TGV glass substrate double-shaft synchronous stable scanning detection method

Abstract

The invention discloses a TGV glass substrate double-axis synchronous stable scanning detection method, which comprises the following steps: firstly, importing hole site coordinate data of a TGV glass substrate, dividing a dense sub-region and a sparse sub-region according to hole site space distribution density, generating a variable density scanning path, and realizing dynamic adaptation of detection speed and resolution; then the double-shaft synchronous driving mechanism drives the substrate to move, and the multi-degree-of-freedom visual scanning mechanism automatically switches the light source state and the camera rotation angle according to the type of the detected target and synchronously collects images; the adsorption pressure is monitored in real time in the whole detection process, and when the pressure is lower than a safety threshold value, automatic speed reduction or alarm is performed to guarantee substrate safety; and finally, defect identification and classification are completed through multi-frame image noise reduction, feature extraction and a support vector machine algorithm. The problems that in traditional detection, efficiency and precision are difficult to consider at the same time, the substrates are prone to damage, and the automation degree is low are solved, efficient, high-precision and low-damage batch detection of the TGV glass substrates is achieved, and the method is suitable for online detection scenes of mass production lines.

Description

Technical Field

[0001] This invention relates to a method for biaxial synchronous stable scanning detection of TGV glass substrates. Background Technology

[0002] With the development of advanced packaging technologies, TGV (Through Glass Via) technology has become a key process for high-density interconnect packaging due to its excellent electrical, thermal, and mechanical properties. The fabrication of TGV glass substrates involves multiple processes such as drilling and metallization, which can easily lead to defects such as hole wall cracks, hole diameter deviations, surface scratches, and incomplete metal filling. Therefore, efficient and high-precision defect detection of TGV glass substrates is crucial during mass production.

[0003] The existing TGV glass substrate inspection technology has the following main shortcomings:

[0004] (1) It is difficult to balance detection efficiency and accuracy. Traditional detection equipment mostly adopts single-axis scanning or static detection methods, which can only detect a local area of ​​a single substrate at a time, making it difficult to meet the batch detection needs of mass production lines. Although some automated detection equipment can achieve continuous scanning, they mostly adopt a uniform scanning strategy and fail to make adaptive adjustments according to the density characteristics of the via distribution on the TGV glass substrate. This results in the scanning speed being too fast in densely populated areas and the image quality deteriorating, or the scanning speed being too slow in sparsely populated areas and the efficiency being low. In addition, the light source configuration and camera shooting angle of existing equipment are relatively fixed, making it difficult to simultaneously meet the different optical requirements of surface defect detection (requiring a uniform surface light source) and via wall defect detection (requiring a specific angle directional light source), resulting in a high rate of missed detection of deep via wall defects.

[0005] (2) There is a contradiction between substrate positioning and protection. TGV glass substrates are fragile and thin (usually 0.1-1mm). Existing positioning mechanisms mostly use rigid clamping or simple vacuum adsorption. Rigid clamping is prone to stress concentration, which can lead to substrate cracking; while vacuum adsorption, if it lacks real-time pressure monitoring and motion linkage protection, can easily cause substrate slippage or deformation damage when scanning vibration or acceleration changes.

[0006] (3) Lack of flexibility in switching detection modes. When detecting surface defects and hole wall defects, existing equipment often requires manual intervention to switch the light source type or adjust the camera angle. It is difficult to automatically identify the current detection area type and adjust the detection parameters in real time based on the hole position coordinate data, resulting in insufficient automation and detection continuity.

[0007] Therefore, there is an urgent need for a scanning and inspection method that can adaptively adjust the scanning strategy according to the hole structure characteristics of TGV glass substrates, take into account the detection requirements of surface and hole wall defects, and have a flexible safety protection mechanism, so as to achieve high-efficiency, high-precision, and low-damage batch inspection. Summary of the Invention

[0008] The present invention provides a method for biaxial synchronous stable scanning detection of TGV glass substrates to solve the problems existing in the prior art.

[0009] The technical solutions adopted in this invention are as follows:

[0010] A method for dual-axis synchronous stable scanning inspection of TGV glass substrates, based on an inspection device comprising a dual-axis synchronous drive mechanism, a substrate support and positioning mechanism, a multi-degree-of-freedom vision scanning mechanism, and a control system, includes the following steps:

[0011] (1) The control system acquires the hole position coordinate data of the TGV glass substrate, divides the scanning area into dense sub-regions and sparse sub-regions according to the spatial distribution density of the hole positions, generates a variable density scanning path, reduces the scanning speed and increases the imaging resolution in the dense sub-regions, and increases the scanning speed and decreases the imaging resolution in the sparse sub-regions.

[0012] (2) The dual-axis synchronous drive mechanism drives the substrate bearing positioning mechanism to move according to the variable density scanning path. The multi-degree-of-freedom visual scanning mechanism adjusts the light source and rotation state according to the type of the detection target: when detecting surface defects, the ring light source in the multi-degree-of-freedom visual scanning mechanism is lit, the strip light source is turned off and the rotation adjustment component maintains the current angle; when detecting hole wall defects, the strip light source in the multi-degree-of-freedom visual scanning mechanism is lit, the ring light source is turned off and the rotation adjustment component drives the linear scan camera to rotate around the Z-axis to adjust the shooting angle. The linear scan camera in the multi-degree-of-freedom visual scanning mechanism synchronously acquires images.

[0013] (3) The flexible adsorption unit in the substrate bearing positioning mechanism collects adsorption pressure data in real time. When the pressure is lower than the preset safety threshold, the control system reduces the current scanning speed to a safe speed or triggers an alarm.

[0014] (4) The control system performs defect identification on the acquired images and outputs the defect judgment results.

[0015] Furthermore, the scanning speed of the dense sub-region is reduced to 50%-70% of the reference speed, and the imaging resolution is increased to 120%-150% of the reference resolution; the scanning speed of the sparse sub-region is increased to 130%-150% of the reference speed, and the imaging resolution is reduced to 60%-80% of the reference resolution.

[0016] Furthermore, the spatial distribution density of the holes is calculated based on the number of holes per unit area. Dense sub-regions are areas where the number of holes per unit area is greater than a preset threshold, and sparse sub-regions are areas where the number of holes per unit area is less than or equal to the preset threshold.

[0017] Furthermore, when detecting defects in the hole wall, the rotation adjustment component drives the linear scan camera to rotate around the Z-axis in an angle range of 0° to 360°, with an angle adjustment accuracy not exceeding ±0.1°.

[0018] Furthermore, the brightness of the ring light source and the strip light source is dynamically adjusted according to the surface reflectivity of the TGV glass substrate. The brightness adjustment range of the ring light source is 1000-5000 lux, and the brightness adjustment range of the strip light source is 500-3000 lux.

[0019] Furthermore, the preset safety threshold is 0.02-0.05 MPa, and the safety speed is 30%-50% of the current scanning speed.

[0020] Furthermore, the defect identification in step (4) includes: performing noise reduction and enhancement preprocessing on the acquired image, extracting defect feature parameters, and classifying the defect type using the support vector machine algorithm.

[0021] Furthermore, the detection target type is determined based on the hole position coordinate data: when the current field of view of the linear scan camera coincides with the hole position coordinates, it is determined to be hole wall defect detection; when the current field of view deviates from the hole position coordinates, it is determined to be surface defect detection.

[0022] Furthermore, the defect identification in step (4) also includes: performing weighted average noise reduction processing on the continuously acquired multi-frame images, and outputting the final defect judgment by fusing the defect identification results of the multi-frame images.

[0023] Furthermore, the defect determination includes the identification and classification of cracks, pore diameter deviations, excessive pore wall roughness, and incomplete metal filling.

[0024] The present invention has the following beneficial effects:

[0025] (1) By generating a variable-density scanning path based on the spatial distribution density of the holes, the scanning speed is reduced and the imaging resolution is increased in densely populated areas, while the scanning speed is increased and the resolution is reduced in sparsely populated areas, thus achieving dynamic optimization of the detection resources. Compared with the traditional uniform-speed scanning method, this method avoids both the blurring and missed detection caused by excessive speed in densely populated areas and the time waste caused by excessively slow speed in sparsely populated areas. It significantly improves the overall detection efficiency while ensuring the detection accuracy of key areas, making it suitable for the online detection needs of mass production lines.

[0026] (2) By establishing a collaborative control strategy for the type of target to be detected, the light source, and the rotation state, differentiated detection is implemented for the different optical characteristics of surface defects and hole wall defects: when detecting surface defects, a ring light source is used for uniform illumination and a fixed angle is maintained; when detecting hole wall defects, a strip light source is used for directional illumination and the shooting angle is adjusted in conjunction with a rotation adjustment component. This collaborative control method effectively solves the problem of insufficient imaging contrast caused by the high light transmittance of TGV glass substrates and the obstruction of hole structures, significantly improves the recognition rate of deep hole defects such as hole wall cracks and hole diameter deviations, and reduces the risk of missing small defects.

[0027] (3) By monitoring the adsorption pressure of the flexible adsorption unit in real time and establishing a safety interlock mechanism between pressure data and scanning speed, the scanning speed is automatically reduced or an alarm is triggered when the pressure is lower than the preset safety threshold, effectively avoiding problems such as substrate slippage, increased vibration, or stress concentration caused by insufficient adsorption force. Compared with traditional fixed-speed scanning or simple shutdown protection methods, the flexible degradation strategy of this invention maintains the continuity of detection while ensuring substrate safety, and significantly reduces the breakage rate of TGV glass substrates during the detection process.

[0028] (4) By automatically determining the type of the detection target based on the hole position coordinate data, the automatic switching between surface detection mode and hole wall detection mode is realized. The light source configuration and camera posture can be automatically adjusted according to the current field of view without manual intervention. At the same time, combined with multi-frame image weighted average noise reduction and defect identification result fusion technology, the accuracy and robustness of defect judgment are further improved. The entire detection process is automatically completed by the control system, reducing the impact of human factors on the detection results and improving the automation level and detection consistency of the production line. Attached Figure Description

[0029] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0030] The invention will now be further described with reference to the accompanying drawings.

[0031] The detection method of the present invention is based on a modular detection device consisting of a dual-axis synchronous drive mechanism, a substrate bearing and positioning mechanism, a multi-degree-of-freedom vision scanning mechanism and a control system. The entire detection process, from equipment initialization, path planning, and substrate positioning to synchronous scanning, safety protection and defect identification, is closely connected to ensure that the detection process is efficient, accurate and low-damage.

[0032] The dual-axis synchronous drive mechanism, the substrate bearing and positioning mechanism, and the multi-degree-of-freedom vision scanning mechanism are all existing conventional structures. The following is only a brief introduction to the structure used in this invention.

[0033] The dual-axis synchronous drive mechanism is an XY dual-axis linear motion platform driven by a servo motor. It adopts high-precision linear guides and synchronous belt transmission to achieve precise positioning and smooth movement of the platform in the horizontal plane.

[0034] The substrate support and positioning mechanism includes an aluminum alloy support platform and a flexible adsorption unit and an edge limiting unit mounted on it. The flexible adsorption unit uses a vacuum adsorption assembly (including a vacuum suction cup, an air pump interface, and a pressure sensor) to fix the substrate by negative pressure adsorption; the edge limiting units are symmetrically arranged on both sides of the support platform for initial positioning of the substrate.

[0035] The multi-degree-of-freedom visual scanning mechanism is mounted on a gantry bracket above the base and includes a Z-axis lifting assembly, a rotation adjustment assembly, a linear scanning camera, and a light source module. The Z-axis lifting assembly is driven by a ball screw to achieve precise vertical positioning of the camera; the rotation adjustment assembly drives the camera to rotate around the Z-axis to adjust the shooting angle; the light source module includes a ring light source and a strip light source, which are used for illuminating surface defects and hole wall defects, respectively.

[0036] The specific process of the biaxial synchronous stable scanning detection method for TGV glass substrates of the present invention is as follows:

[0037] Before the testing process begins, a comprehensive initialization of the equipment and parameters is completed. The control system establishes a connection with the TGV glass substrate design system through a preset communication interface, imports the CAD design file corresponding to the substrate to be tested, and accurately extracts the two-dimensional hole position coordinate data of all TGV holes. This data includes key information such as the center coordinates and diameter range of each hole, providing a precise basis for subsequent scanning area division and target type determination.

[0038] Operators input the basic parameters of the substrate through the human-machine interface, and simultaneously set the baseline scanning speed and baseline imaging resolution. The baseline scanning speed setting needs to be determined in conjunction with the cycle time requirements of the mass production line and the substrate size to ensure that production efficiency is matched without affecting the inspection quality. The baseline imaging resolution is set to be able to clearly identify defects of common sizes, ensuring basic inspection accuracy. At the same time, the substrate bearing positioning mechanism performs a reset operation, and the limiting blocks of the edge limiting unit are adjusted to a distance that matches the size of the substrate to be inspected through the built-in adjusting screw, ensuring that the substrate can be initially positioned when placed; the vacuum suction cup of the flexible adsorption unit is in standby mode, and its matching pressure sensor completes zero-point calibration and accuracy verification to ensure the accuracy of subsequent adsorption pressure detection.

[0039] In the multi-degree-of-freedom visual scanning mechanism, the Z-axis lifting component, driven by a stepper motor, adjusts the linear scanning camera to a preset initial height via ball screw transmission. This height is precisely matched with the substrate thickness and the camera focal length to ensure initial image clarity. The rotation adjustment component drives the camera to reset around the Z-axis to the initial angle. Both the ring light source and the bar light source are in the off state, waiting for the detection command to be triggered.

[0040] After initialization, the scanning path planning stage begins. Based on the extracted hole coordinate data, the control system uses a grid-based method to divide the substrate surface into several equal-area micro-regions. The number of holes within each micro-region is calculated, thus obtaining the number of holes per unit area, which serves as a quantitative indicator of the spatial distribution density of the holes. A preset threshold for the number of holes per unit area is calibrated using extensive experimental data. This threshold must balance defect detection risk and detection efficiency. When the number of holes per unit area in a micro-region exceeds the preset threshold, the region is identified as a densely populated hole sub-region. Due to the concentrated hole structure, these regions have a higher probability of defects such as hole wall cracks, hole diameter deviations, and incomplete metal filling, requiring higher detection accuracy. When the number of holes per unit area is less than or equal to the preset threshold, it is identified as a sparsely populated hole sub-region. These regions have a lower probability of defects, allowing for a more appropriate increase in detection speed to improve overall efficiency.

[0041] Based on the above partitioning results, the control system automatically generates a continuous and interconnected variable density scanning path. During path planning, it ensures that the scanning trajectories of dense and sparse sub-regions do not overlap or omission, and achieves a smooth transition of scanning speed and resolution at the boundary of regions, avoiding imaging fluctuations caused by sudden parameter changes.

[0042] For dense sub-regions, the scanning speed is reduced to 50%-70% of the baseline speed, while the imaging resolution is increased to 120%-150% of the baseline resolution by adjusting the pixel windowing mode of the linear scan camera or the sampling frequency of the image acquisition card, ensuring that minute defects in the hole wall can be clearly captured. For sparse sub-regions, the scanning speed is increased to 130%-150% of the baseline speed, while the imaging resolution is reduced to 60%-80% of the baseline resolution, maximizing the detection efficiency while meeting the surface defect detection requirements.

[0043] After path planning is completed, substrate loading and positioning operations are performed. The operator places the TGV glass substrate smoothly onto the support platform surface. Displacement sensors on the edge limiting unit continuously monitor the contact gap between the substrate edge and the limiting block, ensuring the gap does not exceed a set value. Once proper contact is detected, a positioning confirmation signal is sent to the control system. Upon receiving the signal, the control system activates the vacuum source, delivering negative pressure to the vacuum suction cup via the air pump interface to begin substrate adsorption. Pressure sensors collect adsorption pressure data in real time and transmit it to the control system. When the pressure reaches the preset adsorption pressure range, the control system confirms substrate fixation is complete and sends a start signal to the dual-axis synchronous drive mechanism.

[0044] Upon receiving the start signal, the dual-axis synchronous drive mechanism drives the substrate support and positioning mechanism to move smoothly in the XY plane according to the preset variable density scanning path. The X-axis drive component and Y-axis drive component of the dual-axis synchronous drive mechanism adopt high-precision linear guide rails and synchronous belt drives. The servo motor is linked with the drive pulley group through the synchronous belt, driving the slider to slide along the guide rail, thereby realizing the movement of the support platform.

[0045] The synchronization control module uses a PID algorithm to adjust the speed of the X-axis and Y-axis servo motors in real time, ensuring that the synchronization error between the two is controlled within a very small range. This guarantees smooth movement of the platform and avoids image blurring caused by uneven movement. During the movement of the platform, the control system compares the current field of view of the linear scan camera with the preset hole coordinate data in real time, automatically determining the type of target to be detected: when the current field of view of the camera coincides with a certain hole coordinate, it is determined to be a hole wall defect detection; when the current field of view deviates from all hole coordinates, it is determined to be a surface defect detection.

[0046] Based on the determined type of detection target, the multi-degree-of-freedom vision scanning mechanism automatically adjusts the light source and rotation state. When surface defect detection is detected, the control system triggers the ring light source to illuminate, while the bar light source remains off. The rotation adjustment component maintains the current angle, ensuring that the uniform surface light source generated by the ring light source can fully illuminate the substrate surface, enhancing the imaging contrast between surface scratches, microcracks, and other defects and the substrate surface. Simultaneously, the control system acquires the grayscale value of the scanned image in real time through the image acquisition card and dynamically adjusts the brightness of the ring light source based on the grayscale value feedback. The brightness adjustment range is 1000-5000 lux, ensuring that substrates with different surface reflectivities can obtain clear surface images.

[0047] When a hole wall defect is detected, the control system triggers the lighting of the strip light source and turns off the ring light source. The strip light source uses blue LED beads, whose wavelength can effectively penetrate the glass substrate, enhancing the imaging effect of hole wall and interface defects. At the same time, the rotation adjustment component, driven by the rotary motor and reducer, drives the linear scan camera to rotate around the Z-axis, with a rotation angle range of 0°-360°. The angle encoder detects the camera rotation angle in real time and feeds it back to the control system to ensure that the angle adjustment accuracy does not exceed ±0.1°. By adjusting the camera shooting angle through rotation, 360° scanning imaging of the hole wall without blind spots is achieved. Similarly, based on the grayscale feedback of the hole wall imaging, the brightness of the strip light source is dynamically adjusted, with a brightness adjustment range of 500-3000 lux, to ensure clear imaging of hole wall defects.

[0048] While the light source and angle are adjusted, the linear scan camera starts scanning simultaneously, acquiring images according to the scanning speed and resolution set for the current area. The acquired images are transmitted to the industrial controller in real time through a high-speed image acquisition card. The image acquisition card uses a PCIe interface and has a sampling frequency of no less than 1GHz to ensure that the image data transmission is delayed and without loss.

[0049] Throughout the scanning and testing process, the flexible adsorption unit in the substrate support and positioning mechanism continuously collects adsorption pressure data in real time at a frequency of no less than 100Hz to ensure timely capture of pressure changes. The control system presets a safe threshold range for adsorption pressure of 0.02-0.05MPa. This threshold is determined based on experiments on the load-bearing capacity and adsorption stability of substrates of different thicknesses. When the adsorption pressure is detected to be lower than this safe threshold, it indicates a decrease in substrate adsorption stability and a risk of slippage or vibration. At this time, the control system immediately activates a safety interlock mechanism: firstly, the current scanning speed is reduced to a safe speed, set at 30%-50% of the current scanning speed, to reduce slippage or vibration of the substrate caused by inertial forces; if the adsorption pressure continues to decrease or falls below the minimum safe threshold after the speed is reduced, the control system triggers an audible and visual alarm signal, prompting the operator to check the vacuum suction cup sealing status, vacuum source pressure, etc., to ensure substrate safety while maintaining the continuity of testing as much as possible and avoiding loss of testing efficiency due to simple shutdowns.

[0050] Upon receiving the image data, the industrial controller immediately initiates the defect identification process. The first step is to preprocess the acquired image:

[0051] The median filtering algorithm is used to remove random noise from the image, and the adaptive threshold segmentation algorithm is used to enhance the contrast between the defect contour and the background. At the same time, weighted average noise reduction processing is performed on 3-5 consecutively acquired overlapping images to further reduce noise interference and improve image quality.

[0052] After preprocessing, the edge features of the defects are extracted using the Canny edge detection algorithm. Combined with morphological processing methods, the characteristic parameters of the defects are calculated, including the length and width of the crack, the deviation between the actual size and the design size of the hole, the quantitative index of the hole wall roughness, and the area ratio of the metal filling region.

[0053] Subsequently, the Support Vector Machine (SVM) algorithm is used to classify and identify defect feature parameters. This algorithm establishes a classification model through training on a large number of defect samples in the early stage, and can accurately classify defects into types such as cracks, pore diameter deviations, excessive pore wall roughness, and incomplete metal filling. To improve the accuracy and robustness of defect judgment, the control system also integrates the defect recognition results of multiple frames of images, and verifies defects at the same location multiple times to eliminate misjudgments caused by single imaging errors, and finally outputs accurate defect judgment results.

[0054] After scanning and inspection, the dual-axis synchronous drive mechanism, under the command of the control system, drives the substrate-bearing positioning mechanism back to its initial position along the original path. The vacuum source is turned off, the flexible adsorption unit releases the substrate, and the operator can smoothly remove the inspected substrate. The control system clearly displays the inspection results on the human-machine interface, including whether the substrate is qualified or not, the number of defects, the type of each defect, the specific location coordinates and dimensional parameters, and automatically generates a standardized inspection report. The report can be exported via USB flash drive or uploaded to the production management system via network, facilitating subsequent quality traceability and process optimization. If a non-conforming product is detected, the equipment will continuously emit audible and visual alarm signals to prompt the operator to sort it. After the non-conforming substrate is removed, the equipment automatically resets, ready to enter the next round of inspection, achieving continuous batch inspection.

[0055] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method for dual-axis synchronous stable scanning inspection of TGV glass substrates, based on a detection device comprising a dual-axis synchronous drive mechanism, a substrate support and positioning mechanism, a multi-degree-of-freedom vision scanning mechanism, and a control system, characterized in that: Includes the following steps: (1) The control system acquires the hole position coordinate data of the TGV glass substrate, divides the scanning area into dense sub-regions and sparse sub-regions according to the spatial distribution density of the hole positions, generates a variable density scanning path, reduces the scanning speed and increases the imaging resolution in the dense sub-regions, and increases the scanning speed and decreases the imaging resolution in the sparse sub-regions. (2) The dual-axis synchronous drive mechanism drives the substrate bearing positioning mechanism to move according to the variable density scanning path. The multi-degree-of-freedom visual scanning mechanism adjusts the light source and rotation state according to the type of the detection target: when detecting surface defects, the ring light source in the multi-degree-of-freedom visual scanning mechanism is lit, the strip light source is turned off and the rotation adjustment component maintains the current angle; when detecting hole wall defects, the strip light source in the multi-degree-of-freedom visual scanning mechanism is lit, the ring light source is turned off and the rotation adjustment component drives the linear scan camera to rotate around the Z-axis to adjust the shooting angle. The linear scan camera in the multi-degree-of-freedom visual scanning mechanism synchronously acquires images. (3) The flexible adsorption unit in the substrate bearing positioning mechanism collects adsorption pressure data in real time. When the pressure is lower than the preset safety threshold, the control system reduces the current scanning speed to a safe speed or triggers an alarm. (4) The control system performs defect identification on the acquired images and outputs the defect judgment results.

2. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The scanning speed of the dense sub-region is reduced to 50%-70% of the base speed, and the imaging resolution is increased to 120%-150% of the base resolution. The scanning speed of the sparse sub-region is increased to 130%-150% of the base speed, and the imaging resolution is reduced to 60%-80% of the base resolution.

3. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The spatial distribution density of the pores is calculated based on the number of pores per unit area. Dense sub-regions are areas where the number of pores per unit area is greater than a preset threshold, and sparse sub-regions are areas where the number of pores per unit area is less than or equal to the preset threshold.

4. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: When detecting defects in the hole wall, the rotation adjustment component drives the linear scan camera to rotate around the Z-axis in an angle range of 0° to 360°, with an angle adjustment accuracy of no more than ±0.1°.

5. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The brightness of the ring light source and the bar light source is dynamically adjusted according to the surface reflectivity of the TGV glass substrate. The brightness adjustment range of the ring light source is 1000-5000 lux, and the brightness adjustment range of the bar light source is 500-3000 lux.

6. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The preset safety threshold is 0.02-0.05 MPa, and the safety speed is 30%-50% of the current scanning speed.

7. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The defect identification in step (4) includes: denoising and enhancing the acquired image, extracting defect feature parameters, and classifying the defect type using the support vector machine algorithm.

8. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The detection target type is determined based on the hole position coordinate data: when the current field of view of the linear scan camera coincides with the hole position coordinates, it is determined to be hole wall defect detection; when the current field of view deviates from the hole position coordinates, it is determined to be surface defect detection.

9. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The defect identification in step (4) also includes: performing weighted average noise reduction processing on the continuously acquired multi-frame images, and outputting the final defect judgment by fusing the defect identification results of the multi-frame images.

10. The method for biaxial synchronous stable scanning detection of TGV glass substrates according to claim 1, characterized in that: The defect determination includes the identification and classification of cracks, pore diameter deviations, excessive pore wall roughness, and incomplete metal filling.

Images