KR101046106B1 - How to shape resin stain defect on color filter - Google Patents

Abstract

The present invention relates to a color filter resin smear defect shaping method, comprising: a spectroscopic camera for showing a resin smear defect line region of a color filter as a spectral image, an optical cable for transmitting an illumination light source, and a surface of the color filter connected to the optical cable. It is characterized by using an optical system of a line inspection method comprising a halogen line illumination member for light illumination, and fluorescent line illumination provided under the LCD glass on which the color filter is placed.

According to this, an optical system that can be inspected in the form of a line rather than the point area inspection can reduce the time and cost of defect inspection.

Color filter, spectroscopy, lighting, line, resin, shaping

Description

Configuration method for resin spot defects of color filter

 The present invention relates to a color filter resin spot defect shaping method, and more particularly, to a color filter resin spot defect shaping method using a line inspection method optical system that has improved the point inspection method optical system.

The display market is showing a decrease in production of cathode ray tubes (CRT) and an increase in flat panel display (FPD). In particular, the market size of TFT-LCD is increasing, which is the leading display device. .

   Accordingly, research on flat panel display equipment has been actively conducted in recent years, and research on the design and manufacture of AOI (Automated Optical Inspect) equipment, which is the core of display device inspection equipment, has been actively conducted, and applied to actual lines. It is becoming. Among them, a novel method for mark recognition of LCD panels, a method of effectively inspecting a TFT-LCD surface by combining a line scan camera and an area camera, and a method for spot detection are mainly used.

   The color filter to be inspected in the present invention is an optical component of a TFT-LCD in the form of a thin film that extracts three colors of red, green, and blue from white light and displays them in fine pixel units. The color filter has a glass substrate, a black matrix (BM), red, green, blue, a protective layer, and an ITO layer. 1 is a structure of a general color filter.

  The resin unevenness defect which occurs in a color filter arises in the form of a line because it is not apply | coated uniformly when apply | coating a color to a glass substrate. The characteristic of the resin stain defect can be detected by applying an electrode to the substrate. However, this method is expensive and thus requires inspection to detect the electrode without the electrode. As a device for detecting without an electrode, there is a resin staining device released in Japan, and since the detection rate is only 50%, the need for inspection equipment having a detection rate of 90% or more is increasing.

  Conventional color filter systems have been developed for the inspection equipment using a point spectrometer. However, in this case, there is a disadvantage that the inspection time is remarkably slow because the inspectable area is very small, which is fatal to mass production performance. In order to compensate for this, the present invention proposes a method of shaping a resin stain defect by extending an inspection region to a line by using a spectroscopic camera combining a point spectrometer and a line camera.

In general spectroscopic cameras, the refraction of the prism is different for each wavelength of light, and when the light entering one side comes out to the other side, the angle of declination varies according to the wavelength range. Using this principle, the light from the lens is spectroscopically acquired by the camera to acquire an image. The pixel value in the image represents the intensity for each wavelength region.

   The existing optical system structure was constructed using a point spectrometer as shown in FIG. The lighting used is halogen lighting, which is used because the intensity of light at the visible wavelength is stronger than that of LED lighting.

   In the operation sequence, when light is emitted from halogen light, light passes through a color filter, a test sample, and light is incident on a light receiving unit. The incident light is input to the spectrometer through the optical fiber. In the spectrometer, the input light is separated into a specific wavelength band through the diffraction defect and then the separated light is input to output the data.

   However, the conventional optical system has a drawback in that it takes a long time to inspect a point area, which has a problem in mass productivity.

   An object of the present invention for solving the above-mentioned conventional problems is to reduce the time and cost of defect inspection defects by using an optical system that can be inspected in the form of lines rather than point area inspection.

The optical system used in the resin spot defect shaping method of the color filter according to the present invention proposed in order to achieve the above technical problem is a spectroscopic camera for showing the resin spot defect line region of the color filter as a spectral image; An optical cable for transmitting an illumination light source; A halogen line illumination member connected to the optical cable for linearly illuminating a color filter surface; Fluorescent line illumination provided under the LCD glass on which the color filter is placed; And a controller for storing and managing input data of the spectroscopic camera and the halogen line lighting member and photographed output data of the resin line region of the color filter.

In addition, the optical system used in the method for shaping the resin spot defect of the color filter according to the present invention comprises a spectroscopic camera for showing the resin spot defect line region of the color filter as a spectral image; An optical cable for transmitting an illumination light source; A halogen line illumination member connected to the optical cable for linearly illuminating a color filter surface; And a controller for storing and managing the input data of the spectroscopic camera and the halogen line lighting member and the photographed output data of the resin line region of the color filter, wherein the coaxial white LED line lighting is located under the LCD glass on which the color filter is placed. Any one selected from among green LED lights and red LED lights, or a combination thereof; It is characterized by the addition.

In order to achieve the above technical problem, the resin spot defect shaping method according to the present invention includes an optical system calibration process for performing a spectral axis, a spatial axis, a focus, and an axis correction to analyze accurate light characteristics; Preprocessing to scale image data to minimize illumination unevenness, to use low pass filter to minimize distortion caused by high frequency components, and to normalize and compare color data of spectral region to increase contrast when detecting color difference Difference and max-difference methods that use the process and the comparator baseline data as a result of the difference between the color data of the illumination image and the inspection area color data, and the Kull Back Leibler Distance (KLD) method to indicate the difference between the two data distributions, And a resin stain defect shaping step of shaping a resin stain defect using a Fourier transform method that takes a result value using data of the resin stain defect region and an absolute value of the one-dimensional Fourier transform value in the normal region, respectively. Characterized in that it comprises a.

As described above, the method for shaping the resin stain defect of the color filter according to the present invention is to reduce the time and cost of inspecting the resin stain defect on the surface of the color filter by enabling inspection in the form of lines rather than point area inspection.

 Hereinafter, various embodiments of the present disclosure will be described in detail with reference to the accompanying drawings, photos, graphs, and equations.

3 is a block diagram of an optical system using a spectroscopic camera according to the present invention.

As shown, the configuration of the optical system 100 used in the present invention includes a color filter 200, an LCD glass 300 for arranging and arranging them for inspection of the color filter 200, and a resin stain of the color filter. A spectroscopic camera 110 for showing a defect line region as a spectral image, an optical cable 120 for transmitting an illumination light source generated from the light source generator 190, and an optical cable 120 connected to the color filter ( 200, a halogen line illumination member 130 for illuminating a line light on the surface, a fluorescent line illumination 140 illuminating under the LCD glass 300 on which the color filter 200 is placed, the spectroscopic camera 110 and The control unit 150 stores and manages input data of the halogen line lighting member 130 and photographed output data of the resin line region of the color filter and controls the prism and the spectroscope. In this case, a line reference plate 180 having a straight line shape may be further configured at the alignment position of the color filter 200 on the upper surface of the LCD glass 300. The spectrometer here refers to the spectrophotometer 100.
At this time, the optical system 100 is coaxial white LED line light 131, green LED light 132, red LED under the LCD glass 300, the color filter 200 is placed instead of the fluorescent line light 140 Any one of the lights 133 selected or a combination of these lights may be used.

In this case, the spectroscopic camera 110 and the halogen line lighting member 130 to maintain an appropriate angle as necessary, which will be described in various embodiments below.

Illumination uniformity and spectral characteristics are very important in optical system 100. Ideally, the illumination uniformity would be uniform at all positions and the spectral characteristics would be uniform in light intensity at all wavelengths. 7A to 7C and 8A to 8C are illumination uniformity and spectral characteristics of the optical system according to the present invention proposed in FIGS. 4 to 6, respectively. 7A to 7C are graphs showing illumination uniformity of the optical system according to the present invention, and FIGS. 8A to 8C are graphs showing spectral characteristics of the optical system according to the present invention.

First, FIG. 4 is a first embodiment of an optical system 100 according to the present invention. Here, since the optical system 100 uses the fluorescent line illumination 140, the illumination uniformity is uniform in the center area, but the uniformity decreases toward the end. The spectral characteristics show a general fluorescence illumination distribution.
In the embodiment of FIG. 4, the spectroscopic camera 110 maintains 90 degrees with the horizontal axis of the LCD glass 300, and the halogen line lighting member 130 maintains 30 to 60 degrees with the horizontal axis of the LCD glass 300. It is desirable to.

The optical system 100 of FIG. 5 shows a second embodiment. Here, a coaxial white LED line light 131 was used. The uniformity of the coaxial white LED line light 131 has a problem in that the uniformity is lower than that of a general fluorescent light. Spectral characteristics show white LED line illumination characteristics.

Finally, the optical system 100 of FIG. 6 is a third embodiment of the optical system 100 according to the present invention. The uniformity is similar to that of the optical system 100 of FIG. 5, but the spectral characteristics are the sum of the three spectral characteristics because green 132, red 133, and white LED illumination 131 are used.
5 and 6, the spectroscopic camera 110 maintains 70 to 85 degrees with the horizontal axis of the LCD glass 300, and the halogen line lighting member 130 is 40 with the horizontal axis of the LCD glass 300. It is desirable to maintain to 85 degrees.
Resin spot defect shaping method of the color filter of the present invention using the optical system 100 includes an optical system calibration process for performing the spectral axis, spatial axis, focus, axis correction in order to analyze accurate light characteristics; Scale image data to minimize illumination unevenness, use low pass filter to minimize distortion caused by high frequency components, normalize color data of spectral region to increase contrast when detecting color difference Pretreatment process to analyze; And Difference and max-difference methods that compare the comparative analysis reference data as the difference between the color data of the illumination image and the inspection area color data, and the KLD (Kullback leibler distance) method and color filter to indicate the difference between the two data distributions. A resin stain defect shaping process of shaping a resin stain defect using a Fourier transform method that takes a result value using data in the resin stain defect region and an absolute value of the one-dimensional Fourier transform value in the normal region; Characterized in that consists of.

In order to accurately characterize the light, it is necessary to calibrate the exact location, angle, and spectroscopy. In the present invention, a calibration method for spectral and spatial domains is presented. Hereinafter, the calibration method will be described in detail.

First, the spectral axis calibration requires that the input image represents some wavelength of light and that the resolution of the frequency per pixel must be analyzed and corrected accordingly. A reference light source is required for spectroscopic calibration.

9 is a graph comparing the data extracted from the spectroscopic camera 110 data sheet and the spectroscopic camera 110 and the data extracted from the spectroscope. As a result of the analysis, the data extracted from the spectroscopic camera 110 was compared with the data measured by the fitting data and the data measured by the spectrophotometer. However, when comparing with the data sheet of the spectroscopic camera 110, it can be seen that the other data is the same.

Spatial axis calibration requires accurate calibration such as shooting angle, shooting position, etc. in order to obtain high quality images. Accordingly, a three-step method for calibrating the spatial axis is proposed.

Focus correction uses a thin lens formula to find the focal length when calculating the general focal length. However, the spectrophotometer 110 is not applicable to the thin lens formula because the prism is attached to the front end. In the present invention, a focal length was experimentally obtained and applied. Black and white sample is obtained by the spectroscopic camera 110. The acquired image is as shown in FIG. 10A. The line profile data is extracted from the acquired image in the direction perpendicular to the boundary of the color. After the extracted differential data were obtained for each region, it was determined that the focal length was optimal when the derivative value was the maximum value as shown in FIG. 10B.

And when the camera inspects the object to be inspected, the X- and Z-axis should be calibrated because it must be fixed in the direction perpendicular to the test direction. The calibration method uses the specimen shown in Figure 11a for the X-axis calibration and Figure 11b for the Z-axis calibration.

If the X axis is not inspected along the center line of the specimen, an image as shown in FIG. 12A is acquired, and an image as shown in FIG. 12B is obtained when the correct direction is taken.

For Z-axis calibration, use the specimen as shown in Figure 11b. The distance between the slits in the image is measured to determine the Z-axis angle that represents the most accurate distance.

Next, the following process of the resin spot defect shaping | molding method proposed by this invention is demonstrated in detail. The accompanying drawings are as follows to support this. 12A and 12B are images of equally spaced specimens, FIGS. 13A and 13B are images for comparison before and after applying LPF (low pass filter), FIG. 14 is a test diagram for each algorithm in the first embodiment of FIG. 4, 15A to 15D are images of 2DdiffLPF application results according to the wavelength band in the test of FIG. 14, FIG. 16 is a test diagram for each algorithm in the second embodiment of FIG. 5, and FIGS. 17A to 19D are corresponding to wavelength bands in the test of FIG. 16. 2DdiffLPF application result image, FIG. 20 is a test diagram for each algorithm in the third embodiment of FIG. 6, FIGS. 21A to 21D are 2DdiffLPF application result images according to wavelength bands in the test of FIG. 20, and FIG. 22 is a first image proposed by the present invention. It is a graph which shows the algorithm test result in a 3rd Example.

In order to inspect a large area of glass, the present invention proposes a method of spectroscopic analysis of a target by a line scan method. By using this method, it is possible to expect a shorter inspection time and a reliable inspection for defects that are difficult to visually inspect. That is, the spectral image of the target is acquired using the spectral camera 110 showing the line region of the target as a spectral image and then applied to various algorithms. The result data was recombined to obtain a result image in the form of a 2D image. As a result, the generated image is a 2D area of the scan area, and the displayed image can be generated to have good visibility even for a small color difference that is difficult to see visually. The present invention largely includes two types of pretreatment and shaping.

First, the pretreatment process will be described. The pretreatment process used in the present invention used a method for removing illumination unevenness, a method for removing color filter patterns, and finally a method for normalizing data on color. The illumination nonuniformity elimination method removes the nonuniformity by applying a scale method to remove the illumination nonuniformity because some illumination nonuniformity exists in any illumination. The color filter pattern removal method used a low pass filter to remove the pattern. Finally, the normalization method for color was used to increase the contrast for the color difference.

In addition, when applying the method of inspection by the brightness difference, many parts may vary depending on the illumination used. Therefore, the effects such as unbalance of lighting should be minimized when acquiring images. In order to minimize the unevenness of lighting, a method of scaling image data was used. First, obtain a spectroscopic image of the lighting used, apply projection in the direction of the spatial axis of the image, and then divide the data by the number of pixels on the spatial axis. Finally, the multiplication operation on the input image can minimize the unevenness of lighting.

In addition, when the color filter 200 is photographed using the spectroscopic camera 110, high frequency components appear by the pattern of the color filter 200 as shown in FIG. 13A. Since the influence on the high frequency component has a great influence in inspecting the defects of the resin stain of the color filter having a very small difference, the low pass filter is applied before all inspection to minimize distortion as shown in FIG. 13B.

As a method for increasing the contrast when detecting the color difference, a method of normalizing color data of a spectroscopic image and comparing and analyzing it was used. The application method applies projection in the direction of the spectral axis from the input spectral image, and then divides the total of all projection data by the projection data to obtain NF (normalize factor). When the obtained NF is divided by the original values of the line, when there is no defect, all data values of the line are similarly distributed. However, in the defective area, the data values show a difference and can be recognized as bad. Unlike the scaling method used for the brightness difference analysis, this method is applied when analyzing the color difference.

Subsequently, in order to detect invisible resin stain defects, the shaping process must be performed after the pretreatment process. In the present invention, the four proposed methods were applied to shape the resin stain defects.

First, the applied principle of difference and max-difference refers to a method of taking the reference data as a result of the difference between the color data of the illumination image and the inspection area color data in the comparative analysis. The difference method is a method of obtaining an average after adding all the difference values between two data as in Equation 1, and max-difference is a method of taking the largest value of the difference between two data as in Equation 2 as a result.

(수학식 1)

(Equation 1)

: reference color data N : number of data

: reference color data N: number of data

: object color data

: object color data

(수학식 2)

(Equation 2)

: reference color data N : number of data

: reference color data N: number of data

: object color data

: object color data

The second KLD method is the most widely used method to indicate the degree of difference between two data distributions. The larger the value obtained by applying this method, the more different the distribution is between the two data. The smaller the value, the more similar the distribution. In the present invention, this method was used for color difference analysis. If the PDF (probability density function) of the arbitrary data A is P (x), and the PDF of the arbitrary data B is Q (x), KLD can be expressed as Equation (3).

(수학식 3)

(Equation 3)

Finally, the proposed inspection method uses a single line image of the image for defect analysis. One-line image data at this time can be a one-dimensional Fourier transform. Therefore, the data in the resin uneven defect area and the one-dimensional Fourier transform value in the normal area can be obtained. The difference between these two data was used to shape the resin stain defects. If the value obtained after one-dimensional Fourier transform of any one line data in the normal region is called RFD, and the one-dimensional Fourier transform value of the same line of the inspection region is called TFD, the spectral values of RFD and TFD are expressed as in Equation 4. Can be represented.

(수학식 4)

(Equation 4)

R: real number  I: imaginary number

: normal data

: prosecuting attorney data

: normal data

: Prosecuting attorney data

Subsequently, the RFD and the TFD obtained from Equation 4 are obtained by taking an absolute value as shown in Equation 5 and taking the result.

(수학식 5) real data

(Equation 5)

Hereinafter, an optimal embodiment of the resin stain defect shaping process of the present invention will be described.

In order to evaluate the method proposed in the present invention, the experiment was performed on resin stain defects that cannot be identified in real time. And through the experiments to determine the inspection possibility by shaping the resin stain defects. In the experimental environment, a test bed was made in the dark room to calibrate the frame rate and the moving speed of the spectrophotometer. At this time, the test bed moving speed was fixed at 2cm / s, FPS was fixed at 10, the number of shooting line was determined to be 500 lines.

Table 1. Algorithm Notation Description

sign processing description S preprocessing Image scale to improve lighting unevenness (when checking brightness difference) N Normalize data (when color difference is checked) LPF Low pass filter KLD algorithm Kullback leibler distance Maxdiff Determine the maximum value among the data obtained after finding the difference of the comparison data Diff Mean of the comparison data difference FT How to use the difference between FFT values of 1D spectral data in the normal region and the inspection region

Table 1 describes the notation of the algorithm used in the present invention. For example, NLPFMaxdiff preprocesses the low-pass filter, then normalizes it, and then applies the Max-difference algorithm.
Table 2. Algorithm Types and Descriptions Applied

No. algorithm description One 2DSLPF low pass filter
-scale
intensity difference 2 2DNdiffLPF -normalize
low pass filter
spectrum difference 3 2DdiffLPF low pass filter
spectrum difference 4 2DSdiffLPF -scale
low pass filter
spectrum difference 5 2DMaxdiffLPF low pass filter
spectrum max difference 6 2DNMaxdiffLPF low pass filter
-normalize
spectrum max difference 7 2DSMaxdiffLPF low pass filter
-normalize
spectrum max difference 8 2DKLDLPF low pass filter
KLD algorithm 9 2DNKLDLPF low pass filter
-normalize
KLD algorithm 10 2DSKLDLPF low pass filter
-scale
KLD algorithm 11 2DFTLPF low pass filter
FFT difference 12 2DNFTLPF low pass filter
-normalize
FFT difference 13 2DSFTLPF low pass filter
-scale
FFT difference

delete

In the present invention, experiments were conducted on three optical systems used, and it was determined whether the detection was possible by applying the proposed shaping algorithm as shown in Table 2 in the acquired image.

Among them, the first experimental environment was tested for 20 color filters 200 based on the first optical system using fluorescent and halogen lighting. After the image acquisition, the color filter 200 was formed by applying the 13 algorithms described in Table 2.

14 shows the results of experiments on 20 resin uneven defect samples. As a result of the experiment, when 2DdiffLPF is applied, it is possible to shape about 8 out of 20, so it showed maximum 40% of the detectability.

Next, the test area was tested according to the wavelength range. The inspection area was divided into 300 nm to 400 nm, 400 to 500 nm, 500 nm to 600 nm, and 600 nm to 700 nm, and the 2DdiffLPF algorithm, which is easy to detect in the above experiment, was applied. 15A to 15D show the result of applying the 2DdiffLPF algorithm for each wavelength region.

According to the detection result according to the wavelength region, it was easier to detect resin stain defects in the wavelength region of 600 nm to 700 nm than other wavelength regions.

The second experimental environment was tested in the same way as the first experimental method based on the second optical system using coaxial white LED line illumination 131 and halogen illumination. In this case, as shown in FIG. 16, when three algorithms of 2DdiffLPF, 2DSMaxdiffLPF, and 2DNFTLPF were applied, about eight out of twenty shapes could be formed, and thus showed easy detection of up to 40%.

17A to 19D show results of applying three algorithms that are easy to detect for each wavelength region. The area drawn in the square in the resulting image is a resin stain defect. Among these, FIGS. 17A to 17D show the results of applying 2DdiffLPF, which is best shaped in the region of 500 nm to 600 nm. 18A to 18D show the results of the application of 2DSMaxdiffLPF, which is best shaped in the region of 500 nm to 600 nm. Finally, FIGS. 19A to 19D are images obtained by applying 2DNFTLPF, and are well formed at 500 nm to 600 nm as in FIGS. 17A to 18D.

The final experimental environment was identical to the above experiment based on a third optical system using coaxial white LED line light 131, halogen light, red LED light 133 and green LED light 132.

As shown in FIG. 20, when the 2DdiffLPF algorithm was applied, about 17 out of 20 shapes were possible, and thus showed easy detection of up to 85%.

21A to 21D show results of dividing a wavelength region after applying a 2DdiffLPF algorithm that is easy to detect. As a result of experiment by wavelength region, shaping was the easiest in 600nm ~ 700nm.

A summary of the data experimented in the above three optical systems shows the result as shown in FIG.

As shown, it can be seen that resin stain defects are easily shaped when 2DdiffLPF is applied in three environments. In addition, the ease of detection shows that inspection is more easy when the third optical system is applied than the first and second optical systems. Easily detectable by wavelength range, it is well formed from 600nm to 700nm in the first and third optical systems and from 500nm to 600nm in the second optical system. In other words, in terms of ease of detection, it was confirmed through experiments that shaping is best performed using a wavelength of 600 nm to 700 nm even in the third optical system environment.

As described above, although the present invention has been described by way of limited embodiments and drawings, the present invention is not limited thereto and is intended by those skilled in the art to which the present invention pertains. Of course, various modifications and variations are possible within the scope of equivalents of the claims to be described.

1 is a block diagram showing a general color filter structure.

2 is an optical system configuration diagram using a conventional point spectrometer.

3 is a block diagram of an optical system using a spectroscopic camera according to the present invention.

4 shows a first embodiment of an optical system according to the invention.

5 shows a second embodiment of an optical system according to the invention.

6 shows a third embodiment of an optical system according to the invention.

7a to 7c are graphs showing the illumination uniformity of the optical system according to the invention.

8a to 8c are graphs showing the spectral characteristics of the optical system according to the invention.

9 is a graph of spectral data using several measurement methods.

10A and 10B are graphs illustrating a specimen image and a data analysis method for focus correction, respectively.

11A and 11B are diagrams of specimens used for axis correction.

12A and 12B are images of the specimens taken at equal intervals.

13A and 13B are images for comparison before and after LPF application, respectively.

FIG. 14 is a test diagram for each algorithm in the first embodiment of FIG. 4; FIG.

15A to 15D are images of 2DdiffLPF application results according to wavelength bands in the test of FIG. 14.

FIG. 16 is a test diagram for each algorithm in the second embodiment of FIG. 5; FIG.

17A to 19D are images of 2DdiffLPF application results according to wavelength bands in the test of FIG. 16.

20 is a test diagram for each algorithm in the third embodiment of FIG. 6;

21A to 21D are images of 2DdiffLPF application results according to wavelength bands in the test of FIG. 20.

22 is a graph showing algorithm test results in the first to third embodiments proposed in the present invention.

<Description of Symbols for Main Parts of Drawings>

100: optical system 110: spectroscopic camera

120: optical cable 130: halogen line lighting member

131: coaxial white LED line light 132: green LED light

133: red LED light 140: fluorescent line light

150: control unit 180: line reference plate

190: light source generator 200: color filter
300: LCD glass

Claims (3)

Optical system calibration process that performs spectral, spatial, focus, and axis calibration to accurately characterize light; Scale image data to minimize illumination unevenness, use low pass filter to minimize distortion caused by high frequency components, normalize color data of spectral region to increase contrast when detecting color difference Pretreatment process to analyze; And Difference and max-difference methods that compare the comparative analysis reference data as the difference between the color data of the illumination image and the inspection area color data; A resin stain defect shaping process for shaping a resin stain defect using a Fourier transform method that takes a result value using data of the spot defect region and an absolute value of the one-dimensional Fourier transform value in the normal region; It consists of, The optical system includes a spectroscopic camera for showing a spectral image of a resin stain defect line region of a color filter; An optical cable for transmitting an illumination light source; A halogen line illumination member connected to the optical cable for linearly illuminating a color filter surface; Fluorescent line illumination provided under the LCD glass on which the color filter is placed; And a controller for storing and managing input data of the spectroscopic camera and the halogen line lighting member and photographed output data of the resin line region of the color filter and controlling a prism and a spectroscope. Wherein the spectroscopic camera maintains 90 degrees with the LCD glass horizontal axis, and the halogen line illumination member maintains 30 to 60 degrees with the LCD glass horizontal axis. . Optical system calibration process that performs spectral, spatial, focus, and axis calibration to accurately characterize light; Scale image data to minimize illumination unevenness, use low pass filter to minimize distortion caused by high frequency components, normalize color data of spectral region to increase contrast when detecting color difference Pretreatment process to analyze; And Difference and max-difference methods that compare the comparative analysis reference data as the difference between the color data of the illumination image and the inspection area color data; A resin stain defect shaping process for shaping a resin stain defect using a Fourier transform method that takes a result value using data of the spot defect region and an absolute value of the one-dimensional Fourier transform value in the normal region; It consists of, The optical system includes a spectroscopic camera for showing a spectral image of a resin stain defect line region of a color filter; An optical cable for transmitting an illumination light source; A halogen line illumination member connected to the optical cable for linearly illuminating a color filter surface; A controller for storing and managing input data of the spectroscopic camera and the halogen line lighting member and photographed output data of the resin line region of the color filter and controlling a prism and a spectroscope; And a coaxial white LED line light, a green LED light, a red LED light, or any combination thereof installed below the LCD glass on which the color filter is placed. In this case, the spectroscopic camera maintains 70 to 85 degrees with the LCD glass horizontal axis, and the halogen line lighting member maintains 40 to 85 degrees with the LCD glass horizontal axis. Shaping method. delete

Images