JP2003270087A - Defect inspecting method and apparatus by image processing - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that specific patterns such as each kind of reference mark and test patterns are present on an electrode that is formed on a front plate and a rear plate in addition to a pattern section required for driving a PDP, and this sort of specific pattern being present in an inspection region is detected as 'a defect' in defect inspections and tact in a later review process increases unwillingly, and at the same time, needs to be judged to be non-defective, and reviewing a minute, common defect that may or may not be detected as a defect regardless of 'defect' in terms of resolution when imaging by a CCD camera or a detection limit size by algorithm, and at the same time, reviewing those that are judged to be common defects each time cause the tact to increase. <P>SOLUTION: After a failed portion was detected, a failure judgment step is provided before the review. Then, in the step, the combination between the central coordinates of the failed section and failed elements is compared with that between the central coordinates of the preset defective section and/or a non-defective section for excluding from review targets by a defect-inspecting method and a defect-inspecting apparatus. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plasma display panel member, a plasma display panel defect inspection method, and a defect inspection apparatus.

[0002]

2. Description of the Related Art First, the structure of a general AC type (AC type) plasma display panel (PDP) will be described. The PDP is provided with a pair of regularly arranged electrodes on two opposing glass substrates, and Ne, X
It has a structure in which a gas mainly containing e, He, etc. is enclosed. Then, a voltage is applied between these electrodes to generate a discharge in the minute cells around the electrodes so that each cell emits light to perform display.

The two glass substrates arranged facing each other are called the front plate on the display surface side and the rear plate on the non-display surface side. On the front plate, a composite electrode consisting of a sustain electrode and a bus electrode is arranged in parallel with each other on a glass substrate, a dielectric layer is formed so as to cover it, and a protective layer made of magnesium oxide is formed on the entire surface. There is. On the other hand, on the rear plate, address electrodes are arranged in parallel to each other on a glass substrate, a rear surface dielectric layer is formed so as to cover the address electrodes, and barrier ribs and phosphor layers are formed thereon.

In these manufacturing processes, defects may occur in the formed constituent layers. Regarding electrodes, ribs, and phosphor layers having a pattern shape, pattern thickening or pattern thinning, adjacent electrodes or ribs may be connected by a pattern other than the set pattern, or defects such as chipped or broken patterns may occur. Defects such as holes (pinholes) or cracks may occur in the solid dielectric layer and protective layer.

These defects cause defects such as non-lighting, a decrease in brightness, or constant lighting when the panel is turned on. Therefore, a defect inspection may be performed at the time of forming each constituent layer and correction may be performed to eliminate the defect. Is needed.

As a method of performing defect inspection, visual inspection is difficult because of problems such as the size, variety of defects, and the number of defects. Therefore, a defect inspection device (also called appearance inspection device) is used instead. It has become common to use. The basic principle of this defect inspection apparatus is to illuminate the formation surface, capture an image with a CCD camera, and detect the defect by image processing of the captured image by a specific algorithm. For image pickup, a line sensor or area sensor including a CCD is used, and luminance information is obtained in the case of a monochrome CCD, and luminance information and chromaticity information is obtained in the case of a color CCD. There are several algorithms used for image processing, such as an adjacency comparison method for comparing information between pixels and D for performing pattern expansion / compression.
There are an RC method, a skeleton method for converting a pattern into a line, a line width comparison method for comparing line widths, a trend sensing method for observing a variation in an integrated value between consecutive pixels, and a periodic correlation / autocorrelation method.

In order to smoothly carry out the subsequent defect correction and defect cause investigation, it is necessary to classify the defects according to types after detecting the defects. At this time, a review process is performed as a series of processes in which a person inspects the detected defects in more detail, classifies the defects, determines the necessity of repair, and determines the quality of the substrate. It is common. In recent years, proposals have been made to incorporate a function (Automatic Defect Classification, ADC) capable of automatically classifying defects into a review device or a defect inspection device.

[0008]

On the other hand, the electrodes formed on the front plate and the back plate are often only regular periodic patterns in the display area, but the electrode terminal portion which is the connection portion with the drive circuit is often used. In the non-display area around that, including the pattern portion necessary for driving the PDP, various reference marks and test patterns, which are not necessary for driving the PDP but are necessary for assembling and confirming the characteristics of the PDP. The pattern exists. Such a peculiar pattern is detected as a "defect" by the conventional inspection method, and it is necessary to increase the tact in a later review process and determine that it is not a defect. In addition, there is a problem that such a peculiar pattern is likely to be erroneously determined as a “defect” in an ADC that will be advanced in the future.

Further, due to the resolution of the image picked up by the CCD camera and the detection limit size of the algorithm, there are cases where the defect cannot be detected although it is a "defect". In particular, the pattern breakage of the electrode cannot be repaired after the dielectric layer is formed on the electrode, and PD
When assembled in P, non-lighting defects continue on the line,
Even if the defect size is minute, it becomes a fatal defect. Not only the electrodes but also micro defects often occur when fine foreign matters such as fibers adhere to the screen printing plate at the time of screen printing during the PDP manufacturing process, resulting in printing gaps, or photo processes. This is a case where a fine foreign substance such as a fiber adheres to the inside of the exposure mask to cause exposure blanking. Since these cases occur as common defects on many substrates, it is a big problem to send them to the subsequent process without repairing them. In addition to the small common defects, reviewing the ones that are known to be common defects every time is a cause of increasing the tact.

Therefore, the present invention has been made in view of the above-mentioned problems in the prior art, and its first object is to:
It is to provide a defect inspection method and a defect inspection apparatus which exclude these peculiar patterns from the review target by determining that the peculiar pattern is “not a defect” after the defect detection and before the review. A second object of the present invention is a defect inspection method and a defect inspection method in which a common defect is determined as a prescribed “defect” after a defect detection operation and before a review or by adding defect data, thereby excluding it from a review target. To provide a device.

[0011]

In order to solve the above problems, the present invention has the following constitution. That is, the present invention is
A method for performing a defect inspection by image processing on an image containing an inspected portion, wherein the image processing detects an abnormal portion of the inspected portion, and at least the center of the abnormal portion is detected when the abnormal portion is detected. Abnormal part detection step of outputting coordinates and abnormal element, combination of central coordinates of abnormal part and abnormal element output in abnormal part detection step, and preset central coordinates of defective part and / or non-defective part It is a defect inspection method by image processing characterized by having an abnormality determination step whose determination criterion is to compare a combination of a and an abnormal element, and a defect inspection apparatus for carrying out the method.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION The defect inspection method and the defect inspection apparatus of the present invention are not particularly limited in use, but are preferably used for defect inspection of members for plasma display panels and plasma display panels (PDP).

First, the defect inspection method of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an implementation flow of a defect inspection method of the present invention, and FIG. 2 is a diagram showing an implementation flow of a conventional defect inspection method. The defect inspection method of the present invention is characterized in that an abnormality determination step 2 is provided between the abnormal portion detection step 1 and the review step 3 which are conventional defect inspection methods.

First, the abnormal portion detecting step 1 will be described. In the abnormal portion detection step 1, the surface to be inspected is illuminated, the image is picked up by the CCD camera, and the picked-up image is subjected to image processing by a specific algorithm, which is a portion which is determined to be abnormal or has a different shape, that is, a defect. If the abnormal part that is a candidate is detected and the abnormal part is detected,
As the information regarding the abnormal portion, at least the center coordinates of the abnormal portion and the abnormal element are output.

A line sensor or area sensor composed of a CCD is used for image pickup. In the case of a monochrome CCD, gray scale luminance information is used, and in the case of a color CCD, light such as red, green, and blue is used. The luminance information of each primary color is obtained.

As the illumination, white light such as a halogen lamp or a metal halide lamp is used. Further, when a UV lamp is used for inspection of the phosphor-formed substrate of the PDP, the formed phosphor can be brought into a state of emitting light. Therefore, by using a color CCD for imaging, it is more practical as a PDP. It is possible to carry out the inspection in a close form.

Here, as an image processing method, for example, an adjacency comparison method for comparing information between pixels and a DR for performing pattern expansion / compression as described in Japanese Patent Laid-Open No. 10-219164.
The C method, a skeleton method for converting a pattern into a line, a line width comparison method for comparing line widths, a trend sensing method for observing a variation in an integrated value between consecutive pixels, a periodic correlation / autocorrelation method, and the like can be given. In particular, the DRC method or the skeleton method is preferable for an aperiodic regular pattern such as an electrode terminal portion.

Here, a method of detecting an abnormal portion will be described by taking the adjacent comparison method as an example. First, mutually parallel image data P1 to P6 arranged at equal intervals and at equal pitch corresponding to the pattern to be inspected as shown in FIG. 3 are obtained. Next, among the respective image data, image data obtained from two patterns adjacent to each other are compared, and one image data is subtracted from the other image data to obtain composite data. For example, the composite data D12 obtained from the image data P1 and the image data P2 that are adjacent to each other includes the image data P1 and the image data P2.
Since and have exactly the same shape, no residual data is generated.

However, if there is a pattern defect C as in the image data P3, the combined data D23 and the combined data D34
Remaining data NC are generated at and. Also, the image data P5
If there is the pattern protrusion T as described above, the residual data NT is generated in each of the combined data D45 and the combined data D56. Then, when the residual data NC or the residual data NT is generated, it is detected that an abnormal portion is present in the pattern to be inspected.

Further, in FIG. 3, each pattern to be inspected is a periodic pattern, but a plurality of patterns are set even if each pattern has a different width and a different pitch, for example, three patterns corresponding to RGB are set. Even when a periodic pattern is formed, by treating a region having periodicity as one image data,
An abnormal part can be detected by using the adjacent comparison method.

On the other hand, in the adjacent comparison method, the boundary portion between two periodic patterns as shown in FIG. 4 is detected as an abnormal portion. When such two periodic patterns are designed as normal states, this boundary portion becomes a point that is desired to be excluded from the review target as a peculiar pattern that is not a defect, and is reviewed in the abnormality determination step 2 described later. Take measures to exclude from the target.

Next, a method of detecting an abnormal portion of the skeleton method, which is preferably used for an aperiodic regular pattern such as an electrode terminal portion, will be described. Image data P1 to P6 corresponding to the pattern to be inspected as shown in FIG. 5 are obtained. Next, the pattern width in each image data is multiplied by a predetermined number, for example, 0.
5 times thinning is repeated a predetermined number of times, for example 10 times, to obtain one line data L1 to L6.

At this time, even if the line is bent, a predetermined angle,
For example, an allowable width of ± 45 degrees is set, and the end points that are the line ends of the respective line data are also obtained. In FIG. 5, the end points are indicated by black circles. When there is no abnormality like the image data P4 and P6, the line becomes one line having end points only at both ends of the image data like the line data L4 and L6.

However, if there is a pattern break O as in the image data P3, two lines having the end points close to each other are generated in the line data L3. In addition, the image data P1 and P
When there is a connection S with an adjacent pattern or a pattern protrusion T like 2 and P5, line intersections occur like line data L1, L2 and L5. Then, when an intersection of lines or two lines having end points close to each other are generated, it is detected that an abnormal portion exists in the pattern to be inspected.

In the surrounding non-display area including the electrode terminal portion which is the connection portion with the drive circuit, in addition to the pattern portion necessary for driving the PDP, FIG. 6A and FIG.
There are cases in which there are peculiar patterns such as various reference marks and test patterns as shown in FIG. 4 which are not necessary for driving the PDP but are necessary for assembling the PDP and confirming the characteristics.
Here, FIG. 6A is a diagram showing a normal state of a peculiar pattern (bonding with an adjacent pattern), and FIG. 6B is a defect (pattern break) in a part of the peculiar pattern (bonding with an adjacent pattern). It is a figure showing. Further, FIG. 7A is a diagram showing a normal state of a peculiar pattern (island abnormality between patterns), and FIG. 7B is a defect (adjacent pattern with adjacent pattern) in a peculiar pattern (island abnormality between patterns) part. FIG.

In both the adjacent comparison method and the skeleton method, these peculiar patterns are detected as abnormal parts. These peculiar patterns are also points that are desired to be excluded from the review target as "not a defect", and a measure to exclude them from the review target is taken in the abnormality determination step 2 described later.

The information on the abnormal part obtained by the present invention is as follows:
This information is used as a judgment material in the subsequent abnormality judgment step 2 and review step 3, and as such information,
Examples include central coordinates, abnormal elements, size, and brightness.

Here, the central coordinates indicate the position of the abnormal portion in the inspection area, and the coordinate system thereof is the machine coordinate system, which is generally called a machine coordinate system, and the reference mark in the substrate is the origin. And the coordinate system of the substrate such as using the substrate end point as the origin. In order to improve the determination accuracy in the subsequent abnormality determination step 2, the mechanical coordinate system including the substrate fixed position accuracy error in the device is used. Is preferably the coordinate system of the substrate, and more preferably the reference mark in the substrate is the origin.

The abnormal element indicates an abnormal state, and its name and division boundary vary depending on the image processing method. However, as shown in FIG. This includes binding with adjacent patterns and island-shaped abnormalities between the patterns. In FIG. 8, the PDP electrode pattern when the monochrome image is binarized with a certain threshold value in the luminance gradation is described as an example. here,
The black part shows the electrode pattern, (a) is in a normal state, (b) is a pattern missing, (c) is a pattern cut,
(D) shows the pattern protrusion, (e) shows the coupling with the adjacent pattern, and (f) shows the island-like abnormality between the patterns.

Further, it is also possible to perform a process in which abnormal parts having the same abnormal element that are adjacent to each other are combined and treated as one abnormal part. In this case, it is preferable to handle information about the abnormal part such as the center coordinates and the abnormal element with respect to the final combined abnormal part.

In general, this step is often referred to as a defect detection step. However, as described above, the point detected by the image processing is not always a defect, so in the present invention, it is an actual point. In order to distinguish between a defect and a defect candidate, the former is called a defective part and the latter is called an abnormal part to avoid confusion.

Next, the abnormality determination step 2 will be described. The abnormality determination step 2 checks whether or not the combination of the center coordinates of the abnormal portion and the abnormal element output in the abnormal portion detection step 1 is the same as a preset combination, and the subsequent review This is a step in which the abnormal part (review target) for which the review is executed is sharply output in step 3. Here, the preset combination is a combination of a center coordinate of a unique pattern (which is not a defect) and an abnormal element that is in the inspection region and may be output in the abnormal portion detection step 1 (type A), or , A combination of the center coordinates of points that are output as “abnormal” and an abnormal element (type B), or “defect” even though the defect is not a defect that may occur depending on the method of detecting an abnormal portion. I know that
A combination of the center coordinates of points to be excluded from the review target and the abnormal element (type C) can be cited.

Here, the reason why the combination of the center coordinates of the abnormal portion and the abnormal element is to be compared is that, for example, a pattern which is a different abnormal element is cut off in a "non-defect" peculiar pattern portion which is combined with an adjacent pattern. This is because there may be a "defect" when there is a defect, or there may be a "defect" when there is a combination of patterns that are different abnormal elements in a "non-defect" peculiar pattern portion where the pattern is broken. Therefore, the determination of only the center coordinates is insufficient. (FIGS. 6 and 7) In this step, the combination of the center coordinates of the abnormal part and the abnormal element output in the abnormal part detection step 1 is set to a preset type A or type B.
If it matches with, the review may not be executed in the review step 3 and may be excluded so as not to be output in this step, or a flag (information) indicating “not a defect”.
May be added and output. Further, when the combination of the center coordinates of the abnormal portion and the abnormal element output in the abnormal portion detection step 1 matches the type C, a flag (information) indicating "defect" is added and output. Can be achieved with. Further, when the type C does not exist in the combination group of the center coordinates of the abnormal part and the abnormal element output in the abnormal part detection step 1, the information on the abnormal part of the type C that does not exist is displayed. Achieve by adding and outputting. When adding and outputting this flag (information), a third flag such as “defect determination undecided” is added in order to unify the information series, if the combination does not match the preset combination. Good to do.

As the flow of this step, as shown in FIG. 9, as the first step, the information group regarding the abnormal portion output in the abnormal portion detecting step 1 and the information group regarding the abnormal portion which is "defective" are shown. And the information group about the abnormal part for which these logical sums are taken as the second step is compared with a preset information group about the abnormal part. It is advisable to exclude them from the review target by performing the above processing, and if they do not match, perform the above-described operations and the like and output them as the review target to the review step 3.

The center coordinates of the abnormal portion output in the abnormal portion detection step 1 include a difference in image processing accuracy and a difference in image capturing accuracy by the CCD camera, which may deviate from the center coordinate of the substrate reference. It is preferable to make a comparison using a numerical value on the order of sub-millimeter because the determination accuracy in this abnormality determination step can be improved. For example, 345.
It is to compare with the numerical value of 678 mm as 345.7 mm rounded to one decimal place.

Further, in order to improve the determination accuracy in this abnormality determination step, it is advisable to provide a comparison allowable width with respect to the center coordinates. If the comparison allowable width is too small, it is erroneously determined as “not matching”, and if the comparison allowable width is too large, erroneously determining that the abnormal parts of the same abnormal element that are adjacent are also “matching”.
The comparison allowance is the point or peculiar pattern you want to judge,
The distance from the adjacent pattern is preferably 0.5 to 1.0 times, more preferably 0.6 to 0.8 times. For example,
The distance between a certain peculiar pattern and the adjacent pattern is 100
If it is 0 μm, the comparison allowable width is ± 500 μm to ± 1.
000 μm is preferable, and more preferably ± 600 μm
± 800 μm.

Further, the review step 3 will be described. The review step 3 receives the information on the abnormal portion output from the abnormality determination step 2, and further inspects the abnormal portion determined to be the review target in the abnormality determination step 2 to determine whether or not it is a defect. This is a determination step. Furthermore, the defects may be classified, the necessity of repair, and the quality of the substrate may be determined. As a detailed inspection means, even if a person observes and inspects, the ADC
May be used.

The image used for the review may be an image containing the abnormal portion imaged in the abnormal portion detection step 1 or may be an image in which the abnormal portion is imaged again for this step.

It is preferable that the resolution is higher than that of the CCD camera used in the abnormal portion detection step 1 because a finer shape can be determined. In addition, a color image is preferable because information about colors can be obtained, which enables more detailed classification.

This step may be performed by the defect inspection apparatus, or the review function may be independently provided as a separate apparatus. In the latter case, the abnormal portion detection step 1, the abnormal portion determination step 2 and the review step 3 can be executed independently, so that the total inspection tact can be shortened.

[0041]

According to the defect inspection method of the present invention, the abnormality determination step 2 is provided between the abnormal portion detection step 1 and the review step 3 which are conventional defect inspection methods.
A unique pattern that is not a defect in the inspection area that has been the subject of conventional review, or a point that is detected as "abnormal" despite "not a defect" is determined as "not a defect" in the abnormality determination step 2. Since it is possible to reduce the number of review targets in the subsequent review step 3, it is possible to reduce the takt time in the review step 3. Further, even in an ADC which will be advanced in the future, it is not necessary to judge such a peculiar pattern as a defect, and thus it is possible to eliminate the possibility of being erroneously judged as a “defect”.

Further, in the abnormality judgment step 2, a point to be excluded from the review object which is known to be a "defect" such as a common defect is judged as a prescribed "defect" prior to the review step,
Alternatively, since defect data can be added, the number of review targets in the subsequent review step 3 can be reduced, and the tact of the review step 3 can be reduced.

[Brief description of drawings]

FIG. 1 is a flow chart showing an embodiment of a defect inspection method of the present invention.

FIG. 2 is a flowchart showing a conventional defect inspection method.

FIG. 3 is a pattern diagram illustrating an adjacent comparison method.

FIG. 4 is a diagram showing an example of a peculiar pattern in which an abnormal portion is detected in the adjacent comparison method.

FIG. 5 is a pattern diagram illustrating a skeleton method.

FIG. 6 is a diagram showing an example of a unique pattern.

FIG. 7 is a diagram showing an example of a peculiar pattern different from FIG.

FIG. 8 is a diagram showing an example of an abnormal element.

FIG. 9 is a flowchart showing an embodiment of the abnormality determination step 2 of the present invention.

[Explanation of symbols]

1 Unique pattern part 2 defects P1 to P6 image data D12-D56 synthetic data L1 to L6 line data C pattern missing T pattern protrusion O pattern out S Combining with adjacent pattern NC residual data NT residual data

   ─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 2G051 AA73 AB02 AB03 AB04 AC04                       CA03 CA04 DA15 EA12 EA16                       EA17 EB09 EC01 ED11 ED15                       ED21 ED23                 2G086 EE12                 5B057 AA02 DA03 DB02 DB06 DB09                       DC03 DC06 DC22 DC32                 5C012 AA09 BE03                 5C040 FA01 FA04 GB03 GB14 JA26

Claims (6)

[Claims] 1. A method of performing defect inspection by image processing on an image containing a portion to be inspected, the image processing comprising:
An abnormal part detection step of detecting an abnormal part of the inspected part and outputting at least the center coordinates of the abnormal part and the abnormal element when the abnormal part is detected, and the center coordinates of the abnormal part output in the abnormal part detection step And an abnormal element, and a defect characterized by having an abnormality determination step whose determination criterion is to compare the preset center coordinates of the defective portion and / or the non-defective portion with the combination of the abnormal element. Inspection method. 2. A defect inspection method for a member for a plasma display panel, which uses the defect inspection method according to claim 1. 3. A defect inspection method for a plasma display panel, wherein the defect inspection method by image processing according to claim 1 is used. 4. An apparatus for performing a defect inspection by image processing on an image containing an inspected portion, wherein the image processing means detects an abnormal portion of the inspected portion, and detects the abnormal portion. A defect inspection apparatus comprising at least a means for extracting a center coordinate and an abnormal element, and an abnormality judging means using a combination of these as a judgment criterion. 5. A defect inspection apparatus for a member for a plasma display panel, which uses the defect inspection apparatus according to claim 4. 6. A defect inspection apparatus for a plasma display panel, which uses the defect inspection apparatus according to claim 4.