WO2016043397A1

WO2016043397A1 - Glass defect detection method and apparatus using hyperspectral imaging technique

Info

Publication number: WO2016043397A1
Application number: PCT/KR2015/003634
Authority: WO
Inventors: 홍진광
Original assignee: 한화테크윈 주식회사
Priority date: 2014-09-16
Filing date: 2015-04-10
Publication date: 2016-03-24
Also published as: KR20160032593A; KR102003781B1

Abstract

Disclosed is a glass defect detection apparatus using a hyperspectral imaging technique. The glass defect detection apparatus comprises: a hyperspectral image generation unit for generating a hyperspectral image of glass to be inspected, using an image photographing device and a spectroscope; a hyperspectral image analyzing unit for displaying each defect corresponding to a predetermined inspection item as image data, spectroscopic data, and spatial data values, using the hyperspectral image; and a fault detection unit for detecting a fault among the defects using all of the displayed image data, spectroscopic data, and spatial data values.

Description

Glass defect detection method and device using hyperspectral imaging technique

The present invention relates to a glass defect detection method, and more particularly to a glass defect detection method using a hyperspectral image.

Conventionally, a plurality of images are irradiated at various angles to acquire images from a single camera, thereby obtaining images. In this way, a combination of four or more images obtained from a single camera or a combination of eight to twelve images with two or more cameras disposed at the top and the bottom thereof detects defects and good products. Even if there are multiple images under a large number of lighting conditions, expressing defects and good products in gray values has been limited, and there have been limitations in detecting defects by type during glass and film inspection. In particular, there is a problem that can not maximize the inspection yield because it does not classify the attribute information such as printing or discoloration.

In the present invention, to improve the accuracy of detecting various defects in the inspection of the appearance of the glass and film, and to provide a classification by the type of defects.

In a preferred embodiment of the present invention, a glass defect detection method using a hyperspectral imaging method generates a hyperspectral image of glass, and the hyperspectral image is composed of spectroscopic images of different bands. Selecting at least one spectroscopic image of a band including a defect corresponding to a predetermined test item in the generated hyperspectral image; Measuring a gray value for each defect in the spectroscopic image; Processing defect detection data for each defect in the spectroscopic image; And detecting defects by using a spectroscopic image selected for each band corresponding to the inspection item, measured shade values, and processed defect detection data.

In one preferred embodiment of the present invention has the effect of accurately detecting a variety of defects in the inspection of the appearance of the glass and film. In particular, it accurately detects defects such as printing or discoloration, and accurately classifies floating foreign substances such as dust and stains, thereby maximizing inspection yield.

1 illustrates a glass defect inspection apparatus for generating a hyperspectral image as an exemplary embodiment of the present invention.

FIG. 2 illustrates an example in which defects according to inspection items of the inspection target glass 140 are detected and spectral data for each defect is displayed.

FIG. 3 illustrates an embodiment in which a spectroscopic image of a band in which a defect is detected in a hyperspectral image of the inspection target glass 140 is detected as an exemplary embodiment of the present invention.

4 is a preferred embodiment of the present invention, a defect is detected according to an inspection item of the inspection target glass 140, and the wavelength value measured for each detected defect and a preset reference value are disclosed.

FIG. 5 illustrates, as a preferred embodiment of the present invention, a shade value measurement value and defect detection (Blob) data of a defect detected according to an inspection item of an inspection object glass 140.

6 shows an example of defect detection (Blob) data of a defect as a preferred embodiment of the present invention.

FIG. 7 illustrates an example of detecting a glass defect using all of image data, spectroscopic data, and spatial data as an exemplary embodiment of the present invention.

FIG. 8 illustrates a glass defect detection flowchart using a hyperspectral imaging technique according to an exemplary embodiment of the present invention.

Preferably, the defect detection data for the defect includes at least one or more of the coordinates, the center of gravity coordinates, the area, the circumferential length, and the diagonal length of each defect.

Preferably, the defects are classified by type using band information of a spectroscopic image in which each defect is detected.

According to another preferred embodiment of the present invention, a glass defect detection method using a hyperspectral imaging technique may include generating a hyperspectral image of glass using an imaging apparatus and a spectroscope; Acquiring image data, spectral data, and spatial data measurement values for the defects corresponding to each of the preset inspection items using the generated hyperspectral image; Analyzing preset image data, spectral data, and spatial data reference values and measured values of the image data, spectral data, and spatial data with respect to defects corresponding to each of the inspection items; And classifying a defect corresponding to the inspection item as defective when the image data, the spectral data, and the spatial data measurement values exceed the preset image data, the spectral data, and the spatial data reference values. .

In another preferred embodiment of the present invention, a glass defect detection method using a hyperspectral imaging technique includes: generating a hyperspectral image of an inspection target glass using an imaging apparatus and a spectroscope; A hyperspectral image analysis step of displaying each defect corresponding to a predetermined inspection item in the hyperspectral image as image data, spectral data and spatial data values; and using all of the displayed image data, spectral data and spatial data values Detecting whether each defect is defective.

As the inventive concept allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. Effects and features of the present invention, and methods of achieving them will be apparent with reference to the embodiments described below in detail with reference to the drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various forms.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, and the same or corresponding components will be denoted by the same reference numerals, and redundant description thereof will be omitted. .

In the following embodiments, the terms first, second, etc. are used for the purpose of distinguishing one component from other components rather than a restrictive meaning.

In the following embodiments, the singular forms “a”, “an” and “the” include plural forms unless the context clearly indicates otherwise.

In the following embodiments, terms such as include or have means that the features or components described in the specification are present, and does not preclude the possibility of adding one or more other features or components.

In the drawings, components may be exaggerated or reduced in size for convenience of description. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for convenience of description, and thus the present invention is not necessarily limited to the illustrated.

The glass defect inspection apparatus 100 includes a hyperspectral camera 102 and a hyperspectral image processing apparatus 150. The hyperspectral camera 102 includes image capturing apparatuses 110 and 120, a spectrometer 130, a slit 132, and a focusing lens 134. Image capturing apparatus (110, 120) is composed of a camera 110 and a detector (120).

The glass defect inspection apparatus 100 generates a hyperspectral image of the inspection target glass 140 by using the image photographing apparatus 110 and 120 and the spectrometer 130. Hyperspectral image refers to an image having a very high spectral resolution composed of more than one hundred consecutive bands or channel-specific spectroscopic images.

The spectroscope 130 disperses the light incident from the optical lenses of the imaging apparatuses 110 and 120 into a spatial pixel and a spectral pixel in a hyperspectral camera using a diffraction grating.

Examples of the spectrometer 130 include a reflective grating diffractometer, an acoustic optical bragg grating diffractometer and a volume phase holographic grating diffractometer, and a spectral sensor of a Fabry-Perrot Filter directly applied to a detector. And does not exclude additional spectroscopic components. In addition, examples of the inspection target glass 140 include a bare glass, a print glass, a touch screen pattern, a smartphone cover glass and a film.

The hyperspectral image provides spectroscopic information, spatial information, and image information for each pixel of the image by combining the image photographing apparatus 110 and 120 and the spectrometer 130.

The glass defect inspection apparatus 100 irradiates an illumination light source to the glass 140 to be inspected, and hyperspectrally distributes light information of the glass 140 that is reflected, transmitted, excited, and absorbed from the irradiated light source in the form of a hyperspectral image. The image processing apparatus 150 acquires the image and displays the test result on the screen display unit. The illumination light source generator may be manufactured as a light source having a short wavelength or multiple wavelengths according to the purpose of the inspection object, and may be installed at the upper and lower portions thereof and irradiate the sample. In addition, when the hyperspectral image is acquired, an external unwanted noise light source may be incident to the hyperspectral camera, and thus, the hyperspectral image generator should be manufactured in a dark zone condition to minimize errors. The hyperspectral image processing apparatus 150 may be implemented together with the glass defect inspection apparatus 100 or as a separate device capable of wired and wireless communication with the glass defect inspection apparatus 100. Examples of the hyperspectral image processing apparatus 150 include a thin terminal, a handheld device, a computer, a smartphone, a notebook, and the like.

The hyperspectral image processing apparatus 150 may display the captured hyperspectral image as image data, spectral data, and spatial data values. In addition, the hyperspectral image processing apparatus 150 may detect a defect in the hyperspectral image by using the displayed image data, spectral data, and spatial data values. The hyperspectral image processing apparatus 150 displays the defects corresponding to the preset inspection item by using the photographed hyperspectral image as image data, spectral data, and spatial data values, and displays the displayed image data, spectral data, and spatial data values. All of the defects can be detected using the above.

As an example of the image data, a gray scale value may be used. From the image data of the hyperspectral image generated by the glass defect inspection apparatus 100, one or more images of spectral bands can be selected to display a joke value as a value between 0 and 255 (8 bits), and the expression of the joke value is 0 to 255. It is not limited to (8 bits). In this regard, reference is made to the description related to FIGS. 5 to 6.

The spectral data uses a wavelength value of the hyperspectral image. According to an exemplary embodiment of the present disclosure, the glass defect inspection apparatus 100 may reclassify the detected defects by type by using a characteristic that the detected defects have different wavelengths for each property. In this regard, reference is made to the description of FIGS. 2 to 4.

The spatial data may use (x, y) position or coordinate value of each pixel of the hyperspectral image. For example, coordinate values such as 100 pixels, 100 pixels or 10 mm, 10 mm may be used.

2 illustrates an example in which defects according to inspection items of the inspection object glass 140 are detected and spectral data for each defect is displayed.

The spectral data includes wavelength information of an image for each wavelength band that is continuously obtained for each unit of spectral resolution in the spectral region. For example, when the spectral region is 400 nm to 1,000 nm and the spectral resolution is 2 nm, the spectral data can be obtained from the images obtained in the order of 400 nm, 402 nm, 404 nm, ..., 1000 nm.

When inspecting the defect of the glass sample, the inspection items are print color 210, print discoloration 212, scratches 214, 234, foreign objects 216, 218, 220, 222, imprint 224, chipping 226 , Blobs 228, Mura 230, dents 232, IR or camera hole defects 236.

Referring to FIG. 2, the print color 210 is lambda 1, the print discoloration 212 is lambda 2, and the scratches 214 and 234 are lambda 3 in the window area, and the scratch occurs across the printing unit and the window area. In the case of (234), the spectral data value was measured as λ11.

In the case of the foreign material, the spectral data values of the dust 216 were λ 4, the hair 218 was λ 5, the salinity was (220) λ 6, and the conveyor belt 222 was λ 7.

In addition, the image 224 is λ8, the chipping 226 is λ9, the stain 228 is λ10, the Mura 230 is λ13, the dent 232 is λ12, and IR or camera hole defects. (236) was measured for spectral data as λ14.

FIG. 3 illustrates an embodiment of reclassifying the defects detected by the glass defect inspection apparatus (FIGS. 1 and 100) according to a preferred embodiment of the present invention.

The glass defect inspection apparatus (FIGS. 1 and 100) may reclassify the hyperspectral image generated for the inspection target glass 310 for each type of detected defect. The glass defect inspection apparatus (FIGS. 1 and 100) may classify an image by material and display the hyperspectral image composed of images captured by one or more consecutive bands or channels (310, 312, 314, 316). , 318, 320, 330 and 340).

For example, a defect of bare glass when having a wavelength of λ1 310, a defect of printing when having a wavelength of λ2 312, a defect of printing when having a wavelength of λ2 312, and a floating foreign material defect when having a wavelength of λ3 to λ4 (314, 316). For example, if the wavelength is λ5 (318), it is a fixed foreign matter defect, if it has a wavelength of λ6 (320), IR or a camera hole defect, if it has a wavelength of λ7 (322), the print blot defect, λ6 (320) In the case of having a wavelength of, it is classified as a chipping defect in the case of having a wavelength distributed in the range of λ 10 to λ 30 330, and as a scratch in the case of having a wavelength distributed in the range of λ 20 to λ 40 340. In addition, the criteria of defect classification for each wavelength λ are not limited thereto, and may be applied in various combinations.

FIG. 4 illustrates an embodiment in which spectroscopic data of defects detected in a hyperspectral image of a glass to be inspected is measured as an exemplary embodiment of the present invention.

The hyperspectral image 400 of the smartphone cover glass is largely divided into the print area 401 and the window area 402. The inspection item of the print area 401 includes an IR hole 410, a print color 412, a color change 414, and a scratch 416. Inspection items of the window area 402 include dust 418, conveyor belt 420, stain 422, scratch 424, and stamp 426.

As a result of measuring the hyperspectral image 400 of the smart phone cover glass, in the printing area 401, the IR hole 410 is 630 nm, the printing color 412 is 455 nm, the printing discoloration 414 is 480 nm, and the scratch 416 is The wavelength value of 495 nm was measured.

In a preferred embodiment of the present invention, it is checked whether the wavelength measurement value 430 measured according to each inspection item exceeds the wavelength reference value 440 and the wavelength error range 450.

For example, referring to FIG. 4, in the case of printing discoloration 414 in the printing area (FIGS. 4 and 401), the wavelength measurement value 430 is 480 nm to 484 nm, while the wavelength reference value is 451 nm and the wavelength error is +/−. 5 nm appears to exceed the tolerance. Thus, the print discoloration 414 inspection item is detected as a defect.

Figure 5 is a preferred embodiment of the present invention, as a preferred embodiment of the present invention, measured by inspection items (510, 512, 514, 516, 518, 520, 522, 524, 526) of the inspection target glass The measurement values 530 and 550 of the image data and the spatial data and the predetermined spatial data reference value 552 for each inspection item are shown.

FIG. 6 shows defect detection (Blob) data of the spot 522 detected in the window region 502 of FIG. Defect detection (Blob) data includes the coordinates (x, y) of the inspection item, the center of gravity (600), the area size, the length of the diagonal (S600) and the length of the circumference (S610).

In a preferred embodiment of the present invention, after processing the hyperspectral image, defect detection data for defects in the processed hyperspectral image may be extracted and used. In this case, image processing is performed to detect defects, among which a smoothing filter (averaging, Gaussian, etc.) for removing noise, an edge filter (Sobel filter, Canny filter) for contour detection, a morphological morphology processing technique (Erosion, swelling, removal, filling, etc.), defect detection functions (number of defects, location, center, area, contour information, etc.), image binarization (single binarization, dynamic binarization) However, the present invention is not limited thereto and may include all techniques for recognizing a defect of an object in an image.

In a preferred embodiment of the present invention, after acquiring the hyperspectral image 500 of the smart phone cover glass, as shown in FIG. 3, each spectroscopic image in which a defect is detected is extracted at least or more (FIG. 3, 310, 312, 314, 316, 318, 320, 330 and 340. Thereafter, it is determined whether the wavelength of the defect detected in each spectroscopic image exceeds a predetermined range, and primarily, whether the defect belongs to the defect.

Next, the gray value of the detected defect is measured as in the exemplary embodiment illustrated in FIG. 5 (530). In this process, the image is smoothed using a smoothing filter, the contour detection filter is used to detect the contour of the test object, and the morphological morphology processing technique is used to express the morphological aspects of the image and perform various mathematical operations. Sometimes. In addition, the GV binary value 540 may be performed through a binarization filter according to a purpose before or after performing the above various image processing. Thereafter, the image processing method may be variously modified according to the purpose of detecting a defect, and the present invention may not be limited thereto and may include all techniques for recognizing a defect of an object in an image.

In an exemplary embodiment of the present invention, the defect detection data 550 may be derived using the GV binary value 540 and the defect detection function calculated in FIG. 5.

FIG. 7 illustrates an example of detecting a defect using all of spectral data, image data, and spatial data.

In a preferred embodiment of the present invention, the spectral image is selected using a spectroscopic data including a defect to be detected in the hyperspectral image of the inspection target (see FIG. 3). Thereafter, the gray value of the defect is measured using image data of the defect in the selected spectroscopic image (see FIG. 5). Thereafter, defect detection data is extracted from the binarized image obtained by binarizing the measured gray value using a defect detection function of a defect (see FIG. 5).

Using the spatial data information of the defects provided in the defect detection data (FIGS. 5 and 550), it is determined whether the predetermined area error (FIGS. 5 and 552) deviates, and finally, whether each defect is defective or not.

If described with reference to an embodiment of Figure 7 as follows.

In the hyperspectral image of the inspection target glass, a spectroscopic image including a defect corresponding to the inspection item is selected. A defect corresponding to a predetermined inspection item is detected in each of the print area 701 and the window area 702 in the selected spectroscopic image.

Examples of preset inspection items include IR 710, color 712, color change 714, scratch 716, foreign material 1 718 in the window area 702, and the like. There is foreign material 2 720, stain 722, scratch 724 and chipping 726.

Among the selected spectroscopic images, the bands 730 in which the IR 710 defects are detected are 630 nm, 632 nm, and 634 m, and the shaded values 740 are 51, 55, and 80 in each band. The bands in which the color 712 defect was detected were 450 nm and 452 nm, and the shaded value 740 was measured as 0 and 0 in each band.

Discoloration 714 was detected in the areas of 454nm and 458nm, and the shade value 740 was measured as 10 and 15 in each band. In addition, for the scratch 716, the foreign material 1 718, the foreign material 2 720, the stain 722, the scratch 724 and the chipping 726, the detected band and the shade value 740 in each band, respectively. Measure

The spectral data and the image data obtained by the above method are then subjected to image processing (742, 744, 746). Examples of image processing include at least one of image processing using a filter, mathematical and logical operation processing, and binarization processing.

Thereafter, the defect detection data of the defect is calculated using the gray value 740 value and the defect detection function (Blob) (750). Regarding the defect detection data, refer to the description of FIGS. 5 to 6. By using the spatial data of the defect detection data, it is determined whether the total number of defects for each band of the image, the individual positions of the defects, the individual centers, the individual areas, and the individual contour information pass the preset criteria 760.

The above process compares or calculates the difference with the predetermined standard and classifies it as good or bad, rework, etc. (S780).

8 is a flowchart illustrating a glass defect detection using a hyperspectral imaging technique according to a preferred embodiment of the present invention.

The glass defect inspection apparatus (FIGS. 1 and 100) of the present invention generates a hyperspectral image of the inspection target glass by using a hyperspectral imaging camera (FIGS. 1 and 102) (S800).

Thereafter, in the generated hyperspectral image, a spectroscopic image (see FIG. 3) of a band including a defect corresponding to a predetermined test item is selected (S810). In this case, one or more spectroscopic images may be selected.

The gray value of the defect is measured in the selected spectral image (S820). Thereafter, the selected spectral image is binarized by performing mathematical and logical operations or image capturing methods, and defect detection data for defects in the binarized hyperspectral image is calculated (S830).

Determining whether the specified criteria are met by using wavelength data (spectral band), gray value data, and defect detection data of the defect in the selected spectral image, and then according to the classification method, Are classified as bad (S840).

The present invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention.

Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the equivalent scope will be construed as being included in the present invention.

Claims

       Glass defect detection method using hyperspectral imaging technique,

       A hyperspectral image of glass is generated, and the hyperspectral image is composed of spectroscopic images of different bands, and the spectroscopic image of a band including defects corresponding to a predetermined test item is generated from the generated hyperspectral image. Selecting at least one;

       Measuring a shade value for each defect in the spectroscopic image;

       Calculating defect detection data for each defect in the spectroscopic image;

       And detecting a defect using a spectroscopic image, a measured shade value, and defect detection data selected for each band for each defect corresponding to the inspection item.
The method of claim 1, wherein the defect detection data for the defect is

And at least one of a total number of defects, a coordinate of each defect, a center of gravity coordinate, an area, a perimeter, and a diagonal length.
The method of claim 1, wherein the defects are classified by type using band information of the spectroscopic image in which each of the defects is detected.
       Glass defect detection method using hyperspectral imaging technique,

       Generating a hyperspectral image of the glass by using an imaging apparatus and a spectroscope;

       Measuring image data, spectral data, and spatial data measurement values for defects corresponding to each of the preset inspection items using the generated hyperspectral image;

       Comparing preset image data, spectral data, and spatial data reference values with respect to defects corresponding to each of the inspection items, and measurement values of the image data, spectral data, and spatial data; And

       And classifying the defect corresponding to the inspection item as defective by comparing or calculating the preset image data, the spectral data, and the spatial data measured values with the preset image data, the spectral data, and the spatial data reference values. Characterized in that the method.
The method of claim 4, wherein

And displaying the test items classified as defective on the terminal, wherein the terminal is capable of wired and wireless communication with the image photographing apparatus.
The method of claim 5, wherein the terminal

And reclassifying the defects by type based on the measured spectroscopic data.
The method of claim 4, wherein

The spatial data includes at least one of the total number of defects, coordinates of each defect, center of gravity coordinates, area, circumferential length (contour information), and diagonal length.
The method of claim 4, wherein the test item

At least one of print colors, print discolorations, scratches, foreign objects, imprints, chipping, stains, dents, mura, IR, or camera hole defects.
       Glass defect detection method using hyperspectral imaging technique,

       A hyperspectral image generation step of generating a hyperspectral image of the inspection target glass using an imaging apparatus and a spectroscope;

       A hyperspectral image analysis step of displaying defects corresponding to a predetermined inspection item in the hyperspectral image as image data, spectral data, and spatial data values; and

Detecting whether each of the defects is defective by using all of the displayed image data, spectroscopic data, and spatial data values.
       A glass defect detection device using hyperspectral imaging technique,

       A hyperspectral image generator for generating a hyperspectral image of a glass to be inspected by using an imaging apparatus and a spectroscope;

       A hyperspectral image analyzer for displaying each defect corresponding to a predetermined inspection item using the hyperspectral image as image data, spectral data, and spatial data values; and

       And a defect detector for detecting defects among the defects using all of the displayed image data, spectroscopic data, and spatial data values.
The method of claim 9, wherein the spatial data is

And at least one of a total number of defects, a coordinate of each defect, a center of gravity coordinate, an area, a perimeter, and a diagonal length.
The method of claim 9, wherein the defect detection unit

The wavelength value of each defect is used as the spectral data, the shade value of each defect is used as the image data, and the total number of defects, the coordinates of each defect, the center of gravity coordinate, the area, the perimeter length, and the diagonal length are used as the spatial data. At least one of the devices.
The method of claim 9,

And classifying defects by type using the spectroscopic data.
The method of claim 9, wherein the predetermined test item is

A device comprising at least one of print colors, print discolorations, scratches, foreign objects, imprints, chipping, stains, dents, mura, IR, or camera hole defects.
The method of claim 9, wherein the defect detection unit

And at least one of the displayed image data, spectral data, and spatial data values is determined to be a defect when out of a predetermined value.