JP2017156517A - Defect mode classification system of display device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a defect mode classification system that allows an efficient pixel (defect) repair work to be implemented.SOLUTION: A defect mode classification system classifies a pixel defect mode of a color filter and thin film transistor and is for repairing defects. The defect mode classification system comprises at least: a defect inspection machine for acquiring a size and position information of the defect; a pixel repairing device that is provided with imaging means capable of photographing the defect under a different photographing condition and means for measuring a height of the defect; a mode classifier for identifying a mode of the defect; and a communication instrument that can transfer acquisition data on the detect inspection machine and acquisition data on the pixel repairing device to the mode classifier.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a pixel correction device for a display device and its application, and more particularly to a defect mode classification system.

  The liquid crystal display device generally includes a color filter (CF) and a thin film transistor (TFT) disposed between two transparent substrates.

  In the manufacturing process of a color filter substrate for a liquid crystal display device, defects in the color filter substrate are corrected in order to improve the yield (Patent Documents 1 to 10). Generally, in a color filter for a liquid crystal display device provided on a counter substrate, a black matrix, a colored layer, a transparent conductive layer, a spacer, and an alignment control protrusion are formed on a glass substrate in a plurality of steps. In a color filter manufacturing process, one of typical defects is a defect to which “foreign matter” adheres. There are various types of foreign substances such as a cured piece of resist material, a metal piece caused by wear and deterioration, and dust. When a defect occurs in an intermediate process of manufacturing a color filter, there is a defect correction method in which a defective portion is removed by applying a correction ink corresponding to each constituent layer and curing after removing the defective portion with a laser. Further, as a method for correcting a defect having a color filter height, there is a technique called “tape polishing”. For example, as disclosed in Patent Document 1, tape polishing is a correction method in which protrusions are removed with a polishing tape to remove protrusions and make them non-defective.

  When judging whether or not it is a non-defective product, humans (operators and operators) make judgments based on product information, defect area, defect height, and all information of microscopic observation using transmitted / reflected illumination. The rate drops.

Japanese Examined Patent Publication No. 06-100653 Japanese Patent No. 3381911 JP 2006-145786 A JP 2006-030283 A JP 2007-333972 A JP 2008-180907 A JP 2010-102063 A JP 2010-224359 A JP 2009-271452 A JP 2011-81294 A

  Only a few defects (about 5% of all defects detected by the upstream inspection machine) are actually good products. The method of selecting the protrusion foreign matter from the defects detected by the automatic defect inspection apparatus is that a color filter substrate having defects detected by the automatic defect inspection apparatus is once inserted into the defect correction machine, and a CCD area sensor is used for each defect. Since it is determined whether or not correction is possible after imaging the defective part and confirming the captured image, it takes time to determine whether or not correction is possible, and the operating rate cannot be increased.

  For example, the protruding foreign matter that is a defect is a protruding foreign matter having a diameter of several μm to 10 μm or more and a height of several μm to 3 μm or more. The area of the defect can be detected by the automatic defect detection device, but the height of the protruding foreign matter cannot be detected by the automatic defect detection device. In recent years, an automatic height detection device has also been developed, which has a height discrimination function using depth of focus. However, the time required for height discrimination (judgment time) has become a bottleneck. Not reached.

  Moreover, when a metal foreign material is contained in a defect, since the polishing tape which corrects a projection defect part will be cut | disconnected by a metal foreign material, the defect correction by tape grinding | polishing cannot be performed. Currently, a person determines whether or not correction is possible by checking whether or not a metal foreign object is included in a defect from a microscope image.

  After the correction, the pixel correction device determines whether or not the defect is good (OK, NG) from the defect size (area) detected by the automatic defect detection device and the height of the protruding foreign matter after the correction as described above.

  Furthermore, in order to improve yield and yield rate, the defect manufacturing process is identified by identifying the defect mode from all information of product information, defect area, defect height, microscopic observation with transmitted / reflected illumination. , Feedback is provided to the device.

As described above, in the pixel correction device of the display device, as shown in the following (a) to (c), there are many judgments by humans (operators, operators), and processing is performed even for defects that do not need to be judged. Therefore, there is a problem that the operation rate cannot be increased.
(1) Determination of necessity of correction-Judgment of metal / nonmetal from projection defect size (area), height information and microscopic image → Height of standard NG for nonmetal defect to be corrected (2) Defect quality determination -OK / NG judgment for defect part from size (area), height information and microscopic image (defect mode) of protrusion defect part (3) Mode classification-Size (area) of protrusion defect part, height information and Information obtained from a microscope image (transmission / reflection, magnification switching, etc.) obtained from a selected microscope image by selecting a mode corresponding to all 270 modes includes the following information.
a. Magnification information: magnification information of the eyepiece such as x2, x5, x10, x20, x50 b. Illumination information: Information on how illumination is applied to a transmission / reflection object c. Focus information: Microscope focus position information (distance from the just focus point)
d. Illuminance information: Information on the brightness of the lighting

  The present invention has been made in view of such a problem, and an object of the present invention is to provide a defect mode classification system capable of performing efficient pixel (defect) correction work.

  One aspect of the present invention for solving the above problems is a defect mode classification system for classifying pixel defect modes of color filters and thin film transistors and correcting defects, the defect mode classification system including at least a defect A defect inspecting machine for acquiring size and position information, a pixel correcting device including an imaging means capable of imaging a defect under different imaging conditions and a height measuring means of the defect, and a mode for specifying a defect mode A defect mode classification system comprising: a classifier; a communication device capable of transferring acquired data of a defect inspection machine and acquired data of the pixel correction device to the mode classifier.

  According to the present invention, it is possible to provide a defect mode classification system that enables efficient pixel (defect) correction work.

Figure showing an overview of the mode classification system Operation flow of mode classification system

  FIG. 1 shows an outline of a defect mode classification system according to an embodiment of the present invention. The defect mode classification system is a defect mode classification system for classifying pixel defect modes of color filters and thin film transistors and correcting defects. The defect mode classification system includes at least a defect inspection machine for acquiring defect size and position information, a pixel correction apparatus including an imaging unit capable of imaging a defect under different imaging conditions, and a height measurement unit of the defect, A mode classifier for specifying a defect mode; and a communication device capable of transferring the acquired data of the defect inspection machine and the acquired data of the pixel correction device to the mode classifier.

  As shown in FIG. 1, the pixel correction device can take (download) data (defect coordinates, defect size (area)) of an upstream defect inspection machine from its upper server to its own device.

  1 includes a polishing tape unit, a laser repair unit, a revolver microscope (x2, x5, x10, x20, and x50 (5 magnifications)), an illumination unit (transmission and reflection), and foreign object height measurement. A unit (contact type) and a glass substrate stage are included.

  The illumination unit and the microscope can be switched on the PC or the control BOX, and the defect image captured by the microscope can be projected and observed on the PC. If the observed defect does not include a metal foreign object and has only a height, it can be corrected to a non-defective product by performing tape polishing with a polishing tape unit.

  After the product (glass substrate) is put into the pixel correction device, a focus map is provided at each point on the glass substrate stage (for example, □ 50 mm unit), so that a microscope image can be taken at the just focus position at all points. Become.

  By providing a function of capturing at the just focus position and automatically capturing the captured image (still image), the microscope image can be automatically stored.

The pixel correction apparatus of FIG. 1 can set the imaging conditions (magnification information, illumination information, focus information, illuminance information) of the microscope as follows.
Magnification information: x2, x5, x10, x20, x50 (5 magnifications)
Illumination information: Transmission, reflection (2 optical systems)
Focus information: Just focus position ± 5mm
* The just focus position is determined by the first focus map position setting.
Illuminance information: Volume value of 0 to 9999, and a function that can be combined with all the information shown above.

  The pixel correction device has machine learning and artificial intelligence algorithms for upstream defect inspection machine data (defect coordinates, defect size (area)) and data (microscope image and microscope imaging conditions) acquired by itself. It can be stored and transferred to a device (hereinafter referred to as a mode classifier).

The mode classifier learns the target image and data (teacher image or teacher data) in advance.
At this time, learning is performed with at least N = 10 data to create a model.
After creating the model, the image and data to be classified are put into the model, and the corresponding result (mode) is output. The learning function of the mode classifier can be either inline or offline.

  The mode classifier according to the embodiment of the present invention can create a model adapted to each of (1) correction necessity determination, (2) defect quality determination, and (3) mode classification. (Multiple models can be created.)

FIG. 2 shows an operation flow of the mode classification system. As shown in FIG. 2, the operation of the mode classification system is as follows. First, when a glass substrate is inserted, a defect is imaged with a microscope and automatically stored under arbitrary conditions (S1). Next, the area of the defective part is measured by processing the image of the defect (S2). The height of the defect is measured with a contact-type height measuring machine (S3). It is determined whether correction is necessary (S4). Next, the polishing tape is corrected for the target defect determined to be corrected (S5), and finally the substrate is dispensed. Data exchange at this time can be performed via Ethernet (registered trademark).
Further, regarding (1) correction necessity determination, (2) defect quality determination, and (3) mode classification in FIG. 2, the upstream defect inspection machine data and pixel correction device data are learned by the mode classifier, and learning is performed. Classification is performed by putting the target image or data into the model obtained by the above. By classifying (1), it is possible to identify metals and non-metals, (2) to determine the quality of defects (OK, NG), and (3) to select one mode from 270 modes. Output.
As a result, it is possible to automate items that are manually determined and judged, and to perform efficient pixel (defect) correction work.

The mode can be defined as follows, for example.
(1) Correction necessity determination (3 modes)
(Mode 1) Black metal (When viewed by reflection, metal foreign matter is embedded in black foreign matter)
(Mode 2) White metal (when viewed by reflection, metal foreign matter is embedded in the transparent resin system)
(Mode 3) It is classified into other non-metallic three modes. The necessity of correction is determined according to the classification result.
When identifying a metal defect, for example, a reflection image having a luminance value of 200 or more in 256 gradations can be identified as a metal defect.
(B) Defect quality determination (about 20 modes)
Depending on the required specifications of the product, for example, it can be classified into about 20 modes. Defect quality is determined according to the classification result.
For example, foreign substances, white pinholes, film scratches and the like.
(C) Mode classification (about 270 modes)
It can be arbitrarily classified according to the defect that occurs, and for example, it can be classified by type (first classification), state (second classification), color (third classification), and the like. By classifying in detail, for example, it can be used to improve the manufacturing process.
For example, the defects occurring in the RGB process include a mode of resist residue (first classification) -gel (second classification) -red (R) (third classification).

  The present invention can be applied to a manufacturing process or an inspection process of a display device including a color filter and a thin film transistor.

Claims (9)

A defect mode classification system for classifying pixel defect modes of color filters and thin film transistors and correcting defects,
The defect mode classification system includes at least:
A defect inspection machine for obtaining the size and position information of the defect;
A pixel correction device including an imaging unit capable of imaging the defect under different imaging conditions and a height measurement unit of the defect;
A mode classifier for identifying the mode of the defect;
A defect mode classification system comprising: a communication device capable of transferring acquired data of the defect inspection machine and acquired data of the pixel correction device to the mode classifier. The imaging means of the pixel correction device is
Means for varying the imaging magnification;
Illumination means capable of irradiating with two systems of a transmission optical system and a reflection optical system;
Means for varying the focal position of the imaging means;
The defect mode classification system according to claim 1, further comprising means for varying the illuminance of the illumination means. The means for changing the imaging magnification has a magnification mechanism of x2, x5, x10, x20, x50,
The means for changing the focal position has a mechanism that can be displaced ± 5 mm with respect to the just focus,
The means for varying the illuminance has a variable function from 0 to 9999 in volume value,
The defect mode classification system according to claim 2, which has a function of imaging by combining these imaging conditions.   4. The device according to claim 1, wherein the pixel correction device includes a mechanism that always enables imaging at a just focus position by creating a focus map at each point (□ 50 mm unit) on the stage. Defect mode classification system described in Crab.   The mode classifier is equipped with an algorithm of machine learning or artificial intelligence for defect size and position information acquired by the defect inspection machine, and a captured image and defect height information acquired by the pixel correction device. The defect mode classification system according to any one of claims 1 to 4, wherein the defect mode classification system has a function that can be learned in advance by a device.   The defect mode classification system according to claim 5, wherein the defect mode can be classified from the acquired data to be inspected according to a defect model created by the learning function of the mode classifier.   The defect mode classification system according to claim 6, wherein the defect model is formed by learning at least N = 10 or more image data by the learning function.   The defect mode classification system according to any one of claims 5 to 7, wherein the learning function of the mode classifier can be either inline or offline. The mode classifier is:
The defect mode classification system according to any one of claims 5 to 8, wherein a defect necessity determination, defect quality determination, and defect mode classification are possible for a defective portion.