JPH0713598B2 - Defect inspection method for periodic patterns - Google Patents

Description

The present invention relates to a shadow mask used for a cathode ray tube for a color television, a color separation filter for a color image pickup device, a color filter for a liquid crystal display panel, a mesh electrode used for an electron tube, a VDT. Filters, mesh filters for leak-through devices, rotary encoders, linear encoders, photo masks for ICs, Fresnel lenses, lenticular lenses, etc. Units with certain optical properties and shapes (hereafter unit patterns) are one-dimensional or two-dimensional. Industrial products or unit patterns that are repetitively arranged in a
A method of automatically inspecting defects such as scratches, pinholes, black spots, and dust in the arrayed industrial products by repeating the optical property, shape, and array pitch in the one-dimensional direction and the two-dimensional direction while gradually changing. It is about.

[Background of the Invention]

Conventionally, the defect inspection of industrial products as described above is usually performed visually with a naked eye or a microscope,
It takes a lot of manpower to inspect many products,
Further, since it is a sensory test, there is a problem that the test accuracy and reliability are lacking.

In order to solve such a problem, for an industrial product having a periodic pattern of an equal pitch array, a video signal obtained by a microscopic photographing means capable of sufficiently resolving the array unit and the shape of the defect is examined. Pattern recognition
Alternatively, a method of detecting and inspecting a defect by means such as comparing a pattern having no defect with a signal obtained by similarly photographing is proposed, and in some cases, it is implemented.
Such a method requires mechanical accuracy according to the size of a defect to be detected, and a high-precision device is required to detect a minute defect, which makes the device expensive and microscopic. However, since the size of the screen that can be processed at one time becomes small because of the large number of shots, it takes a lot of time to inspect the entire pattern to be inspected.

For products with periodic apertures, such as mesh-shaped electrodes for electron tubes, an optical Fourier transform spatial filtering method that utilizes the diffraction phenomenon of light due to a periodic pattern when coherent light is irradiated has been proposed. In this method, although the inspection speed and the detection sensitivity are excellent, a space filter must be created for each pattern to be inspected, and a precise optical system is required, which makes the apparatus expensive and further Can be detected, but there are many problems such as the inability to determine the magnitude relationship with the reference value of the defect opening, and a new inspection method has been desired.

[Object of the Invention]

An object of the present invention is to solve the above problems, and to provide an inspection method capable of inspecting a periodic pattern efficiently and with high accuracy, and capable of discriminating black and white of a defect.

[Outline of Invention]

In order to achieve this object, according to the present invention, even if the field of view is large, such that the unit pattern of the periodic pattern is not sufficiently resolved or not resolved at all, defect information is included in the obtained video signal. Random noise reduction effect of video signal by screen addition processing, which is included in the image processing method, and almost no change before and after the movement of the video signal pattern of the defect-free part when the pattern to be inspected is moved by a predetermined method. If the image data before and after the movement is subtracted, the change in the video signal due to the unit pattern of the defect-free portion and the change in the video signal consisting of the fixed noise component due to shading of the imaging system and dust in the optical system are erased. Then, it is found that the image data of a portion having no defective portion becomes substantially 0, and that the value locally changes only in the defective portion, so that the defect can be detected. Specifically, according to the present invention, in a method for inspecting a local defect of a periodic pattern composed of a repeating array of unit patterns, the periodic pattern to be inspected is imaged at a predetermined first position. And the second gradation image data obtained by displacing the periodic pattern from the predetermined position in a predetermined direction by a predetermined distance.
Using the second gradation image data obtained by imaging at the position
By subtracting the second tone image data from the first tone image data, tone image data from which the fixed noise component is removed is obtained, and a defect is generated based on the tone image data from which the fixed noise component is removed. The feature is that the inspection process is performed.

Example of Invention

 Hereinafter, the present invention will be described in detail based on Examples.

FIG. 1 shows an embodiment of a method for inspecting a periodic pattern according to the present invention. As shown in FIG.
The method when applied to the example of inspecting the opening area of the pattern having the above as an example is shown. And the TV camera 1 photographs the inspection area. The image processing device 2 performs A / D conversion of the output signal of the TV camera into digital image data, and performs various kinds of image processing including addition and subtraction of the screen at high speed by a frame memory and a computing unit. A control unit 3 controls the image processing apparatus 2, the pattern moving mechanism including the XY stage 10 and the driving unit 5.

Then, the image processing device 2 and the control unit 3 first set the pattern 6 to be inspected at a predetermined first position, process the gradation image data obtained by picking up the image with the TV camera 1 at this time, and It is stored in the memory as 1 gradation image data.

Next, the pattern 6 to be inspected is displaced from the first position in a predetermined direction by a predetermined distance and set to the second position, where the gradation image data obtained by the image pickup by the TV camera 1 is processed, It is stored in the memory as the gradation image data.

FIG. 2 shows an example of the pattern 6 to be inspected, where 11 is a normal unit pattern and 12 and 13 are defective unit patterns.

Next, an operation of detecting a defect of the pattern 6 according to this embodiment will be described. Here, for simplification of description, a shooting condition in which a change in a video signal by the TV camera 1 due to a unit opening is negligible, for example, a shooting condition such that about 10 unit openings 11 are included in a pattern area corresponding to one pixel. The case where the pattern is moved and displaced is the scanning line direction of the TV camera 1 and the displacement distance of the pattern is an integral multiple of the pixel pitch will be described.

First, FIG. 3 (a) is a diagram showing a light transmittance distribution on a straight line passing through a portion having a pattern defect, and 14 is the pattern 13 of FIG.
As shown in, the change in light transmittance due to a defect (hereinafter referred to as a white dot) having an opening area larger than that of the normal pattern 11 is shown.
2 shows a change in light transmittance due to a defect (hereinafter referred to as a black dot) having an opening area smaller than that of the normal pattern 11 as shown by a pattern 12 in FIG.

Next, FIG. 3 (b) is a diagram showing a video signal scanned on the same line as FIG. 3 (a), showing a gradual signal change (shading) due to uneven illumination of the pattern, uneven sensitivity of the image pickup surface, etc. A fixed noise component consisting of a local change 16 and a random noise generated in the video signal processing circuit are shown in the signal due to dust adhering to the system.

Further, FIG. 3 (c) is a diagram showing the result of the screen addition processing by the image processing device 2, and when the ratio of the random noise components in FIG. 3 (b) is N, It shows that it is decreasing to.

Further, FIG. 3 (d) shows the result of the screen addition processing by displacing the pattern. Although the signal due to the defect on the pattern is moving along with the movement of the pattern, erasing of the imaging system and dust of the optical system, etc. It indicates that the position of signal 16 due to is unchanged.

Then, FIG. 3 (e) is obtained by subtracting the image data of FIG. 3 (d) (second gradation image data) from the image data of FIG. 3 (c) (first gradation image data). The results show that the fixed noise components that do not change with the movement of the pattern, such as shading and local changes 16 included in the data of FIGS. Only the signal due to the change in transmittance and the reduced random noise component remain, and as a result, the signal due to the defect has almost the same value difference from the average value in the vicinity of the number of pixels according to the amount of movement of the pattern. Then, the sign appears to be inverted, and it is understood that the order of inversion is reversed depending on the type of defect (white dot, black dot).

The above is an example of subtracting between the two screen image data subjected to the screen addition processing before and after the pattern displacement, but the screen data after movement is added to the addition screen data before movement in the same frame as addition. Even if the number is subtracted, the result is exactly the same, and the frame memory can be processed with only one surface.

Therefore, the image data that has been subjected to the above processing should be displayed on the TV monitor 4.
When observed with, since the brightness of only the defect part locally changes, the defect can be easily recognized, and the defect type (white dot, black dot) in the order of light and dark reversal with respect to the surroundings at the defect part.
Can be identified.

In addition, from this image data, subtraction of the neighborhood average value, or differential processing, especially when the pattern movement is parallel movement, image processing for computing the difference between corresponding pixels before and after one point on the pattern is moved is performed. Then, as shown in FIG. 3 (f), the gradual change component of the image data is removed, and the defect can be automatically detected by comparison with a predetermined threshold value.

By the way, the above describes the method of erasing signals other than the signal due to the defect by adding the image data, moving the pattern, and subtracting. However, in some cases, the change in the video signal due to the unit pattern cannot be ignored depending on the shooting conditions. As described above, when the pixel pitch is moved by an integral multiple, the unit pattern signal remains in the image data after addition and subtraction as shown in FIG. 4B, and it becomes impossible to detect minute defects. Incidentally, FIG. 4 (a) shows the data before the subtraction.

However, in such a case, for example, in the case of a periodic pattern having an equal pitch arrangement, an integer multiple of the arrangement pitch, in the stripe direction in the stripe pattern, in the concentric circle pattern, the unit pattern is rotated, and the unit pattern is rotated to form a circle. In the arrayed pattern, if the rotation of the array center as the center of rotation and an integer multiple of the array angle is set as a condition for each pattern movement, the image data of the defect-free pattern does not change before and after the movement, and , FIG. 5 (a),
As shown in (b), the image data after addition and subtraction gives the same result as that shown in FIG. 3 (e), and it becomes possible to detect minute defects.

Next, the movement distance of the pattern will be described. FIG. 6 is a view for explaining the relationship between the optical images of the defect patterns 17 and 18, the pixel P, and the relationship between the image data D after addition and subtraction, and FIG. 6A shows that the pattern movement amount is an integer multiple of the pixel pitch, FIG. 6 (b) shows an example when it is not an integral multiple.

Then, in FIG. 6 (a), the image data D of the defective portion is vertically symmetrical with respect to the neighborhood average value, but when it is not an integral multiple, image data by a defect image as shown in the example of FIG. 6 (b). D
Since the change ratio of the difference between adjacent pixels is different before and after the pattern movement, the vertical symmetry with respect to the neighborhood average value is lost, which causes an error in the automatic detection process.

Therefore, it is preferable that the pattern movement is such that the positional relationship between the defect image and the pixel receiving the defect image is the same before and after the pattern movement. For example, when moving in the pixel array direction, an integer multiple of the pixel pitch is used. Setting it will give better results. Therefore, if the change in the video signal due to the unit pattern can be ignored, an integer multiple of the pixel pitch,
If it cannot be ignored, if the moving distance is set to the common multiple of the pattern arrangement pixel and the pixel pitch, both conditions of erasing the image data by the defect-free pattern and making the moving distance an integral multiple of the pixel pitch are satisfied. I can do things.

By the way, the displacement of the pattern required in the inspection method of the present invention is based on the indispensable condition of erasing the information of the pattern having no defect from the image data after addition and subtraction, and under this condition, if possible, the defect image. This means that better results should be obtained by matching the positional relationship between the pixel and the pixel that receives it before and after the movement. Therefore, if these conditions are satisfied, the movement direction does not necessarily have to match the pixel arrangement direction.

Next, as an embodiment of the present invention, a method of reducing the fluctuation of the defect signal level caused by the positional relationship between the defect image and the pixel will be described.

FIG. 7A shows a state in which the image of the defect pattern 17 is in the center of one pixel P, and FIG. 7B shows a state in which the image of the defect pattern 17 is on the contact points of the four pixels P. As described above, even if the same defect is generated, most of the signal changes due to the defect are concentrated on one pixel as shown in FIG. 7 (a) and dispersed in four pixels as shown in FIG. 7 (b). The difference between the case and the case is approximately four times as large as the defect signal level, resulting in a problem that the reproducibility of defect detection is low. Therefore, the distribution of such a defect signal to surrounding pixels is 3 × with the defect image as the center.
Since most of it is contained in the region of 3 pixels and the influence on the outside can be ignored, the image after addition and subtraction of a maximum of 3 × 3 pixel neighboring pixel addition processing is performed by the spatial filter processing often used in image processing. When performed on the data, the total of the defect signals dispersed in the peripheral pixels in the defective portion can be obtained, and the change in the defect signal level can be reduced.

Next, as still another embodiment of the present invention, a method of automatically detecting defects and discriminating types will be described.

In FIG. 8 (a), the direction in which the pattern is displaced is the same as the pixel array direction, the movement distance is set to twice the pixel pitch, and after screen addition processing is performed, the pattern is moved to the right side in the figure,
The same number of frames as addition and an example of image data resulting from subtraction are shown, where 20 indicates a white spot defect, 21 indicates a local change in the image data due to a black spot defect, and the other portions are the unit pattern. The change (low frequency change) of the image data according to the gradual change in the array pitch and size is shown.

Here, looking at the image data of the defective portion in detail, the pixel A (22,22 ') corresponding to the defect at the time of screen addition and the pixel B (23,23') corresponding to the time of subtraction are the pixels having the defect information. The displacement distance of the pattern, that is, the number of pixels (two pixels) corresponding to two pixel pitches, appears, and each pixel data
Let I _C be the neighborhood average value of pixel C (24, 24 ′) corresponding to I _A , I _B , and its midpoint, I _A −I _C and I _B −I _C have almost the same absolute value with opposite signs. The values are the same, and it can be seen that the order of this code corresponds to the type of defect (white dot, black dot), and the absolute value of I _A -I _B is proportional to the size of the defect. It can also be seen that the difference between the average value of I _A and I _B and I _C is a sufficiently small ratio with respect to the value of I _A −I _B. Therefore, image data processing such that I _A −I _B is obtained, for example, I ₍ j ₎ = I ₍ j ₋₁₎ −I ₍ for FIG. 8A).
When (j _{+ 1)} is calculated, it becomes as shown in FIG.
Pixels (25,25 ') where ₍ j _-1) and I ₍ j ₊₁₎ are I _A and I _B , respectively
Since the data indicates the type and size of the defect, and the low frequency change is also reduced, the defect can be detected and the type can be determined by comparing with a constant threshold value.

However, at this time, as shown in FIG. 8 (b), the value sign opposite to both sides of the pixel to be detected as a defect I _A -
Because half of the data (26, 26 ') of I _B is generated, it will be thus also detect false defects for defect having a signal level greater than twice the threshold, the detection and the type of It causes inconvenience in judgment.

Therefore, in I _A , I _B , and I _C , (I _A + I _B ) / 2−I _C is I _A −I _B
On the other hand, by utilizing the fact that the ratio becomes a certain ratio or less, for example, with respect to FIG. 8 (a) (However, Is the neighborhood average value of I ₍ j ₎ ) to obtain FIG. 8 (c), which is compared with a predetermined threshold value to detect (I _A + I _B ) / do it 2-I _C are sorted pixels to be the predetermined value or less, the able to exclude false defects, can be stably defect detection and type of determination.

Next, an example of the defect inspection of the periodic pattern according to the above-mentioned embodiment is shown. The unit pattern has a diameter of 100 μm.
A periodic pattern having an array pitch of 300 μm and having an array as shown in FIG. 2 is illuminated by transmitted light as shown in FIG.
A TV camera 1 using an image pickup tube is used to shoot an area of 300 × 220 mm, and the total of the shooting time and the image processing time is about 5 seconds, and the defect in which the aperture diameter of the unit pattern differs by 5 μm is automatically detected.
It was possible to judge the size of the opening with respect to the surroundings. still,
As the TV photographing device and the illumination method used in the present invention, any one can be used as long as a video signal including defect information can be obtained. For example, as the TV photographing device,
Image pickup tube, TV camera using solid-state image pickup device, image pickup device using image dissector or flying spot tube, X-ray
A TV imager, an electron microscope imager, etc. can be used, and as the illumination method, transmitted light illumination, transmitted darkfield illumination,
Reflective dark-field illumination, specular illumination, etc. are used, and further, as the properties of the illumination light, coherence, spectral characteristics, polarization, etc. are detected such that the signal S / N is high due to the defect to be detected or detected. It can be carried out by any one selected so that the signal level due to an unnecessary defect becomes low.

〔The invention's effect〕

As described above, according to the present invention, minute defects of various industrial products having a periodic pattern are photographed with a wide field of view such that the unit pattern forming the periodic pattern is not resolved. By processing the image data,
It is possible to automatically detect and determine the type (white dot, black dot), and it is possible to obtain effects such as improvement of inspection accuracy, reliability, and efficiency.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a method for inspecting a periodic pattern according to the present invention, FIG. 2 is an explanatory view showing an example of a periodic pattern, FIGS. 3 (a) to (f), and FIG. (A),
(B), FIG. 5 (a), (b), FIG. 6 (a),
(B), FIGS. 7 (a), (b), and FIG. 8 (a)-
(C) is explanatory drawing which shows operation | movement of this invention, respectively. 1 ... TV camera, 2 ... image processing device, 3 ... control unit,
4 ... TV monitor, 5 ... Drive mechanism, 6 ... Inspected object, 7
...... Diffuser, 8 ...... Lamp, 9 ...... DC power supply, 10 ...... Stage, 11 …… Unit pattern, 12,13 …… Defect, 14,15…
… Transmission change due to defects, 16 …… Signal change due to dust in optical system, 17 …… Defect image, 27 …… Threshold value, 28 …… Neighborhood average value.

Claims (10)

[Claims] 1. A method for inspecting a local defect of a periodic pattern composed of a repetitive array of unit patterns, wherein an object having a periodic pattern to be inspected is separated from a predetermined first position by a predetermined distance in a predetermined direction. The first gradation image data obtained by displacing the periodic pattern of the object on a frame-by-frame basis at the first position and displacing it at the second position, and the period of imaging at the first position. Second gradation image data obtained by imaging the same periodic pattern as the periodic pattern at the second position in frame units, and using the first gradation image data to the second gradation image data. The gradation image data from which the fixed noise component has been removed is obtained by subtracting, and the defect inspection processing is performed based on the gradation image data from which this fixed noise component has been removed.
A defect of a periodic pattern, wherein the first and second gradation image data are gradation image data obtained by adding gradation data for each pixel data over a plurality of frames obtained by imaging several times. Inspection method. 2. The periodic pattern according to claim 1, wherein at least one of the shape, size, and arrangement pitch of the unit patterns is changed at a predetermined ratio. Inspection method for periodic patterns. 3. The unit pattern according to claim 1 or 2, wherein the imaging conditions of the first and second gradation image data appear in data of each pixel of the captured gradation data. Is determined so that the influence of each change is sufficiently small and can be ignored, the predetermined direction coincides with the arrangement direction of the pixels, and the predetermined distance is an integral multiple of the pixel pitch. A periodic pattern defect inspection method characterized by being configured. 4. The apparatus according to claim 1 or 2, wherein the predetermined direction coincides with the arrangement direction of the unit patterns and the predetermined distance is an integral multiple of the arrangement pitch of the unit patterns. A periodic pattern defect inspection method characterized by being configured. 5. The apparatus according to claim 1 or 2, wherein the predetermined direction coincides with the arrangement direction of the unit patterns, and the predetermined distance is the arrangement pitch of the unit patterns and the gradation image data. A method of inspecting a defect of a periodic pattern, which is configured to be a common multiple of a pixel pitch. 6. The periodic pattern according to claim 1 or 2, wherein the periodic pattern is a stripe pattern, the predetermined direction is an array direction of each stripe of the stripe pattern, and the predetermined direction is the predetermined direction. A defect inspection method for a periodic pattern, wherein the distance is configured to be an integral multiple of the pixel pitch of the gradation image data. 7. The periodic pattern according to claim 1 or 2, wherein the periodic pattern is a concentric circular pattern, and the predetermined direction is a rotational movement direction with a center of the concentric circular pattern as a rotation axis. A defect inspection method for a periodic pattern, comprising: 8. The circle according to claim 1 or 2, wherein the periodic pattern is a pattern in which unit patterns are arranged on a circle, and the predetermined direction is the same as the arrangement direction of the unit patterns. A method for inspecting a defect of a periodic pattern, which is in a circumferential direction and is configured such that the predetermined distance is an integer multiple of an arrangement pitch on a circumference of the unit pattern. 9. A cycle according to claim 1 or 2, wherein the defect detection processing is configured to include addition processing of neighborhood data of 3 × 3 pixels at maximum. Inspection method for defects in the pattern. 10. The defect detection process according to claim 1 or 2, wherein one point on the periodic pattern corresponds to the gradation image data obtained by the subtraction at the time of screen addition. Pixel is A, pixel corresponding to screen subtraction is B,
When the pixel corresponding to the center of the pixels A and B is C, the difference between the average value of the neighborhood of the pixel C and the average value of the data of the pixels A and B is greater than the difference of the data of the pixels A and B. Is sufficiently small and the difference between the data of the pixels A and B is equal to or higher than a predetermined level is determined as a defect, and the type and size of the defect are recognized from the sign and absolute value of the difference between the data of the pixels A and B. A defect inspection method for a periodic pattern, which is characterized in that: