WO2021124775A1 - Fault classification method, fault classification device, and method for producing glass article - Google Patents

Fault classification method, fault classification device, and method for producing glass article Download PDF

Info

Publication number
WO2021124775A1
WO2021124775A1 PCT/JP2020/043010 JP2020043010W WO2021124775A1 WO 2021124775 A1 WO2021124775 A1 WO 2021124775A1 JP 2020043010 W JP2020043010 W JP 2020043010W WO 2021124775 A1 WO2021124775 A1 WO 2021124775A1
Authority
WO
WIPO (PCT)
Prior art keywords
defect
classification
image data
classifier
defects
Prior art date
Application number
PCT/JP2020/043010
Other languages
French (fr)
Japanese (ja)
Inventor
貴裕 茗原
幹将 北川
厚司 井上
Original Assignee
日本電気硝子株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電気硝子株式会社 filed Critical 日本電気硝子株式会社
Priority to CN202080071666.XA priority Critical patent/CN114556414A/en
Priority to KR1020227010768A priority patent/KR20220113348A/en
Publication of WO2021124775A1 publication Critical patent/WO2021124775A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
    • B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
    • B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
    • B07C5/34—Sorting according to other particular properties
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/22—Measuring arrangements characterised by the use of optical techniques for measuring depth
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G01N21/958—Inspecting transparent materials or objects, e.g. windscreens
    • G—PHYSICS
    • G02—OPTICS
    • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
    • G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
    • G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour 
    • G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour  based on liquid crystals, e.g. single liquid crystal display cells
    • G02F1/133—Constructional arrangements; Operation of liquid crystal cells; Circuit arrangements
    • G02F1/1333—Constructional arrangements; Manufacturing methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8854—Grading and classifying of flaws
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30121—CRT, LCD or plasma display

Definitions

  • the present invention relates to a method and an apparatus for classifying defects contained in a glass article, and a method for manufacturing a glass article by using the method and the apparatus.
  • a glass plate is used for displays such as liquid crystal displays and organic EL displays.
  • glass plates are molded using various manufacturing equipment, but in each case, the glass raw material is heated and melted, the molten glass is homogenized, and then molded into a predetermined shape. It is commonly done. In this manufacturing method, defects may occur on the surface or inside of the glass plate due to various causes. Therefore, the technique of accurately measuring the defect of the glass plate is very important.
  • Patent Document 1 discloses a method for manufacturing a glass plate including an inspection step of measuring the dimensions and depth (distance from the glass surface) of defects existing inside the glass plate with a laser microscope.
  • an inspector operates a laser microscope to measure defects.
  • the present invention has been made in view of the above circumstances, and it is a technical subject to improve the work efficiency of inspection by improving the accuracy of classification of the types of defects related to glass articles with a classifier.
  • the present invention is for solving the above-mentioned problems, and is a method of detecting a defect contained in a glass article as image data and classifying the type of the defect by a defect classification device based on the image data.
  • the types of defects include a first classification and a second classification
  • the defect classification device includes a first classifier created by machine learning the first teacher data related to the first classification, and the first classifier.
  • a second classifier created by machine learning the second teacher data related to the second classification is provided, and whether or not the defect belongs to the first classification is determined by the first classifier based on the image data.
  • the second determination step is provided.
  • the first classifier that performs the first determination step and the second classifier that performs the second determination step are created by machine learning the first teacher data and the second teacher data. , It is possible to accurately classify the types of defects related to the first classification and the second classification.
  • the correct answer rate of the first classifier may be higher than the correct answer rate of the second classifier.
  • the first teacher data according to the first classification includes image data related to defects formed at a predetermined depth of the glass article, and before the first determination step.
  • a depth determination step for determining whether or not the depth of the defect in the image data detected from the glass article is the predetermined depth is provided, and the first determination step is the depth determination step. , Can be executed when the depth of the defect is determined to be the predetermined depth.
  • the image data related to the first teacher data may include only image data related to bubbles, and the image data related to the second teacher data contains only image data related to platinum foreign matter. It may be included.
  • the image data regarding the bubble may include information regarding the shape and / or color of the bubble.
  • information regarding the shape and / or color of the bubble in the first teacher data, it is possible to subdivide the type of the defect related to the bubble into the shape and the color.
  • the image data regarding the platinum foreign substance may include information regarding the shape and / or color of the platinum foreign substance. In this way, by including the information in the shape and / or color of the platinum foreign substance in the second teacher data, it is possible to subdivide the type of the defect related to the platinum foreign substance into the shape and color.
  • the present invention is for solving the above-mentioned problems, and is a method for manufacturing a glass article, which is based on an imaging step of acquiring image data of defects contained in the glass article and any of the above based on the image data. It is characterized by comprising a step of classifying the type of the defect by the defect classification method.
  • the present invention is for solving the above-mentioned problems, and is an apparatus for detecting a defect contained in a glass article as image data and classifying the type of the defect based on the image data.
  • a first classifier created by machine learning the first teacher data related to the first classification, including the first classification and the second classification, and the defect is found based on the image data. It is a first classifier that determines whether or not it belongs to the first classification, and a second classifier created by machine learning the second teacher data related to the second classification, and the defect is the first. It is characterized by including a second classifier that determines whether or not the defect belongs to the second category based on the image data when it is determined that the defect does not belong to one category.
  • the defect classification device by automating the classification of the type of defect by the defect classification device, it becomes possible to efficiently perform the work related to the inspection of the defect. Since the first classifier and the second classifier of the defect classification device according to the present invention are created by machine learning the first teacher data and the second teacher data, the classification of the defect type can be performed accurately. It can be carried out.
  • the present invention it is possible to improve the work efficiency of inspection by accurately classifying the types of defects related to glass articles with a classifier.
  • FIG. 1 to 12 show a method for manufacturing a glass article according to the present invention and an embodiment of an inspection device used in the method.
  • a case of manufacturing a transparent glass plate as an example of a glass article will be described.
  • the inspection device 1 includes an image pickup device 3 that captures an image of a glass plate G mounted on a mounting table 2, and an image processing device that performs image processing on image data acquired by the image pickup device 3. 4 and a defect classification device 5 for classifying the types of defects included in the image data.
  • the image pickup device 3 is provided with an optical microscope, and can acquire a magnified image of the surface S and the inside of the glass plate G as data.
  • the image pickup apparatus 3 is configured to be movable in the vertical direction and the horizontal direction by the moving mechanism 3a.
  • the image pickup apparatus 3 can change the position of the focal point with respect to the thickness direction of the glass plate G by moving in the vertical direction at a predetermined pitch. As a result, the image pickup apparatus 3 can take an image of a defect contained inside the glass plate G.
  • the image pickup device 3 is connected to the image processing device 4.
  • the image pickup device 3 transmits the acquired image data to the image processing device 4.
  • the image processing device 4 includes a PC that incorporates various hardware such as a display 4a and an input interface, in addition to incorporating a CPU, GPU, RAM, ROM, HDD, SDD, and the like. As shown in FIG. 2, the image processing device 4 is located between an arithmetic processing unit 6 composed of a CPU or the like, a storage unit 7 composed of an HDD, an SSD, or the like, an image pickup device 3, and a defect classification device 5. It includes a communication unit 8 for transmitting and receiving data.
  • the image processing device 4 controls the operation of the image pickup device 3 by executing the image analysis program stored in the storage unit 7 by the arithmetic processing unit 6.
  • the image processing device 4 can receive a large number of image data acquired by the image pickup device 3 from the image pickup device 3 and execute various image processes on each image data.
  • the defect classification device 5 includes a PC having built-in hardware such as a CPU, GPU, RAM, ROM, HDD, and SSD. As shown in FIG. 2, the defect classification device 5 transmits / receives data between the arithmetic processing unit 9 composed of a CPU or the like, the storage unit 10 composed of an HDD, SSD, or the like, and the image processing device 4. It is provided with a communication unit 11 for performing.
  • the arithmetic processing unit 9 can classify the types of defects included in the image data acquired by the image pickup apparatus 3 by executing the image analysis program stored in the storage unit 10. Further, the arithmetic processing unit 9 can perform machine learning for classifying the types of defects by using the teacher data stored in the storage unit 10 by executing the image analysis program.
  • the storage unit 10 stores an image analysis program for classifying defects.
  • a deep learning tool program
  • the image analysis program can create an inference model (algorithm) for classifying the types of defects by machine learning (deep learning) based on teacher data.
  • the storage unit 10 includes a first classifier 12 that classifies the types of defects contained in the glass plate G into the first classification, and a second classifier 13 that classifies the types of defects into the second classification.
  • Each of the classifiers 12 and 13 is an inference model created by the arithmetic processing unit 9 by machine learning related to the image analysis program.
  • defects related to "foam” are set to the first category
  • defects related to "platinum foreign matter” are set to the second category.
  • the bubbles as defects include bubbles containing some gas or vacuum bubbles containing no gas.
  • the types of gas are oxygen, carbon dioxide, carbon monoxide, nox (NO X ), nitrogen, chlorine, brom, hydrogen, argon, helium, neon, xenone, steam, socks (SO). X ), sulfurous acid gas, etc.
  • the component at the time of bubble generation may be precipitated as a solid on the inner wall of the bubble in some state.
  • Defects related to "platinum foreign matter” are composed of fine platinum that has shielding properties rather than being transparent to visible light like bubbles. Defects due to "platinum foreign matter” are formed inside the glass plate G, for example, by mixing minute platinum generated from a platinum container for supplying and transferring molten glass for forming the glass plate G into the molten glass. To.
  • the first classifier 12 is an inference model created by performing machine learning on the first teacher data including a plurality of image data related to "bubbles".
  • the first teacher data includes only the image data related to "bubbles" and does not include the image data related to "platinum foreign matter”.
  • the image data of the "bubble" in the first teacher data includes information on the shape and / or color of the bubble.
  • Examples of information related to the shape of bubbles include “horizontal”, “vertical”, and “excessive dimensions”.
  • the shape related to “horizontally long” refers to the shape of a flat bubble extending in the lateral direction as shown in FIG.
  • the shape related to “vertical” refers to the shape of a flat bubble extending in the vertical direction.
  • the shape related to “excessive size” means a “horizontal” shape (or “vertical” shape) in which the entire shape of the bubble is not projected in the range of the screen.
  • An example of information about the "color” of a bubble is "white”. Specifically, each image data of the first teacher data is classified into sub-classifications such as “horizontal”, “vertical”, “excessive size”, and “white”, and each image data is associated with the sub-classification. Machine learning at.
  • the second classifier 13 is an inference model created by performing machine learning on the second teacher data including a plurality of image data related to "platinum foreign matter".
  • the second teacher data includes only the image data related to the "platinum foreign substance” and does not include the image data related to the "bubble".
  • the image data of the "platinum foreign body" in the second teacher data includes information on the shape and / or color of the platinum foreign body.
  • Examples of information on the shape of platinum foreign matter include “triangle”, “polygon”, and “circle”.
  • the shape related to the “triangle” refers to the shape of a platinum foreign substance exhibiting a triangular shape.
  • the shape related to the “polygon” refers to the shape of a polygonal platinum foreign substance excluding a triangle.
  • the shape related to the “circle” refers to the shape of a circular platinum foreign substance.
  • the information regarding the "color” of the platinum foreign substance for example, "white” can be mentioned. Specifically, each image data of the second teacher data is classified into sub-classifications such as “triangle", “polygon”, “circle”, and “white”, and each image data is associated with the sub-classification. Machine learning.
  • this method mainly includes a cutting step S1, an end face processing step S2, a cleaning step S3, and an inspection step S4.
  • the glass original plate supplied to the cutting step S1 is formed by continuously forming a single (strip-shaped) glass ribbon from the molten glass by, for example, a forming apparatus capable of executing an overflow down draw method, and a rectangular glass original plate is formed from the glass ribbon. It is obtained by cutting out.
  • the float method may be used for molding the glass ribbon.
  • the cutting step S1 by cutting the original glass plate, one or a plurality of rectangular glass plates G are cut out to a desired size. Cutting is performed, for example, by forming a scribe line on a glass original plate and then folding it.
  • the glass plate G that has undergone the end face processing step S2 is cleaned.
  • the inspection step S4 includes a first inspection step of inspecting the presence or absence of defects in the glass plate G and a second inspection step of inspecting the defects when the glass plate G contains defects.
  • the glass plate G is inspected for defects by a known inspection device (not shown) arranged on the upstream side of the inspection device 1.
  • a known inspection device for example, an edge light type inspection device is used, but the inspection device is not limited to this.
  • the glass plate G in which defects are detected is further inspected in the second inspection step.
  • Other glass plates G are certified as non-defective products.
  • the second inspection step includes an imaging step S41 for capturing an image of a defect contained in the glass plate G and acquiring the image data, and a measuring step S42 for measuring the state of the defect contained in the image data.
  • a classification step S43 for classifying the types of defects and a quality determination step S44 for determining the quality of the glass plate G (defect) are provided.
  • the image processing device 3 takes a picture of the glass plate G by the operator (inspector) operating the image processing device 4 or by automatic control.
  • the imaging device 3 is moved by the operation of the moving mechanism 3a so that the defect F contained in the glass plate G is located in the field of view.
  • the image pickup apparatus 3 images the surface S and the inside of the glass plate G a plurality of times while moving in the vertical direction at a predetermined pitch.
  • the image data of the glass plate G acquired by the image pickup device 3 is transmitted to the image processing device 4.
  • the arithmetic processing unit 6 of the image processing device 4 stores the received image data in the storage unit 7. Subsequently, the defect F is measured by the image analysis program of the image processing device 4.
  • the size and depth of the defect F are measured.
  • the size (maximum dimension) of the defect F is measured by, for example, an image analysis program.
  • the depth D of the defect F means the distance from the surface S to the defect F in the thickness direction of the glass plate G.
  • the depth D of the defect F is measured, for example, as follows.
  • the most focused image data is selected by the image analysis program of the image processing device 4. Then, the distance from the focal position (coordinates) of the image pickup apparatus 3 to the surface S of the glass plate G in the selected image data is calculated as the depth D of the defect F. Information on the measured size and depth D of the defect F is stored in the storage unit 7 together with the image data.
  • the classification step S43 includes a first classification step of classifying the type of the defect F by the defect classification device 5, and a second classification step of the operator classifying the type of the defect F using the image processing device 4. ..
  • the first classification step includes the first determination step S431 for determining whether or not the defect F belongs to the first classification by the first classifier 12, and whether the defect F belongs to the second classification.
  • a second determination step S432 for determining whether or not to use is provided.
  • the arithmetic processing unit 9 activates the first classifier 12 and determines whether or not the type of the defect F included in the image data belongs to the first classification (“bubble”). In the present embodiment, the arithmetic processing unit 9 determines in the first classifier 12 whether or not the defect F belongs to any of the minor classifications such as "horizontal”, “vertical”, “excessive size", and "white”. .. When the arithmetic processing unit 9 determines that the defect F included in the image data corresponds to the “bubble”, the arithmetic processing unit 9 adds the classification information related to the first classification to the image data and stores it in the storage unit 10. When the first classifier 12 determines that the defect F in the image data is not a "bubble”, the arithmetic processing unit 9 classifies the defect F into "others”, adds the classification information to the image data, and stores the defect F. Store at 10.
  • the arithmetic processing unit 9 activates the second classifier 13, and the image data determined not to belong to the first classification, that is, the image data (defect F) classified as "other" is Determine whether or not it belongs to the second category.
  • the arithmetic processing unit 9 determines in the second classifier 13 whether or not the defect F belongs to any of the sub-classifications such as "triangle", "polygon", “circle", and "white”.
  • the arithmetic processing unit 9 classifies the image data into the second classification.
  • the arithmetic processing unit 9 adds classification information related to the second classification to the image data and stores it in the storage unit 10.
  • the arithmetic processing unit 9 classifies the defect F as "other".
  • the arithmetic processing unit 9 adds the classification information to the image data and stores it in the storage unit 10.
  • the arithmetic processing unit 9 transmits the image data for which the classification has been completed to the image processing device 4.
  • the image processing device 4 stores the classified image data transmitted from the defect classification device 5 in the storage unit 7.
  • the image processing device 4 displays the image data received from the defect classification device 5 on the display 4a by the operation of the operator or by automatic control.
  • the display 4a displays the classification information (information on "bubbles", “platinum foreign matter”, and “others") together with the image data.
  • the operator visually certifies the type of defect F for the defect F classified as "other" in the first classification process. Specifically, the operator visually observes the image data displayed on the display 4a of the image processing device 4, and the type of defect F (for example, "bubble”, “platinum foreign matter”, “foreign matter other than platinum”, “foreign matter other than platinum”, “ “Dirt”, “Scratch”) is certified. The operator adds classification information (certification result) to the image data and stores it in the storage unit 7.
  • the image processing device 4 determines the quality of the glass plate G based on information such as the size (dimensions), depth D, and type of the defect F. Specifically, when the image processing apparatus 4 determines that the defect F adversely affects the surface texture of the surface S of the glass plate G, that is, the defect F causes unevenness or the like on the surface S of the glass plate G. The glass plate G corresponding to the image data is recognized as a defective product. On the other hand, the image processing apparatus 4 recognizes the glass plate G as a non-defective product when the defect F does not adversely affect the surface properties of the surface S of the glass plate G.
  • the type of the defect F included in the image data is classified by the defect classification device 5.
  • the defect classification device 5 By automatically classifying, it is possible to improve the work efficiency of the inspection step S4. Thereby, the manufacturing efficiency of the glass article can be improved.
  • the present invention is not limited to the configuration of the above embodiment, and is not limited to the above-mentioned action and effect.
  • the present invention can be modified in various ways without departing from the gist of the present invention.
  • the inspection device 1 in which the image processing device 4 and the defect classification device 5 are configured by separate devices is illustrated, but the present invention is not limited to this configuration.
  • the image processing device 4 and the defect classification device 5 may be configured by one device.
  • a method and an apparatus for classifying a defect F of a glass plate G as a glass article have been exemplified, but the present invention is not limited to this configuration.
  • the present invention is also applicable, for example, to classify defects contained in tube glass.
  • the present invention is not limited to this configuration.
  • the classification step S43 can be executed only by the defect classification device 5.
  • the first classifier 12 that classifies the types of defects into the first classification (“foam”) and the second classifier 13 that classifies the types of defects into the second classification (“platinum foreign matter”).
  • the defect classification device 5 including the above is illustrated, the present invention is not limited to this configuration.
  • the defect classification device 5 may further include, for example, one or more other classifiers that classify the types of defects into different classifications.
  • the first classification is not limited to "foam” and may be another type of defect
  • the second classification is not limited to "platinum foreign matter" and may be another type of defect.
  • the first classifier 12 and the second classifier 13 are used regardless of the depth of the defect, but the first classifier 12 and the second classifier 13 according to the depth of the defect are prepared.
  • the defects may be classified by the first classifier 12 and the second classifier 13 corresponding to the measured depth of the defect.
  • the first teacher data about "foam” located at the depth around the front surface of the glass plate
  • the first teacher data about "foam” located at the depth around the back surface of the glass plate
  • the middle between the front surface and the back surface prepare the first teacher data about the "bubble” located at the depth.
  • the first classifier for "foam” located around the front surface the first classifier for "foam” located around the back surface, and the "foam” located between the front and back surfaces. Create the first classifier for each.
  • a depth determination step for determining whether or not the defect has a predetermined depth can be performed before the first determination step S431 is executed. Specifically, in the depth determination step, it is determined whether the detected image data defect is located near the front surface, the middle, or the back surface. Then, in the subsequent first determination step S431, the type of defect is classified using the first classifier at the position corresponding to the determination result in the depth determination step. In this case, as for the second classifier, as with the first classifier, a second classifier for "platinum foreign matter" located around the front surface, the middle, and the back surface is created. In the second determination step S432, similarly to the first determination step S431, the type of defect is determined by using the second classifier at the position corresponding to the determination result in the depth determination step.
  • the present inventors conducted a test to confirm an effective machine learning method by a defect classification device in order to accurately classify the types of defects contained in image data.
  • the present inventors have created teacher data in which image data having the attributes of "bubble”, “platinum foreign matter”, “foreign matter other than platinum”, “dirt”, and “scratch” are mixed, and machine learning by a defect classification device. (Deep learning) was carried out.
  • the defect inference model (algorithm) created by this machine learning will be referred to as a comparative example.
  • the present inventors created teacher data (first teacher data) including only image data having the attribute of "bubble", and carried out machine learning by a defect classification device.
  • the defect inference model (first classifier) created by this machine learning will be referred to as Example 1.
  • Example 2 teacher data including only image data having the attribute of "platinum foreign matter", and carried out machine learning by a defect classification device.
  • the defect inference model (second classifier) created by this machine learning will be referred to as Example 2.
  • the correct answer rate of the comparative example was 100%
  • the correct answer rate of the example 1 was 100%
  • the correct answer rate of the comparative example and the correct answer rate of the example 1 were the same.
  • the correct answer rate of Comparative Example was 76%
  • the correct answer rate of Example 2 was 97%, and it was found that the correct answer rate of Example 2 was higher than the correct answer rate of Comparative Example.
  • the classification related to "bubbles” (first determination step S431) is executed first, and then.
  • the classification related to "platinum foreign matter” (second determination step S432), the classification (particularly the classification related to "platinum foreign matter” having a low accuracy rate in the comparative example) can be performed accurately and efficiently. It became clear.

Abstract

A method for classifying a fault in a glass article, the method comprising: a first assessment step S431 in which an assessment is made by a first classifier 12, on the basis of image data, as to whether a fault F belongs in a first classification; and a second assessment step S432 in which, when it has been assessed that the fault F does not belong in the first classification, an assessment is made by a second classifier 13 as to whether the fault F belongs in a second classification.

Description

欠陥分類方法、欠陥分類装置及びガラス物品の製造方法Defect classification method, defect classification device and manufacturing method of glass articles
　本発明は、ガラス物品に含まれる欠陥を分類する方法及び装置、並びに当該方法及び装置を利用することでガラス物品を製造する方法に関する。 The present invention relates to a method and an apparatus for classifying defects contained in a glass article, and a method for manufacturing a glass article by using the method and the apparatus.
　液晶ディスプレイや有機ＥＬディスプレイ等のディスプレイには、ガラス板が使用される。ディスプレイ用のガラス板の製造では、各種の製造装置を使用し、ガラス板が成形されているが、いずれもガラス原料を加熱溶解し、溶融ガラスを均質化した後に所定形状に成形するということが一般に行われている。この製造方法では、種々の原因によってガラス板に表面又は内部に欠陥が生じる場合がある。従って、ガラス板の欠陥を正確に測定する技術は非常に重要なものとなっている。 A glass plate is used for displays such as liquid crystal displays and organic EL displays. In the manufacture of glass plates for displays, glass plates are molded using various manufacturing equipment, but in each case, the glass raw material is heated and melted, the molten glass is homogenized, and then molded into a predetermined shape. It is commonly done. In this manufacturing method, defects may occur on the surface or inside of the glass plate due to various causes. Therefore, the technique of accurately measuring the defect of the glass plate is very important.
　例えば特許文献１には、ガラス板の内部に存在する欠陥の寸法や深さ（ガラス表面からの距離）等をレーザ顕微鏡によって測定する検査工程を備えたガラス板の製造方法が開示されている。この製造方法では、検査工程において、検査員がレーザ顕微鏡を操作することで、欠陥の測定を行っている。 For example, Patent Document 1 discloses a method for manufacturing a glass plate including an inspection step of measuring the dimensions and depth (distance from the glass surface) of defects existing inside the glass plate with a laser microscope. In this manufacturing method, in the inspection process, an inspector operates a laser microscope to measure defects.
特開２０１５－２０５８１１号公報Japanese Unexamined Patent Publication No. 2015-205811
　測定される欠陥には、泡や異物、汚れといった種別が存在し、検査工程で検査員が欠陥の種別を分類する場合がある。この場合、検査における人員の増加や各検査員の負担が増加し、作業効率の低下を招くおそれがあった。また、本発明者らの鋭意研究によれば、分類器に泡や異物、汚れを全て機械学習させ、その分類器で欠陥の種別の分類を行うと、精度が不十分であった。 There are types of defects to be measured, such as bubbles, foreign matter, and dirt, and the inspector may classify the types of defects in the inspection process. In this case, the number of personnel in the inspection increases and the burden on each inspector increases, which may lead to a decrease in work efficiency. In addition, according to the diligent research of the present inventors, when the classifier was made to machine-learn all bubbles, foreign substances, and dirt, and the classifier was used to classify the types of defects, the accuracy was insufficient.
　本発明は上記の事情に鑑みてなされたものであり、ガラス物品に係る欠陥の種別の分類を分類器で精度良くすることで、検査の作業効率を向上させることを技術的課題とする。 The present invention has been made in view of the above circumstances, and it is a technical subject to improve the work efficiency of inspection by improving the accuracy of classification of the types of defects related to glass articles with a classifier.
　本発明は上記の課題を解決するためのものであり、ガラス物品に含まれる欠陥を画像データとして検出し、前記画像データに基づいて前記欠陥の種別を欠陥分類装置によって分類する方法であって、前記欠陥の種別は、第一分類と、第二分類とを含み、前記欠陥分類装置は、前記第一分類に係る第一教師データを機械学習することにより作成した第一分類器と、前記第二分類に係る第二教師データを機械学習することにより作成した第二分類器とを備え、前記画像データに基づいて前記欠陥が前記第一分類に属するか否かを前記第一分類器によって判定する第一判定工程と、前記欠陥が前記第一分類に属さないと判定された場合に、前記画像データに基づいて前記欠陥が前記第二分類に属するか否かを前記第二分類器によって判定する第二判定工程と、を備えることを特徴とする。 The present invention is for solving the above-mentioned problems, and is a method of detecting a defect contained in a glass article as image data and classifying the type of the defect by a defect classification device based on the image data. The types of defects include a first classification and a second classification, and the defect classification device includes a first classifier created by machine learning the first teacher data related to the first classification, and the first classifier. A second classifier created by machine learning the second teacher data related to the second classification is provided, and whether or not the defect belongs to the first classification is determined by the first classifier based on the image data. In the first determination step to be performed, and when it is determined that the defect does not belong to the first classification, whether or not the defect belongs to the second classification is determined by the second classifier based on the image data. The second determination step is provided.
　かかる構成によれば、欠陥分類装置によって欠陥の種別の分類を自動化することで、欠陥の検査に係る作業を効率良く行うことが可能になる。本発明では、第一判定工程を行う第一分類器と、第二判定工程を行う第二分類器は、第一教師データ及び第二教師データを機械学習することにより作成されたものであるため、第一分類及び第二分類に係る欠陥の種別の分類を精度良く行うことができる。 According to such a configuration, by automating the classification of the type of defect by the defect classification device, it becomes possible to efficiently perform the work related to the inspection of the defect. In the present invention, the first classifier that performs the first determination step and the second classifier that performs the second determination step are created by machine learning the first teacher data and the second teacher data. , It is possible to accurately classify the types of defects related to the first classification and the second classification.
　本発明に係る欠陥分類方法において、前記第一分類器の正解率は、前記第二分類器の正解率よりも高くなってもよい。これにより、ガラス物品に含まれる欠陥の種別をさらに精度良く分類できる。 In the defect classification method according to the present invention, the correct answer rate of the first classifier may be higher than the correct answer rate of the second classifier. As a result, the types of defects contained in the glass article can be classified more accurately.
　本発明に係る欠陥分類方法において、前記第一分類に係る前記第一教師データは、前記ガラス物品の所定の深さに形成される欠陥に係る画像データを含み、前記第一判定工程の前に、前記ガラス物品から検出された前記画像データにおける前記欠陥の深さが前記所定の深さである否かを判定する深さ判定工程を備え、前記第一判定工程は、前記深さ判定工程において、前記欠陥の深さが前記所定の深さであると判定された場合に実行され得る。 In the defect classification method according to the present invention, the first teacher data according to the first classification includes image data related to defects formed at a predetermined depth of the glass article, and before the first determination step. A depth determination step for determining whether or not the depth of the defect in the image data detected from the glass article is the predetermined depth is provided, and the first determination step is the depth determination step. , Can be executed when the depth of the defect is determined to be the predetermined depth.
　傷や汚れはガラス物品の表面のみに存在し、中間部分には存在しない。このため、深さ判定工程において、欠陥の深さに関する基準値を設定することで、欠陥の種別を絞り込むことができ、第一分類器及び第二分類器の正解率を向上させることが可能になる。これにより、欠陥の種別をさらに精度良く分類できる。 Scratches and stains exist only on the surface of the glass article, not in the middle part. Therefore, in the depth determination process, by setting a reference value regarding the depth of the defect, the types of defects can be narrowed down, and the accuracy rate of the first classifier and the second classifier can be improved. Become. As a result, the types of defects can be classified more accurately.
　本発明に係る欠陥分類方法において、前記第一教師データに係る前記画像データは、泡に関する画像データのみを含んでもよく、前記第二教師データに係る前記画像データは、白金異物に関する画像データのみを含んでもよい。 In the defect classification method according to the present invention, the image data related to the first teacher data may include only image data related to bubbles, and the image data related to the second teacher data contains only image data related to platinum foreign matter. It may be included.
　本発明者らの鋭意研究により、泡に関する画像データと白金異物に関する画像データとが混在する教師データにより機械学習を行って作成された分類器よりも、泡に関する画像データのみを含む第一教師データと、白金異物に関する画像データのみを含む第二教師データとにより作成された第一分類器及び第二分類器の方が、欠陥の分類に係る正解率が向上することが判明している。これにより、泡及び白金異物に関する欠陥の分類を精度良く行うことが可能になる。 According to the diligent research of the present inventors, the first teacher data containing only the image data related to bubbles rather than the classifier created by machine learning with the teacher data in which the image data related to bubbles and the image data related to platinum foreign matter are mixed. It has been found that the first classifier and the second classifier created by the second teacher data including only the image data related to the platinum foreign matter have a higher accuracy rate related to the defect classification. This makes it possible to accurately classify defects related to bubbles and platinum foreign substances.
　前記泡に関する前記画像データは、前記泡の形状及び／又は色に関する情報を含んでもよい。このように、第一教師データに泡の形状及び／又は色に情報を含ませることで、泡に関する欠陥の種別をその形状、色にまで細分化することが可能になる。 The image data regarding the bubble may include information regarding the shape and / or color of the bubble. In this way, by including the information in the shape and / or color of the bubble in the first teacher data, it is possible to subdivide the type of the defect related to the bubble into the shape and the color.
　前記白金異物に関する前記画像データは、前記白金異物の形状及び／又は色に関する情報を含んでもよい。このように、第二教師データに白金異物の形状及び／又は色に情報を含ませることで、白金異物に関する欠陥の種別をその形状、色にまで細分化することが可能になる。 The image data regarding the platinum foreign substance may include information regarding the shape and / or color of the platinum foreign substance. In this way, by including the information in the shape and / or color of the platinum foreign substance in the second teacher data, it is possible to subdivide the type of the defect related to the platinum foreign substance into the shape and color.
　本発明は上記の課題を解決するためのものであり、ガラス物品の製造方法であって、前記ガラス物品に含まれる欠陥の画像データを取得する撮像工程と、前記画像データに基づいて、上記いずれかの欠陥分類方法によって前記欠陥の種別を分類する工程と、を備えることを特徴とする。 The present invention is for solving the above-mentioned problems, and is a method for manufacturing a glass article, which is based on an imaging step of acquiring image data of defects contained in the glass article and any of the above based on the image data. It is characterized by comprising a step of classifying the type of the defect by the defect classification method.
　かかる構成によれば、撮像工程において取得された画像データに基づいて、上記の欠陥分類方法により欠陥の種別を分類することで、欠陥の検査に係る作業を効率良く行うことが可能になる。これにより、ガラス物品の製造効率を向上させることができる。 According to such a configuration, by classifying the types of defects by the above-mentioned defect classification method based on the image data acquired in the imaging process, it becomes possible to efficiently perform the work related to the defect inspection. Thereby, the manufacturing efficiency of the glass article can be improved.
　本発明は上記の課題を解決するためのものであり、ガラス物品に含まれる欠陥を画像データとして検出し、前記画像データに基づいて前記欠陥の種別を分類する装置であって、前記欠陥の種別は、第一分類と、第二分類とを含み、前記第一分類に係る第一教師データを機械学習することにより作成された第一分類器であって、前記画像データに基づいて前記欠陥が前記第一分類に属するか否かを判定する第一分類器と、前記第二分類に係る第二教師データを機械学習することにより作成された第二分類器であって、前記欠陥が前記第一分類に属さないと判定された場合に、前記画像データに基づいて前記欠陥が前記第二分類に属するか否かを判定する第二分類器と、を備えることを特徴とする。 The present invention is for solving the above-mentioned problems, and is an apparatus for detecting a defect contained in a glass article as image data and classifying the type of the defect based on the image data. Is a first classifier created by machine learning the first teacher data related to the first classification, including the first classification and the second classification, and the defect is found based on the image data. It is a first classifier that determines whether or not it belongs to the first classification, and a second classifier created by machine learning the second teacher data related to the second classification, and the defect is the first. It is characterized by including a second classifier that determines whether or not the defect belongs to the second category based on the image data when it is determined that the defect does not belong to one category.
　かかる構成によれば、欠陥分類装置によって欠陥の種別の分類を自動化することで、欠陥の検査に係る作業を効率良く行うことが可能になる。本発明に係る欠陥分類装置の第一分類器及び第二分類器は、第一教師データ及び第二教師データを機械学習することにより作成されたものであるため、欠陥の種別の分類を精度良く行うことができる。 According to such a configuration, by automating the classification of the type of defect by the defect classification device, it becomes possible to efficiently perform the work related to the inspection of the defect. Since the first classifier and the second classifier of the defect classification device according to the present invention are created by machine learning the first teacher data and the second teacher data, the classification of the defect type can be performed accurately. It can be carried out.
　本発明によれば、ガラス物品に係る欠陥の種別の分類を分類器で精度良くすることで、検査の作業効率を向上させることができる。 According to the present invention, it is possible to improve the work efficiency of inspection by accurately classifying the types of defects related to glass articles with a classifier.
ガラス物品の検査装置を示す側面図である。It is a side view which shows the inspection apparatus of a glass article. ガラス物品の検査装置を示す機能ブロック図である。It is a functional block diagram which shows the inspection apparatus of a glass article. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. 欠陥の画像データの例を示す写真である。It is a photograph which shows the example of the image data of a defect. ガラス物品の製造方法を示すフローチャートである。It is a flowchart which shows the manufacturing method of a glass article. 検査工程を示すフローチャートであるIt is a flowchart which shows an inspection process. 検査工程を示す側面図である。It is a side view which shows the inspection process. 分類工程を示すフローチャートである。It is a flowchart which shows the classification process.
　以下、本発明を実施するための形態について、図面を参照しながら説明する。図１乃至図１２は、本発明に係るガラス物品の製造方法及び当該方法に使用される検査装置の一実施形態を示す。以下では、ガラス物品の例としての透明なガラス板を製造する場合について説明する。 Hereinafter, a mode for carrying out the present invention will be described with reference to the drawings. 1 to 12 show a method for manufacturing a glass article according to the present invention and an embodiment of an inspection device used in the method. In the following, a case of manufacturing a transparent glass plate as an example of a glass article will be described.
　図１に示すように、検査装置１は、載置台２に載置されたガラス板Ｇを撮像する撮像装置３と、撮像装置３によって取得された画像データに対して画像処理を行う画像処理装置４と、当該画像データに含まれる欠陥の種別を分類する欠陥分類装置５と、を備える。 As shown in FIG. 1, the inspection device 1 includes an image pickup device 3 that captures an image of a glass plate G mounted on a mounting table 2, and an image processing device that performs image processing on image data acquired by the image pickup device 3. 4 and a defect classification device 5 for classifying the types of defects included in the image data.
　撮像装置３は、光学顕微鏡を備えており、ガラス板Ｇにおける表面Ｓ及び内部の拡大画像をデータとして取得できる。撮像装置３は、移動機構３ａにより、上下方向及び水平方向に対して移動可能に構成されている。撮像装置３は、所定ピッチで上下方向に移動することで、ガラス板Ｇの厚さ方向に対してその焦点の位置を変更できる。これにより、撮像装置３は、ガラス板Ｇの内部に含まれる欠陥を撮像することができる。撮像装置３は、画像処理装置４に接続されている。撮像装置３は、取得した画像データを画像処理装置４に送信する。 The image pickup device 3 is provided with an optical microscope, and can acquire a magnified image of the surface S and the inside of the glass plate G as data. The image pickup apparatus 3 is configured to be movable in the vertical direction and the horizontal direction by the moving mechanism 3a. The image pickup apparatus 3 can change the position of the focal point with respect to the thickness direction of the glass plate G by moving in the vertical direction at a predetermined pitch. As a result, the image pickup apparatus 3 can take an image of a defect contained inside the glass plate G. The image pickup device 3 is connected to the image processing device 4. The image pickup device 3 transmits the acquired image data to the image processing device 4.
　画像処理装置４は、ＣＰＵ、ＧＰＵ、ＲＡＭ、ＲＯＭ、ＨＤＤ、ＳＤＤ等を内蔵する他、ディスプレイ４ａ、入力インターフェース等の各種ハードウェアを実装するＰＣを含む。図２に示すように、画像処理装置４は、ＣＰＵ等により構成される演算処理部６と、ＨＤＤ、ＳＳＤ等により構成される記憶部７と、撮像装置３、欠陥分類装置５との間でデータの送受信を行う通信部８と、を備える。 The image processing device 4 includes a PC that incorporates various hardware such as a display 4a and an input interface, in addition to incorporating a CPU, GPU, RAM, ROM, HDD, SDD, and the like. As shown in FIG. 2, the image processing device 4 is located between an arithmetic processing unit 6 composed of a CPU or the like, a storage unit 7 composed of an HDD, an SSD, or the like, an image pickup device 3, and a defect classification device 5. It includes a communication unit 8 for transmitting and receiving data.
　画像処理装置４は、記憶部７に記憶される画像解析プログラムを演算処理部６により実行することで、撮像装置３の動作を制御する。画像処理装置４は、撮像装置３により取得される多数の画像データを当該撮像装置３から受信するとともに、各画像データに対して各種の画像処理を実行することができる。 The image processing device 4 controls the operation of the image pickup device 3 by executing the image analysis program stored in the storage unit 7 by the arithmetic processing unit 6. The image processing device 4 can receive a large number of image data acquired by the image pickup device 3 from the image pickup device 3 and execute various image processes on each image data.
　欠陥分類装置５は、ＣＰＵ、ＧＰＵ、ＲＡＭ、ＲＯＭ、ＨＤＤ、ＳＳＤ等のハードウェアを内蔵するＰＣを含む。図２に示すように、欠陥分類装置５は、ＣＰＵ等により構成される演算処理部９と、ＨＤＤ、ＳＳＤ等により構成される記憶部１０と、画像処理装置４との間でデータの送受信を行う通信部１１とを備える。 The defect classification device 5 includes a PC having built-in hardware such as a CPU, GPU, RAM, ROM, HDD, and SSD. As shown in FIG. 2, the defect classification device 5 transmits / receives data between the arithmetic processing unit 9 composed of a CPU or the like, the storage unit 10 composed of an HDD, SSD, or the like, and the image processing device 4. It is provided with a communication unit 11 for performing.
　演算処理部９は、記憶部１０に記憶されている画像解析プログラムを実行することにより、撮像装置３によって取得された画像データに含まれる欠陥の種別を分類することができる。また、演算処理部９は、画像解析プログラムを実行することにより、記憶部１０に保存される教師データを用いて、欠陥の種別を分類するための機械学習を行うことができる。 The arithmetic processing unit 9 can classify the types of defects included in the image data acquired by the image pickup apparatus 3 by executing the image analysis program stored in the storage unit 10. Further, the arithmetic processing unit 9 can perform machine learning for classifying the types of defects by using the teacher data stored in the storage unit 10 by executing the image analysis program.
　記憶部１０は、欠陥の分類を行うための画像解析プログラムを記憶している。画像解析プログラムには、例えばディープラーニングツール（プログラム）が組み込まれている。画像解析プログラムは、教師データに基づく機械学習（ディープラーニング）により、欠陥の種別を分類するための推論モデル（アルゴリズム）を作成することができる。 The storage unit 10 stores an image analysis program for classifying defects. For example, a deep learning tool (program) is incorporated in the image analysis program. The image analysis program can create an inference model (algorithm) for classifying the types of defects by machine learning (deep learning) based on teacher data.
　記憶部１０は、ガラス板Ｇに含まれる欠陥の種別を第一分類に分類する第一分類器１２と、欠陥の種別を第二分類に分類する第二分類器１３とを含む。各分類器１２，１３は、画像解析プログラムに係る機械学習によって演算処理部９が作成する推論モデルである。本実施形態では、欠陥の種別のうち、「泡」に関する欠陥が第一分類に設定され、「白金異物」に関する欠陥が第二分類に設定されている。 The storage unit 10 includes a first classifier 12 that classifies the types of defects contained in the glass plate G into the first classification, and a second classifier 13 that classifies the types of defects into the second classification. Each of the classifiers 12 and 13 is an inference model created by the arithmetic processing unit 9 by machine learning related to the image analysis program. In the present embodiment, among the types of defects, defects related to "foam" are set to the first category, and defects related to "platinum foreign matter" are set to the second category.
　欠陥としての泡は、何らかの気体を含む気泡、或いは気体を含まない真空気泡を含む。気体を含む気泡の場合、その気体の種類としては、酸素、二酸化炭素、一酸化炭素、ノックス（ＮＯX）、窒素、塩素、ブロム、水素、アルゴン、ヘリウム、ネオン、キセノン、水蒸気、ソックス（ＳＯX）、亜硫酸ガス等がある。また、真空気泡である場合には、気泡内壁に気泡生成時の成分が何らかの状態で固体として析出したものとなっている場合がある。 The bubbles as defects include bubbles containing some gas or vacuum bubbles containing no gas. In the case of bubbles containing gas, the types of gas are oxygen, carbon dioxide, carbon monoxide, nox (NO X ), nitrogen, chlorine, brom, hydrogen, argon, helium, neon, xenone, steam, socks (SO). X ), sulfurous acid gas, etc. Further, in the case of a vacuum bubble, the component at the time of bubble generation may be precipitated as a solid on the inner wall of the bubble in some state.
　「白金異物」に係る欠陥は、泡のように可視光に対して透過性を有するのではなく、遮蔽性を有する微細な白金により構成される。「白金異物」による欠陥は、例えばガラス板Ｇを成形するための溶融ガラスを供給及び移送する白金製の容器から生じた微小な白金が溶融ガラスに混入することによりガラス板Ｇの内部に形成される。 Defects related to "platinum foreign matter" are composed of fine platinum that has shielding properties rather than being transparent to visible light like bubbles. Defects due to "platinum foreign matter" are formed inside the glass plate G, for example, by mixing minute platinum generated from a platinum container for supplying and transferring molten glass for forming the glass plate G into the molten glass. To.
　第一分類器１２は、「泡」に関する複数の画像データを含む第一教師データに対して機械学習が行われることにより作成された推論モデルである。第一教師データは、「泡」に関する画像データのみを含み、「白金異物」に関する画像データを含まない。第一教師データにおける「泡」の画像データは、泡の形状及び／又は色に関する情報を含む。 The first classifier 12 is an inference model created by performing machine learning on the first teacher data including a plurality of image data related to "bubbles". The first teacher data includes only the image data related to "bubbles" and does not include the image data related to "platinum foreign matter". The image data of the "bubble" in the first teacher data includes information on the shape and / or color of the bubble.
　泡の形状に係る情報の例としては、例えば、「横長」、「縦長」、「寸法過大」などが挙げられる。「横長」に関する形状は、図３に示すように、横方向に延びる扁平状の泡の形状をいう。「縦長」に関する形状は、図４に示すように、縦方向に延びる扁平状の泡の形状をいう。「寸法過大」に関する形状は、図５に示すように、「横長」形状（又は「縦長」形状）のうち、画面の範囲に泡の全体形状が映し出されていないものを意味する。泡の「色」に関する情報の例としては、例えば「白」が挙げられる。具体的には、第一教師データの各画像データは、「横長」、「縦長」、「寸法過大」、「白」といった小分類に分類され、各画像データは小分類に紐づけられた状態で機械学習される。 Examples of information related to the shape of bubbles include "horizontal", "vertical", and "excessive dimensions". The shape related to "horizontally long" refers to the shape of a flat bubble extending in the lateral direction as shown in FIG. As shown in FIG. 4, the shape related to "vertical" refers to the shape of a flat bubble extending in the vertical direction. As shown in FIG. 5, the shape related to “excessive size” means a “horizontal” shape (or “vertical” shape) in which the entire shape of the bubble is not projected in the range of the screen. An example of information about the "color" of a bubble is "white". Specifically, each image data of the first teacher data is classified into sub-classifications such as "horizontal", "vertical", "excessive size", and "white", and each image data is associated with the sub-classification. Machine learning at.
　第二分類器１３は、「白金異物」に関する複数の画像データを含む第二教師データに対して機械学習が行われることにより作成された推論モデルである。第二教師データは、「白金異物」に関する画像データのみを含み、「泡」に関する画像データを含まない。第二教師データにおける「白金異物」の画像データは、白金異物の形状及び／又は色に関する情報を含む。 The second classifier 13 is an inference model created by performing machine learning on the second teacher data including a plurality of image data related to "platinum foreign matter". The second teacher data includes only the image data related to the "platinum foreign substance" and does not include the image data related to the "bubble". The image data of the "platinum foreign body" in the second teacher data includes information on the shape and / or color of the platinum foreign body.
　白金異物の形状に関する情報の例としては、例えば、「三角」、「多角」、「丸」などが挙げられる。「三角」に関する形状は、図６に示すように、三角形状を呈する白金異物の形状をいう。「多角」に関する形状は、図７に示すように、三角形を除く多角形状の白金異物の形状をいう。「丸」に関する形状は、図８に示すように、円形状の白金異物の形状をいう。また、白金異物の「色」に関する情報の例としては、例えば「白」が挙げられる。具体的には、第二教師データの各画像データは、「三角」、「多角」、「丸」、「白」といった小分類に分類され、各画像データは小分類に紐づけられた状態で機械学習される。 Examples of information on the shape of platinum foreign matter include "triangle", "polygon", and "circle". As shown in FIG. 6, the shape related to the "triangle" refers to the shape of a platinum foreign substance exhibiting a triangular shape. As shown in FIG. 7, the shape related to the “polygon” refers to the shape of a polygonal platinum foreign substance excluding a triangle. As shown in FIG. 8, the shape related to the “circle” refers to the shape of a circular platinum foreign substance. Further, as an example of the information regarding the "color" of the platinum foreign substance, for example, "white" can be mentioned. Specifically, each image data of the second teacher data is classified into sub-classifications such as "triangle", "polygon", "circle", and "white", and each image data is associated with the sub-classification. Machine learning.
　以下、上記の検査装置１を利用してガラス板Ｇを製造する方法について説明する。図９に示すように、本方法は、切断工程Ｓ１と、端面加工工程Ｓ２と、洗浄工程Ｓ３と、検査工程Ｓ４とを主に備える。 Hereinafter, a method of manufacturing the glass plate G using the above-mentioned inspection device 1 will be described. As shown in FIG. 9, this method mainly includes a cutting step S1, an end face processing step S2, a cleaning step S3, and an inspection step S4.
　切断工程Ｓ１に供給されるガラス原板は、例えばオーバーフローダウンドロー法を実行可能な成形装置によって、溶融ガラスから一枚（帯状）のガラスリボンを連続的に成形され、ガラスリボンから矩形状のガラス原板を切り出すことで得られる。なお、ガラスリボンの成形には、フロート法を用いてもよい。 The glass original plate supplied to the cutting step S1 is formed by continuously forming a single (strip-shaped) glass ribbon from the molten glass by, for example, a forming apparatus capable of executing an overflow down draw method, and a rectangular glass original plate is formed from the glass ribbon. It is obtained by cutting out. The float method may be used for molding the glass ribbon.
　切断工程Ｓ１では、ガラス原板を切断することにより、一枚又は複数枚の矩形状のガラス板Ｇが所望の寸法で切り出される。切断は、例えばガラス原板にスクライブ線を形成した後で折り割ることで行われる。 In the cutting step S1, by cutting the original glass plate, one or a plurality of rectangular glass plates G are cut out to a desired size. Cutting is performed, for example, by forming a scribe line on a glass original plate and then folding it.
　端面加工工程Ｓ２では、ガラス板Ｇの端面（切断面）を砥石で研削・研磨することにより、端面のクラックを除去する。 In the end face processing step S2, cracks on the end face are removed by grinding and polishing the end face (cut surface) of the glass plate G with a grindstone.
　洗浄工程Ｓ３では、端面加工工程Ｓ２を経たガラス板Ｇに洗浄処理が施される。 In the cleaning step S3, the glass plate G that has undergone the end face processing step S2 is cleaned.
　検査工程Ｓ４は、ガラス板Ｇにおける欠陥の有無を検査する第一検査工程と、ガラス板Ｇに欠陥が含まれる場合に、その欠陥を検査する第二検査工程とを含む。 The inspection step S4 includes a first inspection step of inspecting the presence or absence of defects in the glass plate G and a second inspection step of inspecting the defects when the glass plate G contains defects.
　第一検査工程では、検査装置１の上流側に配置される公知の検査装置（図示せず）により、ガラス板Ｇに対して欠陥の有無が検査される。この検査装置としては、例えばエッジライト方式の検査装置が使用されるが、これに限定されるものではない。ガラス板Ｇに欠陥が検出された場合、当該欠陥の位置（平面方向における位置）及び大きさに関する情報が画像処理装置４に入力され、その記憶部７に保存される。 In the first inspection step, the glass plate G is inspected for defects by a known inspection device (not shown) arranged on the upstream side of the inspection device 1. As the inspection device, for example, an edge light type inspection device is used, but the inspection device is not limited to this. When a defect is detected in the glass plate G, information on the position (position in the plane direction) and size of the defect is input to the image processing device 4 and stored in the storage unit 7.
　一方、ガラス板Ｇに欠陥が検出されなかった場合、欠陥が存在しない旨の情報が画像処理装置４に入力され、記憶部７に保存される。欠陥が検出されたガラス板Ｇのうちで、当該欠陥の大きさが基準（例えば１０μｍ程度）を超えるガラス板Ｇが第二検査工程においてさらに検査される。その他のガラス板Ｇは、良品と認定される。 On the other hand, when a defect is not detected in the glass plate G, information indicating that the defect does not exist is input to the image processing device 4 and stored in the storage unit 7. Among the glass plates G in which defects are detected, the glass plate G in which the size of the defects exceeds the standard (for example, about 10 μm) is further inspected in the second inspection step. Other glass plates G are certified as non-defective products.
　図１０に示すように、第二検査工程は、ガラス板Ｇに含まれる欠陥を撮像してその画像データを取得する撮像工程Ｓ４１と、画像データに含まれる欠陥の状態を測定する測定工程Ｓ４２と、欠陥の種別を分類する分類工程Ｓ４３と、ガラス板Ｇ（欠陥）の良否を判定する良否判定工程Ｓ４４と、を備える。 As shown in FIG. 10, the second inspection step includes an imaging step S41 for capturing an image of a defect contained in the glass plate G and acquiring the image data, and a measuring step S42 for measuring the state of the defect contained in the image data. A classification step S43 for classifying the types of defects and a quality determination step S44 for determining the quality of the glass plate G (defect) are provided.
　図１１に示すように、撮像工程Ｓ４１では、オペレータ（検査員）が画像処理装置４を操作することにより、又は自動制御により、撮像装置３によるガラス板Ｇの撮影が行われる。撮像工程Ｓ４１では、移動機構３ａの動作により、ガラス板Ｇに含まれている欠陥Ｆが視野内に位置するように、撮像装置３を移動させる。その後、撮像装置３は、所定のピッチで上下方向に移動しつつ、ガラス板Ｇの表面Ｓ及び内部を複数回にわたり撮像する。 As shown in FIG. 11, in the image pickup process S41, the image processing device 3 takes a picture of the glass plate G by the operator (inspector) operating the image processing device 4 or by automatic control. In the imaging step S41, the imaging device 3 is moved by the operation of the moving mechanism 3a so that the defect F contained in the glass plate G is located in the field of view. After that, the image pickup apparatus 3 images the surface S and the inside of the glass plate G a plurality of times while moving in the vertical direction at a predetermined pitch.
　撮像装置３によって取得されたガラス板Ｇの画像データは、画像処理装置４に送信される。画像処理装置４の演算処理部６は、受信した画像データを記憶部７に保存する。続いて、画像処理装置４の画像解析プログラムにより、欠陥Ｆの測定が行われる。 The image data of the glass plate G acquired by the image pickup device 3 is transmitted to the image processing device 4. The arithmetic processing unit 6 of the image processing device 4 stores the received image data in the storage unit 7. Subsequently, the defect F is measured by the image analysis program of the image processing device 4.
　測定工程Ｓ４２では、欠陥Ｆの大きさ及び深さが測定される。欠陥Ｆの大きさ（最大寸法）は、例えば、画像解析プログラムにより測定される。 In the measurement step S42, the size and depth of the defect F are measured. The size (maximum dimension) of the defect F is measured by, for example, an image analysis program.
　図１１に示すように、欠陥Ｆの深さＤは、ガラス板Ｇの厚さ方向において、表面Ｓから当該欠陥Ｆまでの距離を意味する。欠陥Ｆの深さＤは、例えば、以下のようにして測定される。 As shown in FIG. 11, the depth D of the defect F means the distance from the surface S to the defect F in the thickness direction of the glass plate G. The depth D of the defect F is measured, for example, as follows.
　すなわち、複数の焦点位置で撮像された欠陥Ｆに係る複数の画像データのうち、最も焦点の合う画像データを、画像処理装置４の画像解析プログラムにより選定する。そして、選定された画像データにおける撮像装置３の焦点位置（座標）からガラス板Ｇの表面Ｓまでの距離を、欠陥Ｆの深さＤとして算出する。測定された欠陥Ｆの大きさ及び深さＤに関する情報は、画像データとともに記憶部７に保存される。 That is, among the plurality of image data related to the defect F captured at the plurality of focal positions, the most focused image data is selected by the image analysis program of the image processing device 4. Then, the distance from the focal position (coordinates) of the image pickup apparatus 3 to the surface S of the glass plate G in the selected image data is calculated as the depth D of the defect F. Information on the measured size and depth D of the defect F is stored in the storage unit 7 together with the image data.
　分類工程Ｓ４３は、欠陥分類装置５によって欠陥Ｆの種別を分類する第一の分類工程と、オペレータが画像処理装置４を使用して欠陥Ｆの種別を分類する第二の分類工程と、を含む。 The classification step S43 includes a first classification step of classifying the type of the defect F by the defect classification device 5, and a second classification step of the operator classifying the type of the defect F using the image processing device 4. ..
　図１２に示すように、第一の分類工程は、第一分類器１２により欠陥Ｆが第一分類に属するか否かを判定する第一判定工程Ｓ４３１と、欠陥Ｆが第二分類に属するか否かを判定する第二判定工程Ｓ４３２と、を備える。 As shown in FIG. 12, the first classification step includes the first determination step S431 for determining whether or not the defect F belongs to the first classification by the first classifier 12, and whether the defect F belongs to the second classification. A second determination step S432 for determining whether or not to use is provided.
　第一判定工程Ｓ４３１において、演算処理部９は、第一分類器１２を起動させ、画像データに含まれる欠陥Ｆの種別が第一分類（「泡」）に属するか否かを判定する。本実施形態では、演算処理部９は、第一分類器１２で欠陥Ｆが「横長」、「縦長」、「寸法過大」、「白」といった小分類のいずれかに属するか否かを判定する。演算処理部９は、画像データに含まれる欠陥Ｆが「泡」に該当すると判定した場合、当該画像データに第一分類に係る分類情報を付加して記憶部１０に保存する。演算処理部９は、画像データの欠陥Ｆを第一分類器１２が「泡」でないと判定した場合、当該欠陥Ｆを「その他」に分類し、画像データにその分類情報を付加して記憶部１０に保存する。 In the first determination step S431, the arithmetic processing unit 9 activates the first classifier 12 and determines whether or not the type of the defect F included in the image data belongs to the first classification (“bubble”). In the present embodiment, the arithmetic processing unit 9 determines in the first classifier 12 whether or not the defect F belongs to any of the minor classifications such as "horizontal", "vertical", "excessive size", and "white". .. When the arithmetic processing unit 9 determines that the defect F included in the image data corresponds to the “bubble”, the arithmetic processing unit 9 adds the classification information related to the first classification to the image data and stores it in the storage unit 10. When the first classifier 12 determines that the defect F in the image data is not a "bubble", the arithmetic processing unit 9 classifies the defect F into "others", adds the classification information to the image data, and stores the defect F. Store at 10.
　第二判定工程Ｓ４３２において、演算処理部９は、第二分類器１３を起動させ、第一分類に属さないと判定された画像データ、すなわち「その他」に分類された画像データ（欠陥Ｆ）が第二分類に属するか否かを判定する。本実施形態では、演算処理部９は、第二分類器１３で欠陥Ｆが「三角」、「多角」、「丸」、「白」といった小分類のいずれかに属するか否かを判定する。演算処理部９は、第二分類器１３が画像データに含まれる欠陥Ｆを「白金異物」に該当すると判定した場合、当該画像データを第二分類に分類する。演算処理部９は、その画像データに第二分類に係る分類情報を付加して記憶部１０に保存する。 In the second determination step S432, the arithmetic processing unit 9 activates the second classifier 13, and the image data determined not to belong to the first classification, that is, the image data (defect F) classified as "other" is Determine whether or not it belongs to the second category. In the present embodiment, the arithmetic processing unit 9 determines in the second classifier 13 whether or not the defect F belongs to any of the sub-classifications such as "triangle", "polygon", "circle", and "white". When the second classifier 13 determines that the defect F included in the image data corresponds to the "platinum foreign substance", the arithmetic processing unit 9 classifies the image data into the second classification. The arithmetic processing unit 9 adds classification information related to the second classification to the image data and stores it in the storage unit 10.
　演算処理部９は、第二分類器１３が欠陥Ｆを「白金異物」に該当しないと認定した場合に、当該欠陥Ｆを「その他」に分類する。演算処理部９は、その画像データにその分類情報を付加して記憶部１０に保存する。 When the second classifier 13 determines that the defect F does not correspond to "platinum foreign matter", the arithmetic processing unit 9 classifies the defect F as "other". The arithmetic processing unit 9 adds the classification information to the image data and stores it in the storage unit 10.
　その後、演算処理部９は、分類が終了した画像データを画像処理装置４に送信する。画像処理装置４は、欠陥分類装置５から送信された分類済みの画像データを記憶部７に保存する。 After that, the arithmetic processing unit 9 transmits the image data for which the classification has been completed to the image processing device 4. The image processing device 4 stores the classified image data transmitted from the defect classification device 5 in the storage unit 7.
　画像処理装置４は、オペレータの操作により、または自動制御により、欠陥分類装置５から受信した画像データをディスプレイ４ａに表示させる。ディスプレイ４ａには、画像データとともにその分類情報（「泡」、「白金異物」、「その他」に関する情報）が表示される。 The image processing device 4 displays the image data received from the defect classification device 5 on the display 4a by the operation of the operator or by automatic control. The display 4a displays the classification information (information on "bubbles", "platinum foreign matter", and "others") together with the image data.
　第二の分類工程では、第一の分類工程で「その他」に分類された欠陥Ｆについて、オペレータが目視により欠陥Ｆの種別を認定する。具体的には、オペレータは、画像処理装置４のディスプレイ４ａに表示される画像データを目視により観察し、欠陥Ｆの種別（例えば「泡」、「白金異物」、「白金以外の異物」、「汚れ」、「傷」）を認定する。オペレータは、画像データに分類情報（認定結果）を付加して記憶部７に保存する。 In the second classification process, the operator visually certifies the type of defect F for the defect F classified as "other" in the first classification process. Specifically, the operator visually observes the image data displayed on the display 4a of the image processing device 4, and the type of defect F (for example, "bubble", "platinum foreign matter", "foreign matter other than platinum", "foreign matter other than platinum", " "Dirt", "Scratch") is certified. The operator adds classification information (certification result) to the image data and stores it in the storage unit 7.
　良否判定工程Ｓ４４において、画像処理装置４は、欠陥Ｆの大きさ（寸法）、深さＤ及び種別等の情報に基づいて、ガラス板Ｇの良否を認定する。具体的には、画像処理装置４は、欠陥Ｆがガラス板Ｇにおける表面Ｓの面性状に悪影響を及ぼす場合、すなわち、欠陥Ｆがガラス板Ｇの表面Ｓにおける凹凸等の原因になると判断すると、その画像データに対応する当該ガラス板Ｇを不良品と認定する。一方、画像処理装置４は、欠陥Ｆがガラス板Ｇの表面Ｓの面性状に悪影響を及ぼさない場合、当該ガラス板Ｇを良品と認定する。 In the quality determination step S44, the image processing device 4 determines the quality of the glass plate G based on information such as the size (dimensions), depth D, and type of the defect F. Specifically, when the image processing apparatus 4 determines that the defect F adversely affects the surface texture of the surface S of the glass plate G, that is, the defect F causes unevenness or the like on the surface S of the glass plate G. The glass plate G corresponding to the image data is recognized as a defective product. On the other hand, the image processing apparatus 4 recognizes the glass plate G as a non-defective product when the defect F does not adversely affect the surface properties of the surface S of the glass plate G.
　以上説明した本実施形態に係るガラス物品（ガラス板Ｇ）の製造方法、欠陥分類方法及び検査装置１によれば、検査工程Ｓ４において、画像データに含まれる欠陥Ｆの種別を欠陥分類装置５によって自動的に分類することで、検査工程Ｓ４の作業効率を高めることが可能になる。これにより、ガラス物品の製造効率を向上させることができる。 According to the manufacturing method, defect classification method, and inspection device 1 of the glass article (glass plate G) according to the present embodiment described above, in the inspection step S4, the type of the defect F included in the image data is classified by the defect classification device 5. By automatically classifying, it is possible to improve the work efficiency of the inspection step S4. Thereby, the manufacturing efficiency of the glass article can be improved.
　なお、本発明は、上記実施形態の構成に限定されるものではなく、上記した作用効果に限定されるものでもない。本発明は、本発明の要旨を逸脱しない範囲で種々の変更が可能である。 The present invention is not limited to the configuration of the above embodiment, and is not limited to the above-mentioned action and effect. The present invention can be modified in various ways without departing from the gist of the present invention.
　上記の実施形態では、画像処理装置４と欠陥分類装置５とを別々の装置により構成した検査装置１を例示したが、本発明はこの構成に限定されるものではない。例えば、画像処理装置４と欠陥分類装置５とを一台の装置により構成してもよい。 In the above embodiment, the inspection device 1 in which the image processing device 4 and the defect classification device 5 are configured by separate devices is illustrated, but the present invention is not limited to this configuration. For example, the image processing device 4 and the defect classification device 5 may be configured by one device.
　上記の実施形態では、ガラス物品としてガラス板Ｇの欠陥Ｆを分類する方法及び装置を例示したが、本発明はこの構成に限定されるものではない。本発明は、例えば管ガラスに含まれる欠陥を分類する場合にも適用可能である。 In the above embodiment, a method and an apparatus for classifying a defect F of a glass plate G as a glass article have been exemplified, but the present invention is not limited to this configuration. The present invention is also applicable, for example, to classify defects contained in tube glass.
　上記の実施形態では、欠陥分類装置５による第一の分類工程と、オペレータによる第二の分類工程とを実行する欠陥分類方法を例示したが、本発明はこの構成に限定されるものではない。本発明は、欠陥分類装置５のみによって分類工程Ｓ４３を実行することが可能である。 In the above embodiment, a defect classification method for executing the first classification step by the defect classification device 5 and the second classification step by the operator has been illustrated, but the present invention is not limited to this configuration. According to the present invention, the classification step S43 can be executed only by the defect classification device 5.
　上記の実施形態では、欠陥の種別を第一分類（「泡」）に分類する第一分類器１２と、欠陥の種別を第二分類（「白金異物」）に分類する第二分類器１３とを備える欠陥分類装置５を例示したが、本発明はこの構成に限定されるものではない。欠陥分類装置５は、例えば欠陥の種別を別の分類に分類する別の分類器をさらに一以上備えてもよい。また、第一分類は「泡」に限らず、別の欠陥の種別であってもよく、第二分類も「白金異物」に限らず、別の欠陥の種別であってもよい。 In the above embodiment, the first classifier 12 that classifies the types of defects into the first classification (“foam”) and the second classifier 13 that classifies the types of defects into the second classification (“platinum foreign matter”). Although the defect classification device 5 including the above is illustrated, the present invention is not limited to this configuration. The defect classification device 5 may further include, for example, one or more other classifiers that classify the types of defects into different classifications. Further, the first classification is not limited to "foam" and may be another type of defect, and the second classification is not limited to "platinum foreign matter" and may be another type of defect.
　上記の実施形態では、欠陥の深さにかかわらず第一分類器１２及び第二分類器１３を用いたが、欠陥の深さに応じた第一分類器１２及び第二分類器１３を準備し、測定された欠陥の深さに対応する第一分類器１２及び第二分類器１３で欠陥を分類してもよい。例えば、ガラス板の表面周辺の深さに位置する「泡」に関する第一教師データと、ガラス板の裏面周辺の深さに位置する「泡」に関する第一教師データと、表面と裏面の中間の深さに位置する「泡」に関する第一教師データとを準備する。これらの第一教師データをそれぞれ用い、表面周辺に位置する「泡」に関する第一分類器と、裏面周辺に位置する「泡」に関する第一分類器と、表面と裏面の中間に位置する「泡」に関する第一分類器とをそれぞれ作成する。 In the above embodiment, the first classifier 12 and the second classifier 13 are used regardless of the depth of the defect, but the first classifier 12 and the second classifier 13 according to the depth of the defect are prepared. , The defects may be classified by the first classifier 12 and the second classifier 13 corresponding to the measured depth of the defect. For example, the first teacher data about "foam" located at the depth around the front surface of the glass plate, the first teacher data about "foam" located at the depth around the back surface of the glass plate, and the middle between the front surface and the back surface. Prepare the first teacher data about the "bubble" located at the depth. Using these first teacher data, the first classifier for "foam" located around the front surface, the first classifier for "foam" located around the back surface, and the "foam" located between the front and back surfaces. Create the first classifier for each.
　この場合には、第一判定工程Ｓ４３１が実行される前に、欠陥が所定の深さであるか否かを判定する深さ判定工程を行うことができる。具体的には、深さ判定工程では、検出された画像データの欠陥が表面周辺、中間、裏面周辺のいずれに位置するかを判定する。そして、その後の第一判定工程Ｓ４３１では、深さ判定工程での判定結果に対応する位置の第一分類器を用いて欠陥の種別を分類する。この場合、第二分類器についても、第一分類器と同様に表面周辺、中間、裏面周辺に位置する「白金異物」に関する第二分類器を作成する。第二判定工程Ｓ４３２では、第一判定工程Ｓ４３１と同様に、深さ判定工程での判定結果に対応する位置の第二分類器を用いて欠陥の種別を判定する。 In this case, a depth determination step for determining whether or not the defect has a predetermined depth can be performed before the first determination step S431 is executed. Specifically, in the depth determination step, it is determined whether the detected image data defect is located near the front surface, the middle, or the back surface. Then, in the subsequent first determination step S431, the type of defect is classified using the first classifier at the position corresponding to the determination result in the depth determination step. In this case, as for the second classifier, as with the first classifier, a second classifier for "platinum foreign matter" located around the front surface, the middle, and the back surface is created. In the second determination step S432, similarly to the first determination step S431, the type of defect is determined by using the second classifier at the position corresponding to the determination result in the depth determination step.
　本発明者らは、画像データに含まれる欠陥の種別を精度良く分類するために、欠陥分類装置による効果的な機械学習の方法を確認する試験を行った。本発明者らは、「泡」、「白金異物」、「白金以外の異物」、「汚れ」、「傷」の属性を有する画像データが混在する教師データを作成し、欠陥分類装置による機械学習（ディープラーニング）を実施した。以下、この機械学習によって作成された欠陥の推論モデル（アルゴリズム）を比較例という。 The present inventors conducted a test to confirm an effective machine learning method by a defect classification device in order to accurately classify the types of defects contained in image data. The present inventors have created teacher data in which image data having the attributes of "bubble", "platinum foreign matter", "foreign matter other than platinum", "dirt", and "scratch" are mixed, and machine learning by a defect classification device. (Deep learning) was carried out. Hereinafter, the defect inference model (algorithm) created by this machine learning will be referred to as a comparative example.
　本発明者らは、「泡」の属性を有する画像データのみを含む教師データ（第一教師データ）を作成し、欠陥分類装置による機械学習を実施した。以下、この機械学習によって作成された欠陥の推論モデル（第一分類器）を実施例１という。 The present inventors created teacher data (first teacher data) including only image data having the attribute of "bubble", and carried out machine learning by a defect classification device. Hereinafter, the defect inference model (first classifier) created by this machine learning will be referred to as Example 1.
　また、本発明者らは、「白金異物」の属性を有する画像データのみを含む教師データ（第二教師データ）を作成し、欠陥分類装置による機械学習を実施した。以下、この機械学習によって作成された欠陥の推論モデル（第二分類器）を実施例２という。 In addition, the present inventors created teacher data (second teacher data) including only image data having the attribute of "platinum foreign matter", and carried out machine learning by a defect classification device. Hereinafter, the defect inference model (second classifier) created by this machine learning will be referred to as Example 2.
　この試験では、熟練の作業者（オペレータ）によって「泡」と判定された複数の試験用画像データを、比較例及び実施例１によって判定させ、その正解率を比較した。また、熟練の作業者によって「白金異物」と判定された複数の試験用画像データを、比較例及び実施例２によって判定させ、その正解率を比較した。また、「泡」に係る試験用画像データの数（第一教師データのデータ数）と、「白金異物」に係る試験用画像データの数（第二教師データのデータ数）とを同一とし、実施例１によって「泡」を判定させた場合の正解率と、実施例２によって「白金異物」を判定させた場合の正解率とを比較した。 In this test, a plurality of test image data judged as "bubbles" by a skilled worker (operator) were judged by Comparative Example and Example 1, and the correct answer rates were compared. In addition, a plurality of test image data determined to be "platinum foreign matter" by a skilled worker were determined by Comparative Example and Example 2, and the correct answer rates were compared. In addition, the number of test image data related to "bubbles" (the number of data of the first teacher data) and the number of test image data related to "platinum foreign matter" (the number of data of the second teacher data) are made the same. The correct answer rate when "bubbles" were determined according to Example 1 and the correct answer rate when "platinum foreign matter" was determined according to Example 2 were compared.
　試験の結果、「泡」の分類に関し、比較例の正解率は１００％、実施例１の正解率は１００％となり、比較例の正解率と実施例１の正解率が同等であったが、「白金異物」の分類に関し、比較例の正解率は７６％、実施例２の正解率は９７％となり、比較例の正解率よりも実施例２の正解率が高いことが判明した。 As a result of the test, regarding the classification of "foam", the correct answer rate of the comparative example was 100%, the correct answer rate of the example 1 was 100%, and the correct answer rate of the comparative example and the correct answer rate of the example 1 were the same. Regarding the classification of "platinum foreign matter", the correct answer rate of Comparative Example was 76%, and the correct answer rate of Example 2 was 97%, and it was found that the correct answer rate of Example 2 was higher than the correct answer rate of Comparative Example.
　この試験結果から、欠陥分類装置を使用して画像データに対する「泡」及び「白金異物」のによる分類を行う場合に、「泡」に関する分類（第一判定工程Ｓ４３１）を先に実行し、その後に「白金異物」に関する分類（第二判定工程Ｓ４３２）を実行することで、当該分類（特に比較例で正解率が低い「白金異物」に関する分類）を精度良くかつ効率良く行うことが可能となることが明らかとなった。 From this test result, when the image data is classified by "bubbles" and "platinum foreign substances" using the defect classification device, the classification related to "bubbles" (first determination step S431) is executed first, and then. By executing the classification related to "platinum foreign matter" (second determination step S432), the classification (particularly the classification related to "platinum foreign matter" having a low accuracy rate in the comparative example) can be performed accurately and efficiently. It became clear.
　５　　　　　　欠陥分類装置
１２　　　　　　第一分類器
１３　　　　　　第二分類器
　Ｄ　　　　　　欠陥の深さ
　Ｆ　　　　　　欠陥
　Ｇ　　　　　　ガラス板
　Ｓ４１　　　　撮像工程
　Ｓ４３　　　　分類工程
　Ｓ４３１　　　第一判定工程
　Ｓ４３２　　　第二判定工程 5 Defect classifier 12 1st classifier 13 2nd classifier D Depth of defect F Defect G Glass plate S41 Imaging process S43 Classification process S431 1st judgment process S432 2nd judgment process

Claims (8)

  1. 　ガラス物品に含まれる欠陥を画像データとして検出し、前記画像データに基づいて前記欠陥の種別を欠陥分類装置によって分類する方法であって、
    　前記欠陥の種別は、第一分類と、第二分類とを含み、
    　前記欠陥分類装置は、前記第一分類に係る第一教師データを機械学習することにより作成した第一分類器と、前記第二分類に係る第二教師データを機械学習することにより作成した第二分類器とを備え、
    　前記画像データに基づいて前記欠陥が前記第一分類に属するか否かを前記第一分類器によって判定する第一判定工程と、
    　前記欠陥が前記第一分類に属さないと判定された場合に、前記画像データに基づいて前記欠陥が前記第二分類に属するか否かを前記第二分類器によって判定する第二判定工程と、を備えることを特徴とする欠陥分類方法。     A method of detecting a defect contained in a glass article as image data and classifying the type of the defect by a defect classification device based on the image data.
    The types of defects include the first classification and the second classification.
    The defect classification device is a first classifier created by machine learning the first teacher data related to the first classification, and a second classifier created by machine learning the second teacher data related to the second classification. Equipped with a classifier,
    A first determination step of determining whether or not the defect belongs to the first classification based on the image data by the first classifier, and
    When it is determined that the defect does not belong to the first classification, the second determination step of determining whether or not the defect belongs to the second classification based on the image data and the second determination step. A defect classification method characterized by comprising.
  2. 　前記第一分類器の正解率が、前記第二分類器の正解率よりも高い請求項１に記載の欠陥分類方法。 The defect classification method according to claim 1, wherein the correct answer rate of the first classifier is higher than the correct answer rate of the second classifier.
  3. 　前記第一分類に係る前記第一教師データ及び前記第二分類に係る前記第二教師データは、前記ガラス物品の所定の深さに形成される前記欠陥の前記画像データを含み、
    　前記第一判定工程の前に、前記ガラス物品から検出された前記画像データにおける前記欠陥の深さが前記所定の深さである否かを判定する深さ判定工程を備え、
    　前記第一判定工程は、前記深さ判定工程において、前記欠陥の深さが前記所定の深さであると判定された場合に実行される請求項１又は２に記載の欠陥分類方法。     The first teacher data according to the first classification and the second teacher data according to the second classification include the image data of the defect formed at a predetermined depth of the glass article.
    Prior to the first determination step, a depth determination step of determining whether or not the depth of the defect in the image data detected from the glass article is the predetermined depth is provided.
    The defect classification method according to claim 1 or 2, wherein the first determination step is executed when the depth of the defect is determined to be the predetermined depth in the depth determination step.
  4. 　前記第一教師データに係る前記画像データは、泡に関する画像データのみを含み、
    　前記第二教師データに係る前記画像データは、白金異物に関する画像データのみを含む請求項１から３のいずれか一項に記載のガラス物品の欠陥分類方法。     The image data related to the first teacher data includes only image data related to bubbles, and includes only image data.
    The defect classification method for a glass article according to any one of claims 1 to 3, wherein the image data according to the second teacher data includes only image data relating to a platinum foreign substance.
  5. 　前記泡に関する前記画像データは、前記泡の形状及び／又は色に関する情報を含む請求項４に記載の欠陥分類方法。 The defect classification method according to claim 4, wherein the image data regarding the bubbles includes information regarding the shape and / or color of the bubbles.
  6. 　前記白金異物に関する前記画像データは、前記白金異物の形状及び／又は色に関する情報を含む請求項４又は５に記載の欠陥分類方法。 The defect classification method according to claim 4 or 5, wherein the image data relating to the platinum foreign substance includes information regarding the shape and / or color of the platinum foreign substance.
  7. 　ガラス物品の製造方法であって、
    　前記ガラス物品に含まれる欠陥の画像データを取得する撮像工程と、
    　前記画像データに基づいて、請求項１から６のいずれか一項に記載の欠陥分類方法によって前記欠陥の種別を分類する工程と、を備えることを特徴とするガラス物品の製造方法。     It is a manufacturing method of glass articles.
    An imaging process for acquiring image data of defects contained in the glass article, and
    A method for manufacturing a glass article, which comprises a step of classifying the type of the defect by the defect classification method according to any one of claims 1 to 6 based on the image data.
  8. 　ガラス物品に含まれる欠陥を画像データとして検出し、前記画像データに基づいて前記欠陥の種別を分類する装置であって、
    　前記欠陥の種別は、第一分類と、第二分類とを含み、
    　前記第一分類に係る第一教師データを機械学習することにより作成された第一分類器であって、前記画像データに基づいて前記欠陥が前記第一分類に属するか否かを判定する第一分類器と、
    　前記第二分類に係る第二教師データを機械学習することにより作成された第二分類器であって、前記欠陥が前記第一分類に属さないと判定された場合に、前記画像データに基づいて前記欠陥が前記第二分類に属するか否かを判定する第二分類器と、を備えることを特徴とする欠陥分類装置。     A device that detects defects contained in a glass article as image data and classifies the types of the defects based on the image data.
    The types of defects include the first classification and the second classification.
    A first classifier created by machine learning the first teacher data related to the first classification, and determining whether or not the defect belongs to the first classification based on the image data. With a classifier
    It is a second classifier created by machine learning the second teacher data related to the second classification, and when it is determined that the defect does not belong to the first classification, it is based on the image data. A defect classification device including a second classifier for determining whether or not the defect belongs to the second classification.
PCT/JP2020/043010 2019-12-17 2020-11-18 Fault classification method, fault classification device, and method for producing glass article WO2021124775A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202080071666.XA CN114556414A (en) 2019-12-17 2020-11-18 Defect classification method, defect classification device, and method for manufacturing glass article
KR1020227010768A KR20220113348A (en) 2019-12-17 2020-11-18 Defect classification method, defect classification apparatus, and method of manufacturing a glass article

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2019227239A JP7349066B2 (en) 2019-12-17 2019-12-17 Defect classification method, defect classification device and method for manufacturing glass articles
JP2019-227239 2019-12-17

Publications (1)

Publication Number Publication Date
WO2021124775A1 true WO2021124775A1 (en) 2021-06-24

Family

ID=76431082

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/043010 WO2021124775A1 (en) 2019-12-17 2020-11-18 Fault classification method, fault classification device, and method for producing glass article

Country Status (5)

Country Link
JP (1) JP7349066B2 (en)
KR (1) KR20220113348A (en)
CN (1) CN114556414A (en)
TW (1) TW202132766A (en)
WO (1) WO2021124775A1 (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023095756A1 (en) * 2021-11-29 2023-06-01 日本電気硝子株式会社 Method for producing glass plate

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004354251A (en) * 2003-05-29 2004-12-16 Nidek Co Ltd Defect inspection device
JP2009103508A (en) * 2007-10-22 2009-05-14 Hitachi Ltd Defect classification method and device thereof
JP2012026982A (en) * 2010-07-27 2012-02-09 Panasonic Electric Works Sunx Co Ltd Inspection device
JP2015038441A (en) * 2013-08-19 2015-02-26 株式会社Ｓｃｒｅｅｎホールディングス Classifier acquisition method, defect classification method, defect classification device, and program
JP2017054239A (en) * 2015-09-08 2017-03-16 株式会社Ｓｃｒｅｅｎホールディングス Image classification device and image classification method
CN106645180A (en) * 2017-02-06 2017-05-10 东旭科技集团有限公司 Method for checking defects of substrate glass, field terminal and server
CN109635856A (en) * 2018-11-29 2019-04-16 上海集成电路研发中心有限公司 A kind of producing line defect image intelligent classification system and classification method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP6067777B2 (en) 2015-04-27 2017-01-25 ＡｖａｎＳｔｒａｔｅ株式会社 Glass substrate for flat panel display, manufacturing method thereof, and liquid crystal display

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004354251A (en) * 2003-05-29 2004-12-16 Nidek Co Ltd Defect inspection device
JP2009103508A (en) * 2007-10-22 2009-05-14 Hitachi Ltd Defect classification method and device thereof
JP2012026982A (en) * 2010-07-27 2012-02-09 Panasonic Electric Works Sunx Co Ltd Inspection device
JP2015038441A (en) * 2013-08-19 2015-02-26 株式会社Ｓｃｒｅｅｎホールディングス Classifier acquisition method, defect classification method, defect classification device, and program
JP2017054239A (en) * 2015-09-08 2017-03-16 株式会社Ｓｃｒｅｅｎホールディングス Image classification device and image classification method
CN106645180A (en) * 2017-02-06 2017-05-10 东旭科技集团有限公司 Method for checking defects of substrate glass, field terminal and server
CN109635856A (en) * 2018-11-29 2019-04-16 上海集成电路研发中心有限公司 A kind of producing line defect image intelligent classification system and classification method

Also Published As

Publication number Publication date
JP7349066B2 (en) 2023-09-22
JP2021096151A (en) 2021-06-24
KR20220113348A (en) 2022-08-12
CN114556414A (en) 2022-05-27
TW202132766A (en) 2021-09-01

Similar Documents

Publication Publication Date Title
TWI637144B (en) Methods for determining a shape of a substantially cylindrical specular reflective surface
CN111696077A (en) Wafer defect detection method based on wafer Det network
JP2007134498A (en) Defective data processing and review device
TW200305545A (en) Manufacturing methods of glass substrate for liquid crystal display device and mother glass thereof, and mother glass inspection apparatus
CN108445020B (en) Glass substrate defect aggregation identification method
CN105092473B (en) A kind of quality determining method and system of polysilicon membrane
JP6788837B2 (en) Glass plate inspection method and its manufacturing method and glass plate inspection equipment
JP2006234771A (en) Method of inspecting surface defect of metal roll, and its device
WO2015014067A1 (en) Quality detection method, apparatus and device for liquid crystal screen
WO2021124775A1 (en) Fault classification method, fault classification device, and method for producing glass article
CN115035092A (en) Image-based bottle detection method, device, equipment and storage medium
CN205593939U (en) Surface form defect automatic checkout device of lens
TW201239347A (en) Method for checking membrane surface and apparatus using the same
CN105665919B (en) The system and method for on-line automatic repairing substrate defect
TW201415011A (en) Defect inspection method of plate-like body and defect inspection device
JP5247664B2 (en) Substrate inspection apparatus and measurement operation system thereof
JP2020085774A (en) Method for inspecting tube glass, method for learning, and tube glass inspection device
CN104297259A (en) Apparent defect detection method for touch screen protective glass
TW201617605A (en) Defect inspection method and apparatus thereof
JP2007057278A (en) Evaluator for cut sheet piece and disposition determining method for cut sheet piece
CN116997927A (en) Curved substrate bubble detection method and detection system
CN110998298B (en) Method for manufacturing plate-shaped glass
TW201901146A (en) Glass plate, method for inspecting end surface of glass plate, and method for manufacturing glass plate
CN215179799U (en) Float glass defect detection device
JP7309225B2 (en) Contact lens defect analysis and tracking system

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20903212

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20903212

Country of ref document: EP

Kind code of ref document: A1