection target area of the semiconductor wafer falls within the visual field of the ultraviolet light objective lens 38.

Next, in step S1-12, the control computer 61 drives the ultraviolet auto focus control section 39 to perform automatic focus positioning of the ultraviolet light objective lens.

Next, in step S1-13, an image of the semiconductor wafer is taken by the ultraviolet CCD camera 31, and the taken ultraviolet image is sent to the image processing computer 60. The ultraviolet image captured here is an image at the defect position coordinates indicated by the defect position coordinate file, that is, a defect image. In addition, imaging of the defect image here is performed with higher resolution than imaging using visible light, using ultraviolet light that is light having a shorter wavelength than visible light.

Next, in step S1-14, the X stage 15 and the Y stage 16 are driven by the control computer 61 to move the semiconductor wafer to the reference position coordinates. 38. Here, the reference area is
A region other than the inspection target region of the semiconductor wafer, in which a device pattern similar to the device pattern in the inspection target region of the semiconductor wafer is formed.

Next, in step S1-15, the control computer 61 drives the ultraviolet light autofocus control section 39 to perform automatic focusing of the ultraviolet light objective lens.

Next, in step S1-16, an image of the semiconductor wafer is captured by the ultraviolet CCD camera 31, and the captured ultraviolet image is sent to the image processing computer 60. The ultraviolet image captured here is an image of a region where a device pattern similar to the device pattern in the inspection target region of the semiconductor wafer is formed, that is, a reference image. The imaging of the reference image here is performed at a higher resolution than when visible light is used by using ultraviolet light that is light having a shorter wavelength than visible light.

Next, in step S1-17, the image processing computer 60 compares the defect image captured in step S1-13 with the reference image captured in step S1-16, and detects a defect from the defect image. I do.
If a defect can be detected, step S1-1
Proceeding to step S8, if no defect is detected, proceeding to step S1-19.

In step S1-18, the image processing computer 60 checks what the detected defect is and classifies it. If the defect classification has been completed, the process proceeds to step S1-10, and the defect classification result is stored as described above. On the other hand, if the defect cannot be classified, the process proceeds to step S1-19.

In step S1-19, information indicating that the defect cannot be classified is stored. Here, information indicating that the defect cannot be classified is stored, for example, in a storage device connected to the image processing computer 60 or the control computer 61. Note that this information may be transferred to another computer connected to the image processing computer 60 or the control computer 61 via a network and stored.

According to the above-described procedure, first, the image picked up by the visible light CCD camera 30 is processed and analyzed to inspect the semiconductor wafer at a low resolution, and to detect a defect by visible light and When classification was not possible,
Next, the semiconductor wafer is inspected with high resolution by processing and analyzing the image captured by the ultraviolet CCD camera 31.

Here, a method of detecting a defect from a reference image and a defect image picked up by the CCD cameras 30 and 31 will be described with reference to FIG.

FIG. 7A shows an example of an image of a reference area where a device pattern similar to the device pattern in the inspection target area is formed, that is, an example of the reference image. FIG. 7B shows an example of an image of a region to be inspected which is assumed to have a defect, that is, an example of a defect image.

When detecting a defect from such a reference image and a defect image, a device pattern is extracted from the reference image as shown in FIG. 7C based on color information, density information, and the like. Further, a difference image is obtained from the reference image and the defect image, and a portion having a large difference is extracted as a defect as shown in FIG.

Then, as shown in FIG.
An image obtained by superimposing the image of the device pattern extraction result shown in FIG. 7C and the image of the defect extraction result shown in FIG. 7D is obtained, and the rate at which the defect exists in the device pattern is determined. It is extracted as a feature value.

According to the above method, the CCD camera 3
The image processing computer 60 processes and analyzes the reference image and the defect image captured by 0 and 31 to detect a defect and inspect the semiconductor wafer.

As described above, the image recognition device 1 first inspects the semiconductor wafer at a low resolution by processing and analyzing the image picked up by the visible light CCD camera 30, and performs inspection with the visible light. If a defect cannot be detected or classified, the semiconductor wafer is inspected at a high resolution by processing and analyzing the image captured by the ultraviolet CCD camera 31. , Compared to detecting and classifying defects using only visible light,
Detection and classification of finer defects can be performed.

However, imaging at a low resolution using visible light has a wider area that can be imaged at a time, so if a defect is sufficiently large, inspection of a semiconductor wafer at a low resolution using visible light is possible. Is more efficient. Therefore,
Rather than using ultraviolet light to inspect and classify defects from the beginning, as described above, inspecting and classifying defects using visible light first enables more efficient semiconductor wafers. Inspection can be performed.

In the image recognition apparatus 1, the device pattern of the semiconductor wafer to be inspected is extremely fine, for example, about 0.18 μm. It is important to make the position correspond to the actual position of this defect on the semiconductor wafer with high accuracy.

Therefore, in the image recognition apparatus 1, as described above, the inspection stage 14 and the optical unit 21 are calibrated as described below before performing the defect inspection and classification. The coordinate system of the provided support 11 (hereinafter, referred to as mechanical coordinate system).
And a coordinate system (hereinafter, referred to as a camera coordinate system) in an image captured by the optical unit 21.

In the following description, a coordinate system is defined as shown in FIG. That is, as shown in FIG. 8, the mechanical coordinate origin coordinates (X _M -Y _M) and O _M, the origin coordinate of the coordinate system of the inspection stage 14 in the mechanical coordinate system (X _S -Y _S) , O _S at the rotation center of the θ stage 17. In the algorithm of the calibration to be described below, it is not necessary to point O _S is necessarily coincide with the center of rotation of the θ stage 17, if the obvious fixed point position from the center of rotation of the θ stage 17 Good. The origin coordinate of the semiconductor wafer fixed and supported on the inspection stage 14 is defined as O _W.
It should be noted that, this point O _W does not have to match the origin O _S of the inspection stage 14. The camera coordinate system in the mechanical coordinate system the origin coordinates (X _A -Y _A) to O _A.

[0147] At this time, the P of coordinates arbitrary point on a semiconductor wafer in a mechanical coordinate system and P _M, the coordinates in the camera coordinate system of the point P and P _A, linear transformation from the mechanical coordinate system to the camera coordinate system If the matrix is A, then P
Relationship between _A and P _M can be expressed by Equation 10 below.

[0148]

[Equation 19]

The primary transformation matrix A includes transformation of the inclination and orthogonality between the mechanical coordinate system and the camera coordinate system. The origin coordinate O _A of the linear transformation matrix A and the camera coordinate system, the light receiving lens in the optical unit 21, i.e., the visible light objective lens 34 and the ultraviolet objective lens 3
8 is a unique matrix. Therefore, in the calibration described below, these primary transformation matrices A
And means to determine the origin coordinate O _A of the camera coordinate system.

Hereinafter, first, a case will be described in which the image recognition device 1 performs calibration on an image captured using the visible light objective lens 34.

First, the operation of the image conversion apparatus 1 in the case of obtaining the first-order transformation matrix A in the above equation (10) will be described below.

First, an arbitrary point P is selected on the semiconductor wafer fixed and supported on the inspection stage 14, and X is set so that this point P is within the visual field of the visible light objective lens 34.
The stage 15 and the Y stage 16 are moved. As the point P, for example, as shown in FIG. 9, an alignment mark formed on a semiconductor wafer may be identified by the image processing computer 60, and one point on the alignment mark may be set as a point P. In FIG. 9,
This shows a case where the alignment mark is formed as a character "K".

Next, within the visual field of the objective lens for visible light 34, the position of the point P is translated by the X stage 15 and the Y stage 16. Then, the image processing computer 60 obtains the coordinates of the three points P in the mechanical coordinate system and the camera coordinate system, respectively. At this time, the coordinates of the three points in the mechanical coordinate system are represented by _PM1 , _PM2 , _PM3 ,
The coordinates of the three points in the camera coordinate system are defined as P _A1 , P _A2 , and P _A3 .

Then, the following Expression 11 to Expression 13 can be expressed by using Expression 10 described above.

[0155]

(Equation 20)

Then, when Equation 11 is divided from Equations 12 and 13, respectively, the following Equations 14 and 15 are obtained.

[0157]

(Equation 21)

When these equations 14 and 15 are modified, the following equations 16 and 17 are obtained.

[0159]

(Equation 22)

Here, the definition is made as follows.

[0161]

(Equation 23)

When the primary transformation matrix A is defined and solved as described above, the following equation 1 is obtained. Thus, the primary transformation matrix A can be obtained from the coordinates of three points.

[0163]

(Equation 24)

It is assumed that the following equation 18 is used instead of the above equation 15.

[0165]

(Equation 25)

In this case, the following definition is made.
Solving for the primary transformation matrix A yields the same result as Equation 1.

[0167]

(Equation 26)

[0168] Next origin coordinate O _A of the camera coordinate system, the case of obtaining the origin coordinate O _A of the camera coordinate system in Equation 10 described above, the operation of the image recognition apparatus 1 will be described. The method of obtaining the origin coordinate O _A, there are two methods.

[0169] The first method can be center coordinates O _S of the inspection stage 14 in the mechanical coordinate system, it applied to a case that can be identified by a jig provided in the inspection stage 14. In this case, first, so that the center coordinates O _S of the inspection stage 14 is within the field of view of the visible light objective lens 34, X stage 15 and Y stage 1
Move 6. Then, the center coordinates of the inspection stage 14, acquired by the camera coordinate system, the coordinate and P _A.
Then, the following Expression 2 can be expressed using Expression 10 described above.

[0170]

[Equation 27]

[0171] Then, the image processing computer 60, by solving Equation 2, it is possible to determine the origin coordinate O _A of the camera coordinate system in the mechanical coordinate system.

The second method can be applied when the inspection stage 14 is not provided with a jig indicating the central coordinates in the mechanical coordinate system. In this case,
The point P is moved by the θ stage 17 to the objective lens 3 for visible light.
Rotate and move in the field of view of 4. At this time, the image processing computer 60 calculates the coordinates of the point P before and after the rotational movement in the mechanical coordinate system and the camera coordinate system, respectively.
_M1 and P _M2 and P _A1 and P _A2 are acquired. FIG.
As shown in, the position vector of the point P from the origin coordinate O _S of the inspection stage 14 in the mechanical coordinate system, before and after the rotational movement, respectively, and P _P1, P _P2. At this time, a relationship such as the following Expression 20 is established between P _P1 and P _P2 .

[0173]

[Equation 28]

Therefore, P _M1 and P _M2 are calculated by the above equation (10).
Expression 20 and Expression 20 can be expressed as Expressions 21 and 22 below.

[0175]

(Equation 29)

Next, when P _P1 is deleted from Equations 21 and 22, and O _A is solved, the following Equation 3 is obtained.

[0177]

[Equation 30]

[0178] Thus, it can be determined by the origin coordinate O _A of the camera coordinate system to obtain points two points on the semiconductor wafer in a mechanical coordinate system. That is, the image processing computer 60, by solving Equation 3, it is possible to determine the origin coordinate O _A.

By the way, as described above, the origin coordinate O
_{When obtaining A} , since the point P is rotated in the image, it is difficult to increase the rotation angle θ. However, the accuracy of the origin coordinate O _A is highly dependent on the rotational angle theta. Therefore, in order to determine the origin coordinate O _A with high accuracy, it is effective to increase the rotational movement of the position of the point P to the outside of the image. Thus, when rotating moving the position of the point P to the outside of the image, the method described below, can be determined origin coordinate O _A.

In this case, after the point P is rotated by the θ stage 17 out of the range of the visual field of the visible light objective lens 34, the point P moved out of the range of the image is again moved into the range of the image. The X stage 15 and the Y stage 16 are moved in parallel so that

At this time, the image processing computer 60
The coordinates of the point P before and after the movement, respectively mechanical coordinate system and the camera coordinate system, P _M1, P _M2 and P _A1,
To get as P _A2. Further, as shown in FIG. 11, the origin coordinates O of the inspection stage 14 in the mechanical coordinate system.
The position vector of the point P from _S is calculated before and after the movement.
Let them be P _P1 and P _P2 respectively. Further, the coordinates of the rotation center of the inspection stage in the mechanical coordinate system after the movement are represented by O
Get as _S2 .

Then, the above equation (22) is replaced by the following equation (23).
Can be expressed as

[0183]

(Equation 31)

Therefore, from the above equations 21 and 23, P
Clear the _P1, it is solved for O _A, so that the equation 4 below.

[0185]

(Equation 32)

[0186] Then, the image processing computer 60, by solving this equation 4, it is possible to determine the origin coordinate O _A. Therefore, large rotates moving the position of the point P to the outside of the image, it is possible to determine the origin coordinate O _A high precision.

As described above, the linear transformation matrix A of the movement amount of the point P from the mechanical coordinate system to the camera coordinate system is obtained, and the mechanical coordinate system and the camera coordinate system are correlated by using the above-described equation (10). By doing so, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

Next, a method for performing calibration between a plurality of lenses in the image recognition apparatus 1 will be described.

The image recognition device 1 includes an objective lens 3 for visible light.
4 and an ultraviolet light objective lens 38. Also,
Objective lenses having different magnifications may be interchangeably disposed on each lens. When performing the calibration in the image recognition apparatus 1, since the linear transformation matrix A and the origin coordinates O _A of the camera coordinate system is a unique matrix for each lens in the optical unit 21,
Respect plurality each lens was, respectively, it is necessary to determine the linear transformation matrix A and the origin coordinate O _A.

[0190] At this time, by performing the operations described above for each lens, it may be obtained linear transformation matrix A and the origin coordinate O _A, but the method described below, have already obtained the predetermined lens ( hereinafter, a linear transformation matrix in that.) the first lens a _1, the origin coordinates as O _A1, on the basis of these linear transformation matrix a ₁ and origin coordinate O _A1, another lens (hereinafter referred to as a second lens .) of may determine the linear transformation matrix a ₂ and the origin coordinate O _A2. As a result, the time required for calibration can be reduced, and the accuracy of calibration can be improved.

Correction of Primary Transformation Matrix between Lenses For example, when only the objective lens to be used is switched using the same imaging camera, the magnification accuracy of the lenses is relatively low at about 5%. The components other than the magnification in, that is, the rotational direction component and the orthogonality are all equal. Therefore, the above-described linear transformation matrices A ₁ and A ₂ have a relationship represented by the following Expression 24.

[0192]

[Equation 33]

However, in the equation (24), k is a constant.

Then, first, in the optical unit 21, the point P is identified by using the first lens,
The linear transformation matrix A ₁ in the first lens, determined by the method described above.

Next, the lens used in the optical unit 21 is replaced with a second lens. Then, by moving the coordinates of the point P in parallel within the range of the image captured by using the second lens, the coordinates of the point P are changed to two points, and the image processing computer 60 uses the mechanical coordinate system and the camera. Acquire with coordinate system. At this time, the coordinates of two points in the mechanical coordinate system are P _M1 and P _M2 , and the coordinates of the two points in the camera coordinate system are P _A1 and P _A2, and are defined as follows.

[0196]

(Equation 34)

Then, from Equation 16 described above, the primary transformation matrix A ₁ can be expressed as in Equation 25 below.

[0198]

(Equation 35)

Therefore, by solving equation (25) to find the constant k and using equation (24), the primary conversion matrix A ₂ can be obtained based on the primary conversion matrix A ₁ .

The image recognizing device 1 obtains the primary transformation matrix A ₂ by the image processing computer 60 by the above-described method, and based on the primary transformation matrix A ₁ of a predetermined lens, obtains the primary transformation matrix A ₂ of another lens. in the transformation matrix a ₂ only to obtain two points the coordinates of the point P, it can be determined, the time of shortening required for calibration, it is possible to improve the calibration accuracy.

Correction of the origin coordinates between the lenses Next, the origin coordinates of the first lens, which have already been obtained, are set to O _A1 between a plurality of lenses, and the origin coordinates O
A case in which the origin coordinate _OA2 of the second lens is obtained based on _A1 will be described.

[0202] In the image recognition apparatus 1, when the accuracy of the origin coordinate O _A in each lens is not very high, when obtaining the origin coordinate O _A for each lens, when switching the lens, in captured image There is a possibility that the position on the semiconductor wafer is significantly shifted. Therefore, based on the origin coordinate O _A1 of the first lens that has already determined, it is desirable to determine the origin coordinate O _A2 in the second lens. Accordingly, the relative position accuracy of the origin coordinates between the lenses does not depend on the absolute position accuracy of the origin coordinates of each lens. In the image recognition device 1,
Since the position of the semiconductor wafer mounted on the inspection stage 14 always includes an error, the relative position accuracy of the origin coordinates between the lenses is better than the absolute position accuracy of the origin coordinates for each lens. Is important.

Therefore, first, in the optical unit 21, the point P is identified by using the first lens,
The origin coordinate O _A1 of the first lens is obtained by the method as described above. At this time, in the image captured by the first lens, as shown in FIG. 12, the linear transformation matrix is A _1, origin coordinates is assumed to be O _A1. Further, it is assumed that the coordinates of the point P in the camera coordinate system of the first lens are P _A1 .

Next, the lens used in the optical unit 21 is replaced with a second lens. At this time, as shown in FIG. 12, in the image captured by the second lens, it is assumed that the primary transformation matrix is A ₂ and the coordinates of the point P in the camera coordinate system of the second lens are P _{A2. Assuming} that the coordinates of the point P in the mechanical coordinate system are PM, the following Expression 26 and Expression 27 can be expressed from Expression 10 described above.

[0205]

[Equation 36]

[0206] Therefore, clearing the P _M by the equation 26 and equation 27, as shown in the following expression 6, based on the O _A1, O _A2
Can be requested.

[0207]

(37)

By the way, when the lens used in the optical unit 21 is replaced with the second lens, there may be a case where the point P does not exist in the image picked up by the second lens.
In this case, as shown in FIG. 13, after the replacement with the second lens, the inspection stage 14 is moved in parallel so that the point P falls within the image. At this time, the point P before the movement
The coordinates of a mechanical coordinate system and P _M1, when the coordinates of the mechanical coordinate system of the point P after the movement and P _M2, as in the following Equation 7, based on the O _A1, be obtained O _A2 it can.

[0209]

(38)

As described above, even in the case where the image recognition apparatus 1 includes a plurality of lenses, the image recognition apparatus 1 performs other operations based on the primary transformation matrix A ₁ and the origin coordinates O _A1 of the predetermined lens already obtained. by obtaining the linear transformation matrix a ₂ and origin coordinate O _A2 of the lens, the time of shortening required for calibration, it is possible to improve the calibration accuracy.

[0211] The image processing apparatus 1, obtained as described above, a linear transformation matrix A, based on the origin coordinates O _A of the camera coordinate system in the mechanical coordinate system, the mechanical coordinates by using the relational expression shown in Equation 10 Image recognition is performed by associating the system with the camera coordinate system. Therefore, the image recognition device 1 can associate the mechanical coordinate system with the camera coordinate system based on the image captured by the optical unit 21 without using a special jig. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

In the above description, the image recognition apparatus 1 to which the present invention is applied has been used for checking what the defect is in the semiconductor wafer. However, the application of the image recognition device 1 according to the present invention can also be used for applications other than defect identification of a semiconductor wafer. That is, the image recognition device 1 according to the present invention can be used, for example, to inspect whether a device pattern formed on a semiconductor wafer is formed in an appropriate shape as a desired pattern. Further, the application of the image recognition device 1 according to the present invention is not limited to inspection of a semiconductor wafer, and the image recognition device 1 according to the present invention is widely applicable to inspection of fine patterns. It is also effective for inspection of a flat panel display on which a fine pattern is formed.

[0213]

As described above, the image recognition apparatus according to the present invention makes the mechanical coordinate system and the camera coordinate system correspond to each other based on the image taken by the image pickup means without using a special jig. Can be. Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily regardless of the mechanical and optical positional accuracy of the device.

Further, when the image recognition device according to the present invention includes a plurality of light receiving lenses for receiving the imaging light, based on the primary conversion matrix A ₁ of the predetermined light receiving lens, the primary conversion matrix A of the other light receiving lenses is obtained. You can ask for ₂ . Therefore, the mechanical coordinate system and the camera coordinate system can be made to correspond to each other without using a special jig based on the image captured by the image capturing means. The coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily.

Further, when the image recognition device according to the present invention includes a plurality of light receiving lenses for receiving the imaging light, the image recognition device is based on the origin coordinate O _A1 of the camera coordinate system in the mechanical coordinate system when a predetermined light receiving lens is used. Te, it is possible to determine the origin coordinate O _A2 of the camera coordinate system when using other light-receiving lens. Therefore, the mechanical coordinate system and the camera coordinate system can be associated with each other without using a special jig based on the image captured by the image capturing unit.
The mechanical coordinate system and the camera coordinate system can be made to correspond with high accuracy and easily without depending on the optical position accuracy.

Therefore, according to the image recognition device of the present invention, it is possible to realize an image recognition device capable of performing image recognition easily, at low cost and with high accuracy.

[Brief description of the drawings]

FIG. 1 is a perspective view showing the appearance of an image recognition device to which the present invention is applied.

FIG. 2 is a diagram showing a state where the apparatus main body disposed inside a clean box of the image recognition apparatus is viewed from the direction of arrow A1 in FIG.

FIG. 3 is a plan view schematically showing a state where the apparatus main body of the image recognition apparatus is viewed from above.

FIG. 4 is a block diagram illustrating a configuration example of the image recognition device.

FIG. 5 is a diagram illustrating a configuration example of an optical system of an optical unit of the image recognition device.

FIG. 6 is a flowchart illustrating an example of a procedure when inspecting a semiconductor wafer by the image recognition device.

FIG. 7 is a diagram for explaining a method of detecting a defect from a reference image and a defect image.

FIG. 8 is a schematic diagram for explaining a coordinate system in the image recognition device.

FIG. 9 is a schematic diagram for explaining an alignment mark formed on a semiconductor wafer to be inspected by the image recognition device.

FIG. 10 is a schematic diagram illustrating a coordinate system when the inspection stage is rotated in the image recognition device.

FIG. 11 is another schematic diagram for explaining a coordinate system when the inspection stage is rotated and moved in the image recognition device.

FIG. 12 is a schematic diagram illustrating a coordinate system when a lens is switched in the image recognition device.

FIG. 13 is another schematic diagram for explaining a coordinate system when a lens is switched in the image recognition device.

[Explanation of symbols]

1 image recognition device, 2 clean unit, 3 clean box, 4 clean air unit, 5a, 5b
Blower, 10 main unit, 14 inspection stage, 21
Optical unit, 22 elevator, 23 transfer robot, 24 pre-aligner, 30 visible light CCD camera, 31 ultraviolet light CCD camera, 32 halogen lamp, 33 visible light optical system, 34 visible light objective lens, 36 ultraviolet light laser Light source, 37 UV light optical system, 3
8 Objective lens for ultraviolet light, 50 external units, 51,5
2 display device, 53 input device, 60 computer for image processing, 61 computer for control

 ──────────────────────────────────────────────────続 き Continued on front page F term (reference) 2F065 AA03 AA14 AA56 BB02 BB03 CC19 FF04 GG02 GG04 JJ03 JJ05 JJ26 LL04 PP12 QQ31 UU04 UU05 4M106 AA01 BA07 CA39 CA41 DB04 DB07 DB08 DB12 DB13 DB16 DJ04 DJ05 DJ06 DJ06 DJ07 5L096 BA03 BA20 CA02 CA14 EA14 HA07 9A001 BB05 HH20 HH24 JJ49 KK16 KK31 KK54

Claims (7)

[Claims] 1. A base, fixed support means provided on the base, fixedly supporting an object to be inspected, and enabling the object to be inspected to be movable in parallel within a predetermined plane, and An image pickup means provided on a table, wherein the fixed support means picks up the object to be inspected from outside the surface in which the object to be inspected is movable; and performs image processing by processing an image picked up by the image pickup means to perform image recognition An image recognizing unit, wherein the image recognizing unit translates an arbitrary point P on the inspection object in an image captured by the image capturing unit by the fixed support unit, and sets the coordinates of the point P to 3 point, obtained in the coordinate P _M in the mechanical coordinate system is the base coordinate system, respectively, the coordinates P _a in the camera coordinate system is a coordinate system in the image captured by the imaging unit, 3 in the mechanical coordinate system The coordinates of the point _M1, P _M2, P
_M3 , the coordinates of the three points in the camera coordinate system are P _A1 , P _A2 , P
_{When A3} is set, As, calculated by Equation 1 below linear transformation matrix A movement amount of the point P from the mechanical coordinate system to the camera coordinate system, equation P _M when the origin coordinates of the camera coordinate system and O _A in mechanical coordinate system = A ^-1 P _A + O _A , wherein the image recognition is performed by associating the mechanical coordinate system with the camera coordinate system. (Equation 2) 2. A base, which is provided on the base and fixedly supports an object to be inspected, so that the object to be inspected can be moved in parallel within a predetermined plane, and Fixed support means rotatable about a vertical axis and provided with an index indicating the center of this rotation, and an out-of-plane provided on the base and allowing the fixed support means to move the inspection object Image capturing means for capturing an image of the object to be inspected, and image recognizing means for recognizing an image captured by the image capturing means by performing arithmetic processing. The image recognizing means includes mechanical coordinates that are a coordinate system of the base. coordinates P _M of arbitrary point P on the inspection object in the system, the camera coordinate coordinates in the camera coordinate system is a coordinate system in the image captured by the imaging means of the point P P _a, the mechanical coordinate system Of point P to the system The linear transformation matrix of the dynamic quantity is taken as A, by imaging the indicator by the image pickup means, to get the coordinates O _S of the rotation center of said fixed support means in a mechanical coordinate system, the camera coordinate system in the mechanical coordinate system seeking the origin coordinate O _a by equation 2 below, equation P _M
= A ^-1 P _A + O _A , wherein the image recognition is performed by associating the mechanical coordinate system with the camera coordinate system. (Equation 3) 3. A base, provided on the base, fixedly supporting an object to be inspected, enabling the object to be moved in parallel within a predetermined plane, and Fixed support means rotatable around a vertical axis, and image pickup means provided on the base, wherein the fixed support means images the test object from outside the plane in which the test object is movable. Image recognition means for performing image processing on the image picked up by the image pickup means and recognizing the image, wherein the image recognition means comprises an arbitrary point on the inspection object in a mechanical coordinate system which is a coordinate system of the base. The coordinate of P is P _M , the coordinate of this point P in the camera coordinate system, which is the coordinate system in the image captured by the image capturing means, is P _A , and the primary displacement of the movement of the point P from the mechanical coordinate system to the camera coordinate system. Matrix is A, mechanical The coordinates of the center of rotation of the fixed support means when the O _S in the target system, obtains two points on the camera coordinate system is rotated moving the coordinates of the point P within the image which the image pickup means takes an image, acquires the coordinates when the P _A1, P _A2, determined by equation 3 below the origin coordinates O _a of the camera coordinate system in the mechanical coordinate system, mechanical with equation ^{_{P M = a -1 P a +}} O a An image recognition device for performing image recognition by associating a coordinate system with a camera coordinate system. (Equation 4) 4. A base, which is provided on the base and fixedly supports the object to be inspected so that the object to be inspected can move in parallel within a predetermined plane. Fixed support means rotatable around a vertical axis, and image pickup means provided on the base, wherein the fixed support means images the test object from outside the plane in which the test object is movable. Image recognition means for performing image processing on the image picked up by the image pickup means and recognizing the image, wherein the image recognition means comprises an arbitrary point on the inspection object in a mechanical coordinate system which is a coordinate system of the base. The coordinate of P is P _M , the coordinate of this point P in the camera coordinate system, which is the coordinate system in the image captured by the image capturing means, is P _A , and the primary displacement of the movement of the point P from the mechanical coordinate system to the camera coordinate system. When the matrix is A Acquires the coordinate O _S1 of the rotation center of said fixed support means in the coordinate P _A1 and the mechanical coordinate system of the point P in the camera coordinate system is rotated moving the position of the point P to the outside of the image which the imaging means for imaging after the, the fixed support means to this point P is within the image is moved in parallel, the coordinate O _S2 of the rotation center of said fixed support means in the coordinate P _A2 and the mechanical coordinate system of the point P in the camera coordinate system Obtain and set the origin coordinate of the camera coordinate system in the mechanical coordinate system to O _A
An image recognition apparatus characterized in that image recognition is performed by associating a mechanical coordinate system with a camera coordinate system using a relational expression P _M = A ^-1 P _A + O _A. (Equation 5) 5. A base, fixed support means provided on the base, fixedly supporting an object to be inspected, and enabling the object to be inspected to be movable in parallel within a predetermined plane; An imaging unit provided on a table, wherein the fixed support unit captures an image of the object to be inspected from outside a plane that allows the object to be moved, and an imaging unit that includes at least two light receiving lenses that receive imaging light; Image recognizing means for recognizing the image picked up by the image pick-up means by performing arithmetic processing, wherein the image recognizing means is provided for an arbitrary point P on the inspection object in a mechanical coordinate system which is a coordinate system of the base. coordinates P _M, of the point P coordinates are coordinates of P _a in the camera coordinate system is in the image captured by the imaging unit, the origin coordinates of the camera coordinate system in the mechanical coordinate system and O _a, of the image pickup means 2 Between the light receiving lens, the linear transformation when the linear transformation matrix point movement amount P from the mechanical coordinate system to the camera coordinate system when using the first light receiving lens with A _1, a second light receiving lens A matrix is set to A _2, and the coordinates of the point P are two points by moving the coordinates of the point P in parallel within the range of the image picked up by using the second light receiving lens. When the coordinates of two points in the mechanical coordinate system are P _M1 and P _M2 and the coordinates of the two points in the camera coordinate system are P _A1 and P _A2 , The primary transformation matrix A ₂ when the second light receiving lens is used is obtained by the following equation 5, and the relational expression P _M = A ⁻¹ P _A
+ O _A image recognition apparatus characterized by performing image recognition by associating the mechanical coordinate system and the camera coordinate system used. (Equation 7) 6. A base, fixed support means provided on the base, fixedly supporting an object to be inspected, and enabling the object to be inspected to be movable in parallel within a predetermined plane, and An imaging unit provided on a table, wherein the fixed support unit captures an image of the object to be inspected from outside a plane that allows the object to be moved, and an imaging unit that includes at least two light receiving lenses that receive imaging light; Image recognizing means for recognizing the image picked up by the image pick-up means by performing arithmetic processing, wherein the image recognizing means is provided for an arbitrary point P on the inspection object in a mechanical coordinate system which is a coordinate system of the base. The coordinates are P _M , the coordinates of this point P in the camera coordinate system, which is the coordinate system in the image captured by the image capturing means, are P _A, and the camera coordinate system when the first light receiving lens is used in the image capturing means. Of point P at And P _A1, a linear transformation matrix A ₁ of the amount of movement of the point P from the mechanical coordinate system to the camera coordinate system,
Origin coordinate O _A1 of camera coordinate system in mechanical coordinate system
And the primary transformation of the coordinates P _A2 of the point P in the camera coordinate system and the amount of movement of the point P from the mechanical coordinate system to the camera coordinate system when the second light receiving lens is used in the imaging means. A matrix A ₂ is obtained, and the origin coordinate O _A2 of the camera coordinate system in the mechanical coordinate system when the second light receiving lens is used is calculated between the two light receiving lenses in the imaging means by the following equation (6).
Is calculated based on the origin coordinate O _A1 of the light receiving lens of
An image recognition device, wherein image recognition is performed by associating a mechanical coordinate system with a camera coordinate system using _M = A ^-1 P _A + O _A. (Equation 8) 7. A base, fixed support means provided on the base, fixedly supporting an object to be inspected, and enabling the object to be inspected to be movable in parallel within a predetermined plane, and An imaging unit provided on a table, wherein the fixed support unit captures an image of the object to be inspected from outside a plane that allows the object to be moved, and an imaging unit that includes at least two light receiving lenses that receive imaging light; Image recognizing means for recognizing the image picked up by the image pick-up means by performing arithmetic processing, wherein the image recognizing means is provided for an arbitrary point P on the inspection object in a mechanical coordinate system which is a coordinate system of the base. the coordinates P _M, the coordinates in the camera coordinate system is a coordinate system in the image captured by the imaging means of the point P and P _a, in the case of using the first light receiving lens by the imaging means, a mechanical coordinate system Point P at The coordinates P _M1, and the coordinates P _A1 point P in the camera coordinate system, a linear transformation matrix A ₁ of the amount of movement of the point P from the mechanical coordinate system to the camera coordinate system, the origin coordinates of the camera coordinate system in the mechanical coordinate system O _A1 is obtained, and the point P
Is moved out of the range of the image captured by the imaging unit, the fixed support unit is moved in parallel until the point P falls within the image, and the coordinates P _M2 of the point P in the mechanical coordinate system are The coordinates P _A2 of the point P in the camera coordinate system and the primary transformation matrix A ₂ of the movement amount of the point P from the mechanical coordinate system to the camera coordinate system. by equation 7 below the origin coordinates O _A2 of the camera coordinate system in the mechanical coordinate system when using the second light receiving lens first
Is calculated based on the origin coordinate O _A1 of the light receiving lens of
An image recognition device, wherein image recognition is performed by associating a mechanical coordinate system with a camera coordinate system using _M = A ^-1 P _A + O _A. (Equation 9)