JP2000028535A - Defect inspecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize defect inspection of high reliability by comparing a plurality of image data photographed by having a substance to be detected moved relatively. SOLUTION: A wafer 7 placed on an XY-stage 6 is illuminated by a diffusing illuminating light from an illuminating unit 1. An image caused by a positive reflection light from the wafer 7 is photographed by a CCD camera 10. At this time, since an illuminating surface of a diffusion plate 4 has an area of a predetermined size, an image having no variation of a photograph condition caused by a warpage of the water 7 is obtained in the CCD camera. The image signal is captured in an image processing device 20, and a first sheet of image data is stored. Next, a control device 23 drives and controls a driving device 22 so as to deviate it against an arrangement direction of a chip pattern by one-chip portion to move the XY-stage 6. Then, an image signal of the wafer 7 after the movement is taken in the image processing device 20, and a second sheet of image data is memorized. The image processing device 20 carries out a defect detection by comparing two sheets of image data stored.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a defect inspection apparatus for inspecting a defect on a surface of a test object (work) such as a semiconductor wafer or liquid crystal glass, and an illumination apparatus for the defect inspection.

[0002]

2. Description of the Related Art In a process of manufacturing a semiconductor wafer, liquid crystal glass, or the like, a so-called macro inspection for inspecting a defect macroscopically is performed. For example, in a macro inspection of a semiconductor wafer (hereinafter, simply referred to as a wafer), a visual inspection is mainly performed by an operator by the following two methods.

One is to irradiate the wafer surface with strong illumination light and observe the dark field while tilting or rotating the wafer at various angles. If there is a defect, it is observed as a difference in brightness or color under certain conditions of inclination and angle.

[0004] Second, bright field observation is performed using diffuse illumination and mainly using specularly reflected light. If there is a defect, it is observed as a slight light / dark or color difference.

[0005]

However, such a macro inspection visually performed by the operator makes it difficult for the operator to cause contamination and maintain cleanliness due to approaching the wafer, which hinders quality improvement. . Also,
Operators have variations and oversight of inspection capabilities,
There are difficulties in securing stable quality, and securing and training excellent human resources also require enormous costs and time. Therefore, automation of macro inspection is desired.

As a method of automating the macro inspection, a method of imaging a test object with an image pickup device such as a CCD and extracting a defect by image processing the image data is conceivable. However, there is the following problem only by taking an image instead of visual observation by an operator.

[0007] In dark-field observation, since inspection is performed using diffracted light, an observation angle at which a defect can be confirmed differs depending on the type of the test object. Therefore, a mechanism for adjusting the angle is required.

In the bright field observation, specular reflection light is used. However, depending on the size of the area of the diffused illumination plate, the imaging condition is affected by the magic mirror phenomenon and the diffraction effect due to the warpage of the test object. Pseudo defects occur even in a defect-free test object.

Further, when a wafer having a repetitive pattern is normally imaged by the image pickup device, moire occurs due to the relationship between the pixel period of the image pickup device and the pattern period, which causes a pseudo defect.

The present invention has been made in view of the above prior art and provides a defect inspection apparatus and a defect inspection system capable of performing a highly reliable defect inspection without complicating the mechanism and eliminating the effects of warpage and diffracted light of a test object. It is an object of the present invention to provide a lighting device for a vehicle.

[0011]

Means for Solving the Problems In order to solve the above problems, the present invention is characterized by having the following configuration.

(1) In a defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, a substantially uniform illumination surface having a size determined based on the degree of warpage expected on the test surface of the test object. An illumination optical system that illuminates the test object by irradiating the diffused light of luminance, and an imaging optical system that has an imaging element that images the test object by specularly reflected light from the test object illuminated by the diffuse illumination light, Defect detecting means for detecting a defect on the test object based on an image picked up by the image pickup element, and moving means for moving the test object relative to the image pickup optical system; The test object is relatively moved with respect to the imaging optical system so that a pattern at a certain first position and a pattern at a second position on the test object are substantially the same. Moving and imaged It is characterized in that a defect on a test object is detected based on comparison of at least two pieces of image data.

(2) In a defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, a substantially uniform illumination surface having a size determined based on the degree of warpage expected on the test surface of the test object. An illumination optical system that illuminates the test object by irradiating the diffused light of luminance, and an imaging optical system that has an imaging element that images the test object by specularly reflected light from the test object illuminated by the diffuse illumination light, A defect detecting means for detecting a defect on the test object based on an image picked up by the image pickup element; a storage means for storing image data of the test object having a defect-free pattern; Positioning means for relatively moving the test object so as to position the test object so that a pattern imaged with respect to the imaging optical system has a predetermined positional relationship, wherein the image stored in the storage means is provided. Data and images of the subject The method is characterized in that a defect on a test object is detected based on comparison with data.

(3) In a defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, a substantially uniform illumination surface having a size determined based on the degree of warpage expected on the test surface of the test object. An illumination optical system that illuminates the test object by irradiating the diffused light of luminance, and an imaging optical system that has an imaging element that images the test object by specularly reflected light from the test object illuminated by the diffuse illumination light, Defect detection means for detecting a defect on a test object based on an image picked up by the image pickup device; and a position of a pattern of a first position and a pattern of a second position on the test object with respect to pixels of the image pickup device Moving means for relatively moving the test object with respect to the imaging optical system so that the relationship becomes substantially the same, wherein the defect detecting means compares at least two pieces of image data imaged by the movement of the moving means. Subject based on First defect detection means for detecting the above defect, storage means for storing, as a reference image, image data of the test object which has been made defect-free by the first defect detection means,
Detecting a defect on the object based on a comparison between the image data obtained by imaging the object positioned so that the pattern imaged with respect to the imaging optical system is in a predetermined positional relationship and the reference image data; And a second defect detecting means.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration of a defect inspection apparatus according to the present invention.

Reference numeral 1 denotes an illumination unit constituting an illumination optical system, which includes a halogen lamp 2 as a light source and diffusion plates 3 and 4. The size of the illumination surface of the diffusion plate 4 at the final emission end is determined in consideration of the degree of warpage of the wafer 7 as a test object, the diffraction effect of the pattern of the wafer 7, and the like (described later). The light of the halogen lamp 2 is diffused by the two diffusing plates 3 and 4 to provide diffused illumination light having sufficiently uniform luminance. The diffused illumination light emitted from the diffuser plate 4 is made substantially parallel by the collimator lens 5, and then irradiates the wafer 7 placed on the XY stripe 6.

Reference numeral 10 denotes a CCD camera constituting an image pickup optical system, 11 denotes an aperture, 12 denotes a photographing lens, 13 denotes an image pickup device, and the image pickup optical axis of the illumination optical system is sandwiched by the optical axis of the collimator lens 5. They are arranged so as to be symmetrical with the optical axis. The specularly reflected light from the wafer 7 illuminated by the illumination optical system is converged by the collimator lens 5 and
After that, an image of almost the entire surface of the wafer 7 is formed on the image sensor 13 by the photographing lens 12. Note that the illumination optical system and the imaging optical system may be coaxial by using a half mirror.

Reference numeral 20 denotes an image processing device, which performs predetermined processing such as A / D conversion on a video signal from the CCD camera 10 and captures it, and then performs necessary preprocessing such as noise removal and sensitivity correction of an image sensor. To detect defects. 20a is a storage device of the image processing apparatus 20. Reference numeral 21 denotes a display on which an image captured by the image processing device 20 is displayed. 22 is a driving device for moving the XY stage 6, 23
Is a control device for controlling the entire defect inspection apparatus.

Next, determination of the size of the diffusion plate 4 will be described. Substantially planar work (specimen) with specular reflection characteristics
In the case of inspecting the color or brightness of an image (in the case of inspection by bright field illumination), an image is usually taken with a surface light source slightly larger than the inspection area of the workpiece as a background. However, when an extremely fine repetitive pattern is formed as in a wafer, light is diffracted by using a large surface light source, and an image is captured as if the color was locally applied.

In order not to be affected by the diffraction effect, a point light source is used, and the light is condensed and illuminated by a lens which is slightly larger than the work so that the image can be captured only by the specularly reflected light. It is possible. However, some wafer surfaces in the manufacturing process have a slight warp (incline), and if there is such a warp, a light-dark pattern is generated due to the magic mirror phenomenon. In order to prevent light and dark from occurring, it is conceivable to increase the aperture of the taking lens.However, in this case, the place where light passes through the taking lens changes depending on the warping state, and the imaging characteristics change due to aberration. I will.

Therefore, the size of the illuminating surface of the diffuser plate 4 is set to a size excluding the effects of the magic mirror phenomenon and the diffractive action in the following manner in consideration of the degree of warpage (inclination) expected for the work. .

First, the size for avoiding the influence of the magic mirror phenomenon will be described. Now, as schematically shown in FIG. 2 (for ease of understanding, the drawing is made without the refraction by the collimating lens 5), a CCD camera having a shooting aperture S having an aperture diameter D is used. Assume that the entire work (test object) W is imaged. If the workpiece W has no warp and the inspection surface is parallel to the reference surface, light from the diffused illumination surface M that is specularly reflected at the surface is in a region d around the illumination optical axis.
Light emitted from No. 1 enters a photographic stop S having an aperture diameter D, and the light enters an image sensor of a camera. Here, when the workpiece W is inclined by ± θ with respect to the reference plane, the areas of the regular reflection light from the illumination surface M incident on the photographing stop S are d2 and d.
Change to 3.

Therefore, when the maximum inclination expected for the work W is ± θ, the size of the illumination surface M is set to a region larger than the region D ′ covering d2 and d3, and
If the in-plane luminance is made sufficiently uniform, it is possible to avoid the magic mirror phenomenon that causes pseudo light and dark (the amount of specularly reflected light incident on the image pickup device of the camera can be made substantially the same). ). Therefore, in consideration of the maximum inclination in all directions expected on the work surface, when the total area when the photographing aperture S is projected on the illumination surface is A1, the size of the diffuse illumination surface is larger than A1. In this case, it is sufficient that the illumination is performed at a sufficiently uniform luminance.

Next, a size for avoiding the influence of the diffracted light will be described. As in the case of FIG. 2, it is assumed that the whole work W is imaged by the CCD camera having the photographing aperture S having the aperture diameter D, and the work W is inclined with respect to the reference plane as shown in FIG. Suppose you have At this time, as for light from the illumination surface M that is regularly reflected on the surface of the work W, light emitted from the region d4 enters the imaging stop S. Here, assuming that the minimum diffraction angle due to the pattern formed on the work W is θ2, if the viewing angle when viewing the illumination surface from the work W is an illumination surface area D ″ smaller than the diffraction angle θ2, diffraction Light does not enter the shooting aperture S (when the degree of warpage of the work is small,
It can be said that the area D ″ can be expanded to a range not exceeding 2θ2).

Therefore, in consideration of the maximum inclination in all directions expected to the work W, even if the light is emitted from any part of the illumination surface, the maximum area where the diffracted light by the pattern of the work W does not enter the photographing lens. Is A2, the size of the illumination surface may be smaller than A2 (a sufficient condition is that the aperture angle of the illumination viewed from the position of the work W is smaller than the minimum diffraction angle. good). The minimum diffraction angle θ2 is, for example, when the average pitch of the pattern on the work W is 2 μm and the shortest wavelength included in the illumination light is 0.4 μm, θ2 = sin ⁻¹ (0.4 / 2) = 11 °.

In the area A2, even when the pattern on the workpiece W has a pattern with a partially large pitch without completely eliminating the influence of the diffracted light due to the pattern, the amount of the diffracted light is small. Within an allowable error range for measurement (a range where the amount of diffracted light from the pattern does not exceed a predetermined ratio with respect to the amount of light received by the image sensor)
It is practically sufficient to determine the size with. This allowable range may be determined in design according to the degree of warpage of the work and the detection sensitivity.

From these, the size of the illuminating surface of the diffusion plate 4 is determined in consideration of the degree of warpage (inclination) expected for the work.
If the area A1 is larger than the area A1 determined from the design of the optical arrangement and is smaller than the area A2 determined from the pattern on the wafer 7 and the wavelength of the illumination light (the sensitivity wavelength of the CCD camera), the magic mirror phenomenon and the diffraction action can be prevented. The effects can be removed. If A1 is larger than A2, the short wavelength component of the illumination light may be cut, for example,
The problem can be solved by providing a filter for cutting the short wavelength component in the illumination optical system or the imaging optical system.

According to the present inventor, when inspecting a wafer having a diameter of about 200 mm, the warpage (inclination angle) of the wafer is 1 mm.
It was confirmed that practically enough if the degree was expected.

Next, detection of defects such as scratches and dust using specularly reflected light and the size of the diffusion plate 4 for this purpose will be described.

When inspecting defects having scattering characteristics such as scratches and dust by bright-field observation using specular reflection light, the amount of light lost due to scattering is detected. If the size of the defect is larger than the resolution of the image pickup device, the defect can be easily recognized by a large light amount difference from the non-defective portion. However, when the size of the defect is smaller than the resolution of the image sensor, a small change in the amount of light is observed. If the light amount changes due to factors other than the defect, the defect cannot be detected unless the change is larger than this, and the sensitivity is deteriorated. On the other hand, as described above, it is not affected by the magic mirror phenomenon due to the warp or the cause of the pseudo defect due to the diffracted light (further, the moiré due to the pattern described later and the cause of the pseudo defect due to the characteristic difference between the pixels of the imaging device) This makes it possible to detect flaws, dust, and the like using specularly reflected light with high sensitivity.

In defect detection using specularly reflected light, it is ideal that scattered light does not enter the imaging element surface. For this reason, the defect can be detected with higher sensitivity as the illumination surface that emits diffuse illumination light is smaller (the larger the size of the illumination surface, the higher the rate of scattered light incident on the imaging element surface and the lower the detection sensitivity). . For example, if the defect is a linear flaw, the distribution of the scattered light is fan-shaped with the flaw as an axis. When the defect is irradiated with illumination light from all directions of 180 degrees, the total amount of scattered light incident on the imaging element surface becomes the same as the regular reflection light, so that the amount of light lost due to scattering is not detected. . Assuming that the intensity distribution of the scattered light due to the defect is uniform, the angle of illumination viewed from the defect divided by 180 degrees is the light reduction rate due to the defect. It is better that the reduction rate is large, but what is actually detected is the reduction amount.
The amount of decrease between / 10 and 1 / ∞ is 0.9 and 1.0, and the detection sensitivity differs only by 10%. Therefore, a reduction rate of about 1/10 is sufficient for practical use. That is, in the macro inspection of the wafer, it can be said that if the size of the illumination surface is such that the angle of illumination viewed from the defect is smaller than about 18 degrees, it is possible to detect a defect sufficiently enough for practical use (actual scattering of actual scattered light). Since the intensity distribution is generally not uniform, it is desirable that the intensity distribution is smaller.)

As described above, the size of the illuminating surface of the diffusion plate 4 is set to be smaller than a predetermined angle (an angle determined by the relationship with the inspection accuracy) as described above while the angle of the illuminating surface viewed from the defect is as small as possible. If the size is minimized so as not to be affected by the magic mirror phenomenon, a defect having a scattering characteristic such as a flaw or dust can be detected with high sensitivity only by specular reflection light.

Next, the operation of the defect inspection according to the present embodiment will be described (here, it is assumed that the wafer 7 has a repetitive pattern). First, the wafer 7 mounted on the XY stage 6 is illuminated by the diffuse illumination light from the illumination unit 1. With this illumination, an image of the regular reflection light from the wafer 7 is captured by the CCD camera 10. At this time,
Since the illumination surface of the diffusion plate 4 has an area of the size determined as described above, the CCD camera 10 can obtain an image in which the imaging condition does not change due to the warpage of the wafer 7. The video signal from the CCD camera 10 is taken into the image processing device 20 and the first image data is stored.

When the image data of the first sheet can be stored, the controller 23 connected to the image processing apparatus 20 controls one chip (or one chip) in the arrangement direction of the chip pattern formed on the wafer 7. The XY stage 6 is moved by controlling the driving of the driving device 22 so that the XY stage 6 is shifted by a positive number of times of the pattern. In this case, the image data from the CCD camera 10 is processed to determine the arrangement direction of the chip pattern, and the movement may be performed based on the data of the arrangement direction and the pattern pitch.
The image signal of the moved wafer 7 is taken into the image processing device 30 and the second image data is stored. The image processing device 20 performs defect detection by comparing the stored two pieces of image data.

The defect detection of the repetitive pattern can be performed by a method of calculating the difference between the images of the adjacent patterns (pattern matching). Is rough, and moire occurs in the captured image. For this reason,
In the differential processing using one image data, a pseudo defect is detected due to the effect of moire. However, as described above, by using two image data in which the relationship between the pattern and the pixel of the image sensor is matched. In addition, it is possible to detect only the original defect while eliminating the influence of moire. Hereinafter, the reason will be described with reference to FIG.

FIG. 4A shows CC before the wafer is moved.
FIG. 4B is a diagram schematically illustrating a positional relationship between patterns of two adjacent chips 1 and 2 with respect to a pixel of the D camera, and FIG. 4B is a diagram illustrating a luminance signal of image data at that time. . (C) and (d) are diagrams schematically showing the same after the movement (the same as when the optical system is relatively moved). Here chip 1
It is assumed that there is a defect pattern 50 in the pattern on the side. FIG. 4C shows the positional relationship of the pixels with respect to the pattern of the chip 1 in FIG.
The pattern of the chip 2 and the positional relationship between the pixels are the same. Accordingly, the same moiré is generated in the image data of the pattern of the chip 1 in FIG. 4B and the image data of the pattern of the chip 2 in FIG. 4D, respectively. If the difference processing is performed, FIG.
As shown in (e), only the defect data 50 remains, and a defect inspection without generation of a pseudo defect due to moire or the like can be performed.
If it is necessary to specify where the defect exists, the method described in Japanese Patent Application No. 9-369221 filed by the present applicant can be used.

According to the method for detecting a defect of an object having a repetitive pattern described above, the same image can be obtained if the same pattern exists at the same position, even if the pattern of the object is fine. It is not necessary to use a special high-resolution imaging means, and an expensive and high-speed image processing apparatus does not need to be used.

In order to make the pattern positional relationship with respect to the pixels of two pieces of image data the same, the number of pixels in one chip pattern can be increased without moving one chip or a positive multiple thereof. Depending on whether it is on top (percentage of pixels), it can also be moved with a minimal amount of movement. For example, 1
If the repetition pitch of one chip is 20.5 pixels,
With a small movement of 0.5 pixel, the positional relationship between the pixel and the pattern becomes substantially the same.

In the defect inspection by the CCD camera 10 configured as described above, the size of the diffusion plate 4 is determined while taking into consideration the degree of warpage expected for the test object while securing the area A1 described above. If it is sufficiently small, scratches and dust can be detected as black defects with high sensitivity only by the specular reflection light from the diffused illumination (specular reflection light from scratches and dust having scattering characteristics is greatly attenuated. It is detected with high sensitivity as a defect). For this reason, it is not necessary to provide a spot illumination system and perform dark field observation, and it is possible to perform a necessary inspection only with the present illumination system.

In detecting a defect of a test object having a pattern, in addition to comparing (differential processing) images of adjacent patterns with respect to one work, an image of a reference work having no defect (other than raw image data). , Which may be processed image data) in the storage device 2 of the image processing apparatus.
0a is stored in advance, and the defect can be detected by comparing this with the image of the work to be inspected. In this case, when storing the image data of the work to be inspected, the work to be inspected is placed so that the positional relationship with respect to the pixels of the CCD when the reference work stored in the storage device is imaged is the same. The XY stage 6 to be moved is controlled for alignment. This can be achieved by performing image processing on image data obtained from the CCD camera 10 with respect to the image data of the reference work to obtain the movement information. By doing so, even in the case of comparison with the reference work, the defect inspection can be performed while eliminating the influence of moiré, and only one piece of image data needs to be obtained for the work to be inspected, thereby increasing the inspection throughput.

Further, when inspecting a large number of workpieces having the same pattern formed in one carrier, such as a semiconductor wafer, the following is convenient. First, as shown in the previous example, a defect inspection is performed by obtaining image data before and after movement of a certain work, and if there is a work determined to be defect-free, this is stored as image data of a reference work. Keep it. Thereafter, defect detection is performed by comparing the image of the reference work and the image of the work to be inspected in the same manner as described above. Using this method, it is detected that the lot of the work on which the same pattern is formed has changed (for example, it is only necessary to know that the carrier has changed), and whether or not the image data of the reference work is required by the detection is determined. If the judgment is made, the automation of the inspection can be further advanced.

Defect detection of a test object having no repetitive pattern (for example, defect inspection such as scratches and dust on a wafer before forming a pattern) can be detected from one piece of image data. In the case where there is a characteristic difference between pixels of the imaging device of the CCD camera, which becomes a pseudo defect and affects the detection sensitivity, the following may be performed. In other words, it has moved by a moving amount of a positive multiple of the pixel of the image sensor.
The comparison processing (difference processing) of the image data of the sheets is performed. Or
A comparison process is performed between non-defective reference image data stored in advance and image data obtained as an inspection target. In this case, the defect inspection can be performed with high sensitivity without being affected by the characteristic difference between the pixels.

[0043]

As described above, according to the present invention,
Without being affected by the warpage of the test object and diffraction by the pattern,
Stable and highly reliable inspection can be performed. As a result, the macro inspection can be automated, and it is possible to stabilize the product and save labor in the inspection. In addition, since it is sufficient to image the entire test object only by specular reflection light (only by bright-field observation) without changing the imaging angle of the test object, high-speed inspection can be performed without complicating the inspection mechanism. When an image of a defect-free test object is used as a reference image, only one image needs to be captured thereafter, and only one difference process is required.
Inspection can be performed more efficiently, and substantial throughput can be increased.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a defect inspection apparatus according to the present invention.

FIG. 2 is a schematic diagram for explaining the size of an illumination surface that avoids the influence of the magic mirror phenomenon.

FIG. 3 is a schematic diagram for explaining the size of an illumination surface that avoids the influence of diffracted light.

FIG. 4 is a diagram for explaining a method of detecting only an original defect while eliminating the influence of moiré.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Illumination unit 3, 4 Diffusion plate 5 Collimator lens 7 Wafer 10 CCD camera 13 Image sensor 20 Image processing device

Claims (3)

[Claims] 1. A defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, wherein a substantially uniform luminance is obtained from an illumination surface having a size determined based on a degree of warpage expected on a test surface of the test object. An illumination optical system that illuminates the object by irradiating the object with diffused light, an imaging optical system that has an imaging element that images the object with specularly reflected light from the object illuminated by the diffused illumination light, Defect detecting means for detecting a defect on the test object based on an image picked up by the image pickup device; and moving means for moving the test object relative to the image pickup optical system; The test object is relatively moved with respect to the imaging optical system so that the positional relationship between the pattern at a certain first position and the pattern at the second position on the inspection object is substantially the same. At least moved and imaged A defect inspection apparatus for detecting a defect on a test object based on a comparison between two image data. 2. A defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, wherein a substantially uniform luminance is obtained from an illumination surface having a size determined based on a degree of warpage expected on a test surface of the test object. An illumination optical system that illuminates the object by irradiating the object with diffused light, an imaging optical system that has an imaging element that images the object with specularly reflected light from the object illuminated by the diffused illumination light, Defect detection means for detecting a defect on a test object based on an image picked up by an image pickup device, storage means for storing image data of the test object having a defect-free pattern, and an image pickup optical system. Positioning means for relatively moving the test object and positioning the test object so that a pattern imaged with respect to the imaging optical system has a predetermined positional relationship;
A defect inspection apparatus for detecting a defect on the test object based on a comparison between the image data stored in the storage unit and the image data obtained by imaging the test object. 3. A defect inspection apparatus for inspecting a defect of a test object having a repetitive pattern, wherein a substantially uniform luminance is obtained from an illumination surface having a size determined based on a degree of warpage expected on a test surface of the test object. An illumination optical system that illuminates the object by irradiating the object with diffused light, an imaging optical system that has an imaging element that images the object with specularly reflected light from the object illuminated by the diffused illumination light, Defect detection means for detecting a defect on a test object based on an image picked up by an image pickup device; and a positional relationship between a pattern at a certain first position and a pattern at a second position on the test object with respect to pixels of the image pickup device Moving means for relatively moving the test object with respect to the imaging optical system so that the two are substantially the same, and the defect detecting means is configured to compare at least two pieces of image data imaged by the movement of the moving means. Missing on the test object First defect detection means for detecting a defect;
Storage means for storing, as a reference image, image data of the test object which has been determined to be defect-free by the first defect detection means; and a positioning means for positioning the pattern imaged with respect to the imaging optical system in a predetermined positional relationship. And a second defect detecting means for detecting a defect on the test object based on a comparison between the image data obtained by imaging the test object and the reference image data.