Liquid crystal glass defect detection device and control method

Abstract

This application discloses a liquid crystal glass defect detection device and control method. The detection device includes a transmission mechanism, a gantry, and an optical imaging assembly. Specifically, the camera angle is adjusted by the imaging adjustment assembly to satisfy Scherm's law, solving the depth-of-field limitation of the inclined plane and achieving single-shot full-clear imaging. In the illumination system, the side-entry optical coupling assembly uses a linear drive and a floating suspension mechanism to drive the optical coupling head, which includes a rigid air-float block and a silicone pad. The air-float block forms a fixed-distance air film, and the silicone pad flexibly adheres, achieving frictionless floating relative to the side surface of the glass. This liquid crystal glass defect detection device effectively solves the coupling difficulties and collision damage caused by fluctuations in the glass edge position during high-speed transmission, ensuring efficient and stable light coupling to enhance the contrast of internal defects and guaranteeing high-precision and high-efficiency online detection.

Description

Technical Field

[0001] This invention relates to a glass defect detection device, and more particularly to a liquid crystal glass defect detection device and control method. Background Technology

[0002] In the manufacturing process of modern flat panel displays, the liquid crystal glass substrate is a crucial basic material, and its edge quality directly affects the mechanical strength and lifespan of the finished panel. After being cut into sheets, liquid crystal glass typically undergoes an edge grinding and chamfering process to create specific beveled surfaces (chamfered surfaces) and straight side surfaces. This eliminates cutting stress and prevents chipping or breakage during subsequent handling and assembly. As display technology advances towards ultra-thinness and narrow bezels, the requirements for the processing quality of glass edges are becoming increasingly stringent. Therefore, using automated optical inspection equipment for high-precision, all-around online inspection of the edges of liquid crystal glass has become an indispensable key process in the production line.

[0003] Existing glass edge inspection devices typically consist of a transmission mechanism, an image acquisition unit, and an illumination unit. A typical structure is as follows: the transmission mechanism moves the liquid crystal glass horizontally; the image acquisition unit (such as an industrial camera with a telecentric lens) is mounted above the glass, with the lens optical axis usually perpendicular to the large surface of the glass or aligned at a fixed angle to the edge; the illumination unit often uses coaxial light or a strip light source to illuminate the glass surface. During inspection, the camera captures images of the illuminated edges, and image processing algorithms identify notches, scratches, or dimensional deviations at the edges. For the detection of internal defects, some existing technologies attempt to use side lighting, i.e., placing a light source on the side of the glass to try to guide light into the glass to illuminate internal cracks.

[0004] However, the aforementioned existing technologies face significant technical bottlenecks when performing high-precision detection of LCD glass edges. First, the chamfered edge of the glass is a slope with a depth difference. Conventional vertical imaging is limited by the depth of field of the optical lens, making it difficult to simultaneously capture clear images of both the top and bottom of the slope in a single exposure, resulting in blurred images or requiring multi-layer scanning at the expense of detection efficiency. Second, and more critically, during the high-speed movement of the glass driven by the transmission mechanism, the side surface of the glass edge often exhibits micron-level lateral wavy position fluctuations relative to the detection device due to transmission jitter or the glass's own cutting tolerances. Existing side-illuminated solutions, if using non-contact illumination, suffer from severe light reflection at the air-glass interface, making it difficult to penetrate the glass and form effective total internal reflection illumination, leading to low contrast for internal microcracks. If contact-coupled light guides are used, the lack of an effective floating and following mechanism makes the rigid contact prone to hard collisions when the glass exhibits wavy fluctuations, causing glass scratches or equipment damage, making it difficult to adapt to edge position changes while ensuring efficient light coupling. Summary of the Invention

[0005] The present invention aims to provide a liquid crystal glass defect detection device and control method to solve the technical problems of existing technologies that are limited by depth of field and cannot achieve full-clear imaging of oblique surfaces in a single shot, and cannot effectively detect internal defects and are difficult to distinguish from surface dirt, thereby achieving full-depth clear imaging and accurate identification of multi-dimensional defects in a single exposure.

[0006] The technical solution adopted by the present invention to solve the above problems is: a liquid crystal glass defect detection device, wherein the liquid crystal glass includes an edge bevel and a side straight surface, comprising:

[0007] The bracket has a transmission mechanism at its top, which has a transmission plane for carrying and driving the liquid crystal glass to move along a preset direction.

[0008] A gantry frame is mounted on top of the support and spans the conveying plane, and the gantry frame is equipped with a controllable movable end;

[0009] An optical imaging component is connected to the mobile terminal. The optical imaging component includes a detection camera and an optical lens. The detection camera has a photosensitive target surface, and the optical lens has a main plane.

[0010] An imaging adjustment component is connected between the detection camera and the optical lens to adjust the angle of the detection camera relative to the optical lens, so that the extended surfaces of the photosensitive target surface, the main plane, and the edge slope of the liquid crystal glass to be detected intersect on the same straight line when the detection device is in the detection state.

[0011] Lighting system, including:

[0012] Surface lighting module for illuminating the edge bevel;

[0013] A side-entry optical coupling assembly for detecting internal defects near the edge of a liquid crystal glass includes a linear drive assembly, a floating suspension mechanism, and an optical coupling head. The linear drive assembly is connected to the bracket or the gantry frame and drives the optical coupling head to move forward and backward toward the side surface of the liquid crystal glass to be inspected. The floating suspension mechanism is disposed between the linear drive assembly and the optical coupling head and is configured to allow the optical coupling head to float frictionlessly relative to the linear drive assembly in a direction perpendicular to the side surface. The optical coupling head includes a rigid air float, a light guide prism, and a silicone pad. The rigid air float is configured to form a fixed-distance air film between the detection device in the detection state and the side surface of the liquid crystal glass to be inspected. The silicone pad is configured to be compressed and adhere to the side surface within the distance defined by the fixed-distance air film.

[0014] Preferably, the floating suspension mechanism includes a floating shaft, an air bushing, and a connecting structure; the floating shaft is disposed at the moving end of the linear drive assembly along a direction pointing towards the side face; the air bushing is sleeved outside the floating shaft and connected to the optical coupling head through the connecting structure; the air bushing includes a bushing housing and a porous bushing disposed inside the bushing housing, and a preset gap for gas flow is formed between the porous bushing and the floating shaft to form an air film supporting the free sliding of the air bushing in the ventilated state.

[0015] Preferably, the optical coupling head has a stepped contact surface structure: the rigid air-float block has an internal receiving through hole, the light guide prism is disposed in the receiving through hole, and the rigid air-float block protrudes from the light-emitting surface of the light guide prism. The silicone pad is located in the receiving through hole and is in contact with the light-emitting surface of the light guide prism. In its natural state, the silicone pad protrudes from the outer side of the rigid air-float block. When the detection device is in the detection state, the rigid air-float block is locked at a distance from the liquid crystal glass by the air film, and the protruding silicone pad is compressed to fit tightly against the side surface.

[0016] Preferably, the side-entry optical coupling assembly further includes a constant force thrust mechanism, which is disposed on the floating shaft. The constant force thrust mechanism includes a pushing end connected to the air bushing. The constant force thrust mechanism is configured to apply a constant thrust pointing towards the side surface of the liquid crystal glass to the air bushing when the detection device is in the detection state, so as to balance the rebound force when the silicone pad is compressed and the air film repulsion force generated by the rigid air float.

[0017] Preferably, the imaging adjustment component includes an electric adjustment mechanism and a controller; the controller is configured to acquire the edge bevel angle parameter of the liquid crystal glass to be inspected, and control the electric adjustment mechanism to automatically adjust the relative angle between the detection camera and the optical lens according to the angle parameter, until the extended surfaces of the photosensitive target surface, the main plane and the edge bevel of the liquid crystal glass to be inspected intersect on the same straight line when the detection device is in the detection state.

[0018] Preferably, the surface illumination module includes a main illumination source and at least two auxiliary illumination sources; the main illumination source provides coaxial illumination of a first wavelength; the two auxiliary illumination sources are symmetrically arranged on both sides of the main illumination source to provide wide-angle diffuse illumination of a second wavelength; the light source of the side-entry optical coupling component is configured to emit light of a third wavelength; wherein the first wavelength, the second wavelength, and the third wavelength are different from each other, and the detection camera is configured as a color camera capable of simultaneously acquiring wavelength information of the first wavelength, the second wavelength, and the third wavelength.

[0019] Preferably, the liquid crystal glass defect detection device further includes an image processing unit; the image processing unit is configured to receive a composite image acquired by the detection camera in a single exposure, and separate it into independent channel images corresponding to the first wavelength, the second wavelength and the third wavelength; the image processing unit uses the first wavelength channel image to identify surface micro-defects, uses the second wavelength channel image to identify edge contour deviations, and uses the third wavelength channel image to identify internal defects of the glass.

[0020] Preferably, the rigid air-float block has a guide slope on the upstream side relative to the moving direction of the liquid crystal glass; the guide slope is configured to push the optical coupling head away from the liquid crystal glass through physical contact when the liquid crystal glass moves relative to the optical coupling head and its edge has a protrusion.

[0021] Preferably, the linear drive assembly further includes an advance / retreat control unit; the advance / retreat control unit is configured to control the optical coupler to perform an active avoidance action: holding the optical coupler in an avoidance position before the head of the liquid crystal glass to be detected reaches the optical coupler; pushing the optical coupler into a detection position after detecting that the head of the liquid crystal glass to be detected has passed the optical coupler; and retracting the optical coupler to the avoidance position before the tail of the liquid crystal glass to be detected leaves the optical coupler.

[0022] In particular, a control method for a liquid crystal glass defect detection device as described above includes the following steps:

[0023] Based on the edge slope angle of the liquid crystal glass to be tested, the imaging adjustment component is controlled to adjust the angle of the detection camera to meet the imaging conditions of Scherm's law.

[0024] The linear drive assembly is controlled to push the optical coupling head toward the liquid crystal glass to be tested, the floating suspension mechanism is activated and a coupling thrust is applied, so that the rigid air float block forms the fixed-distance air film on the surface of the liquid crystal glass, and the silicone pad is compressed and adhered to the side surface of the liquid crystal glass to be tested within the distance defined by the fixed-distance air film.

[0025] Activate the surface illumination module and the side-entry optical coupling component, and control the detection camera to perform a single exposure acquisition on the edge slope of the liquid crystal glass to be detected, and acquire a composite image containing the first wavelength, the second wavelength and the third wavelength information;

[0026] The composite image is separated into a first wavelength channel image, a second wavelength channel image, and a third wavelength channel image to determine surface defects, contour deviations, and internal defects, respectively.

[0027] The beneficial effects of the embodiments of the present invention are as follows:

[0028] 1. Because this invention uses an imaging adjustment component to adjust the angle of the detection camera relative to the optical lens to meet the imaging conditions of Scherm's law, and is combined with a side-entry optical coupling component including a linear drive component, a floating suspension mechanism, and an optical coupling head, especially by using the floating suspension mechanism to achieve frictionless floating of the optical coupling head relative to the linear drive component, and by using a combination of a rigid air float and a silicone pad to achieve flexible bonding and light guiding under a fixed-distance air film; therefore, it effectively solves the technical problems in the prior art where the oblique surface imaging is blurred due to the lens depth of field limitation, and the side-entry illumination light is difficult to couple efficiently or is prone to collision damage due to the wavy position fluctuation of the glass edge during high-speed transmission; thus, it can obtain a clear image with full depth of field in a single shot, and can maintain a constant micro-gap and tight fit between the optical coupling head and the glass side surface while adapting to the micron-level position fluctuation of the glass edge, ensuring that the light enters the glass interior efficiently and stably to form total internal reflection illumination, significantly improving the imaging contrast and detection safety of internal hidden crack defects, and ensuring high-precision and high-efficiency online detection of liquid crystal glass.

[0029] 2. By employing an imaging adjustment component that adjusts according to the angle of the inclined plane to meet the imaging conditions of Scherm's Law, controlling the linear drive component and the floating suspension mechanism to form a fixed-distance air film with the rigid air-floating block and press the silicone pad to achieve flexible bonding and light guiding, and utilizing multi-wavelength synchronous illumination for single-exposure acquisition and separation of channel data, this technology effectively solves the technical problems in existing technologies, such as blurred inclined plane imaging due to lens depth of field limitations, unstable side-incident light coupling due to glass edge position fluctuations, or collision damage caused by rigid contact, as well as low detection efficiency and inability to accurately distinguish between surface and internal defects due to a single light source or time-division shooting. This allows for the acquisition of a clear image with full depth of field in a single shooting action. While ensuring that the optical coupling head follows the glass edge fluctuations for non-destructive and efficient light guiding, spectral separation technology accurately determines surface micro-defects, contour deviations, and internal microcracks at the same time, significantly improving the overall efficiency, robustness, and classification accuracy of automated detection. Attached Figure Description

[0030] Figure 1 A schematic structure of a liquid crystal glass defect detection device according to an embodiment of the present invention is shown. Figure 1 .

[0031] Figure 2 A schematic structure of a liquid crystal glass defect detection device according to an embodiment of the present invention is shown. Figure 2 .

[0032] Figure 3 A schematic structure of a liquid crystal glass defect detection device according to an embodiment of the present invention is shown. Figure 3 .

[0033] Figure 4 A schematic structure of a liquid crystal glass defect detection device according to an embodiment of the present invention is shown. Figure 4 .

[0034] Figure 5 A schematic structure of a liquid crystal glass defect detection device according to an embodiment of the present invention is shown. Figure 5 .

[0035] Figure 6 It shows Figure 5 A schematic magnified view of point A in the middle.

[0036] Figure 7 A schematic cross-sectional view of a side-entry optical coupling assembly according to an embodiment of the present invention is shown.

[0037] Figure 8 A schematic diagram of the optical path according to a Scham's law proposed in an embodiment of the present invention is shown.

[0038] The components include: 1. Bracket; 2. Transmission mechanism; 3. Positioning component; 4. LCD glass; 410. Side straight surface; 420. Edge bevel; 5. Optical coupling head; 510. Light guide prism; 520. Rigid air float block; 530. Silicone pad; 6. Gantry frame; 7. Imaging adjustment assembly; 8. Inspection camera; 9. Optical lens; 10. Illumination system; 1010. Main illumination source; 1020. Secondary illumination source; 11. Linear drive assembly; 12. Floating suspension mechanism; 1210. Floating shaft; 1220. Bushing housing; 1230. Perforated bushing; 13. Constant force thrust mechanism; 14. Connection structure. Detailed Implementation

[0039] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0040] See Figures 1 to 8 In a preferred embodiment of this application, a defect detection device for liquid crystal glass 4 is proposed. This detection device is specifically designed to detect the edge area of ​​liquid crystal glass 4, and in particular to perform comprehensive detection of the edge bevel 420 (chamfered surface) and side straight surface 410 (vertical cut surface) of liquid crystal glass 4.

[0041] The testing device includes a support 1, a gantry 6, an optical imaging assembly, an imaging adjustment assembly 7, and an illumination system 10. The support 1 has a transmission mechanism 2 at its top, which has a transmission plane for carrying and moving the liquid crystal glass 4 along a preset direction. The gantry 6 is located at the top of the support 1 and spans the transmission plane, and has a controllable movable end. The optical imaging assembly is connected to the movable end and includes a detection camera 8 and an optical lens 9. The detection camera 8 has a photosensitive target surface, and the optical lens 9 has a main plane. The imaging adjustment assembly 7 is connected between the detection camera 8 and the optical lens 9, and is used to adjust the angle of the detection camera 8 relative to the optical lens 9, so that the extended surfaces of the photosensitive target surface, the main plane, and the edge slope 420 of the liquid crystal glass 4 intersect on the same straight line when the testing device is in the testing state. The illumination system 10 includes a surface illumination module for illuminating the edge slope 420 and a module for detecting the liquid crystal glass 4 near its edge. The side-entry optical coupling assembly for detecting internal defects includes a linear drive assembly 11, a floating suspension mechanism 12, and an optical coupling head 5. The linear drive assembly 11 is connected to the bracket 1 or the gantry 6 and is used to drive the optical coupling head 5 to move forward and backward toward the side surface 410 of the liquid crystal glass 4 to be tested. The floating suspension mechanism 12 is disposed between the linear drive assembly 11 and the optical coupling head 5 and is configured to enable the optical coupling head 5 to float frictionlessly relative to the linear drive assembly 11 in a direction perpendicular to the side surface 410. The optical coupling head 5 includes a rigid air float 520, a light guide prism 510, and a silicone pad 530. The rigid air float 520 is configured to form a fixed-distance air film between the detection device in the detection state and the side surface 410 of the liquid crystal glass 4 to be tested. The silicone pad 530 is configured to be compressed and adhered to the side surface 410 at a distance defined by the fixed-distance air film.

[0042] The bracket 1, serving as the fundamental support component of the entire device, is typically made of high-rigidity metal to provide a stable mounting reference. A transmission mechanism 2 is positioned at the top of the bracket 1, creating a flat transmission plane. The transmission mechanism 2 is configured to carry the liquid crystal glass 4 to be tested and to move the glass stably along a preset process direction (typically a horizontal straight line). The transmission mechanism 2 can employ precision belt conveying, roller conveying, or air-bearing conveying to ensure the smoothness of the glass during movement and reduce vibration. In some embodiments, a positioning element 3 is also provided on the transmission plane to constrain the liquid crystal glass 4. The positioning element 3 allows the liquid crystal glass 4 to move synchronously with the transmission plane. It should be noted that the positioning element 3 is prior art and does not affect the continuous contact between the side-entry optical coupling assembly and the side surface 410 of the liquid crystal glass 4.

[0043] The gantry 6 is securely mounted on top of the support 1 and spans across the conveying plane, forming a detection gantry structure. The gantry 6 has a controllable moving end, typically driven by a linear motor, lead screw module, or synchronous belt, capable of precise position adjustment perpendicular to the glass conveying direction or perpendicular to the ground. This allows the detection assembly to flexibly adjust its detection position according to different sizes or specifications of LCD glass 4.

[0044] The optical imaging assembly is physically connected to the moving end of the gantry 6 and moves with the gantry for positioning. This assembly consists of a detection camera 8 and an optical lens 9. The detection camera 8 has a photosensitive target surface (i.e., the surface of the image sensor chip) inside, used to receive light signals and convert them into electrical signals; the optical lens 9 is mounted at the front of the camera and has an optical principal plane. The optical imaging assembly is configured to capture images of the edge area of ​​the liquid crystal glass 4.

[0045] The imaging adjustment assembly 7 is a key mechanical structure connecting the inspection camera 8 and the optical lens 9. It features angle adjustment capabilities and can be a wedge-shaped adjustment flange, a tilting displacement stage, or a flexible bellows. Its core function is to adjust the angle of the inspection camera 8 relative to the optical lens 9 (i.e., change the relative orientation of the photosensitive target surface and the lens's principal plane). Through adjustment, in the inspection state, the extended planes of the geometric plane containing the photosensitive target surface, the principal plane of the optical lens 9, and the geometric plane containing the inclined surface 420 of the edge of the liquid crystal glass 4 under inspection can intersect along the same straight line in space. This structural design satisfies Scherm's Law, thereby achieving clear focus across the entire plane when photographing inclined edges with depth differences. The imaging adjustment assembly 7 is an existing product, and its specific structure will not be described in detail here.

[0046] The illumination system 10 comprises two parts: a surface illumination module and a side-lit optical coupling assembly, to address different inspection needs. The surface illumination module is positioned near the optical imaging assembly to directly illuminate the beveled edge 420 of the liquid crystal glass 4, enabling the camera to capture surface defects such as scratches and dirt. The side-lit optical coupling assembly is the core component for detecting internal defects (such as microcracks and bubbles) near the edge of the liquid crystal glass 4.

[0047] The side-entry optical coupling assembly includes a linear drive assembly 11, a floating suspension mechanism 12, and an optical coupling head 5.

[0048] The linear drive assembly 11 is connected to the fixed end of the bracket 1 or the gantry 6 and serves as a macroscopic drive unit (such as a cylinder or electric push rod) to drive the entire optocoupler head 5 to move forward or backward toward the side surface 410 of the liquid crystal glass 4 to be tested (i.e., to enter the working position or to retract to the safe position).

[0049] A floating suspension mechanism 12 is disposed between the linear drive assembly 11 and the optical coupler 5. It is configured as a flexible connection or air-bearing guide structure, allowing the optical coupler 5 to float frictionlessly relative to the linear drive assembly 11 in a direction perpendicular to the glass side surface 410. This design allows the optical coupler 5 to "follow" at a microscale to adapt to positional fluctuations at the glass edge.

[0050] The optical coupling head 5 is a component that interacts directly with the glass and includes a rigid air flotation block 520, a light guide prism 510, and a silicone pad 530.

[0051] The rigid air flotation block 520 is made of a rigid material with porous structure and is configured to form a constant thickness air film (fixed distance air film) between itself and the side surface 410 of the liquid crystal glass 4 in the ventilated working state, thereby locking the reference distance between the coupling head and the glass.

[0052] The light guide prism 510 is used to transmit detection light of a specific wavelength.

[0053] The silicone pad 530 is disposed on the light guide path and is configured to be compressed and tightly adhered to the side surface 410 of the glass at the distance defined by the aforementioned fixed-distance air film, in order to eliminate air gaps and achieve efficient coupling of light.

[0054] The device operates based on a deep integration of opto-mechatronics and physical optics. Before detection begins or during model change, the imaging adjustment component 7 is activated. The tilt angle of the detection camera 8 is adjusted according to the angle of the inclined surface 420° at the edge of the liquid crystal glass 4 until the photosensitive target surface, the lens principal plane, and the inclined glass surface satisfy the collinearity condition of Scherm's law. This step ensures that the lens can achieve a large depth of field covering the entire inclined surface without stopping down the aperture, solving the problem of blurred top and sharp bottom or vice versa in inclined surface imaging.

[0055] When the transmission mechanism 2 moves the liquid crystal glass 4 to the detection area, the linear drive assembly 11 activates, pushing the photocoupler 5 closer to the glass side surface 410. During this process, the rigid air float 520 first approaches the glass and forms a highly rigid, fixed-distance gas film on the surface of the glass side surface 410 by ejecting gas, suspending the photocoupler 5 in a fixed position very close to the glass. At the same time, the silicone pad 530 on the photocoupler 5, due to its slightly protruding design size from the air float, is compressed and deformed under the fixed-distance effect of the gas film, tightly adhering to the glass side surface 410. The light emitted by the light source passes through the light guide prism 510, through the compressed silicone pad 530, and is coupled into the interior of the liquid crystal glass 4 without damage, undergoing total internal reflection transmission between the upper and lower surfaces of the glass to illuminate internal defects.

[0056] The liquid crystal glass 4 moves at high speed along the transmission direction. Due to potential micron-level undulations or transmission jitter on the glass sides, the floating suspension mechanism 12 comes into play. It allows the light coupling head 5 to float with extremely low resistance in a direction perpendicular to the side surface 410, achieving frictionless movement. The rigid air-bearing block 520, relying on air film repulsion, consistently follows the undulations of the glass surface, causing the silicone pad 530 to maintain a constant compression and fit, ensuring the continuity and stability of light coupling while preventing direct impact of rigid components onto the glass.

[0057] This technical solution is applicable to the online inspection process in the flat panel display manufacturing industry, particularly in the quality control process after glass substrate edge grinding. The device should be installed in an industrial environment with certain cleanliness requirements to prevent dust from interfering with imaging or damaging the air flotation components. This device is suitable for inspecting LCD glass substrates of various thicknesses and sizes, especially in scenarios with large edge chamfering angles and high requirements for the detection rate of internal micro-cracks. A stable compressed air source is required for the floating suspension mechanism 12 and the rigid air flotation block 520 to maintain the formation of the air film.

[0058] In some alternative embodiments, the transmission mechanism 2 may employ a transmission method using an air-bearing platform in conjunction with gripping fingers, in addition to a conveyor belt. The imaging adjustment component 7 may be a manual differential head adjustment mechanism or a motor-driven automatic adjustment mechanism. The linear drive component 11 may be a pneumatic slide or an electric linear module.

[0059] In this embodiment, the imaging adjustment component 7 is used to adjust the angle of the detection camera 8 relative to the optical lens 9 to meet the imaging conditions of Scherm's law. This is combined with a side-entry optical coupling component including a linear drive component 11, a floating suspension mechanism 12, and an optical coupling head 5. Specifically, the floating suspension mechanism 12 enables frictionless floating of the optical coupling head 5 relative to the linear drive component 11, and the combination of a rigid air float 520 and a silicone pad 530 achieves flexible bonding and light guiding under a fixed-distance air film. Therefore, this effectively solves the technical problems in the prior art, such as blurred imaging on oblique surfaces due to lens depth-of-field limitations, and the difficulty in efficient coupling of side-entry illumination light or the risk of collision damage due to rigid contact caused by the wavy position fluctuations of the glass edge during high-speed transmission. This allows for obtaining a clear image with full depth of field in a single shot, while maintaining a constant micro-gap and tight fit between the optical coupling head 5 and the glass side surface 410, adapting to micron-level position fluctuations at the glass edge. This ensures efficient and stable light entry into the glass to form total internal reflection illumination, significantly improving the imaging contrast and detection safety of internal hidden crack defects, and guaranteeing high-precision and high-efficiency online detection of the liquid crystal glass 4.

[0060] Further, see Figures 3 to 7In some embodiments, a specific structural design was implemented for the floating suspension mechanism 12 in the side-entry optical coupling assembly. This floating suspension mechanism 12 mainly consists of a floating shaft 1210, an air bushing, and a connecting structure 14, aiming to construct a zero-friction follow-up guidance system. The floating shaft 1210 is positioned at the moving end of the linear drive assembly 11 along a direction pointing towards the side surface 410; the air bushing is fitted outside the floating shaft 1210 and connected to the optical coupling head 5 via the connecting structure 14; the air bushing includes a bushing housing 1220 and a porous bushing 1230 disposed inside the bushing housing 1220, with a preset gap between the porous bushing 1230 and the floating shaft 1210 for gas flow, thereby forming an air film supporting the free sliding of the air bushing in the ventilated state.

[0061] The floating shaft 1210 is a precision-ground and surface-hardened guide shaft, typically cylindrical in shape. It is preferably made of ceramic, stainless steel, or hard-anodized aluminum alloy to achieve extremely high surface finish and wear resistance. One end of the floating shaft 1210 is rigidly mounted on the moving end of the linear drive assembly 11. Spatially, the axial direction of the floating shaft 1210 is strictly set to point towards the side face 410 of the liquid crystal glass 4 to be tested, i.e., perpendicular to the normal direction of the glass edge. As the stator of the entire suspension mechanism, the floating shaft 1210 provides a reference guide for the forward and backward floating of the optocoupler 5.

[0062] The air bushing is fitted onto the outside of the floating shaft 1210 and serves as the moving part, capable of reciprocating along the axial direction of the floating shaft 1210. The air bushing adopts a composite structure, mainly consisting of an outer bushing housing 1220 and an inner porous bushing 1230.

[0063] The bushing housing 1220 is the external support frame of the air bushing, usually made of aluminum alloy or stainless steel, and has a certain mechanical strength. An air inlet port is provided on the housing for connecting to an external compressed air source.

[0064] The porous bushing 1230 is tightly embedded or bonded to the inner wall of the bushing housing 1220, directly surrounding the floating shaft 1210. The porous bushing 1230 is preferably made of porous carbon graphite material or sintered metal material. This material contains countless micron-sized micropores, which allow compressed air to permeate evenly.

[0065] An extremely small physical gap (typically a few micrometers to tens of micrometers) is left between the inner wall surface of the porous bushing 1230 and the outer peripheral surface of the floating shaft 1210 during manufacturing and assembly. This gap constitutes the space for gas flow and gas film formation.

[0066] The connecting structure 14 is mounted on the air bushing and is used to rigidly connect the air bushing to the optical coupler head 5. The connecting structure 14 needs to be lightweight to reduce the moving mass and improve the follow-up response speed.

[0067] It should be noted that the inner wall of the bushing housing 1220 is provided with an annular pressure equalizing groove, which extends circumferentially and communicates with the air inlet. The porous bushing 1230 is fixed to the bushing housing 1220 by interference fit or bonding, and both ends of the porous bushing 1230 are sealed to the inner wall of the bushing housing 1220, so that the outer peripheral surface of the porous bushing 1230 and the annular pressure equalizing groove form a closed gas diffusion cavity. During operation, external compressed gas enters the gas diffusion cavity through the air inlet and evenly coats the outer peripheral surface of the porous bushing 1230; subsequently, under pressure, the gas permeates through the microporous channels inside the porous bushing 1230 and finally overflows evenly from the inner surface of the porous bushing 1230, thereby forming a uniform and highly rigid gas film between the porous bushing 1230 and the floating shaft 1210.

[0068] The floating suspension mechanism 12 operates based on the principle of static pressure air buoyancy in aerodynamics. When clean compressed air enters the bushing housing 1220 through the air inlet, the high-pressure gas fills the chamber between the housing and the bushing and permeates evenly into the porous bushing 1230 through its microporous walls. When the gas enters the preset gap between the porous bushing 1230 and the floating shaft 1210, a high-pressure gas film is established within the gap due to the throttling effect of the narrow gap. This gas film has extremely high rigidity, enabling it to stably suspend the air bushing (and the connected optical coupling head 5) on the floating shaft 1210, completely isolating the physical contact between the bushing and the shaft. When the optical coupling head 5 needs to move due to fluctuations in the glass edge position, the air bushing slides axially along the floating shaft 1210 under the lubrication of the gas film. Since there is no solid contact, the frictional resistance during the sliding process is almost zero, and there is no "creeping" or hysteresis caused by static friction.

[0069] This technical solution is suitable for precision testing environments that require extremely high stability of motion and high response sensitivity.

[0070] In this embodiment, the floating suspension mechanism 12, including a floating shaft 1210 arranged along the direction of the pointing side straight surface 410 and an air bushing sleeved outside it, is adopted. In particular, the porous bushing 1230 arranged inside the bushing housing 1220 forms a preset gap between the floating shaft 1210 for gas flow and forms an air film supporting the free sliding of the air bushing in the ventilated state. Therefore, it effectively solves the technical problems of large static friction, motion lag, difficulty in responding to micron-level high-frequency fluctuations at the glass edge, and long-term contact wear leading to decreased accuracy in the prior art using mechanical slide rails or ordinary sliding bearings. As a result, it achieves zero-friction, lag-free, and high-sensitivity floating following of the optical coupling head 5 relative to the linear drive assembly 11, ensuring that the optical coupling head 5 can adapt to the position changes of the edge of the liquid crystal glass 4 in real time and accurately, greatly improving the stability of the detection process and the durability of the equipment.

[0071] Further, see Figures 6 to 7 In some embodiments, the internal spatial layout and contact surface geometry of the optical coupling head 5 are designed, employing a stepped contact surface structure. This structure aims to achieve functional separation and coordination between optical components and mechanical distance-fixing components through a stepped configuration of geometric dimensions. The optical coupling head 5 has a stepped contact surface structure: the rigid air-float block 520 has an internal receiving through hole, the light guide prism 510 is disposed within the receiving through hole, and the rigid air-float block 520 protrudes beyond the light-emitting surface of the light guide prism 510. The silicone pad 530 is partially located within the receiving through hole, and the silicone pad 530 is in contact with the light-emitting surface of the light guide prism 510. In its natural state, the silicone pad 530 protrudes beyond the outer surface of the rigid air-float block 520. When the detection device is in the detection state, the rigid air-float block 520 is locked at a distance from the liquid crystal glass 4 by the air film, while the protruding silicone pad 530 is compressed to fit tightly against the side surface 410.

[0072] The main frame of the optical coupling head 5 is composed of a rigid air-bearing block 520. A through-hole or semi-through-hole is formed in the central region of the rigid air-bearing block 520, the cross-sectional shape of which matches the cross-sectional shape of the light guide prism 510. The light guide prism 510 is installed deep within this through-hole. Spatially, the light-emitting surface (i.e., the end face from which light rays are emitted) of the light guide prism 510 is not flush with the outer surface of the rigid air-bearing block 520, but rather recessed inwards, causing the end face of the rigid air-bearing block 520 to protrude beyond the light-emitting surface of the light guide prism 510. This design creates a recessed cavity of a predetermined depth between the end face of the rigid air-bearing block 520 and the light-emitting surface of the light guide prism 510.

[0073] The silicone pad 530 serves as a flexible light-guiding medium, partially filling and fixing itself within the aforementioned recessed cavity. Its bottom surface is tightly fitted or optically bonded to the light-emitting surface of the light-guiding prism 510. The thickness of the silicone pad 530 is designed to be greater than the depth of the recessed cavity. Therefore, in its natural state (i.e., without external compression), the top surface of the silicone pad 530 will further protrude beyond the outer surface of the rigid air-bearing block 520 (i.e., the air-bearing working surface).

[0074] This creates a three-tiered, stepped structure: the light guide prism 510 is the deepest, the rigid air float 520 is in the middle, and the silicone pad 530 is the most prominent. The outer surface of the rigid air float 520 serves as a mechanical limiting reference surface, while the protruding silicone pad 530 serves as a flexible contact working surface.

[0075] The working principle of this stepped structure is based on the synergistic mechanism of air-bearing distance and fixed-compression. When the detection device enters the detection state, the optical coupling head 5 approaches the side surface 410 of the liquid crystal glass 4 to be detected. Since the silicone pad 530 protrudes from the rigid air-bearing block 520 in its natural state, the surface of the silicone pad 530 will first contact the side surface 410 of the liquid crystal glass 4. As the optical coupling head 5 continues to advance, the silicone pad 530 begins to undergo elastic compression deformation due to the reaction force of the glass sidewall, filling the microscopic unevenness of the glass surface and expelling air between the contact interfaces. When the optical coupling head 5 advances to the preset position, compressed gas is introduced into the rigid air-bearing block 520. A layer of highly rigid distance-fixed air film is formed between the rigid air-bearing block 520 and the side surface 410 of the liquid crystal glass 4. This air film generates a huge repulsive force, preventing the rigid air-bearing block 520 from continuing to approach the glass, locking it at a constant height of micrometers above the glass surface.

[0076] At this point, the system reaches equilibrium. Although the rigid air flotation block 520 is not in direct contact with the glass, its distance from the glass is locked by the air film. Since the natural protrusion height of the silicone pad 530 is fixed, and the air film thickness is also fixed, the silicone pad 530 is forcibly compressed to a fixed depth (the compression amount equals the natural protrusion height of the silicone pad 530 minus the air film thickness). This mechanism ensures that the compression ratio of the silicone pad 530 remains constant regardless of the number of operations, preventing damage due to overpressure or poor coupling due to underpressure.

[0077] In this embodiment, a stepped contact surface structure is adopted, in which a rigid air float block 520 has an internal through-hole to accommodate the light guide prism 510, and the rigid air float block 520 protrudes beyond the light-emitting surface of the prism, while the silicone pad 530 protrudes further beyond the outer side of the rigid air float block 520 in its natural state. Therefore, this effectively solves the technical problems in the prior art where direct rigid contact causes glass scratches or simple flexible contact leads to uncontrollable compression, resulting in fluctuations in light coupling efficiency and excessive wear of silicone. Furthermore, it achieves the precise limitation of the compression of the silicone pad 530 by using the air film of the rigid air float block 520 as a high-precision position reference, ensuring efficient and lossless coupling of light into the glass while maximizing the service life of the silicone pad 530 and protecting the safety of the edge of the liquid crystal glass 4.

[0078] Further, see Figures 3 to 5 and Figure 7 In some embodiments, the side-entry optical coupling assembly is also equipped with a key constant force thrust mechanism 13, which is disposed on the floating shaft 1210 of the floating suspension mechanism 12. The constant force thrust mechanism 13 is designed to provide a precise and controllable axial power source for the air bushing. The constant force thrust mechanism 13 is disposed on the floating shaft 1210 and includes a push end connected to the air bushing. The constant force thrust mechanism 13 is configured to apply a constant thrust pointing towards the side surface 410 of the liquid crystal glass 4 to the air bushing when the detection device is in the detection state, so as to balance the rebound force when the silicone pad 530 is compressed and the air film repulsion force generated by the rigid air float 520.

[0079] Specifically, the constant force thrust mechanism 13 preferably uses a voice coil motor as the core actuator. A voice coil motor is a direct drive device with characteristics of no hysteresis, zero friction, and high response speed. Moreover, its output force has a strictly linear proportional relationship with the input current, and it can output constant thrust regardless of the stroke position.

[0080] The stator of the voice coil motor is fixedly mounted on the fixed end of the floating shaft 1210 and remains stationary.

[0081] The mover (coil assembly) of the voice coil motor, acting as the driving end, extends coaxially or parallel to the rear end of the air bushing (the side away from the LCD glass 4). The driving end and the air bushing are connected via a non-rigid contact or a flexible connection. For example, the driving end can directly abut against the end face of the air bushing, or through a miniature ball joint, to ensure that the thrust is transmitted only axially without introducing radial lateral forces that could interfere with the air bushing's levitation.

[0082] The constant force thrust mechanism 13 operates based on a precise mechanical balance mechanism. When the detection device enters the detection state, the controller inputs a set constant current to the constant force thrust mechanism 13 (such as a voice coil motor). According to Ampere's law, the pushing end generates a constant thrust pointing towards the side surface 410 of the liquid crystal glass 4 to be tested. This thrust pushes the air bushing to slide along the floating shaft 1210 towards the glass. When the rigid air float 520 at the front end of the optical coupling head 5 approaches the glass surface and establishes an air film, and at the same time the silicone pad 530 contacts the side surface 410 of the glass and is compressed, the system instantly establishes a dynamic mechanical balance system. At this time, the forward constant thrust applied by the constant force thrust mechanism 13 is balanced by the rebound force of the two backward silicone pads 530 and the air film repulsion force. The rebound force of the silicone pad 530 is the elastic recovery force generated by the compression of the silicone pad 530, and the air film repulsion force is the hydrostatic pressure generated by the high-pressure air film between the rigid air float 520 and the glass.

[0083] Furthermore, since the input thrust is constant, and the compressive rebound force of the silicone pad 530 is related to the compression amount, while the air film repulsion force is related to the air film thickness, the system automatically finds and locks into a unique equilibrium position. At this equilibrium position, the silicone pad 530 is compressed to a fixed depth (e.g., tens of micrometers), while the rigid air flotation block 520 is suspended at a fixed micrometer-level height from the glass. Regardless of fluctuations in the glass edge position, as long as the thrust remains constant, the geometric relationship between this air film thickness and the silicone compression amount remains unchanged.

[0084] In this embodiment, a constant force thrust mechanism 13 is installed on the floating shaft 1210 and applies a constant thrust to the air bushing pointing towards the side surface 410 of the liquid crystal glass 4 through the pushing end. In particular, this constant thrust is used to dynamically balance the rebound force of the silicone pad 530 when it is compressed and the air film repulsion force generated by the rigid air float 520. Therefore, it effectively solves the technical problem in the prior art where the contact pressure changes drastically with the glass position due to the use of rigid connection or ordinary spring, which leads to unstable silicone compression, causing fluctuations in optical coupling efficiency or instability of the air film gap, resulting in collisions. Thus, when the position fluctuates at the edge of the glass, the optical coupling head 5 can always maintain a constant air film gap and a constant silicone compression, ensuring the extreme stability of the optical coupling interface and the safety of the detection process.

[0085] Furthermore, in some embodiments aimed at achieving production line automation and flexible manufacturing, the imaging adjustment component 7 has been intelligently upgraded, mainly consisting of two core hardware components: an electric adjustment mechanism and a controller. The controller is configured to acquire the angle parameters of the edge bevel 420 of the liquid crystal glass 4 to be inspected, and control the electric adjustment mechanism to automatically adjust the relative angle between the detection camera 8 and the optical lens 9 according to the angle parameters, until the extended surfaces of the photosensitive target surface, the main plane, and the edge bevel 420 of the liquid crystal glass 4 to be inspected intersect on the same straight line when the detection device is in the detection state.

[0086] The electric adjustment mechanism is a high-precision mechatronic device that is physically connected between the detection camera 8 and the optical lens 9, or serves as a mounting base for the detection camera 8. Its mechanical body typically employs a precision electric angle measuring stage or an electric tilting flange structure.

[0087] The drive unit preferably employs a micro stepper motor or servo motor, coupled with a high-reduction-ratio worm gear transmission mechanism or a piezoelectric ceramic actuator. This transmission method not only provides micron-level displacement resolution but also features a power-off self-locking function to prevent angular drift of the camera under gravity. The motor's output shaft drives a movable platform to rotate around a virtual pivot point. The detection camera 8 is fixed on this movable platform, thereby causing its photosensitive target surface to tilt relative to the main plane of the optical lens 9.

[0088] The electric adjustment mechanism integrates a high-precision angle encoder or grating ruler to monitor the actual tilt angle of the moving platform in real time and feed the position signal back to the controller to form a closed-loop control.

[0089] The controller can be a standalone programmable logic controller, an embedded motion control card, or a software control module integrated into an industrial computer. The controller has a data communication interface for receiving external commands and sending drive pulses. Internally, the controller stores a computational model based on geometric optics, specifically the mathematical formulas for solving Scherm's law.

[0090] The working principle of this automated adjustment system is based on parameter-driven geometric calculation and position servo control. Before inspecting different types of LCD glass 4, the controller first obtains the "edge bevel angle parameter 420" of the LCD glass 4 to be inspected through human-machine interface input, barcode scanning, or communication with the host computer. This parameter reflects the theoretical chamfer angle of the glass edging process. Based on the built-in Schahm's Law algorithm model, combined with known system optical parameters (such as the focal length, magnification, and working distance of the optical lens 9), the controller performs real-time calculations using the input edge bevel angle parameter 420. The algorithm calculates the target tilt angle value that the photosensitive target surface of the detection camera 8 theoretically needs to deflect in order to meet the condition that the photosensitive target surface plane, the lens main plane, and the glass bevel surface plane are aligned. After the calculation is completed, the controller sends the corresponding pulse signal to the drive motor of the electric adjustment mechanism. The motor drives the moving platform to precisely deflect the detection camera 8. During the movement, the angle encoder provides real-time feedback of the current angle. The controller compares the feedback value with the target value and eliminates errors through a PID control algorithm until the detection camera 8 accurately reaches the target tilt angle and locks in. At this point, the optical system automatically completed the panoramic depth-of-focus setting for the current glass slope.

[0091] The above algorithm is based on existing, classic geometric optics principles. It is based on the mathematical expression of Scherm's Law, and its core formula is a standard trigonometric function relationship. The Scherm's Law algorithm model stored internally in the controller is based on classic geometric optics imaging principles. Specifically, to achieve clear imaging across the entire plane, the tilt angle of the photosensitive target surface must satisfy the following mathematical relationship (i.e., the Scherm's Law formula):

[0092]

[0093] in:

[0094] : Represents the target tilt angle (i.e., the angle that needs to be adjusted by the motor) of the photosensitive target surface of the detection camera 8 relative to the main plane of the optical lens 9.

[0095] : Represents the lateral optical magnification of the optical imaging component at the current working distance (this parameter is determined by the selected optical lens 9 and the working distance, and is a known fixed value or a system calibration value).

[0096] : Represents the tilt angle of the plane containing the beveled edge 420 of the liquid crystal glass 4 relative to the main plane of the optical lens 9.

[0097] In the actual control process, the controller first reads the process parameters of the liquid crystal glass 4 to be tested (e.g., the chamfer angle is...). (e.g., 45°). Since the optical lens 9 is usually mounted perpendicular to the large surface of the liquid crystal glass 4, the angle of inclination of the edge bevel 420 relative to the main plane of the lens is... It is usually directly equal to or related to the chamfer angle of the glass. (For example Or directly for (This depends on the definition of the coordinate system).

[0098] The controller will use the known magnification. and obtained Substituting into the tangent formula above, we can calculate... The value of is then obtained through the arctangent function. Calculate the final target angle .

[0099] For example, if the optical magnification is The angle of the glass chamfer leads to (Right now Then the calculation yields This allows us to calculate the required tilt of the photosensitive target surface. The controller then converts the angle value into a pulse count, driving the electric adjustment mechanism to perform the action.

[0100] Wherein, formula This is the most standard simplified formula of Scherm's Law. It tells us that the tangent of the image plane (sensor) tilt angle is equal to the magnification multiplied by the tangent of the object plane (glass inclined plane). Magnification The same 45-degree slope, if photographed with a high-magnification lens ( (Larger), the camera needs to be tilted even more; use a low-magnification lens to shoot ( (Small), just tilt the camera slightly.

[0101] In this embodiment, the use of an imaging adjustment component 7, which includes an electric adjustment mechanism and a controller, and the automatic calculation and driving of the electric adjustment mechanism by the controller based on the acquired edge slope angle parameter of 420° to adjust the relative angle between the detection camera 8 and the optical lens 9 until the imaging conditions of Schahm's law are met, effectively solves the technical problems of low efficiency, experience-dependent accuracy, and difficulty in adapting to frequent production line changes leading to long downtime for debugging in the prior art. Furthermore, it achieves millisecond-level rapid adaptive adjustment of the optical system to glass with different chamfer angles, ensuring that each product can obtain a clear full-depth imaging in a single inspection, significantly improving the automation level and production efficiency of the flexible production line.

[0102] Further, see Figures 1 to 4In some preferred embodiments designed to achieve simultaneous multidimensional defect detection, the illumination system 10 employs a spatial distribution architecture of a composite spectrum, primarily composed of a surface illumination module and the light source portion of a side-entry optical coupling assembly, and works in conjunction with a color inspection camera 8. The surface illumination module includes a main illumination source 1010 and at least two secondary illumination sources 1020; the main illumination source 1010 provides coaxial illumination of a first wavelength; the two secondary illumination sources 1020 are symmetrically arranged on either side of the main illumination source 1010, providing large-angle diffuse illumination of a second wavelength; the light source of the side-entry optical coupling assembly is configured to emit light of a third wavelength; wherein the first, second, and third wavelengths are different from each other, and the inspection camera 8 is configured as a color camera capable of simultaneously acquiring wavelength information of the first, second, and third wavelengths.

[0103] The main illumination source 1010 is located at the center of the surface illumination module, and its mounting optical axis forms a specific angle with the optical axis of the optical imaging component (usually coaxial or nearly coaxial). This light source is designed as a coaxial illumination structure and is typically equipped with a semi-transparent mirror or collimating lens group, capable of emitting a narrow beam with extremely high parallelism, specifically for simulating specular reflection conditions. In terms of spectral configuration, the main illumination source 1010 is set to emit light of a first wavelength, preferably short-wavelength blue light, utilizing the high scattering rate of short-wavelength light to keenly capture the microscopic morphology of the glass surface.

[0104] At least two secondary illumination sources 1020 are symmetrically mounted on both sides of the main illumination source 1010, forming a wing-like encircling structure. The front end of each secondary illumination source 1020 is fitted with a high-haze diffuser or diffuser lens, causing the emitted light to be no longer parallel light, but rather flexible diffused light with a large divergence angle. In terms of spectral configuration, the secondary illumination source 1020 is configured to emit light of a second wavelength, preferably long-wavelength red light, different from the first wavelength. Its illumination angle range is designed to be very wide, specifically covering the range of primary reflection angle offsets that may occur due to manufacturing process tolerances (such as cutting angle jitter) on the edge bevel 420 of the liquid crystal glass 4.

[0105] The side-entry optical coupling component serves as the light source for internal defect detection. It is configured to emit a third wavelength of light, preferably green or infrared light, which is highly distinguishable from the first two wavelengths. This light source is coupled into the glass interior via a light guide structure to create a dark-field illumination environment. The first, second, and third wavelengths do not overlap or interfere with each other spectrally.

[0106] The detection camera 8 selected in the optical imaging assembly is a single-sensor color industrial camera, such as a CMOS camera equipped with a Bayer array filter or a multi-sensor spectrophotometer. This camera has the ability to simultaneously acquire and distinguish the spectral information of the first wavelength, the second wavelength, and the third wavelength, and can acquire a composite image containing three light field information within one exposure cycle.

[0107] The multi-wavelength illumination system 10 operates based on the logic of spectral multiplexing and light field fusion. During the detection process, when the liquid crystal glass 4 moves to the detection area, three light sources of different wavelengths are activated simultaneously:

[0108] The first wavelength (such as blue light) illuminates the inclined plane coaxially. If the inclined plane surface is smooth and flat, the light is specularly reflected into the lens; if there are tiny scratches on the surface, the light is scattered. This constructs a bright-field model for microscopic surface defects.

[0109] The second wavelength (such as red light) illuminates the slope in a wide-angle diffuse manner. Due to the large coverage and long wavelength of the diffused light, it is not sensitive to minor surface imperfections, but can stably illuminate the overall contour and edge corners of the slope. Even if there are minor tolerances in the glass cutting angle that cause the coaxial light reflection to deviate, the diffused light can still ensure that the slope is illuminated, thus constructing a stable background model for the contour and tolerances.

[0110] A third wavelength (such as green light) propagates inside the glass via side-entry coupling. In defect-free regions, the light is confined by total internal reflection, scattering and escaping only at internal cracks or bubbles. This allows for the construction of a high signal-to-noise ratio dark-field model for internal defects.

[0111] The color inspection camera 8 performs a single exposure on the glass edge under the aforementioned composite light field, and the photosensitive target surface simultaneously records light signals of three wavelengths. Subsequently, the image processing unit uses a channel separation algorithm to decompose the composite image into independent channel images corresponding to the three wavelengths, which are used to analyze surface, contour, and internal features respectively.

[0112] To ensure the purity of channel separation, it is recommended that the equipment be installed in a darkroom environment with a light-shielding enclosure to prevent external full-spectrum white light from interfering with the color camera's color reproduction accuracy. The specific selection of each wavelength must consider the transmittance of the glass material and the camera's quantum efficiency response curve to ensure balanced signal strength across all channels.

[0113] In this embodiment, a composite spectral architecture is adopted, consisting of a main illumination source 1010 providing coaxial illumination of the first wavelength, symmetrically arranged secondary illumination sources 1020 providing diffuse illumination of the second wavelength, and a side-entered optical coupling component providing illumination of the third wavelength. This is combined with a color inspection camera 8 to simultaneously acquire information from the three wavelengths. Therefore, this effectively solves the technical problems in the prior art, such as the difficulty of simultaneously detecting surface scratches and macroscopic contours with a single light source, the loss of main light reflection due to glass processing angle jitter causing missed detections, and the low detection efficiency caused by the need for multiple time-division shooting of different features. As a result, multi-dimensional detection is achieved, which simultaneously completes the identification of surface micro-defects, edge contour tolerance monitoring, and internal defect detection in a single exposure, greatly enhancing the system's adaptability to incoming material processing tolerances and its detection throughput.

[0114] Furthermore, in some embodiments designed to achieve intelligent automatic judgment, the detection device is equipped with a high-performance image processing unit. The image processing unit is configured to receive a composite image acquired by the detection camera 8 in a single exposure and separate it into independent channel images corresponding to the first wavelength, the second wavelength, and the third wavelength. The image processing unit uses the first wavelength channel image to identify surface micro-defects, uses the second wavelength channel image to identify edge contour deviations, and uses the third wavelength channel image to identify internal glass defects.

[0115] The image processing unit typically consists of a high-performance industrial control computer, an embedded processor, or an edge computing module equipped with a graphics processing unit. Its physical interface is directly connected to the inspection camera 8 via a high-speed data transmission bus for real-time reception of massive amounts of image data. Internally, the image processing unit integrates multiple functionally independent algorithm modules, primarily including a spectral channel separation module for deconstructing the original composite image, a surface defect analysis module specifically for processing the first wavelength channel image, a contour tolerance analysis module specifically for processing the second wavelength channel image, and an internal defect analysis module specifically for processing the third wavelength channel image. These modules work collaboratively through a parallel computing architecture, sharing memory resources but independently executing detection logic.

[0116] The image processing unit operates based on a single-image input, multi-path parallel processing, and feature decoupling mechanism. After the detection camera 8 completes a single exposure, the image processing unit first receives the raw composite image data via a high-speed interface. Because the illumination system 10 uses simultaneous illumination at different wavelengths (first wavelength, second wavelength, and third wavelength), this raw image contains mixed spectral information. The spectral channel separation module utilizes the filtering characteristics of the detection camera 8 sensor (such as a Bayer array) or the multi-sensor spectral splitting characteristics to decompose this composite image into three independent grayscale images in logical memory: a first-channel image corresponding to the first wavelength, a second-channel image corresponding to the second wavelength, and a third-channel image corresponding to the third wavelength. These three images each carry completely different physical feature information.

[0117] The image processing unit calls the surface defect analysis module to process the first channel image. Since the first wavelength (such as blue light) mainly forms bright field reflections on the surface, this module first performs background homogenization processing on the image to eliminate uneven illumination, and then uses a high-pass filtering algorithm or texture suppression algorithm to extract high-frequency signals. The algorithm searches for tiny regions in the image where the brightness changes drastically (such as dark spots or abnormal bright lines), and determines them as microscopic scratches, pits, or foreign objects on the surface based on the geometric characteristics (area, aspect ratio) of the connected regions.

[0118] The image processing unit calls the contour tolerance analysis module to process the second channel image. Since the second wavelength (e.g., red light) provides diffuse illumination covering the tolerance band, this module uses a sub-pixel-level edge detection algorithm (e.g., the Canny operator or the Sobel operator) to extract the geometric contour of the glass edge. Subsequently, the algorithm fits and compares the extracted actual contour with a preset standard template or reference line, calculating the offset and angle change of the edge position relative to the theoretical reference. If the offset exceeds a preset tolerance threshold, it is determined to be an edge contour deviation or a processing angle exceeding tolerance.

[0119] The image processing unit calls the internal defect analysis module to process the third channel image. Based on the dark field principle of side-entry light coupling, the ideal background for this channel is pure black. The module uses a high-sensitivity spot detection algorithm or spot clustering algorithm to globally search for pixel regions with abnormally high grayscale values ​​in the black background. Since surface dust usually cannot disrupt total internal reflection to generate bright signals, the algorithm can directly identify the extracted bright spots as internal cracks, chipping, or bubbles in the glass, achieving physical-level separation between internal defects and surface interference.

[0120] In this embodiment, an image processing unit receives a composite image from a single exposure and separates it into independent channel images corresponding to different wavelengths. The first wavelength channel is used to identify surface defects, the second wavelength channel to identify contour deviations, and the third wavelength channel to identify internal defects. Therefore, this effectively solves the technical problems in existing technologies, such as the inability to accurately distinguish defect types due to mixed information in a single image, the slow system cycle due to time-division multiplexing, and the difficulty in quantifying internal microcrack characteristics due to manual visual inspection. This enables a comprehensive automated assessment of surface quality, geometric dimensions, and internal structural integrity within a single inspection cycle, significantly improving the classification accuracy, processing efficiency, and data traceability of defect detection.

[0121] Furthermore, in some embodiments that emphasize system safety and high fault tolerance, the geometry of the rigid air-bearing block 520 in the optical coupler 5 has been specifically designed. The rigid air-bearing block 520 has a guide slope on its upstream side relative to the moving direction of the liquid crystal glass 4 (i.e., the windward side of the liquid crystal glass 4 entering the detection area); the guide slope is configured to push the optical coupler 5 away from the liquid crystal glass 4 through physical contact when the liquid crystal glass 4 moves relative to the optical coupler 5 and its edge has protrusions.

[0122] The guide ramp is located between the side end face of the rigid air float 520 and the working plane facing the liquid crystal glass 4, forming a transition area. Geometrically, the guide ramp is not parallel to the side straight surface 410 of the liquid crystal glass 4, but rather presents a preset tilt angle relative to the side straight surface 410, thus forming a flared structure that gradually converges along the glass movement direction. The surface of the guide ramp is subjected to high-precision grinding and polishing, or coated with a wear-resistant coating with a low coefficient of friction (such as a diamond-like carbon coating or a polytetrafluoroethylene coating) to ensure minimal frictional resistance in the event of accidental physical contact. The tilt angle of the ramp is designed to be relatively gentle to facilitate the smooth conversion of lateral impact forces into normal components.

[0123] The working principle of the guide ramp is based on the mechanical cam effect and the vector decomposition of forces. During normal testing, the working plane of the rigid air-float block 520 is suspended by an air film at a tiny gap from the side surface 410 of the liquid crystal glass 4. The guide ramp is in a non-contact state, does not participate in light guiding, and does not affect air floating. However, when the edge quality of the liquid crystal glass 4 is poor, with sudden protrusions or burrs, or when the instantaneous lateral displacement of the glass edge exceeds the thickness of the air floating film due to transmission jitter, the protruding part of the glass edge will first contact the guide ramp located on the upstream side. At this time, the liquid crystal glass 4 continues to move at high speed along the transmission direction, and the protrusions on its edge press against the guide ramp like a slider. Due to the presence of the ramp, the forward horizontal thrust of the glass is decomposed: one component is perpendicular to the side surface 410 of the liquid crystal glass 4 and points outward. This vertical component acts on the rigid air-float block 520, overcoming the preload or inertia of the floating suspension mechanism 12, forcing the floating photocoupler head 5 to quickly retreat away from the liquid crystal glass 4. Through this process, the optical coupling head 5 is physically pushed away before a head-on collision occurs, thus avoiding the protruding part of the glass edge and allowing it to pass smoothly through the detection area.

[0124] The guide slope can be a straight chamfered surface or a convex arc surface with a certain radius of curvature. The arc surface design can further reduce the contact stress at the moment of contact, making the pushing action smoother. Furthermore, a two-stage guide structure can be designed, with the first stage having a larger angle to cope with a large range of deviations, and the second stage having a smaller angle for precise fine adjustment, gradually guiding the optical coupling head 5 into the working position.

[0125] In this embodiment, by employing a guide slope on the upstream side of the rigid air flotation block 520 relative to the moving direction of the liquid crystal glass 4, the technical problem in the prior art of damage to the equipment or glass edge breakage caused by a sudden protrusion or processing step on the glass edge leading to a rigid forward collision between the precision air flotation component and the glass is effectively solved. Furthermore, it provides passive mechanical safety protection in extreme working conditions where the active avoidance control response is not timely or the air film fails. By using physical contact to convert the lateral impact force into a normal yielding force, the optical coupling head 5 is forced to automatically avoid obstacles, which greatly improves the fault tolerance of the system and the operational safety of the core components.

[0126] Furthermore, in some preferred embodiments that emphasize equipment safety and automated protection, the control architecture of the linear drive component 11 has been deeply optimized, and an advance / retreat control unit has been added. This advance / retreat control unit is not a single mechanical component, but a comprehensive control subsystem integrating sensing and monitoring, logic operation, and execution drive. The advance / retreat control unit is configured to control the optical coupler head 5 to perform active avoidance actions: holding the optical coupler head 5 in an avoidance position before the head of the liquid crystal glass 4 to be detected reaches the optical coupler head 5; pushing the optical coupler head 5 into a detection position after detecting that the head of the liquid crystal glass 4 to be detected has passed the optical coupler head 5; and retracting the optical coupler head 5 to the avoidance position before the tail of the liquid crystal glass 4 to be detected leaves the optical coupler head 5.

[0127] In terms of hardware configuration, the forward and backward control unit includes a position monitoring sensor (such as a fiber optic sensor or a laser beam sensor) set on the transmission path of the liquid crystal glass 4, a drive controller (such as a solenoid valve controller or a servo driver) connected to the linear drive assembly 11, and a core logic processing module.

[0128] Position monitoring sensors are arranged upstream and downstream of the optical coupler 5, or integrated into the transmission mechanism 2, to monitor the real-time position of the head (leading edge) and tail (rear edge) of the liquid crystal glass 4 under test relative to the optical coupler 5.

[0129] The logic processing module receives signals from the sensors and calculates the trigger time for the forward and backward movements based on preset timing logic or encoder position feedback.

[0130] This embodiment defines two key physical position states. One is the avoidance position, in which the optical coupling head 5 is fully retracted, and its foremost tip maintains a sufficient safe distance from the transmission path of the liquid crystal glass 4 to ensure that it will not make contact even if the glass vibrates laterally; the other is the detection position, in which the optical coupling head 5 is extended and in a working state that can fit against the side surface 410 of the liquid crystal glass 4.

[0131] The forward / backward control unit executes active avoidance logic to direct the optical coupler 5 to perform precise throughput movements in the detection cycle. The specific process is as follows:

[0132] When the transmission mechanism 2 is activated but the liquid crystal glass 4 has not yet reached the detection area, or when the head of the liquid crystal glass 4 to be detected is approaching the optocoupler 5 but has not yet completely passed through, the advance / retreat control unit controls the linear drive assembly 11 to remain stationary and locked in the avoidance position. During this stage, the optocoupler 5 is moved away from the transmission line. The purpose of this logic is to prevent the sharp front corner of the liquid crystal glass 4 from directly impacting the optocoupler 5, thus avoiding the shattering of the rigid air float 520 or the tearing of the silicone pad 530.

[0133] Once the position monitoring sensor confirms that the head of the LCD glass 4 has completely passed the physical center line of the optocoupler 5 by a preset safe distance, the advance / retreat control unit issues a feed command. The linear drive assembly 11 quickly moves, pushing the optocoupler 5 from the avoidance position into the detection position. At this time, since the side surface 410 of the glass is directly in front of the optocoupler 5, the optocoupler 5 approaches and adheres to the side surface of the glass vertically, rather than colliding with the corner of the glass. Subsequently, the optocoupler 5 enters the normal follow-up detection state.

[0134] As the detection progresses, the LCD glass 4 continues to move along the preset direction. The advance / retreat control unit continuously monitors the position of the glass's tail. At the critical moment when the tail of the LCD glass 4 is about to reach the optical coupler head 5, but has not yet completely left the contact surface of the optical coupler head 5 (i.e., the preset lead time before the tail arrives), the advance / retreat control unit immediately issues a retraction command. The linear drive assembly 11 responds quickly, forcibly pulling the optical coupler head 5 from the detection position to the avoidance position. This action occurs before the glass completely leaves, preventing the optical coupler head 5 from falling and hitting the glass due to loss of support the moment it slides out of the tail, and also preventing the optical coupler head 5 from being struck by the head of the next glass piece that follows.

[0135] This proactive obstacle avoidance solution is suitable for continuous automated production lines, especially in scenarios with high transmission speeds and sharp glass edges.

[0136] In this embodiment, the active avoidance technology of the linear drive assembly 11, controlled by the advance and retreat control unit, is effectively solved by using a method that allows the head of the liquid crystal glass 4 to avoid the object before it arrives, push it into the detection position after the head passes, and retract to avoid the object before the tail leaves. This effectively solves the technical problems in the prior art, such as equipment damage or glass chipping caused by the continuous extension of the optical coupling head 5 and its rigid collision with the sharp corner of the liquid crystal glass 4, as well as the falling impact caused by the slippage of the tail of the glass. This achieves safe operation protection for the optical coupling assembly throughout its entire life cycle, ensuring efficient optical coupling detection while maximizing the service life of the core contact components and improving the operational reliability of the system.

[0137] In particular, to ensure good defect detection results, this embodiment proposes a control method for the above-mentioned liquid crystal glass 4 defect detection device, which specifically includes the following steps:

[0138] Step S100: Based on the 420° angle of the edge slope of the liquid crystal glass 4 to be tested, control the imaging adjustment component 7 to adjust the angle of the detection camera 8 to meet the imaging conditions of Schahm's law;

[0139] Step S200: Control the linear drive assembly 11 to push the optical coupling head 5 toward the liquid crystal glass 4 to be tested, activate the floating suspension mechanism 12 and apply coupling thrust, so that the rigid air float 520 forms the fixed-distance air film on the surface of the liquid crystal glass 4, and the silicone pad 530 is compressed and adheres to the side surface 410 of the liquid crystal glass 4 to be tested at the distance defined by the fixed-distance air film.

[0140] Step S300: Activate the surface illumination module and the side-entry light coupling component, and control the detection camera 8 to perform a single exposure acquisition on the edge slope 420 of the liquid crystal glass 4 to be detected, and acquire a composite image containing the first wavelength, the second wavelength and the third wavelength information;

[0141] Step S400: Separate the composite image into the first wavelength channel image, the second wavelength channel image, and the third wavelength channel image to determine surface defects, contour deviations, and internal defects respectively.

[0142] The specific implementation process of step S100 is as follows:

[0143] Step S110: Obtain process parameters and system constants. The controller first obtains the edge bevel angle parameter of the current LCD glass 4 to be inspected (denoted as 420°) through the human-machine interface, production line management system communication, or upstream barcode scanner. This parameter represents the theoretical chamfer angle (e.g., 45°, 60°, etc.) set for the glass edging process. Simultaneously, the controller retrieves pre-stored optical system constants from the memory unit, primarily including the lateral magnification of the optical lens 9 (denoted as...). This magnification is a known fixed value determined by the selected industrial lens and the current working distance (WD).

[0144] Step S120: Execute Scham's law geometric solution. The controller uses a preset geometric optics algorithm model to perform real-time solution.

[0145] The algorithm is based on the mathematical expression of Scham's Law: .

[0146] in: This refers to the tilt angle of the object's plane relative to the lens's principal plane. Since the optical lens 9 is typically mounted perpendicular to the horizontal surface of the liquid crystal glass 4, therefore... Directly related to the 420° angle parameter of the edge bevel (For example, in a specific coordinate system) Or take directly ). This refers to the target tilt angle of the photosensitive target surface relative to the lens's principal plane. The controller will acquire the edge slope angle parameters. Convert to And combined with magnification Substitute into the formula to calculate Then, by using inverse trigonometric functions, the target angle that the detection camera's 8-sensor surface theoretically needs to deflect can be calculated. ).

[0147] Step S130: After the closed-loop position control calculation of the drive mechanism is completed, the controller generates the corresponding motion control command to drive the electric adjustment mechanism (such as an electric angle measuring stage or an electric tilting flange) in the imaging adjustment assembly 7. The motor drives the detection camera 8 to rotate around a preset pivot point. During the rotation, the angle encoder inside the imaging adjustment assembly 7 feeds back the current actual angle value to the controller in real time. The controller uses a PID closed-loop control algorithm to compare the actual angle with the target angle value (θtarget) and correct the error until the detection camera 8 accurately reaches the target position and locks in, completing the coarse adjustment.

[0148] The specific implementation process of step S200 is as follows:

[0149] Step S210: Pre-activation and self-check of the air flotation system Before performing any mechanical movement, the controller first opens the air circuit control valve to supply air to the air bushing of the floating suspension mechanism 12 and the rigid air flotation block 520 of the optical coupling head 5.

[0150] Compressed air enters the porous bushing 1230 of the air bushing, forming a lubricating air film between the floating shaft 1210 and the bushing, so that the optical coupling head 5 is in a zero-friction suspended standby state, eliminating static friction.

[0151] The air pressure sensor monitors the air supply circuit pressure in real time. Only when the air pressure reaches the preset working threshold and stabilizes will the controller allow subsequent actions to be executed, in order to prevent the rigid air flotation block 520 from being scratched by dry friction with the glass due to insufficient air pressure.

[0152] Step S220: Macroscopic feed (coarse positioning) of linear drive assembly 11. The controller sends a feed command to the linear drive assembly 11 (such as a cylinder or electric slide). The linear drive assembly 11 drives the entire side-entry optocoupler assembly to move rapidly from the safety clearance position toward the straight surface 410 of the liquid crystal glass 4 side.

[0153] The linear drive assembly 11 does not directly push the optical coupling head 5 to the contact glass, but instead moves to a preset near-field ready position and stops and locks. This position maintains a millimeter-level safety clearance (e.g., 1mm to 2mm) from the theoretical position of the side surface 410 of the liquid crystal glass 4, ensuring that even if there is a large positioning error in the glass, the rigid linear drive mechanism will not directly impact the glass.

[0154] Step S230: Micro soft landing (precise contact) of constant force thrust mechanism 13 After the linear drive assembly 11 is locked in the near field ready position, the controller activates the constant force thrust mechanism 13 (such as voice coil motor) on the floating suspension mechanism 12.

[0155] The controller controls the voice coil motor to output a pre-set small constant current, which generates a constant thrust pointing towards the glass side surface 410.

[0156] Under this thrust, the optical coupling head 5, relying on the zero-friction characteristics of the air bushing, slides along the floating shaft 1210 relative to the stationary linear drive assembly 11 toward the glass surface. Due to the constant and gentle thrust, the optical coupling head 5 approaches the glass at a controlled low speed, achieving flexible contact.

[0157] Step S240: Air film spacing and mechanical balance locking as the optical coupling head 5 continues to approach the glass.

[0158] The front end of the rigid air flotation block 520 first senses the back pressure of the glass surface, and the airflow is obstructed. A high-rigidity fixed-distance air film (thickness locked at, for example, 10 micrometers) is quickly formed between the rigid air flotation block 520 and the glass side surface 410.

[0159] At the same time, the silicone pad 530 protruding from the surface of the rigid air flotation block 520 comes into contact with the glass and is squeezed.

[0160] When the thrust of the constant force mechanism 13 equals the elastic rebound force of the silicone pad 530 plus the fluid repulsion force of the air film, the system automatically reaches a state of mechanical equilibrium. At this time, the rigid air flotation block 520 is precisely locked by the air film at a height of micrometers above the glass, while the silicone pad 530 is forced to maintain a constant compression, tightly adhering to the glass side surface 410, thus completing the preparation for efficient coupling of the optical path.

[0161] In step S300, single-exposure acquisition refers to the electronic shutter of the detection camera 8 opening only once, simultaneously receiving and converting all target wavelength light signals within a continuous integration time window. Unlike existing time-division stroboscopic shooting (i.e., taking one picture with the red light on first, then another with the blue light on), this method does not require mechanical movement or multiple data readouts, thus greatly improving the detection speed and eliminating image position registration errors that may occur between multiple shots.

[0162] The steps for acquiring a composite image include:

[0163] Step S310: The synchronous excitation controller of the multidimensional light field sends a synchronous trigger signal to the lighting system 10.

[0164] The main illumination source 1010 (emitting a first wavelength, such as blue light), the secondary illumination source 1020 (emitting a second wavelength, such as red light), and the light source of the side-entered optical coupling component (emitting a third wavelength, such as green light) in the surface illumination module are simultaneously lit (or kept on during the same exposure cycle of the camera).

[0165] At this time, three light fields are physically superimposed on the edge slope 420 of the liquid crystal glass 4: a blue coaxial reflection light field, a red diffuse contour light field, and a green internal scattering light field. These three light rays, carrying their respective characteristic information, simultaneously enter the optical lens 9.

[0166] Step S320: After receiving the trigger signal, the spectral spatial sampling detection camera 8 based on the filter array opens the electronic shutter for exposure. The detection camera 8 is preferably a single-chip color camera equipped with a Bayer filter array.

[0167] The light entering the lens reaches the photosensitive target surface. The photosensitive target surface is covered with a micron-sized mosaic filter (R, G, B filter array).

[0168] The blue filter unit on the photosensitive target surface only allows the first wavelength (blue light) from the main illumination source 1010 to pass through, blocking red and green light. Therefore, the B-channel pixel physically records only the reflected energy from the microscopic scratches on the surface.

[0169] The red filter unit on the photosensitive target surface only allows the second wavelength (red light) from the secondary illumination source 1020 to pass through. Therefore, the R channel pixel physically records only the diffuse energy of the edge contour.

[0170] The green filter unit on the photosensitive target surface only allows the third wavelength (green light) from the side-entry component to pass through. Therefore, the G-channel pixel physically records only the scattered energy of internal defects.

[0171] Each pixel simultaneously converts the energy of photons of a specific wavelength it receives into a charge signal, enabling parallel sampling of three physical features on the same sensor plane at the same time.

[0172] Step S330: RAW data readout and composite image generation After exposure, the analog-to-digital converter inside the camera converts the charge signal of each pixel into a digital grayscale value.

[0173] The camera doesn't distinguish what defects these lights represent; it directly packages the grayscale data of the R, G, and B pixels into a standard color image data format. This image is the composite image. To the human eye, it may appear as a colorful picture, but at the data level, it is a three-dimensional matrix that strictly corresponds to the intensity distribution information of the three wavelengths of light, providing the raw data foundation for the channel separation in the subsequent S400 step.

[0174] The specific implementation process of step S400 is as follows:

[0175] Step S410: The image processing unit receives single-frame color composite image data transmitted from the detection camera 8. Since the illumination system 10 uses physically isolated spectra, and the detection camera 8 employs a Bayer array filter, the image processing unit executes the following separation logic:

[0176] If the input is raw data, first execute the demosaic interpolation algorithm to generate an RGB full-color image; if the input is already in RGB format, then directly perform in-memory data splitting.

[0177] The RGB data matrix of the composite image is decomposed into three independent two-dimensional grayscale matrices:

[0178] B-channel grayscale image (first wavelength image): Mapped to a surface defect detection map, containing only reflection information in the approximately 450-470nm wavelength band;

[0179] R-channel grayscale image (second wavelength image): mapped to a contour tolerance detection map, containing only diffuse information in the approximately 620-640nm band;

[0180] G-channel grayscale image (third wavelength image): Mapped to an internal defect detection map, containing only scattering information in the approximately 520-540nm wavelength band.

[0181] Step S420: Surface micro-defect determination based on the first wavelength channel. For the separated B-channel grayscale image (first wavelength), the image processing unit performs the following analysis:

[0182] Gaussian smoothing filter is used to estimate the background illumination distribution. The original image is divided by the background image to eliminate low-frequency illumination changes caused by uneven reflection from the glass slope and enhance local contrast.

[0183] Apply a high-pass filter (such as the Laplacian operator or the Sobel operator) to the homogenized image to suppress smooth background areas and highlight edge signals where brightness changes drastically.

[0184] Perform dynamic thresholding on the filtered image. Extract connected components of pixels whose gray values ​​are higher (or lower) than the background noise floor. Calculate the geometric features of the connected components (such as aspect ratio and compactness).

[0185] If the connected region exhibits linear characteristics and its aspect ratio exceeds a preset value, it is identified as a "scratch"; if it is circular and has a small area, it is identified as a "pockmark" or "foreign object".

[0186] Step S430: Contour Deviation Determination Based on the Second Wavelength Channel. For the separated R-channel grayscale image (second wavelength), the image processing unit performs the following analysis:

[0187] In the stable spot image formed by red diffuse illumination, the upper and lower physical boundary contours of the glass edge slope 420 are accurately extracted by using the Canny edge detection operator combined with the sub-pixel interpolation algorithm.

[0188] The extracted contour points are fitted to geometric straight lines using the least squares method. Pre-set standard sample baseline contour data is then retrieved.

[0189] Calculate the Euclidean distance (positional deviation) and the included angle (angular deviation) between the current fitted line and the reference line.

[0190] If the calculated positional or angular deviation exceeds the preset process tolerance threshold (e.g., ±0.1mm or ±0.5°), it is determined as "contour deviation" or "chamfer angle out of tolerance", and an NG signal is output.

[0191] Step S440: Internal Defect Judgment Based on the Third Wavelength Channel. For the separated G-channel grayscale image (third wavelength), the image processing unit performs the following analysis:

[0192] Since this channel is based on the principle of total internal reflection, the ideal background is pure black. The algorithm first calculates the grayscale histogram of the entire image to determine the basis threshold of the background noise.

[0193] A speckle detection algorithm is used to globally search for pixel clustering regions in the image where the gray value is significantly higher than the background noise threshold.

[0194] Calculate the area, brightness, and position of the extracted highlighted areas.

[0195] Since surface dust does not disrupt total internal reflection (does not emit light), while internal cracks disrupt total internal reflection (emitting light), any bright spot appearing in this channel, and whose area is larger than the minimum resolution limit (e.g., 5 pixels), is directly identified as an "internal crack," "edge chipping," or "bubble" in the glass.

[0196] In summary, by executing steps S100 to S400, the control method proposed in this embodiment deeply integrates physical-level optomechanical precision adjustment (Schahm's law attitude correction, air-bearing follower light coupling) with algorithm-level multidimensional spectral analysis (composite image channel separation). This detection process not only physically eliminates the depth-of-field blurring of inclined imaging and the mechanical risks of contact light guiding, but also logically achieves complete decoupling of surface, contour, and internal features. Therefore, through the above steps, the device can accurately and efficiently detect specific defects such as microscopic scratches, geometric contour processing errors, and deep internal cracks on the inclined surface 420 of the LCD glass 4 edge within a single detection cycle, thereby meeting the stringent requirements of high-generation LCD panel production lines for full edge quality inspection.

[0197] The above description is merely illustrative of the invention. Those skilled in the art can make various modifications or additions to the described specific embodiments or use similar methods to replace them, as long as they do not depart from the content of this specification or exceed the scope defined by the claims, all of which should fall within the protection scope of this invention.

Claims

1. A liquid crystal glass defect detection apparatus, wherein, The liquid crystal glass includes an edge bevel and a side straight surface, characterized in that it includes: The bracket has a transmission mechanism at its top, which has a transmission plane for carrying and driving the liquid crystal glass to move along a preset direction. A gantry frame is mounted on top of the support and spans the conveying plane, and the gantry frame is equipped with a controllable movable end; An optical imaging component is connected to the mobile terminal. The optical imaging component includes a detection camera and an optical lens. The detection camera has a photosensitive target surface, and the optical lens has a main plane. An imaging adjustment component is connected between the detection camera and the optical lens to adjust the angle of the detection camera relative to the optical lens, so that the extended surfaces of the photosensitive target surface, the main plane, and the edge slope of the liquid crystal glass to be detected intersect on the same straight line when the detection device is in the detection state. Lighting system, including: Surface lighting module for illuminating the edge bevel; A side-entry optical coupling assembly for detecting internal defects near the edge of a liquid crystal glass includes a linear drive assembly, a floating suspension mechanism, and an optical coupling head. The linear drive assembly is connected to the bracket or the gantry frame and drives the optical coupling head to move forward and backward toward the side surface of the liquid crystal glass to be inspected. The floating suspension mechanism is disposed between the linear drive assembly and the optical coupling head and is configured to allow the optical coupling head to float frictionlessly relative to the linear drive assembly in a direction perpendicular to the side surface. The optical coupling head includes a rigid air float, a light guide prism, and a silicone pad. The rigid air float is configured to form a fixed-distance air film between the detection device in the detection state and the side surface of the liquid crystal glass to be inspected. The silicone pad is configured to be compressed and adhere to the side surface within the distance defined by the fixed-distance air film. The floating suspension mechanism includes a floating shaft, an air bushing, and a connecting structure; the floating shaft is disposed at the moving end of the linear drive assembly along a direction pointing towards the side face; the air bushing is sleeved outside the floating shaft and connected to the optical coupling head through the connecting structure; the air bushing includes a bushing housing and a porous bushing disposed inside the bushing housing, and a preset gap for gas flow is formed between the porous bushing and the floating shaft to form an air film supporting the free sliding of the air bushing in the ventilated state; The optical coupling head has a stepped contact surface structure: the rigid air-float block has an internal receiving through hole, the light guide prism is disposed in the receiving through hole, and the rigid air-float block protrudes from the light-emitting surface of the light guide prism. The silicone pad is located in the receiving through hole and is in contact with the light-emitting surface of the light guide prism. The silicone pad protrudes from the outer side of the rigid air-float block in its natural state. The rigid air-float block has a guide slope on its upstream side relative to the moving direction of the liquid crystal glass. When the detection device is in the detection state, the rigid air-float block locks the distance between itself and the liquid crystal glass through the air film, and the protruding silicone pad is compressed to fit tightly against the side surface.

2. The liquid crystal glass defect detection apparatus according to claim 1, characterized by The side-entry optical coupling assembly further includes a constant force thrust mechanism, which is disposed on the floating shaft. The constant force thrust mechanism includes a pushing end, which is connected to the air bushing. The constant force thrust mechanism is configured to apply a constant thrust pointing towards the side surface of the liquid crystal glass to the air bushing when the detection device is in the detection state, so as to balance the rebound force when the silicone pad is compressed and the air film repulsion force generated by the rigid air float.

3. The liquid crystal glass defect detection device according to claim 1, characterized in that, The imaging adjustment assembly includes an electric adjustment mechanism and a controller; the controller is configured to acquire the edge bevel angle parameter of the liquid crystal glass to be inspected, and control the electric adjustment mechanism to automatically adjust the relative angle between the detection camera and the optical lens according to the angle parameter, until the extended surfaces of the photosensitive target surface, the main plane and the edge bevel of the liquid crystal glass to be inspected intersect on the same straight line when the detection device is in the detection state.

4. The liquid crystal glass defect detection device according to claim 1, characterized in that, The surface lighting module includes a main lighting source and at least two secondary lighting sources; The main illumination source provides coaxial illumination of a first wavelength; two secondary illumination sources are symmetrically arranged on both sides of the main illumination source to provide wide-angle diffuse illumination of a second wavelength; the light source of the side-entry optical coupling component is configured to emit light of a third wavelength; wherein the first wavelength, the second wavelength, and the third wavelength are different from each other, and the detection camera is configured as a color camera capable of simultaneously acquiring wavelength information of the first wavelength, the second wavelength, and the third wavelength.

5. The liquid crystal glass defect detection device according to claim 4, characterized in that, It also includes an image processing unit; the image processing unit is configured to receive a composite image acquired by the detection camera in a single exposure, and separate it into independent channel images corresponding to the first wavelength, the second wavelength and the third wavelength; the image processing unit uses the first wavelength channel image to identify surface micro-defects, uses the second wavelength channel image to identify edge contour deviations, and uses the third wavelength channel image to identify internal defects in the glass.

6. The liquid crystal glass defect detection device according to claim 5, characterized in that, The guide ramp is configured to push the optical coupling head away from the liquid crystal glass through physical contact when the liquid crystal glass moves relative to the optical coupling head and its edge has a protrusion.

7. The liquid crystal glass defect detection device according to claim 6, characterized in that, The linear drive assembly further includes an advance / retreat control unit; the advance / retreat control unit is configured to control the optical coupler to perform an active avoidance action: holding the optical coupler in an avoidance position before the head of the liquid crystal glass to be detected reaches the optical coupler; pushing the optical coupler into a detection position after detecting that the head of the liquid crystal glass to be detected has passed the optical coupler; and retracting the optical coupler to the avoidance position before the tail of the liquid crystal glass to be detected leaves the optical coupler.

8. A control method for the liquid crystal glass defect detection device as described in any one of claims 4-7, characterized in that, Includes the following steps: Based on the edge slope angle of the liquid crystal glass to be tested, the imaging adjustment component is controlled to adjust the angle of the detection camera to meet the imaging conditions of Scherm's law. The linear drive assembly is controlled to push the optical coupling head toward the liquid crystal glass to be tested, the floating suspension mechanism is activated and a coupling thrust is applied, so that the rigid air float block forms the fixed-distance air film on the surface of the liquid crystal glass, and the silicone pad is compressed and adhered to the side surface of the liquid crystal glass to be tested within the distance defined by the fixed-distance air film. Activate the surface illumination module and the side-entry optical coupling component, and control the detection camera to perform a single exposure acquisition on the edge slope of the liquid crystal glass to be detected, and obtain a composite image containing information of the first wavelength, the second wavelength and the third wavelength. The composite image is separated into a first wavelength channel image, a second wavelength channel image, and a third wavelength channel image to determine surface defects, contour deviations, and internal defects, respectively.

Images