KR100939424B1 - How to inspect defects on glass substrates - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect inspection method of a glass substrate which acquires an image of a defect inspection target region of the glass substrate and checks the degree of change in brightness of the obtained image to detect a defect present in the glass substrate. Acquiring an image at a predetermined interval depth, accumulating the brightness variation between the acquired images, generating a new image, converting the image into a binary image, and analyzing the converted binary image to detect a defect prediction region; Extracting defect information of the defect predicted area based on the contour of the defect predicted area and the position and quantity of pixels constituting the interior of the defect predicted area, and setting short and short axes in the defect predicted area based on the extracted defect information; ; Detecting a plurality of defect long axis outline points and a plurality of defect short edge outline points using defect information, an outline of the defect predicted area, and a long and short axis set in the defect predicted area; By determining the ratio of the connection length between the plurality of defect long axis outer points and the connection length between the plurality of short axis outer points to determine the shape of the defect based on the identified ratio size, even in the case of a thick large glass substrate It is possible to automatically check the existence of the defect, the location, size, shape and type of the defect, there is an effect that can more accurately perform the defect inspection of the glass substrate.

Description

Method for Defect Inspection of Glass Substrate {Method for Detecting Defect in Glass Panel}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect inspection method of a glass substrate, and more particularly, to obtain an image of a defect inspection target region of a glass substrate, and to check a degree of change in brightness of the obtained image to detect defects present in the glass substrate. It relates to a defect inspection method of the substrate.

In general, a visual inspection method, which depends on the experience and visual inspection of the inspector, is widely used as a defect inspection method of the glass substrate in manufacturing a glass substrate, which is a display equipment such as LCD and PDP.

However, the visual inspection method is difficult to achieve the consistency of the test results according to the experience and condition of the inspector, it is difficult to standardize and quantify, and the inspector must check visually, so as the area and thickness of the glass substrate increases, It shows the limitation in the time required.

Therefore, recently, a method for automatically inspecting defects in glass substrates has been increasingly used in combination or selectively with a visual inspection method using a microscope.

In the microscopic inspection method, a plurality of coordinates are set for each specific position of a glass substrate, inputted into a computer, and then a specific position on the glass substrate that passes through the inspection table with a microscope provided in the camera head is enlarged at a predetermined ratio to obtain an image through the camera. Examine the defects of the glass substrate by acquiring and confirming.

However, the microscopic inspection method is also limited due to the thickness of the glass substrate, which is being enlarged in recent years, so that it can be extended to the inner depth of the glass substrate and confirmed with an image. In this case, there is a problem that a defect of the glass substrate cannot be detected.

Furthermore, even if the defects present in the glass substrate are detected, the position of the detected defects cannot be accurately determined, and the size and type of the detected defects cannot be confirmed, so that defect inspection of the glass substrate can be precisely performed. There is a problem that cannot be done.

The present invention has been made to solve the problems described above, the image obtained by the depth of the substrate in the defect inspection target region of the glass substrate, and the brightness of each image obtained by measuring the brightness of each image, the brightness measured for each image Detects the presence and location of defects on the glass substrate based on the degree of change of the glass substrate, and analyzes the brightness distribution of the image acquired at the depth corresponding to the defect presence position to detect the size of the defect and the type of the defect. It is an object of the present invention to provide a defect inspection method of a substrate.

In the defect inspection method of the glass substrate according to an embodiment of the present invention for achieving the above object, the image is obtained by a predetermined interval depth of the glass substrate, and a new image is accumulated by accumulating the amount of brightness change between the obtained images Generating and converting the binary image, and analyzing the converted binary image to detect a defect prediction region; A second extracting defect information of the defect predicted area based on the contour of the defect predicted area and the position and quantity of pixels constituting the inside of the contour, and setting a short and short axis in the defect predicted area based on the extracted defect information; Process; A third step of detecting a plurality of defect long axis outer points and a plurality of defect short axis outer points using defect information, an outline of the defect predicted area, and a long short axis set in the defect predicted area; And a fourth process of determining a shape of a defect based on the identified ratio size by checking a ratio of the connection length between the plurality of defect long axis outer points and the connection length between the plurality of short axis outer points.

According to the defect inspection method of the glass substrate of this invention, an image is acquired by the depth of a board | substrate in the defect inspection object area | region of a glass substrate, and the brightness | luminance of each acquired image is measured, and the degree of the change of the brightness measured for each image is measured. Detects the presence and location of defects on the glass substrate, and analyzes the brightness distribution of the image obtained at the depth corresponding to the defect presence position to detect the size of the defect and the type of the defect. It is possible to automatically check the existence of the defect, the location, size, shape and type of the defect, there is an effect that can more accurately perform the defect inspection of the glass substrate.

Hereinafter, with reference to the accompanying drawings will be described in detail with respect to the defect inspection system and method of the glass substrate according to an embodiment of the present invention.

1 is a view schematically showing a system configuration for defect inspection of a glass substrate according to an embodiment of the present invention.

Referring to FIG. 1, a system configuration for defect inspection of a glass substrate according to an exemplary embodiment of the present invention includes an imaging unit 100, a driving unit 200, and a control unit 300.

The height of the image capturing unit 100 is changed by the driving unit 200, and acquires an image of an image focal plane located at the depth of the glass substrate corresponding to the changed height and transmits the image to the control unit 300. For example, the imaging unit 100 may include an optical system to obtain an optical image of the imaging focal plane.

The driver 200 changes the height of the imaging unit 100 under the control of the controller 300 to vertically change the imaging focal plane of the imaging unit 100.

The controller 300 controls the driving unit 200 to change the height of the imaging focal plane of the imaging unit 100 at regular intervals within the glass substrate 10, while controlling the imaging unit 100 to control the depth of the glass substrate. Shoot the image very much.

For example, the defect inspection system of the glass substrate according to the embodiment of the present invention is configured as shown in Figure 2 to take the image for each depth of the substrate 10. On the other hand, when the defect 20 is located deep in the center of the glass substrate 10, depth-specific images as shown in FIG. 3 are photographed.

Thereafter, the controller 300 calculates a brightness change amount between the captured images and generates a new image by accumulating the calculated brightness change amount. For example, as shown in FIG. 4, the controller 300 accumulates the amount of brightness variation between the images to generate a new image.

In this case, the controller 300 calculates the amount of change in brightness of the entire images photographed for each depth of the glass substrate 10 by adding the absolute values of the brightness differences between the captured images. For example, the controller 300 may calculate an amount of change in brightness of the images photographed for each depth of the glass substrate 10 using Equation 1 below.

('i' and 'j' are pixel positions, 'k' is the image number, 'I' is the brightness of the image, and 'F' is the cumulative value of the amount of change in brightness between the images)

That is, the controller 300 calculates the brightness change amount between the pixels constituting each image for the entire image, and generates a new image having the sum of the calculated brightness change amounts as the brightness of each pixel. In this case, it is preferable that the pixels of the new image appear to have a greater degree of brightness as the cumulative value of the brightness change amount is larger.

In addition, the controller 300 smoothes the new image, converts the brightness of each pixel to a binary image by adjusting the brightness of each pixel to be lighter or darker according to whether the brightness of each pixel constituting the smoothed new image is greater than or equal to the first threshold value, and converts the converted image into a binary image. After that, an area composed of bright pixels is determined as a defect predicted area.

Here, the controller 300 converts the new image into a binary image by displaying brightly when the brightness of the pixel is greater than or equal to the first threshold value and by darkly displaying the brightness of the pixel below the first threshold value.

For example, as illustrated in FIG. 5, the controller 300 binarizes a new image to convert the image into a binary image, and determines the region of bright pixels after the conversion as the defect predicted region 30.

In addition, the controller 300 extracts the outline of the defect prediction region 30 by using a chain code algorithm. Here, the chain code algorithm refers to an implementation of a technique for extracting the contour of a specific region by using a chain coding method.

For example, as illustrated in FIG. 6, the controller 300 may extract the outline of the defect prediction region 30 using a chain code algorithm.

On the other hand, the controller 300 checks the size and location of the defect prediction area 30 by using the quantity and location of the pixels constituting the defect prediction area 30, so that the size of the identified defect prediction area 30 is determined. If it is less than the second threshold value, it is determined that the defect prediction area 30 does not correspond to the defect 20, and even when the position of the defect prediction area 30 is located at the boundary of the image, the defect prediction area 30 It is preferable to determine that it does not correspond to the defect 20, and to exclude it from an inspection object.

In addition, the controller 300 checks the number of the detected defect prediction areas 30, and when the number of the identified defect prediction areas 30 is equal to or less than the reference quantity, the contour of each defect prediction area 30 and the contour of the contour are determined. Defect information of the defect prediction region 30 is extracted based on the position and quantity of pixels constituting the inside.

Here, the defect information of the defect predicted region 30 may include, for example, the position, area, perimeter, center of gravity, maximum boundary region (MBR) of the defect 20, MBR based on the long axis in the rotated state, It is preferable to include the amount of distortion, the length in the statistical long and short axis direction, the length of the long and short axis, and the like.

On the other hand, when the number of the defect | prediction area | region 30 exceeds the reference quantity, the control part 300 determines that it is a total defect of the glass substrate 10, and it is preferable to complete the defect inspection of the glass substrate 10. FIG. .

In addition, the controller 300 sets the short and short axis in the defect predicted area 30 based on information on the outermost area, the length in the statistical long and short axis direction, and the length of the long and short axis among the defect information in the defect predicted area 30. . Here, the long and short axes are preferably set to be orthogonal to each other.

For example, as illustrated in FIG. 7A, the controller 300 sets the short and short axes l and s in the defect predicted region 30.

Thereafter, the control unit 300 intersects the set long axis l and short axis s among the pixels constituting the defect prediction region 30 and is located at the outermost point, that is, the plurality of first long axis axes. The outer points BP _lp and BP _ln and the plurality of first shortest outermost points BP _sp and BP _sn are detected.

For example, as illustrated in FIG. 7B, the plurality of first long axis outermost points are the outermost of the pixels defining the positive (+) direction of the long axis l and the defect prediction region 30. The position of the pixel located at the point (BP _lp ), the negative (-) direction of the major axis (l), and the position of the pixel located at the outermost point of the pixels that outline the defect prediction area 30 (BP _ln ); The plurality of first shortest outermost points are the position (BP _sp ) of the pixel located at the outermost point among the pixels defining the positive (+) direction of the short axis s and the defect prediction region 30, and the short axis (s ) And the position BP _sn of the pixel located at the outermost point among the pixels defining the negative (-) direction and the defect prediction region 30.

Subsequently, the controller 300 searches for an image having the greatest brightness degree of the pixels constituting the defect prediction region 30 among the images acquired through the imaging unit 100, and searches the image for the defect prediction region 30. After setting the outline and setting the defect candidate area including the outline of the set defect prediction area 30, the image of the set defect candidate area is extracted.

For example, as illustrated in FIG. 8, the controller 300 sets the defect candidate region 40 in a quadrangular shape to the image having the greatest brightness degree of the pixels constituting the defect prediction region 30, and sets the defect candidate. The image corresponding to the area 40 is extracted. In this case, the controller 300 uses the height of the imaging unit 100 at the time of capturing an image having the greatest brightness degree of pixels constituting the found defect prediction region 30 among the images acquired through the imaging unit 100. The presence depth of the defect predicted region 30 in the glass substrate 10 can also be measured.

In addition, as shown in FIG. 9, the control unit 300 sets the long and short axes l and s at a position in the defect candidate area 40 that is the same as the position of the short and short axis set in the defect predicted area 30, and is set. The brightness distribution (IP _{l , s} ) of the long and short axes projected on the long and short axes ( _{l , s} ) is measured through the imaging unit 100, and the measured brightness distribution (IP _{l , s} ) for each of the long and short axes is differentiated. A point having a minimum brightness value for each long and short axis, that is, a plurality of second longest outermost points IP _ln and IP _lp and a plurality of second shortest outermost points IPsn and IPsp are detected.

In addition, as shown in FIG. 10, the controller 300 controls the brightness of each of the plurality of axes having a predetermined offset from the long axis set at a position in the defect candidate area 40 that is the same as the position of the long axis set in the defect predicted area 30. The average value of (I _2l , I _3l ) is calculated, and the brightness distribution (ID _l ) on the long axis is calculated by subtracting the brightness (I _1l ) of the long axis from the calculated average of the brightness (I _2l , I _3l ) of each of the plurality of axes. The average value of the brightnesses I _{2 s} and I _{3 s} of each of the plurality of axes having a constant offset from the short axis set at the position in the defect candidate area 40 equal to the position of the short axis set in the defect predicted area 30 is calculated. After calculating the brightness distribution (ID _s ) of the short axis by subtracting the brightness (I _1s ) of the short axis from the average value of the brightness (I _2s , I _3s ) of each of the plurality of axes, the calculated brightness distribution (ID _{l , s} ) Differentiate the outermost points of the 3rd major axis (ID _ln , ID _lp) ) And a plurality of third shortest outermost points ID _sn and ID _sp . For example, the controller 300 may calculate brightness distributions ID _{l and s} on the short and short axes using Equation 2 below.

Subsequently, the controller 300 may include a connection length between a plurality of first longest axis points BP _lp and BP _ln , a connection length between a plurality of second longest axis points IP _lp and IP _ln , and a plurality of third long axis points. By comparing the connection lengths between the outermost points (ID _lp , ID _ln ), the plurality of longest axis outermost points having the largest length are finally determined as the plurality of defect long axis outer points, and the plurality of first shorter axis outermost points BP _sp , BP _sn ), the length of the connection between the plurality of second shortest outermost point (IP _sp , IP _sn ) and the length of the connection between the plurality of third shortest outermost point (ID _sp , ID _sn ) After determining the largest plurality of short axis outermost points as the plurality of defect shortening outer points, the ratio of the connection length between the determined plurality of long axis outer edges and the plurality of defect shortening outer points is checked. Determine the shape of the defect 20 based on do.

That is, the controller 300 may determine the defect 20 when the ratio of the connection length between the plurality of defect long axis outer points and the connection length ratio between the plurality of defect short edges is greater than or equal to a third threshold (eg, 1.5 times). It is preferable to discriminate to an ellipse shape, and to distinguish the defect 20 to circular shape, when it is below a 3rd threshold value.

In addition, the controller 300 calculates a long-axis defect type determination value D _l by taking an absolute value between the difference between the brightness I _1l of the long axis and the brightness distribution IP _l projected on the long axis, and calculates the brightness I of the short axis. _1s ) and an absolute value of the difference between the brightness distribution (IP _s ) projected on the short axis is calculated to calculate the short-term defect type determination value (D _s ), and then the calculated defect type discrimination value (D _{l, s} ) for each short and short axis The type of the defect 20 is determined. For example, the controller 300 calculates a defect type determination value using Equation 3 below.

11 is a view sequentially showing a defect inspection method of a glass substrate according to an embodiment of the present invention.

According to FIG. 11, the defect inspection method of the glass substrate which concerns on one Example of this invention is as follows.

First, the controller 300 controls the driving unit 200 and the imaging unit 100 to obtain images at predetermined intervals in the central portion of the glass substrate 10, and calculates and calculates the brightness variation between the acquired images. A new image is generated by accumulating the changed amount of brightness (S500).

In step S500, the controller 300 may calculate an amount of change in brightness of the images photographed for each depth of the glass substrate 10 by adding the absolute values of the brightness differences between the acquired images.

It is preferable that the pixels of the new image generated through the above-described step S500 appear to have a greater degree of brightness as the cumulative value of the brightness change amount is larger.

Thereafter, the controller 300 smoothes the new image generated in step S500 by 3 S3, and brightens or darkens the brightness of each pixel according to whether the brightness of each pixel constituting the smoothed new image is greater than or equal to the first threshold value. By adjusting, the image is converted into a binary image, and the area of bright pixels after the conversion is determined as the defect predicted area 30 (S510).

In operation S510, the controller 300 may display a bright image when the brightness of the pixel is greater than or equal to the first threshold value and display the image dark when the pixel brightness is less than the first threshold value, thereby converting the new image into a binary image.

Subsequently, the controller 300 extracts the outline of the defect predicted area 30 determined through the above-described step S510 using a chain code algorithm (S520), and then outlines the extracted defect predicted area 30. By determining the size and position of the defect prediction region 30 using the quantity and position of the pixels to determine whether or not the defect 20 in the defect prediction region 30 (S530), if not the defect 20 While the defect predicted area 30 is excluded from the inspection target (S540), the number of the defect predicted areas 30 determined as the defect 20 is checked, and the number of the defect predicted areas 30 is equal to or less than the reference quantity. Check (S550).

In step S530, when the size of the defect prediction area 30 is less than the second threshold value, the controller 300 determines that the defect prediction area 30 does not correspond to the defect 20, and the defect prediction area 30 is determined. It is preferable to determine that the defect predicted area 30 does not correspond to the defect 20 even when the position of) is located at the boundary of the image.

Subsequently, when the number of the defect prediction areas 30 identified through the above-described step S550 is less than or equal to the reference quantity, the controller 300 positions and quantities of the contours of the respective defect prediction areas 30 and the pixels constituting the interior of the contour. Based on this, defect information of the defect predicted area 30 is extracted (S560).

The defect information extracted through the above step S560 is, for example, the position, area, circumference, center of gravity, maximum boundary region (MBR) of the defect 20, the MBR based on the long axis in the rotated state, and the distortion. It is preferable to include the amount, the length in the statistical long and short axis direction, the length of the long and short axis and the like.

On the other hand, the control unit 300 determines that the total defect of the glass substrate 10 when the number of the defect expected regions 30 confirmed through the above-described step S550 exceeds the reference quantity, the defect of the substrate 10. It is preferable to end the inspection.

Subsequently, the controller 300 determines the outermost region, the length in the direction of the statistical long and short axes l and s, and the length of the long and short axes l and s, among the defect information of the defect predicted region 30 extracted through the above-described step S560. The long and short axes (l, s) are set in the defect predicted area 30 based on the information on (S570), and the long and short axes set through the above-described step S570 among the pixels constituting the defect predicted area 30. The position point of the pixel located at the outermost point while intersecting with, is detected, that is, the plurality of first longest axis outermost points (BP _lp , BP _ln ) and the plurality of first shortest axis outermost points (BP _sp , BP _sn ) (S580). ).

It is preferable that the long and short axes l and s set through the step S570 are set to be orthogonal to each other.

Specifically, the plurality of first longest axis outermost points (BP _lp , BP _ln ) detected through step S580 intersects with the positive (p) direction p of the long axis l set through step S570. The position (BP _lp ) of the pixel located at the outermost point among the pixels constituting the defect prediction region 30 and the negative (-) direction (n) of the long axis (l) of the defect prediction region 30 A position BP _ln of the pixel located at the outermost point among the contoured pixels, and the plurality of first shortest outermost points BP _sp and BP _sn are positive (+) directions of the short axis s (p). ) and the position of the pixel located in the outermost point of the cross-pixels constituting the contour of the defect estimated area 30 in which (BP _sp), shortening (s) tone (a-) fault estimated area intersecting the direction (n) ( 30 includes the position BP _sn of the pixel located at the outermost point among the pixels defining the outline.

Subsequently, the controller 300 searches for an image having the largest brightness degree of pixels constituting the defect prediction region 30 among the images acquired through step S500, and predicts a defect extracted through the S520 in the retrieved image. After setting the outline of the area 30, setting the defect candidate area 40 including the outline of the set defect prediction area 30, and extracting an image corresponding to the set defect candidate area 40 (S590). .

Thereafter, the controller 300 sets the long and short axes l and s at a position in the defect candidate region 40 extracted by the step S590, which is the same as the position of the long and short axis set through step S580 (S600). , The brightness distribution (IP _{l , s} ) of the long and short axis (l, s) projected on the set long and short axis (l, s) is measured through the imaging unit 100, and the measured long and short axis (l, s) Differentiate the brightness distribution (IP _{l , s} ), respectively, to the point having the minimum brightness value for each long and short axis (l, s), that is, the plurality of second longest axis outermost points (IP _ln , IP _lp ) and the plurality of second shortest axes The outer point (IP _sn , IP _sp ) is detected (S610).

Meanwhile, the controller 300 calculates an average value of the brightnesses I _2l and I _{3l of the} plurality of axes having a predetermined offset from the long axis l set through the above-described step S600, and calculates the brightness I of the plurality of axes. _2l , I _3l ) by subtracting the brightness (I _1l ) of the long axis to calculate the brightness distribution (ID _l ) on the long axis, and the brightness (I _2s , I _3s ) of each of the plurality of axes having a constant offset from the short axis (s) Calculate the brightness distribution (ID _s ) of the short axis by subtracting the brightness (I _{1 s} ) of the short axis from the calculated average values of the brightness (I _{2 s} , I _{3 s} ) of each of the plurality of axes. The brightness distributions ID _{l and s} on the short axis are differentiated to detect a plurality of third longest axis outermost points ID _ln and ID _lp and a plurality of third shorter axis outermost points ID _sn and ID _sp (S620). .

Thereafter, the controller 300 connects the lengths of the plurality of first long axis outermost points BP _lp and BP _ln detected through step S580 and the plurality of second long axis outermost points detected through step S610. A plurality of outermost points having the longest length by comparing the connection length between (IP _ln , IP _lp ) and the connection lengths between the plurality of third longest axis IDs _ln , ID _lp detected through step S620 described above. Is finally determined as a plurality of defect long axis outer points, and the connection length between the plurality of first short axis outermost points BP _sp and BP _sn detected through step S580 and a plurality of detected through step S610. The connection length between the second shortest outermost point IP _sn , IP _sp is compared with the connection length between the plurality of third shortest outermost points ID _sn , ID _sp detected through the above step S620. Ends multiple long outermost points with multiple fault shortening outer points After determining (S630), the ratio of the determined connection length between the plurality of defect long axis outer points and the connection length between the plurality of short axis outer points is checked to determine the shape of the defect 20 based on the determined ratio size (S640).

In step S640, the controller 300 determines that the ratio of the connection length between the plurality of defect long axis outer points and the connection length between the plurality of defect short edges is greater than or equal to a third threshold (for example, 1.5 times). It is preferable to discriminate the 20 in an elliptic shape and to determine the defect 20 in a circular shape when it is less than the second threshold.

On the other hand, the controller 300 takes the absolute value of the difference between the brightness I _1l of the long axis in step S620 and the brightness distribution IP _l projected on the long axis in step S610, and determines the long-term defect type determination value ( D ₁ ) is calculated, and an absolute value is taken as the difference between the brightness I _{1 s} of the short axis in step S620 and the brightness distribution IP _s projected on the short axis in step S610, and the short axis defect type determination value ( After calculating D _s ), the type of the defect 20 is determined using the calculated short-term short-term defect type determination values D _{l, s} (S650).

In step S650, the control unit 300 uses the change amount according to the long axis of the long axis defect type determination value D _l as shown in FIG. 12 to determine the maximum value P _l of the long axis defect type determination value D _l . ) for calculating and longitudinal defect type determination value (D _l) by integration of the graph formed by calculating a long-axis defect type determination area of the value amount of change (area _l), the calculated area if ever (area _l) finalized The average area (A _l ) is calculated by dividing by the connection length (N _l ) between a plurality of defect long axis edges, and the shortened defect type discrimination value (D _s ) using the change amount according to the shortening of the shortened defect type discrimination value (D _s) ) the maximum value (P _s), the calculation, and the speed and calculating a defect type determination value (D _s) the integration by reducing the defect area of the determination value variation graph (area _s) formed by the calculated area (area _s of ) the connection length between the end-determined plurality of defect shortened outer point (divide by N _s) Mean area (A _s) and then calculates the, chapter the calculated speed defect type determination value (D _{l, s)} maximum values (P _{l, s)} is the greater than the third threshold value, or the average area of the calculation of (A _{l , s} ) is larger than the fourth threshold value, and is determined as the bubble defect 20, and the maximum value P _{l , s of the} short-term short-term defect type determination value D _{l, s} is smaller than or equal to the third threshold value. If the averaged area _{Al and s} is smaller than the fourth threshold value, it is preferable to discriminate the defects 20 other than bubbles.

According to the defect inspection method of the glass substrate of this invention, an image is acquired by the depth of a board | substrate in the defect inspection object area | region of a glass substrate, and the brightness | luminance of each acquired image is measured, and the degree of the change of the brightness measured for each image is measured. Detects the presence and location of defects on the glass substrate, and analyzes the brightness distribution of the image obtained at the depth corresponding to the defect presence position to detect the size of the defect and the type of the defect. The presence, defect location, size, shape and type of defects can be automatically checked, and defect inspection of the glass substrate can be performed more precisely.

1 is a view schematically showing a system configuration for defect inspection of a glass substrate according to an embodiment of the present invention.

FIG. 2 exemplarily shows a defect inspection system of a glass substrate in FIG. 1; FIG.

FIG. 3 exemplarily shows images taken by depths of a glass substrate for defect inspection of the glass substrate of FIG. 1.

FIG. 4 exemplarily illustrates a new image generated by accumulating an amount of change in brightness between images in FIG. 3; FIG.

FIG. 5 exemplarily shows a binary image generated by binarizing a new image in FIG. 4; FIG.

6 is a diagram exemplarily illustrating a contour of a defect prediction region extracted by using a chain code algorithm according to the present invention;

FIG. 7 is a diagram exemplarily illustrating a position point of a pixel located at an outermost point while intersecting with a set long axis and a short axis among pixels that form a long and short axis set in a defect predicted area according to the present invention; FIG.

8 is a diagram illustrating a defect candidate area set including the outline of a defect predicted area according to the present invention.

9 is a diagram showing the brightness distribution projected on the long and short axis set in the defect candidate area according to the present invention.

FIG. 10 is a diagram illustrating a brightness distribution of long and short axes set in a defect candidate area and a brightness distribution of axes having a predetermined offset in FIG. 9. FIG.

11 is a view sequentially showing a defect inspection method of a glass substrate according to an embodiment of the present invention.

12 is a diagram showing the amount of change in long-term short-term defect type determination value according to the present invention illustratively.

*** Explanation of symbols for the main parts of the drawing ***

10: glass substrate 20: defect

30: defect prediction area 40: defect candidate area

100: imaging unit 200: driving unit

300: control unit

Claims (8)

A first process of acquiring an image at a predetermined interval depth of the glass substrate, accumulating a change in brightness between the acquired images, generating a new image, converting the image into a binary image, and analyzing the converted binary image to detect a defect prediction region; ; A second extracting defect information of the defect predicted area based on the contour of the defect predicted area and the position and quantity of pixels constituting the inside of the contour, and setting a short and short axis in the defect predicted area based on the extracted defect information; Process; A third step of detecting a plurality of defect long axis outer points and a plurality of defect short axis outer points using defect information, an outline of the defect predicted area, and a long short axis set in the defect predicted area; And a fourth process of determining a shape of a defect based on the identified ratio size by checking a ratio of the connection length between the plurality of defect long axis outer points and the connection length between the plurality of short axis outer points. Method of defect inspection of the board. The method of claim 1, The first process, Smoothing the new image and adjusting the brightness of each pixel to be a binary image according to whether the brightness of each pixel constituting the smoothed new image is equal to or greater than a first threshold value; And detecting a region consisting of bright pixels after the binary image conversion as a defect predicted region. The method of claim 1, The second process, Extracting a contour of the detected defect predicted area using a chain code algorithm; Determining whether the defect prediction region is defective by checking the size and position of the defect prediction region using the quantity and the position of the pixels constituting the defect prediction region; If the defect prediction region does not correspond to a defect as a result of determining whether the defect is present, excluding the defect prediction region from an inspection object; Determining whether the number of the defect prediction areas is equal to or less than a reference quantity, when the defect prediction area corresponds to a defect; Extracting the defect information of the defect prediction area based on the contour of each defect prediction area and the position and quantity of pixels constituting the interior of the contour if the number of defect prediction areas is equal to or less than a reference quantity; And determining the total defect of the glass substrate when the number of the defect predicted regions exceeds the reference quantity, and ending the defect inspection. The method of claim 3, Determining whether the defect is, When the size of the defect prediction area is less than the second threshold value and when the position of the defect prediction area is located at the boundary of the image, the defect prediction area is determined to not correspond to a defect. How to check for defects. The method of claim 1, The third process, Detecting a plurality of first long axis outermost points and a plurality of first short axis outermost points that intersect the short and short axes set in the defect predicted area among the pixels constituting the defect predicted area; Among the obtained glass substrate depth-specific images, an outline of the defect predicted region is set to an image having the greatest brightness degree of pixels constituting the defect predicted region, and a defect candidate region including the outline of the set defect predicted region is selected. Setting up; A plurality of second longest axis outermost points having a minimum brightness value among the brightness distributions of the long and short axes projected to the set long and short axes, the long and short axes being set at the same position as the long and short axes set in the defect predicted area within the defect candidate area; Detecting a plurality of second shortest outermost points; The brightness distribution on the long axis is calculated by subtracting the brightness of the long axis from the average value of the brightness of the plurality of axes having a constant offset from the long axis set in the defect candidate region, and the short axis is obtained from the average value of the brightness of the plurality of axes having the constant offset from the short axis. Calculating the brightness distribution on the short axis by subtracting the brightness, and then detecting the plurality of third longest outermost points and the plurality of third shortest outermost points using the calculated brightness distribution on the long and short axes; A plurality of outermost points having the longest length by comparing the connection length between the plurality of first longest outermost points, the connection length between the plurality of second longest outermost points and the connection length between the plurality of third longest outer points Final determining a plurality of defect long axis edges; A plurality of outermost edges having the longest length by comparing a connection length between the plurality of first shortest outermost points, a connection length between the plurality of second shortest outermost points and a connection length between the plurality of third shortest outermost points Finally determining a point as a plurality of defect shortening outlines; Determining a shape of a defect on the basis of the identified ratio size by checking a ratio of the connection length between the plurality of defect long axis outer points and the connection length between the plurality of short axis outer points. How to check for defects. The method according to any one of claims 1, 3 and 5, The defect information, Information about the location, area, perimeter, center of gravity, maximum boundary region (MBR) of the defect, the MBR relative to the long axis in the rotated state, the amount of distortion, the length in the statistical short and short axis direction, and the length of the short and short axis. Defect inspection method of the glass substrate comprising a. The method of claim 5, Calculating a long-axis defect type discrimination value by taking an absolute value of a difference between the brightness of the long axis and the brightness distribution projected on the long axis; Calculating an uniaxial defect type determination value by taking an absolute value between the brightness of the short axis and the brightness distribution projected on the short axis; And determining the type of the defect by using the calculated long axis and short axis defect type discrimination values. The method of claim 7, wherein The step of determining the type of the defect, Calculating a maximum value of the long-axis defect type determination value by using the amount of change along the long axis of the long-axis defect type determination value; Calculating an area of the long-axis defect type determination value change amount by integrating a graph formed by the long-axis defect type determination value; Calculating an average area by dividing the calculated area by a connection length between a plurality of final determined long axis edges; Calculating a maximum value of the shortened defect type determination value by using the amount of change according to the shortening of the shortened defect type determination value; Integrating a graph formed by the shortened defect type determination value to calculate an area of the shortened defect determination value change amount; Calculating an average area by dividing by a connection length between the plurality of determined short short edges; Determining the bubble defect if the maximum value of the calculated short-term defect type is greater than a third threshold value or the calculated average area is greater than a fourth threshold value; And determining the defect as a non-bubble if the maximum value of the short-term short-term defect type determination value is smaller than the third threshold value or the calculated average area is smaller than the fourth threshold value. How to check for defects.

Images