Laser-based inclusion detection system and method

Abstract

Apparatus and methods for detecting inclusions in glass are described. The apparatus and methods employ a laser configured to project a laser sheet at a first angle from one side of a glass sheet, and a camera configured to capture images from the other side of the glass sheet at a second angle. While the camera captures images, the glass sheet moves past the laser sheet. One or more processing devices execute image processing algorithms to identify areas of the glass sheet containing inclusions based on the captured images. In some examples, the identified areas of the glass sheet are revisited to confirm that they contain inclusions.

Description

Technical Field

[0001] This application claims priority under patent law to U.S. Provisional Application No. 62 / 947,800, filed on December 13, 2019, the contents of which are relied upon and incorporated herein by reference in their entirety.

[0002] The present application relates to the detection of inclusions in glass; more particularly, the present application relates to apparatus and methods for detecting inclusions in thin, textured glass. Background Art

[0003] Glass sheets are used in a variety of applications. For example, glass sheets can be used in glass display panels, such as those found in mobile devices, laptops, tablet computers, computer monitors, and television displays. However, when glass sheets are produced, they may contain defects, such as inclusions or surface discontinuities. When certain defects appear on the surface of the glass, they may be referred to as "bumps." These bumps may be raised features that protrude above the surface of the surrounding (e.g., polished) glass. In some examples, inclusions may appear within the glass sheet. Glass manufacturers, for example, inspect glass sheets for quality control or sorting purposes in an attempt to detect these defects. Inclusions in the glass can result in functional (e.g., strength) defects or surface (e.g., visual appearance) defects.

[0004] In some conventional examples, human glass inspectors attempt to detect inclusions in glass sheets. In this implementation, a sheet of glass is illuminated from its edge (e.g., using darkfield illumination) while a highly trained inspector wearing a pair of magnifying glasses manually tilts and tilts the glass against a black background. In a time-consuming process, the inspector attempts to distinguish scattering centers within the glass volume from a pool of scattering centers caused by the rough glass surface. For example, with thin glass with a textured surface, light scattering due to the surface texture can produce a high density of false positives. In some examples, inclusions can be very small (e.g., as small as 10 μm in size), making detection more challenging. Furthermore, performance can vary significantly between inspectors due to differences in training, experience, and vision. Even for the same person performing such a demanding task, inclusion detection can degrade over time. Therefore, there is an opportunity to improve the detection of defects in glass sheets. Summary of the Invention

[0005] Features disclosed herein allow for the detection of defects such as inclusions in thin, textured glass, such as rolled sheet glass. In some examples, a laser sheet is projected onto the glass sheet at an oblique angle. An area scan camera is mounted on the other side of the glass sheet at another oblique angle. The scan camera captures an image of the intersection of the laser sheet and the glass. The glass passes through the laser sheet as the camera captures a series of images. Image processing algorithms allow for the detection of suspicious areas in the glass that may contain inclusions from the image data while ignoring noise that may be caused by surface defects in the glass. In some examples, the suspicious areas are re-imaged using high-resolution techniques to confirm and characterize the inclusions in the glass.

[0006] Among other advantages, embodiments may allow for the detection of small inclusions (e.g., <10 μm) in textured glass. Additionally, embodiments may reduce inspection turnaround time and significantly simplify conventional glass inspection processes. Furthermore, embodiments may allow for the detection of thin glass (e.g., ) and can suppress surface noise, such as speckle, texture, and contaminants, that conventional systems may detect and lead to erroneous inclusion detections (e.g., false positives). Furthermore, embodiments can allow glass sheets to be inspected faster than conventional methods. Additional benefits will be recognized by those skilled in the art having the benefit of this disclosure.

[0007] In some examples, the apparatus includes: a laser configured to project a laser sheet onto a first side of a glass sheet; and a camera configured to capture an image of the glass sheet from a second side of the glass sheet, wherein the camera captures the image using darkfield illumination. In some examples, the apparatus includes a motion stage configured to move the glass sheet past the laser sheet.

[0008] In some examples, the apparatus includes at least one processor configured to determine an area of ​​relatively high light intensity in the captured image. In some examples, the at least one processor is configured to identify inclusions in the glass sheet based on the area of ​​relatively high light intensity in the captured image.

[0009] In some examples, the at least one processor is configured to determine an area of ​​relatively higher light intensity in the captured image by identifying a top line of higher light intensity in the first image, identifying a bottom line of higher light intensity in the first image, and identifying a first area of ​​higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity in the first image. The at least one processor is also configured to determine inclusions in the first area based on the light intensities of the top line, the bottom line, and the first area.

[0010] In some examples, at least one processor is configured to determine an area of ​​relatively higher light intensity in the captured image by determining a first distance from a first area to a top line, and determining a second area in the second image based on the first distance, wherein the first area overlaps the top line of higher light intensity in the second image. The at least one processor is also configured to determine that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image.

[0011] In some examples, at least one processor is configured to determine an area of ​​relatively higher light intensity in the captured image by determining a second distance from the first area to a baseline, determining a third area in the third image based on the second distance, wherein the first area overlaps the baseline of higher light intensity in the third image, and determining that a first light intensity of the first area in the first image is greater than a third light intensity of the third area in the third image.

[0012] In some examples, at least one processor is configured to identify an area of ​​relatively high light intensity in a captured image by determining a first distance from the first area to a top line. The at least one processor is also configured to determine, for each of a first plurality of captured images, a first expected location of the first area in each image based on the first distance. Furthermore, the at least one processor is configured to determine a second distance from the first area to the bottom line. Furthermore, the at least one processor is configured to determine, for each of a second plurality of captured images, a second expected location of the first area in each image based on the second distance. The at least one processor is configured to determine the light intensity at each first expected location and at each second expected location. Furthermore, the at least one processor is configured to execute a machine learning algorithm to classify the light intensities and determine inclusions in the first area based on the classified light intensities.

[0013] In some examples, the apparatus includes: a laser configured to project a laser sheet onto a first side of a glass sheet; a first camera configured to capture an image of the glass sheet from a second side of the glass sheet; a motion stage configured to move the glass sheet past the laser sheet; and a microscope imaging camera configured to view the glass sheet from the second side of the glass sheet. In some examples, the first camera can capture the image using darkfield illumination.

[0014] In some examples, the apparatus includes: a laser configured to project a laser sheet onto a first side of a glass sheet; a first camera configured to capture an image of the glass sheet from a second side of the glass sheet; a motion stage configured to move the glass sheet past the laser sheet; and a second camera configured to capture a second image of the glass sheet from the first side of the glass sheet. In some examples, the first camera may capture the image using darkfield illumination, and the second camera may capture the second image using brightfield illumination. In some examples, the apparatus also includes a diffuse blue light emitting diode configured to provide light to the first side of the glass sheet, wherein the first camera includes a filter that blocks blue light. In some examples, the laser is a red line laser. In some examples, the first camera and the second camera are configured to capture the first image and the second image, respectively, simultaneously.

[0015] In some examples, an apparatus includes: a laser configured to project a laser sheet onto a first side of a glass sheet; and a first camera configured to capture an image of the glass sheet from a second side of the glass sheet, wherein the first camera uses darkfield illumination to capture the image. The apparatus also includes a motion stage configured to move the glass sheet past the laser sheet. The apparatus further includes: a backlight, such as a brightfield backlight, configured to project light onto the first side of the glass sheet; and a microscope configured to view the second side of the glass sheet.

[0016] In some examples, the motion stage is configured to move the glass sheet a predetermined distance past the laser sheet. In some examples, the predetermined distance is less than (or equal to) the width of the laser sheet (eg, the thickness of the laser line).

[0017] In some examples, a method using a processing device includes capturing an image of a glass sheet. The method may also include identifying a top line of higher light intensity in a first image of the captured image, and identifying a bottom line of higher light intensity in the first image. The method may further include identifying a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity in the first image. The method may also include determining an inclusion in the first region based on the light intensities of the top line, the bottom line, and the first region.

[0018] In some examples, the method includes determining a first distance from a first area to a top line; determining a second area in a second image of the captured image based on the first distance, wherein the first area overlaps the top line of higher light intensity in the second image; and determining that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image.

[0019] In some examples, the method includes determining a second distance from the first region to the baseline; determining a third region in the third image based on the second distance, wherein the first region overlaps the baseline of higher light intensity in the third image; and determining that a first light intensity of the first region in the first image is greater than a third light intensity of the third region in the third image.

[0020] In some examples, a method using a processing device includes capturing images of a glass sheet, identifying a top line of higher light intensity in a first image of the captured images, and identifying a bottom line of higher light intensity in the first image. The method further includes identifying a first region of higher light intensity between the top line of higher light intensity and a bottom line of higher light intensity in the first image. The method also includes determining a first distance from the first region to the top line. Furthermore, the method includes determining, for each of a first plurality of captured images, a first expected location of the first region in each image based on the first distance. The method also includes determining a second distance from the first region to the bottom line. The method further includes determining, for each of a second plurality of captured images, a second expected location of the first region in each image based on the second distance. The method also includes determining light intensities at each first expected location and each second expected location. Furthermore, the method includes executing a machine learning algorithm to classify the light intensities, and determining inclusions in the first region based on the classified light intensities.

[0021] In some examples, a non-transitory computer-readable medium has instructions stored thereon, wherein when executed by at least one processor, the instructions cause a computing device to perform operations including capturing an image of a glass sheet, identifying a top line of higher light intensity in a first image of the captured images, and identifying a bottom line of higher light intensity in the first image. The operations further include identifying a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity in the first image. The operations also include determining an inclusion in the first region based on the light intensities of the top line, the bottom line, and the first region.

[0022] In some examples, the operations include determining a first distance from a first area to a top line, determining a second area in a second image of the captured image based on the first distance, wherein the first area overlaps the top line of higher light intensity in the second image, and determining that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image.

[0023] In some examples, the operations include determining a second distance from the first area to the bottom line, determining a third area in the third image based on the second distance, wherein the first area overlaps the bottom line of higher light intensity in the third image, and determining that a first light intensity of the first area in the first image is greater than a third light intensity of the third area in the third image.

[0024] In some examples, a non-transitory computer-readable medium has instructions stored thereon, wherein when executed by at least one processor, the instructions cause a computing device to perform operations including: capturing an image of a glass sheet, identifying a top line of higher light intensity in a first image of the captured images, and identifying a bottom line of higher light intensity in the first image. The operations further include: identifying a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity in the first image. The operations also include: determining a first distance from the first region to the top line. Furthermore, the operations include: determining, for each of a first plurality of captured images, a first expected position of the first region in each image based on the first distance. The operations also include: determining a second distance from the first region to the bottom line. The method further includes: determining, for each of a second plurality of captured images, a second expected position of the first region in each image based on the second distance. The operations also include: determining the light intensity at each first expected position and each second expected position. Additionally, the operations include classifying the light intensities based on execution of a machine learning algorithm, and determining inclusions in the first region based on the classified light intensities. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above [Summary] and the following [Implementations] of illustrative embodiments should be read in conjunction with the accompanying drawings. The accompanying drawings illustrate some illustrative embodiments discussed herein. As further explained below, the claims are not limited to the illustrative embodiments. For clarity and ease of reading, the drawings may omit views of certain features.

[0026] Figure 1 An exemplary glass inclusion detection apparatus according to some examples is schematically illustrated.

[0027] Figure 2 is a block diagram of inclusion detection performed by an exemplary glass inclusion detection apparatus according to some examples.

[0028] Figure 3 Shows that when Figure 2 The glass inclusion detection equipment detects inclusions based on the light scattering at the location of the inclusions when they pass through the laser sheet.

[0029] Figure 4 Shown with Figure 3 The image of the light scattering intensity corresponding to the light scattering caused by the inclusions.

[0030] Figure 5 Shown is a comparison of light intensity graphs when inclusions are present in the glass and when surface texture is present in the glass.

[0031] Figure 6 is a block diagram of an exemplary glass inclusion detection apparatus with a water bath system for reducing surface scattering, according to some examples.

[0032] Figure 7A 、 Figure 7B and Figure 7C is a block diagram of an exemplary glass inclusion detection apparatus that employs a laser dark field system to further evaluate suspicious areas detected by a macro camera in a first pass, according to some examples.

[0033] Figure 8 An image of inclusions detected by the laser dark field system of FIG. 7 according to some examples is shown.

[0034] Figure 9 is a block diagram of an exemplary glass inclusion detection system employing a dark field camera and illustrating light intensity distribution when an inclusion passes through a laser sheet, according to some examples.

[0035] Figure 10A Shown are some examples of Figure 9 An exemplary glass inclusion detection device captures an image of inclusions in glass.

[0036] Figure 10B Shown are some examples of Figure 9 An exemplary glass inclusion detection device captures an image of surface defects on glass.

[0037] Figure 11 is a block diagram of an exemplary glass inclusion detection apparatus employing a dark-field camera, according to some examples.

[0038] Figure 12A Shown are some examples of Figure 11 Dynamic tracking images captured by an exemplary glass inclusion detection device.

[0039] Figure 12B Shown are some examples of Figure 11 A static tracking image captured by an exemplary glass inclusion detection device.

[0040] Figure 13 is a block diagram of an exemplary glass inclusion detection apparatus employing a brightfield camera, according to some examples.

[0041] Figure 14A Shown are some examples of Figure 13 Dynamic tracking images captured by an exemplary glass inclusion detection device.

[0042] Figure 14B Shown are some examples of Figure 13 A static tracking image captured by an exemplary glass inclusion detection device.

[0043] Figure 15 The use of the Sham principle in a camera system according to some examples is shown.

[0044] Figure 16 is a block diagram of an exemplary glass inclusion detection apparatus that employs a brightfield camera setup based on the Schaum principle to achieve an extended depth of field, according to some examples.

[0045] Figure 17 is a block diagram of a color confocal imaging system according to some examples that can be combined with any glass inclusion detection equipment to simultaneously scan glass for inclusions.

[0046] Figure 18 is a block diagram of a chromatic confocal sensor having surface and height depth measurement capabilities that can be combined with any glass inclusion detection equipment to simultaneously scan the glass for inclusions, according to some examples.

[0047] Figure 19 Exemplary methods that may be performed by a glass inclusion detection apparatus according to some examples are shown.

[0048] Figure 20 Another exemplary method that may be performed by a glass inclusion detection apparatus according to some examples is shown. DETAILED DESCRIPTION

[0049] This application discloses illustrative (i.e., exemplary) embodiments. This application is not limited to the illustrative embodiments. Therefore, many embodiments of the claims will differ from the illustrative embodiments. Various modifications may be made to the claims without departing from the spirit and scope of this application. The claims are intended to cover embodiments having such modifications.

[0050] Sometimes, this application uses directional terms (e.g., front, back, up, down, left, right, etc.) to provide context for the reader when reading the drawings. However, the claims are not limited to the directions shown in the drawings. Any absolute terms (e.g., upper and lower, etc.) should be understood as disclosing the corresponding relative terms (e.g., higher and lower, etc.).

[0051] This application proposes an apparatus and method for detecting inclusions in glass (such as thin glass with a textured surface). Inclusions in glass often lead to functional defects (such as strength defects) or surface defects (such as those that affect visual appearance).

[0052] Among other advantages, embodiments can reduce inspection turnaround time and significantly simplify conventional glass inspection processes. Additionally, embodiments can allow for the detection of small inclusions (e.g., <10 μm) in textured glass. Additionally, embodiments can allow for the inspection of thin glass (e.g., ), and embodiments can suppress surface noise, such as speckle, texture, and contaminants, that conventional systems may detect and lead to erroneous inclusion detections (e.g., false positives). Furthermore, embodiments can allow for faster inspection of glass sheets compared to conventional methods. Additional benefits will be recognized by those skilled in the art having the benefit of this disclosure.

[0053] In some examples, a glass inclusion detection apparatus includes a laser sheet generator (also known as a laser line generator), a camera with a lens, and a motion stage. The laser sheet can be a laser beam extending along one axis (e.g., a flat laser beam). For example, the laser sheet can be a laser beam having a length and a width, wherein the length is greater than the width. The laser sheet generator is configured to project the laser sheet onto one side (e.g., the top) of a glass sheet at an angle (e.g., angle α) relative to the glass sheet. For example, the angle between the laser sheet and the normal direction of the glass sheet can be adjustable. The camera is configured to capture an image from the other side (e.g., the bottom) of the glass sheet, wherein the camera is configured to capture the image at an angle (e.g., angle β) relative to the glass sheet. For example, the angle can be measured between the optical axis of the lens and the normal direction of the glass sheet. The angle of the laser relative to the glass sheet (e.g., angle α) and the angle of the camera relative to the glass sheet (e.g., angle β) can be adjusted to achieve optimal sensitivity in each inspection area captured in the image. In some examples, the angle of the laser relative to the glass sheet is greater than the angle of the camera relative to the glass sheet.

[0054] In some examples, a camera captures images using darkfield illumination (e.g., a darkfield camera). A darkfield camera can, for example, be a camera used in a darkfield imaging system in which the illumination source (whether directly from the illumination source to the camera or directly specularly reflected from an object) is not directly captured by the imaging system. Instead, the illumination source is used to illuminate the object. In some examples, as described herein, the illumination source is used to illuminate the glass sheet, and the camera captures only the scattered light from the glass sheet.

[0055] The motion stage is configured to move the glass sheet past the laser sheet. As the motion stage moves the glass sheet past the laser sheet, a camera captures images. The intersection between the glass and the laser sheet can be located at the center of the camera's field of view (FOV). If an inclusion is present, it scatters light from the laser sheet as it passes through it. The camera captures this scattered light.

[0056] In some examples, the motion stage moves the glass (e.g., horizontally) through the FOV of the laser sheet and the camera. In some examples, the motion stage includes an encoder (e.g., a processor) that is configured to send electrical pulses to the camera at periodic intervals (e.g., each time the glass sheet moves a predetermined distance). In some examples, the predetermined distance is less than the thickness of the laser sheet. Each time the camera receives an electrical pulse from the motion stage encoder, the camera captures an image. Images can be captured at a default exposure time. In some examples, the image captured by the camera shows an upper bright area (upper bright line area) due to scattering from the top surface, and a lower bright area (lower bright line area) due to scattering from the bottom surface. The scattering from the top surface may be caused by glass surface defects on the top and bottom surfaces, respectively. The area between the upper bright area and the lower bright area is referred to herein as an inspection region of interest (ROI). The captured image can be stored as image data in a memory such as a hard disk.

[0057] In some cases, the glass sheet is wider than the laser sheet. To inspect other areas of the glass, the motion stage can move the glass laterally a certain distance, then move the glass sheet horizontally across the laser sheet. In this way, glass sheets of various widths can be inspected.

[0058] Once the image is captured, a processing device, such as a processor, digital signal processor (DSP), computer, or server, can execute image processing algorithms to detect potential inclusions in the image data. As the glass traverses the laser sheet, inclusions that intersect the laser sheet will scatter light. These inclusions may appear as bright spots within the inspection ROI (Restricted Area of ​​Interest) of the image. The processing device can detect these bright areas in the image. Sufficiently large surface particles or surface specks can also be illuminated by surface scattering (e.g., light scattered by surface imperfections). In some cases, these surface particles can also cause relatively bright scattered signals (e.g., bright areas) to appear within the inspection ROI (Restricted Area of ​​Interest) of the image. These can be false detections (also known as "surface noise") due to surface imperfections rather than inclusions within the glass. Because surface specks can be orders of magnitude larger than actual inclusions, a significant portion of the detections may be false.

[0059] The processing device may filter out false detection results based on executing one or more filtering algorithms. The filtering algorithms may determine whether a bright spot should appear in the top surface scattering or bottom surface scattering of the image, and determine whether the bright spot is more intense in the ROI of the original image or the top surface scattering or bottom surface scattering of the determined image.

[0060] For example, assume that for a given image, a bright spot is detected in an image ROI at image coordinates defined by (C_n, R_n), where C represents image rows and R represents image columns. A static model can be employed, whereby the processing device calculates integers p and q based on the distances of the bright spot to surface scattering (top and bottom surface scattering). At image n+p and image nq, the scattering source of the bright spot is expected to overlap with the top and bottom surface scattering, respectively. The co-ordinated expected image coordinates (C_n+p, R_n+p) representing the expected bright spot in image n+p are also calculated. At image nq, the scattering source of the expected bright spot overlaps with the bottom surface scattering. The co-ordinated expected image coordinates (C_n-q, R_n-q) representing the expected bright spot in image np are also calculated.

[0061] The intensity of a bright spot in each of three images (e.g., image n, image n+p, and image nq) is determined. For example, a first intensity is determined based on the bright spot at coordinates (C_n, R_n) in the original image (image n). A second intensity is determined based on the expected coordinates (C_n+p, R_n+p) in image n+p, and a third intensity is determined based on the expected coordinates (C_n-q, R_n-q) in image nq.

[0062] In some examples, if a first intensity (representing a bright spot at coordinates (C_n, R_n) in the ROI of image n) is greater than both the second intensity and the third intensity, the bright spot is determined to be caused by an inclusion (e.g., an inclusion is detected). In some examples, the bright spot is determined to be caused by an inclusion when the first intensity is greater than the second intensity by at least a first threshold (e.g., first intensity > (second intensity + first threshold)) and the first intensity is greater than the third intensity by at least a second threshold (e.g., first intensity > (third intensity + second threshold)). In some examples, the first threshold and the second threshold are the same.

[0063] As another example, a dynamic model can be employed in which a sequence of integers [nqm, nq-m+1, ..., n+p+m] is generated, where m is a predefined integer used as a tuning parameter. The processing device calculates the expected coordinates (C_n-qm, R_n-qm), (C_n-q-m+1, R_n-q-m+1), ..., (C_n+p+m, R_n+p+m) of the scatter source in each image [nqm, nq-m+1, ..., n+p+m]. The processing device can then generate an array of intensity traces, where for i in [nqm, nq-m+1, ..., n+p+m], each element represents the intensity at (C_i, R_i) in image i. In some examples, the array of intensity traces can be a one-dimensional array of size (p+q+2*m+1).

[0064] The processing device can then employ a machine learning algorithm, such as a supervised classification algorithm (e.g., based on a support vector machine, a neural network, or a deep learning-based approach), to classify the array of intensity traces as inclusions or surface noise. The machine learning algorithm can be trained on supervised data that identifies light scattering caused by inclusions and light scattering caused by surface defects.

[0065] In some examples, both static and dynamic models may be employed to increase suppression of surface-related defects.

[0066] In some examples, the glass can be immersed in a liquid, such as water, to suppress or reduce surface noise. This option is advantageous for heavily textured glass. The liquid can partially match the refractive index of the glass surface, further reducing surface noise. In some examples, precise matching of the liquid and glass refractive indices is avoided because surface scattering, such as top and bottom surface scattering, is relied upon to locate inclusions. In addition to suppressing surface scattering, light is less bent by the glass surface. This allows for further separation of surface scattering and creates a wider inspection ROI.

[0067] In some examples, after filtering surface noise, inclusions can be revisited by one of a number of methods.

[0068] Microscope Revisited

[0069] In one example, a high-resolution microscope can be used to revisit each identified inclusion. The microscope can more precisely locate the inclusion within the glass. For example, the microscope's field of view is moved to the inclusion's location based on the inclusion's coordinates. The microscope is positioned so that the inclusion appears near the center of its field of view, where the microscope's optical axis is perpendicular to the glass. A brightfield backlight is positioned on the other side of the glass. Under the backlight, the microscope gradually makes various adjustments to shift the depth of field (DOF) from the bottom surface of the glass to the top surface (or vice versa) while the camera captures the image. The image can be stored in memory.

[0070] Once the image is captured, the processing device can determine the image in which the surface is most in focus (e.g., by executing an image processing algorithm) and can focus on any bright spots (due to brightfield backlighting) between the two surfaces. If any bright spots are detected between the two surfaces, the bright spots are determined to be caused by inclusions. The locations of the inclusions can be recorded. Machine learning models (e.g., deep learning-based algorithms) can be used to classify the inclusion type, and the size of the inclusions can be measured from the image. If no bright spots are detected between the two surfaces, the bright spots are determined to be caused by surface noise.

[0071] Laser Dark-Field-Assisted Microscopy Revisited

[0072] In another example, after filtering for surface noise, inclusions are revisited using laser darkfield. The laser is turned on, and the glass is then moved to a position where the inclusion scatters the laser light within the camera's field of view. A high-resolution microscope is positioned so that its field of view is focused on the glass from the other side of the glass (not the side from which the laser projects the laser sheet). The microscope is then focused on the scattering-based inclusion (e.g., until the scattering is most concentrated). The laser is then turned off, and a brightfield backlight is turned on. The brightfield backlight projects light onto the glass from the same side from which the laser projects the laser sheet. Inclusions are then inspected using the microscope as described above.

[0073] In some examples, such as where glass sheets (single or continuous) are transported on a linear conveyor (e.g., along the x-axis), embodiments can be adapted to perform a complete inspection (e.g., 100%) of the glass sheets. For example, an array of cameras and lasers can be positioned along the y-axis to cover the entire width of the glass, thereby inspecting all of the glass along the x-axis as it is moved by a motion stage.

[0074] In some examples, the glass inclusion detection apparatus may include a second camera. The second camera may be directed to the same surface of the glass as the laser directs the laser sheet. The second camera may be angled to the glass at the same angle (e.g., β) as the first camera from the other side of the glass. The first and second cameras may be calibrated to image the same area of ​​the glass (albeit from opposite sides). Inclusions may be determined from the image captured by the second camera, as described above for the first camera. The results may then be compared. For example, inclusions in images determined by both cameras may be identified, while inclusions in images determined by only one of the cameras may be discarded (e.g., ignored).

[0075] In some examples, solvents with varying degrees of residual residue (e.g., high-purity IPA, low-purity IPA, glass cleaner, or surface spray) can be used to modify (e.g., increase or decrease) the surface scattering amplitude and / or improve the surface scattering uniformity. Selecting and applying the appropriate surface treatment material prior to inclusion detection allows this technique to be applied to a variety of surface textures.

[0076] In some cases, the glass includes inclusions oriented in the same plane as the glass sheet. For example, 2D specular inclusions, which can include metal crystals as small as 5 μm, can be oriented in this manner in the glass. To increase the detection of these types of inclusions, several methods have been proposed.

[0077] Diffuse coaxial reflectance bright field (BF)

[0078] In one example, an optical system including a lens and a coaxial light (which incorporates a blue diffuse light emitting diode (LED)) is added to the glass inclusion detection device described above. The lens can be a 1X telecentric lens with a variable aperture (e.g., a depth of field of 0.5-1.5 mm). A blue diffuse LED (e.g., a diffuse coaxial LED with a wavelength of 1) is used for bright field, while a red laser (e.g., a line laser with a wavelength of 2) is used for dark field. The dark field camera has a filter that blocks blue light, which allows simultaneous scanning of the dark field and bright field without the LED interfering with the dark field measurement. The diffuse blue LED provides a more uniform reflection signal from a textured glass surface (e.g., a gray field). As a result, highly reflective local features such as metal crystals provide a significantly higher reflection signal than surfaces with good contrast and visibility.

[0079] While scanning the glass in steps (e.g., the same scan used for dark field), a dark field laser line is used during bright field imaging so that reflection events can be distinguished as surface noise or inclusions. The bright field image is used to identify and track reflection events (candidate inclusions or potential inclusions) between the laser lines. The signal is tracked (e.g., recorded) as the signal approaches and passes through both the top surface interception and the bottom surface interception of the glass as the laser sheet approaches. If the signal is reflected from a metallic inclusion, the reflection signal will not change significantly (e.g., become significantly brighter or darker) at these interceptions. In addition, the reflection signal will tend not to be significantly brighter than the adjacent laser line surface signal. Conversely, if the signal is reflected from a surface particle, the reflection signal will become significantly stronger at the top or bottom surface laser interception point, depending on which surface of the glass the particle is located.

[0080] Diffuse on-axis reflected brightfield (BF) with extended depth of field

[0081] In some examples, the Scham principle is applied to optical systems. This approach extends the system's depth of field and allows for focused imaging of metallic crystals regardless of their depth within the glass. For example, the system's image plane is tilted in the scanning direction, with the top surface laser line at one end of the image plane and the bottom surface at the other. For a glass thickness of 1 mm and a 45-degree laser angle, an example would be to tilt the image plane at a depth of 1 mm between the top and bottom laser lines over a width of approximately 0.53 mm. This results in a tilt angle of approximately 62 degrees. When scanning the glass, any metallic defects will be scanned through this tilted image plane. When a defect intersects the image plane, it will be clearly focused by the microscope. This device and method offer several benefits, including allowing for an increased depth of field, thus enabling scanning of thicker glass. Additionally, since defects appear in sharper focus, image quality and contrast can be improved. Furthermore, because the image plane is tilted between the top and bottom surfaces of the glass, and because the defect's location can be identified within the microscope's field of view in optimal focus, the depth of the inclusion can be estimated.

[0082] Color confocal imager

[0083] In this example, a color confocal imaging system images reflective inclusions in textured glass at a large depth of field (e.g., 3 mm). The system captures a sharp, focused image of the defect regardless of its depth. In some examples, color confocal imaging is added to the aforementioned glass inclusion detection equipment to simultaneously scan glass for all inclusion types (e.g., color confocal imaging can be used to identify specular reflective inclusions, while the glass inclusion detection equipment can be used to detect other inclusions).

[0084] Chromatic confocal sensor with surface depth measurement system

[0085] In this example, the chromatic confocal sensor detects reflective events and provides a depth measurement of the events. In this example, the depth of the defect is directly measured. In some examples, a chromatic confocal sensor with a surface depth measurement system is added to the above-described glass inclusion detection apparatus to simultaneously scan the glass for all inclusion types (e.g., the chromatic confocal sensor with a surface depth measurement system can be used to identify specular reflection inclusions, while the glass inclusion detection apparatus can be used to detect other inclusions).

[0086] Reference Figure 1 Glass inclusion detection apparatus 100 includes a laser 102, a camera 104, and a motion stage 108 supporting a glass sheet 106. Laser 102 is operable to provide a laser beam to a portion of a first side of glass sheet 106. The laser beam may have a wavelength between violet and infrared, for example. The laser beam may cover, for example, the width of glass sheet 106. Laser 102 may provide the laser beam at an angle (e.g., α) perpendicular to the first side of the glass sheet, and the angle may be adjustable.

[0087] Camera 104 can be operated to capture an image from the second side of glass sheet 106. Camera 104 can be, for example, a dark field camera. Camera 104 can be configured to capture images at an angle (e.g., β) between an optical axis of camera 104 and a normal direction of glass sheet 106. In some examples, laser 102 provides a laser sheet from the first side of glass sheet 106 at an angle greater than the angle at which camera 104 captures images from the second side of glass sheet 106 (e.g., angle α is greater than angle β).

[0088] The motion stage 108 can be operated to move the glass sheet 106 laterally in the x-direction (e.g., as shown in the illustration) across the field of view of the laser sheet and camera. In some examples, the motion stage 108 can be operated to move the glass sheet 106 in the y-direction. For example, the motion stage 108 can move the glass sheet 106 laterally in the x-direction. When the end of the glass sheet 106 is reached, the motion stage 108 can move the glass sheet 106 in the y-direction and then resume moving the glass sheet 106 in the x-direction. In this manner, the laser 102 can cover the entire area of ​​the glass sheet 106.

[0089] In some examples, motion stage 108 moves the glass sheet a predetermined distance in the x-direction. Camera 104 then captures an image, and motion stage 108 again moves the glass sheet a predetermined distance in the x-direction. For example, motion stage 108 may include an encoder that sends an electrical signal (e.g., a pulse) to camera 104 each time motion stage 108 moves glass sheet 106 a predetermined distance. When the camera receives the electrical signal from the motion stage 108 encoder, camera 104 captures an image at a default exposure time. In some examples, the predetermined distance is less than the thickness of the laser line (e.g., the width of the laser sheet).

[0090] Figure 2 A block diagram of a glass inclusion detection apparatus 100 is shown. As shown, a laser 102 provides a laser sheet 202 at an angle α 204 relative to the normal direction of the glass 106. Similarly, the optical axis of the camera 104 is located at an angle β relative to the normal direction of the glass sheet 106. The laser 102 and the camera 104 are on opposite sides of the glass sheet 106. As a motion stage (not shown in this figure) moves along the glass sheet 106, an inclusion 220 will eventually intersect the laser sheet 202. Furthermore, as the inclusion 220 moves across the laser sheet 202, the inclusion 220 moves "downward" within the field of view of the camera 104.

[0091] When inclusion 220 intersects laser sheet 202, laser sheet 202 scatters. Camera 104 can capture an image showing the scattering of laser sheet 202. Laser sheet 202 can also scatter when it intersects top surface defects 210 (top surface scattering) or bottom surface defects 212 (bottom surface scattering). Camera 104 can also capture an image showing this top and bottom surface scattering. Figure 2 Also shown are lower intensity laser sheet reflections 214 representing reflections of the laser sheet 202 from the inner portions of the top and bottom surface layers of the glass plate 106 .

[0092] Figure 3 The relative amount of scattering when laser sheet 202 intersects an inclusion is shown compared to when laser sheet 202 does not intersect the inclusion. For example, column 302 shows that for inclusion 310 on the top surface of glass sheet 106, the top surface scattering is stronger (e.g., greater) when inclusion 310 is in the path of laser sheet 202 compared to when inclusion 310 is outside the path of laser sheet 202. For example, when glass sheet 106 moves to a position where a particle on the top surface of glass sheet 106 is at the intersection between laser sheet 202 and the top surface of glass sheet 106, the particle will be brightest. Similarly, column 304 shows that for inclusion 312 embedded within glass sheet 106, the scattering is stronger when inclusion 312 is in the path of laser sheet 202 compared to when inclusion 312 appears before or after laser sheet 202. Column 306 shows that for inclusion 314 on the bottom surface of glass sheet 106 , the bottom layer scattering is stronger when inclusion 314 is in the path of laser sheet 202 than when inclusion 314 is out of the path of laser sheet 202 .

[0093] Figure 4Images of top surface scattering, scattering in a region of interest (ROI), and bottom surface scattering are shown in each of rows 410, 412, and 414. The image in row 410 represents an image captured by camera 104 when a particle or indentation is located on the top surface of glass sheet 106. In the image in column 402, bright areas can be seen, indicating top surface scattering caused by top surface particles. In the image in column 404, very little, if any, scattering is exhibited. Similarly, in the image in column 406, some bottom surface scattering is exhibited. Because the scattering intensity in the ROI, as shown in the image in column 404, is not greater than the top surface scattering in column 402, a processing device executing the static model described above will determine that this is not an inclusion.

[0094] The images in row 412 represent images captured by camera 104 when an inclusion within the glass is located within the ROI. Images in columns 402 and 406, respectively, exhibit very little top or bottom surface scattering (if any). However, column 404 shows a brighter area within the ROI. Because the scattering intensity within the ROI, as shown in the image in column 404, is greater than both the top surface scattering shown in the image in column 402 and the bottom surface scattering shown in the image in column 406, a processing device executing the aforementioned static model would determine that this is an inclusion.

[0095] The images in row 414 represent images captured by camera 104 when a particle or dent is located on the bottom surface of glass sheet 106. In the images in column 402, a slightly brighter area can be seen, indicating some top surface scattering. In the images in column 404, very little, if any, scattering is exhibited. However, in the images in column 406, stronger bottom surface scattering is exhibited. Because the scattering intensity within the ROI, as shown in the images in column 404, is no greater than the bottom surface scattering, as shown in the images in column 406, a processing device executing the static model described above would determine that this is not an inclusion.

[0096] Figure 5 Graph 500 and graph 550 are shown. Each of graphs 500 and 550 includes intensity along each vertical axis and vertical position along each horizontal axis (measured from the top column of each image). Graph 500 shows an intensity plot when inclusions are present, while graph 550 shows an intensity plot when surface defects (e.g., surface noise) are present. For example, graph 500 shows peak intensity 510; peak intensity 510 is located somewhere between top and bottom surface scattering. However, graph 550 shows peak intensity 512 located at the top surface scattering. The intensities of graphs 500 and 550 can be represented using the array of intensity traces described above with respect to the dynamic model.

[0097] Figure 6A glass inclusion detection apparatus 600 is shown; the glass inclusion detection apparatus 600 includes a bath container 610 filled with a liquid (e.g., water) 612. A glass sheet 106 is immersed in the liquid 162. For example, for heavier glass, the liquid 612 can partially match the refractive index of the glass surface to further reduce surface noise. The glass sheet 106 is moved within the bath container 610 by a motion stage 108 (not shown). The bath container 610 has two transparent windows 602 and 604 for illumination by the laser sheet 202 and observation by the camera 104. In addition to surface scattering suppression, light is less bent at the glass surface; therefore, the glass surface scattering is further separated and produces a wider inspection ROI. In this example, the angle β 206 is less than or equal to the angle α 204.

[0098] Figure 7A 、 Figure 7B and Figure 7C Examples of microscope-reviewed equipment and processes are shown. Figure 7A In FIG, the glass sheet 106 is moved to the position where the inclusion was previously found so that the inclusion scatters the laser sheet 202 in the field of view of the camera 104. Figure 7B In FIG. 1 , a microscope 704 (e.g., one with a field of view) is focused below the inclusion. The microscope 704 may have a higher magnification than the camera 104 and may be positioned at an angle relative to the glass sheet 106 that is different from the angle of the camera 104 relative to the glass sheet 106. The microscope 704 may locate the inclusion based on the scattering of the laser sheet 202 caused by the inclusion. Next, as Figure 7C As shown, laser darkfield microscopy is used. Here, laser 102 is turned off, and brightfield backlight 706 is turned on. The brightfield backlight is on the same side of the glass sheet as laser 102. With the brightfield backlight on, inclusions can be easily located using microscope 704. The location of the inclusion can be marked or recorded.

[0099] Figure 8 Image 802 in Figure 8 illustrates a re-examination without laser dark field. The inclusion in this image 802 is more difficult to detect than with laser dark field, as shown in image 804. Similarly, the small inclusion in image 806 is more difficult to detect without laser dark field. In image 808, the same inclusion is easily detected with laser dark field. With laser dark field, as shown in image 810, microscope 704 can be easily focused on the inclusion.

[0100] Figure 9A glass inclusion detection system 900 is shown; the glass inclusion detection system 900 includes the laser 102, a darkfield camera 902, and a re-interview camera 904. The re-interview camera 904 can be a higher-resolution microscopic imaging camera to verify the inclusions located by the darkfield camera 902. As shown, the glass sheet 106 is moved along a horizontal axis (via a motion stage, not shown in this figure). As the inclusion 910 moves across the laser sheet 202, the inclusion 910 moves "downward" from the field of view of the darkfield camera 902. When the inclusion 910 intersects the laser sheet 202, the darkfield camera 902 captures an image with a (e.g., most intense) bright spot (e.g., an image area with higher light intensity). For example, the light intensity curve 912 shows that when the inclusion 910 intersects the laser sheet 202, a peak intensity 914 is captured in the image.

[0101] For example, Figure 10A As shown in image 1002 , inclusions 910 may be displayed in the image captured by darkfield camera 902 ; inclusions 910 are shown as circled bright areas.

[0102] The dark field camera 902 may also capture an image having bright areas due to top surface scattering (caused by top surface defects 920 ) and bright areas due to bottom surface scattering (caused by bottom surface defects 922 ).

[0103] For example, Figure 10B As shown in image 1004 of FIG1004 , a top surface defect may appear in the image captured by darkfield camera 902, such as a bright area indicated by the circled bright area located near the top of image 1004. A "top line" of the bright area (top surface scattered lines) may be seen due to the top surface defect of glass sheet 104. Similarly, a bottom surface defect may appear in the image captured by darkfield camera 902, such as a bright area indicated by the circled bright area located near the bottom of image 1004. A "bottom line" of the bright area (bottom surface scattered lines) may be seen due to the bottom surface defect of glass sheet 104.

[0104] Return Reference Figure 10A , a circled bright area caused by inclusion 910 can be seen in the region of interest (ROI) between the top and bottom surface scattered lines. In this way, inclusions can be distinguished from surface events.

[0105] Figure 11A glass inclusion detection apparatus 1100 is shown; it includes a darkfield camera 904, a laser 102, a motion stage 108, and a glass sheet 106. The motion stage 108 can move the glass sheet 106 through and intersect the laser beam projected by the laser 102, and can pass through the field of view of the camera 902. For example, the laser 102 is turned on and the laser beam is projected onto the top side of the glass sheet 106; the motion stage 108 can move the glass sheet 106 horizontally (e.g., in the X direction) a predetermined distance (e.g., 26 μm). The predetermined distance can be less than (or equal to) the width of the laser beam (e.g., the width of the laser beam in the X direction). This ensures that the entire glass volume is scanned for inclusions, because regardless of the inclusion's position along the scan axis, the inclusion will always be intercepted by a portion of the laser beam and generate a scattered signal. The darkfield camera 902 can then be configured to capture an image of the glass sheet 106 from its bottom side. The motion stage 108 can then again move the glass sheet 106 a predetermined distance, and the dark field camera 902 can again capture an image. This process can continue until an image of the entire length of the glass sheet 106 is captured.

[0106] In some examples, motion stage 108 may move glass sheet 106 in a lateral direction (e.g., in the Y direction), and the above process may be repeated to capture images. For example, glass sheet 106 may be moved laterally by an amount less than (or equal to) the length of the laser sheet (e.g., the length of the laser sheet in the Y direction). In this manner, an image of the entire area of ​​glass sheet 106 (e.g., an image of the entire area defined by the length and width of glass sheet 106) may be captured.

[0107] Figure 12A An image 1202 composed of various frames is shown. Image 1202 shows bright areas caused by inclusions and detected by a processing device, such as an image processing device, that processes the image captured by darkfield camera 902. The box highlights the bright areas caused by the inclusions. Image 1202 also shows other bright areas, such as top scattered lines 1204 and bottom scattered lines 1206. In this example, the processing device employs the aforementioned dynamic model to identify the inclusions.

[0108] Figure 12BAlso shown is image 1250, comprised of various frames. In this example, the processing device detected an inclusion that resulted in the bright area (circled) in the center of the image. In this example, the processing device executed the static model described above. Image 1250 also illustrates the brightness of the inclusion captured when it entered the field of view of darkfield camera 902 (as glass sheet 106 was moved by motion stage 108), as indicated by the top circle. Image 1254 further illustrates the brightness of the inclusion captured when it left the field of view of darkfield camera 902, as indicated by the bottom circle. As shown, the circled area in the center of image 1204 is brighter than the circled areas at the top and bottom of image 1204. This is because the detected inclusion did not intersect the laser sheet until some time after entering the field of view of darkfield camera 902 and left the area covered by the laser sheet some time before leaving the field of view of darkfield camera 902.

[0109] Figure 13 A glass inclusion detection apparatus 1300 is shown, which includes a red line laser 1302, a dark field camera 902, and a bright field camera 1304. The red line laser 1302 generates a red laser sheet 1320 that intersects the glass sheet 106. The dark field camera 902 includes a red pass filter 904 that allows red light to pass. In this way, the dark field camera 902 can capture scattering events caused by inclusions in the glass sheet 106 that intersect the red laser sheet 1320.

[0110] Brightfield camera 1304 can capture images using brightfield illumination. For example, brightfield camera 1304 can be a camera used in a brightfield imaging system, in which the illumination source is captured directly by the imaging system, either by shining the illumination source directly onto the camera (e.g., transmissive brightfield) or by specular reflection of the illumination source from an object onto the camera (e.g., reflective brightfield). In the case of transmissive brightfield, light from the illumination source is transmitted through the object, and the camera directly images the light transmitted through the object. In the case of reflective brightfield, the illumination source is used to illuminate the object, and specular reflections from the object are directly captured by the camera. As described further below, brightfield camera 1304 is used to more reliably capture reflective metallic inclusions, which may produce little or no scattered signal (e.g., due to laser sheet 202). Brightfield camera 1304 is located on the opposite side of glass sheet 106 from darkfield camera 902. For example, brightfield camera 1304 is located on the same side as red line laser 1302. In this example, reflective brightfield is used. In other examples, the glass inclusion detection device 1300 may use transmission bright field.

[0111] The glass inclusion detection apparatus 1300 also includes a beam splitter 1308 having a diffuse blue LED 1308. The diffuse blue LED 1308 serves as the light source for the brightfield camera 1304. The beam splitter 1308 directs light from the diffuse blue LED 1308 toward the glass sheet 106. Specifically, the blue light from the diffuse blue LED 1308 is directed by the beam splitter 1308 toward the top surface of the glass sheet 106 (e.g., at a normal angle relative to the top surface of the glass sheet 106). The blue light can reflect off the glass sheet 106, and the brightfield camera 1304 can capture the reflection of the blue light. The diffuse blue LED 1308 provides a more uniform reflection signal from the textured glass surface of the glass sheet 106. Highly reflective local features, such as metal crystals, provide a much higher reflection signal than the reflection signal provided by the top surface of the glass sheet 106. Because the darkfield camera 902 includes a red pass filter 904, the blue light from the diffuse blue LED 1308 is blocked and is not seen by the darkfield camera 902. This allows scanning with the brightfield camera 1304 and the darkfield camera 902 to be performed simultaneously.

[0112] In some examples, darkfield camera 902 and brightfield camera 1304 are used simultaneously to distinguish whether a reflection event is a surface defect or an inclusion. For example, images captured by brightfield camera 1304 can be used to identify and track reflection events that occur between laser lines of red laser sheet 1320. As the reflection signal captured by brightfield camera 1304 (e.g., from diffuse blue LED 1308) approaches and passes through both the top and bottom surface interceptions of red laser sheet 1320 on glass sheet 106, the reflection signal is tracked. If the reflection signal is reflected from a metallic inclusion, the reflection signal will not change significantly at these interception points. For example, the reflection signal will not be significantly brighter than the reflection signal reflected from the adjacent laser line surface signal. If the reflection signal is instead reflected from a surface particle (e.g., a defect on the top or bottom surface of glass sheet 106), the reflection signal will become significantly stronger at the top or bottom surface laser interception point, depending on which surface of glass sheet 106 the particle is located. For example, if the particle is on the top surface of the glass sheet 106, the reflected signal will become significantly stronger at the top surface laser interception point, and if the particle is on the bottom surface of the glass sheet 106, the reflected signal will become significantly stronger at the bottom surface laser interception point.

[0113] Figure 14AAn image 1402 composed of various frames captured by brightfield camera 1304 is shown. Image 1404 shows a bright area, indicated by the box, caused by a metallic inclusion reflecting light from a diffuse blue LED 1308. As shown, because the signal is reflected from the metallic inclusion, the reflected signal does not significantly change between the top surface scattered line 1404 and the bottom surface scattered line 1406. In this example, the processing device detects the metallic inclusion by executing the aforementioned dynamic model.

[0114] Figure 14B An image 1450 is shown, composed of various frames and also captured by brightfield camera 1304. In this example, the processing device detected an inclusion that resulted in the bright area in the center of the image, which is circled. Image 1450 also shows the brightness of the inclusion captured when it entered the field of view of darkfield camera 902 (as motion stage 108 moved glass sheet 106) (as indicated by the top circle). Image 1450 also shows the brightness of the inclusion captured when it left the field of view of darkfield camera 902 (as indicated by the bottom circle). In this example, the processing device detected the metallic inclusion by executing the static model described above.

[0115] Figure 15 The use of the Sham principle in a camera system is illustrated. An image 1502 of an object 1506 is captured through a lens 1504. Image 1502 appears along an image plane 1512, and object 1506 is positioned along an object plane 1518. Lens 1504 is positioned along a lens plane 1510 and has a lens axis 1514. Image plane 1512, lens plane 1510, and object plane 1518 intersect at a Sham intersection 1508. As shown, image plane 1512 and lens plane 1510 are not parallel to each other. Additionally, object plane 1518 is not perpendicular to lens axis 1514. As a result, an extended depth of field 1520 of object 1506 can be achieved.

[0116] Figure 16A glass inclusion detection apparatus 1600 is shown; similar to the glass inclusion detection apparatus 1300, it utilizes the Sham principle. Consequently, compared to the glass inclusion detection apparatus 1300, the glass inclusion detection apparatus 1600 can provide a greater depth of field when capturing images using a brightfield camera 1304. In this example, the image plane of the brightfield camera 1304 is angled (e.g., tilted) toward the red laser sheet 1320, such that the top surface laser line of the red laser sheet 1320 (provided by the red line laser 1302) is at one end of the image plane, while the bottom surface laser line of the red laser sheet 1320 is at the other end. As the glass sheet 106 is scanned by the brightfield camera 1304, metal defects will reflect light from the diffuse blue LED 1308 and be received by the brightfield camera 1304 through this tilted image plane. When the metal defect intersects the image plane of the brightfield camera 1304, the metal defect will be in sharp focus.

[0117] Figure 17 A color confocal imaging system 1700 is shown; the color confocal imaging system 1700 can be combined with any of the glass inclusion detection devices described herein to simultaneously scan for inclusions in the glass. The color confocal imaging system 1700 includes a charge-coupled device (CCD) and / or complementary metal oxide semiconductor (CMOS) detector 1702, which can capture an image of an object 1708 over a large depth of field 1710. A light source (not shown) provides light through a slit 1704, and a beam splitter 1706 directs the light toward the object 1708. The light reflects from the object 1708 back to the beam splitter 1706, which provides the light to the CCD / CMOS detector 1702. The image captured by the CCD / CMOS detector 1702 can be stored, for example, in a memory, and can be processed by a processing device.

[0118] In some examples, the color confocal imaging system 1700 can be used with the glass inclusion detection apparatus 100 to verify suspected inclusions. For example, the glass inclusion detection apparatus 100 can identify areas of the glass sheet 106 that have suspected inclusions, and the color confocal imaging system 1700 can be used to return to those areas to verify whether the findings are actually inclusions.

[0119] Figure 18 A chromatic confocal imaging system 1800 including depth measurement is shown. In this example, a light source 1802 provides light to a glass sheet 106. As shown, the light is provided at varying intensities to reach various depths within the glass sheet 106. The light is reflected from the glass sheet 106 and captured by a spectral camera 1804. Based on the wavelength of the captured light, the depth of the inclusion can be determined. For example, graph 1810 shows the intensity of light captured from the top surface of the glass sheet 106, from an inclusion in a region of interest (ROI) of the glass sheet 106, and from the bottom surface of the glass sheet 106.

[0120] In some examples, the color confocal imaging system 1800 can be used with the glass inclusion detection apparatus 100 to verify suspected inclusions. For example, the glass inclusion detection apparatus 100 can identify areas of the glass sheet 106 that have suspected inclusions, and the color confocal imaging system 1800 can be used to return to those areas to verify whether the findings are actually inclusions.

[0121] Figure 19 An exemplary method 1900 is shown; the exemplary method 1900 can be performed by a glass inclusion detection apparatus described herein, such as the glass inclusion detection apparatus 100. Beginning at step 1902, a laser sheet is projected onto a first side of a glass sheet at a first angle. At step 1904, an image is captured at a second angle by a camera directed toward a second side of the glass sheet. The second side of the glass sheet is opposite the first side of the glass sheet. In some examples, the first angle is greater than or equal to the second angle. At step 1906, the captured image is stored in memory.

[0122] Proceeding to step 1908, the glass sheet is moved a predetermined distance past the laser sheet and camera field of view. For example, motion stage 108 may move glass sheet 106 a distance less than (or equal to) the width of laser sheet 202. At step 1910, a determination is made as to whether the length of the glass sheet has been scanned. If the length of the glass sheet has not been scanned, the method returns to step 1904, where another image of the glass sheet is captured. Otherwise, if the length of the glass sheet has been scanned, the method ends.

[0123] Figure 20 An exemplary method 2000 is shown that can be performed by one or more computing devices (e.g., a processing device as described herein). Beginning at step 2002, a first image captured by a camera directed toward one side of a glass sheet is obtained. For example, the processing device may obtain the first image from a database. At step 2004, a bright spot is detected in the image at a first coordinate location (which is between a top scattering region and a bottom scattering region). At step 2006, a distance from the first coordinate location to the top of the top scattering region is determined. At step 2008, a second image is determined based on the top distance. The second image is an image in which the bright spot is expected to overlap the top scattering region at the second coordinate location.

[0124] Proceeding to step 2010, a bottom distance from the first coordinate position to the bottom scattering region is determined. At step 2012, a third image is determined based on the bottom distance. The third image is an image in which the expected bright spot covers the bottom scattering region at the third coordinate position.

[0125] At step 2014, a first light intensity is determined at a first coordinate position in the first image. At step 2016, a second light intensity is determined at a second coordinate position in the second image. At step 2018, a third light intensity is determined at a third coordinate position in the third image.

[0126] The method proceeds to step 2020, where it is determined whether the first light intensity is greater than the second and third light intensities. If the first light intensity is not greater than the second and third light intensities, the method proceeds to step 2022, where no inclusion is identified. For example, it is suspected that there is no inclusion at the first coordinate in the first image. However, if the first light intensity is greater than the second and third light intensities, the method proceeds to step 2024, where an inclusion is identified. For example, it is suspected that there is an inclusion at the first coordinate in the first image. The method then ends.

[0127] Although the method described above is with reference to the illustrated flowchart, it will be understood that many other ways of performing the actions associated with the method may be used. For example, the order of some operations may be changed, and some operations described may be optional.

[0128] Additionally, the methods and systems described herein may be embodied, at least in part, in the form of computer-implemented processes and devices for implementing those processes. The disclosed methods may also be embodied, at least in part, in the form of a tangible, non-transitory, machine-readable storage medium encoded with computer program code. For example, the steps of the method may be embodied in hardware, executable instructions (e.g., software) executed by a processor, or a combination of both. The medium may include, for example, RAM, ROM, CD-ROM, DVD-ROM, BD-ROM, hard drive, flash memory, or any other non-transitory, machine-readable storage medium. When the computer program code is loaded into a computer and executed by the computer, the computer becomes a device for implementing the method. The method may also be embodied, at least in part, in the form of a computer that loads or executes the computer program code, making the computer a special-purpose computer for implementing the method. When the computer program code segments are implemented on a general-purpose processor, the computer program code segments configure the processor to create specific logic circuits. The method may alternatively be embodied, at least in part, in a dedicated integrated circuit for performing the method.

[0129] The foregoing is provided to illustrate, explain, and describe the embodiments of the present application. Modifications and adaptations to these embodiments will be apparent to those skilled in the art and may be made without departing from the scope or spirit of the present application.

Claims

1. A laser-based inclusion detection device comprising: a laser configured to project a laser sheet onto the first side of the glass sheet; a first camera configured to capture a first image of the glass sheet from a second side of the glass sheet, wherein the first camera captures the first image using darkfield illumination; and at least one processor configured to determine areas of relatively higher light intensity in the first captured image, Wherein determining an area of ​​relatively higher light intensity in the captured first image comprises: In the first image, identify the top line of higher light intensity; In the first image, identifying a baseline of higher light intensity; identifying, in the first image, a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity; and determining inclusions in the first area based on the top line, the bottom line, and the light intensity of the first area, Wherein determining an area of ​​relatively higher light intensity in the captured first image further comprises: determining a first distance from the first area to the top line; determining a second area in a second image based on the first distance, wherein the first area overlaps the top line of the higher light intensity in the second image; and It is determined that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image.

2. The apparatus of claim 1, further comprising a motion stage configured to move the glass sheet past the laser sheet.

3. The apparatus of claim 1, wherein the at least one processor is configured to identify inclusions in the glass sheet based on areas of relatively higher light intensity in the captured first image.

4. The apparatus of claim 1 , wherein determining an area of ​​relatively higher light intensity in the captured first image further comprises: determining a second distance from the first area to the bottom line; determining a third area in a third image based on the second distance, wherein the first area overlaps the bottom line of the higher light intensity in the third image; and It is determined that the first light intensity of the first area in the first image is greater than a third light intensity of the third area in the third image.

5. The apparatus of claim 1 , wherein determining an area of ​​relatively higher light intensity in the captured first image further comprises: determining a first distance from the first area to the top line; determining, for each of a first plurality of captured first images, a first expected position of the first area in each image based on the first distance; determining a second distance from the first area to the bottom line; determining, for each of a second plurality of captured first images, a second expected position of the first area in each image based on the second distance; determining a light intensity at each first expected location and each second expected location; executing a machine learning algorithm to classify the light intensity; and The inclusions in the first region are determined based on the classified light intensities.

6. The apparatus of claim 1, comprising a microscopic imaging camera configured to view the glass sheet from the second side of the glass sheet.

7. The apparatus of claim 1, comprising a second camera configured to capture a second image of the glass sheet from the first side of the glass sheet, wherein the second camera captures the second image using brightfield illumination.

8. The apparatus of claim 7, comprising a diffuse blue light emitting diode configured to provide light to the first side of the glass sheet, and wherein the first camera comprises a filter that blocks blue light.

9. The apparatus of claim 8, wherein the laser is a red line laser.

10. The apparatus of claim 8, wherein the first camera and the second camera are configured to capture the first image and the second image, respectively, simultaneously.

11. The apparatus of claim 1 , comprising: A chromatic confocal sensor is configured to capture a second image of events reflected from the glass sheet from the first side of the glass sheet.

12. A laser-based inclusion detection system comprising: a laser configured to project a laser sheet onto the first side of the glass sheet; a camera configured to capture an image of the glass sheet from a second side of the glass sheet, wherein the camera captures the image using darkfield illumination; a motion stage configured to move the glass plate past the laser sheet; a backlight configured to project light onto the first side of the glass sheet; a microscope configured to inspect the second side of the glass sheet; and at least one processor configured to determine areas of relatively higher light intensity in the captured first image, Wherein determining an area of ​​relatively higher light intensity in the captured first image comprises: In the first image, identify the top line of higher light intensity; In the first image, identifying a baseline of higher light intensity; identifying, in the first image, a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity; and determining inclusions in the first area based on the top line, the bottom line, and the light intensity of the first area, Wherein determining an area of ​​relatively higher light intensity in the captured first image further comprises: determining a first distance from the first area to the top line; determining a second area in a second image based on the first distance, wherein the first area overlaps the top line of the higher light intensity in the second image; and It is determined that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image.

13. The system of claim 12, wherein the motion stage is configured to move the glass sheet a predetermined distance past the laser sheet.

14. The system of claim 13, wherein the predetermined distance is less than or equal to a width of the laser sheet.

15. A laser-based inclusion detection method comprising: capturing an image of the glass plate; In a first image of the captured images, identifying a top line of higher light intensity; In the first image, identifying a baseline of higher light intensity; identifying, in the first image, a first region of higher light intensity between the top line of higher light intensity and the bottom line of higher light intensity; and determining a first distance from the first area to the top line; determining a second area in a second image of the captured image based on the first distance, wherein the first area overlaps a top line of the higher light intensity in the second image; and determining that a first light intensity of the first area in the first image is greater than a second light intensity of the second area in the second image; Inclusions in the first region are determined based on the top line, the bottom line, and the light intensity of the first region.

16. The method of claim 15, further comprising: determining a second distance from the first area to the bottom line; determining a third area in a third image based on the second distance, wherein the first area covers the bottom line of the higher light intensity in the third image; and It is determined that a first light intensity of the first area in the first image is greater than a third light intensity of the third area in the third image.

17. The method of claim 15, further comprising: determining a first distance from the first area to the top line; determining, for each of a first plurality of the captured images, a first expected position of the first area in each image based on the first distance; determining a second distance from the first area to the bottom line; determining, for each of a second plurality of the captured images, a second expected position of the first area in each image based on the second distance; determining a light intensity at each first expected location and each second expected location; executing a machine learning algorithm to classify the light intensity; and The inclusions in the first region are determined based on the classified light intensities.

Images