A glass defect layering inspection apparatus based on multiple imaging planes
By designing a multi-image-plane combined acquisition and fusion analysis module, the shortcomings of existing glass defect detection equipment in multi-angle image acquisition and result fusion are solved, achieving efficient and accurate glass defect detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HEFEI SHIZHAN TECHNOLOGY CO LTD
- Filing Date
- 2026-03-12
- Publication Date
- 2026-07-21
AI Technical Summary
Existing glass defect detection equipment has shortcomings in multi-angle image acquisition and result fusion, resulting in weak anti-interference ability, high false alarm rate, low detection efficiency and insufficient result accuracy.
A multi-image-plane glass defect layering detection device is adopted, which acquires images by combining front linear array camera, side area array camera and side edge area array camera, and uses fusion analysis module to perform spatial correlation discrimination and credibility verification of defect detection results, and outputs unified detection results.
It improves the accuracy and comprehensiveness of glass defect detection, reduces false alarm rate and duplicate statistics, and ensures the reliability and integrity of test results.
Smart Images
Figure CN121830503B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of glass defect detection technology, and in particular to a glass defect layer detection device based on multiple imaging planes. Background Technology
[0002] During glass production and processing, defects such as scratches, bubbles, and chips are easily generated on the glass surface and sides. Due to the high transparency, complex reflection and refraction, and low defect contrast of glass materials, defect detection has always been a technical challenge in the glass manufacturing industry. Therefore, some glass defect detection equipment has appeared on the market, such as the glass defect detection equipment with announcement numbers CN114034720B and CN110694922A. However, most of these glass defect detection devices are based on the analysis of front images, with little design for the hardware components of side images. Furthermore, the defect detection methods attempt to use a unified algorithm to process different imaging surfaces, resulting in the following problems when using such glass defect detection equipment: lack of multi-angle glass image acquisition, weak anti-interference ability, and high false alarm rate; lack of effective fusion mechanism between results from multiple detection surfaces, low detection efficiency, easy duplication of defect statistics, and insufficient accuracy of detection results. Summary of the Invention
[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a glass defect layering detection device based on multiple imaging planes that can overcome or at least partially solve the above problems.
[0004] To achieve the above objectives, the present invention adopts the following technical solution: A glass defect layering detection device based on multiple imaging planes includes: a conveying module, a scanning module, and a fusion analysis module. The scanning module is mounted on the conveying module, and the fusion analysis module is connected to the scanning module via a wiring harness. The scanning module includes a front linear array camera, a side area array camera, and a side edge area array camera. The front linear array camera and the side area array camera are mounted on the conveying module and are used to acquire front and side images of the glass. The side area array camera is mounted on both sides of the conveying module and is used to acquire side images of the glass. The fusion analysis module can perform defect detection on images from different imaging surfaces, spatially correlate defects detected on each imaging surface, and combine spatially correlated defects with imaging surface priority and credibility verification mechanisms for fusion judgment, outputting a unified glass defect detection result. This eliminates duplicate defect identification and misjudgment caused by multi-imaging-surface detection, and improves the accuracy and comprehensiveness of glass defect detection.
[0005] Furthermore, the method for spatial correlation discrimination of defects detected on each imaging surface is as follows: Step 1: Load the reference data of the glass being tested and identify the actual dimensions of the glass: length l0, width b0, height h0; Step 2: On the images of different imaging surfaces, establish a three-dimensional rectangular coordinate system with the same corner of the glass as the origin, the side containing the length as the x-axis, the side containing the width as the y-axis, and the side containing the height as the z-axis. Step 3: Calculate the coordinates of the defect center point based on its position in the image, and then register the defect center point coordinates. Step 4: Calculate the distance between the center points of defects on two different imaging surfaces. If the distance between the center points of defects on two different imaging surfaces is less than the preset distance threshold δ, it indicates that the two defects are spatially related; otherwise, the two defects are not spatially related.
[0006] Preferably, the method for registering the coordinates of the defect center point in step three is as follows: identify the length l1, width b1, and height h1 of the glass in the image, and calculate the registration coefficients γ for the length, width, and height respectively. l γ b γ h : γ l =l0 / l1;γ b =b0 / b1; γ h =h0 / h1; For the coordinates (x, y) of the i-th defect in the image i ,y i ,z i The registered coordinates are (γ) l x i ,γ b y i ,γ h z i ).
[0007] Preferably, the method for calculating the distance between the center points of defects on two different imaging planes in step four is as follows: For the spatial location of the defect on two different imaging planes (γ) l1 x1,γ b1 y1,γ h1 z1) and (γ) l2 x2,γ b2 y2,γ h2 z2), where γ l1 and γ l2 γ is the registration coefficient for the lengths of the two different imaging planes of the glass being tested. b1 and γ b2 γ is the registration factor for the width of the two different imaging planes of the tested glass. h1 and γ h2The registration coefficient is high on two different imaging planes of the tested glass, and the distance between the center points of these two defects is: .
[0008] Furthermore, the fusion analysis method of the fusion analysis module includes: Step 1: Label the detection results from different imaging surfaces; Step II: Determine whether the defects detected by different imaging surfaces are spatially correlated; Step 3: For defects with spatial correlation, perform priority determination based on the imaging surface type; Step IV: Introduce a credibility verification mechanism during the priority determination process; Step V: Finally, output the fused defect results.
[0009] Preferably, in step I, the labeling of the detection results from different imaging surfaces includes the type of imaging surface, defect category, severity level, score, spatial location of the defect, and area range.
[0010] Preferably, the credibility verification mechanism in step IV includes: whether the defect categories are consistent or have a reasonable mapping relationship; whether there is a significant conflict in the defect severity level; and whether the defect area range is consistent.
[0011] Preferably, the output of the fused defect result in step V includes the fused defect category and severity level, the source identifier of the imaging plane to which the defect belongs, and the confidence mark of the fusion judgment.
[0012] Furthermore, the conveying module is equipped with a rotating wheel and a first sensor group and a second sensor group mounted on the rotating wheel, wherein, The first sensor group is located in the left half of the rotating wheel, including sensor one and sensor two. Sensor one is located at the front end of the rotating wheel, and sensor two is located at the rear end of the rotating wheel. The second sensor group is symmetrically arranged in the right half of the rotating wheel, including sensor three and sensor four. Sensor three is located at the front end of the rotating wheel, and sensor four is located at the rear end of the rotating wheel. By setting up the first and second sensor groups, the horizontal state of the glass on the conveying module can be detected.
[0013] Preferably, the conveying module is further provided with a motor, which is connected to the rotating shaft through a reducer. The rotating shaft is connected to the rotating wheel through a coupling assembly, which includes a coupling and a limiting retaining ring. The coupling assembly includes a coupling and a retaining ring. The retaining ring is a grooved ring located at the head of the coupling. The coupling has a corresponding retaining block, which is fitted into the groove of the retaining ring.
[0014] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: 1. This invention, through the combination of a front linear array camera, a side array camera, and a side edge array camera, can acquire images of the detected glass from the front, side, and side layers, providing multi-angle support for the judgment of glass defects, avoiding missed acquisition blind spots, and preventing incomplete defect statistics.
[0015] 2. This invention, through the design of a fusion analysis method, integrates and analyzes the detection results from different imaging surfaces, enabling spatial correlation determination of the detection results from different imaging surfaces, comparative analysis to obtain detection results with higher reliability, reducing false alarm rate and avoiding duplicate statistics.
[0016] 3. By utilizing the design of the first and second sensor groups on the rotating wheel, the time difference between the different sensors sensing the glass is used to determine whether the glass being tested is in a horizontal state. This avoids the glass tilting, which would cause deviations in the detection results, making the glass defect detection inaccurate and affecting the detection results.
[0017] In summary, this invention provides comprehensive image support for glass defect detection through the design of the scanning module, making the detection results more accurate; the design of the fusion analysis method can reduce the false alarm rate and avoid duplicate statistics; and the design of the first and second sensor groups on the rotating wheel can detect whether the glass being detected is in a horizontal state, further ensuring the accuracy of the detection results. Attached Figure Description
[0018] Figure 1 This is a three-dimensional structural diagram of a glass defect layer detection device based on multiple imaging surfaces proposed in this invention. Figure 2 This is a schematic diagram of the internal structure of the scanning module in a glass defect layer detection device based on multiple imaging surfaces proposed in this invention; Figure 3 This is a logical schematic diagram of the fusion analysis method in a glass defect layer detection device based on multiple imaging surfaces proposed in this invention; Figure 4 This is a partial structural diagram of the rotating wheel with an added first and second sensor group in a glass defect layer detection device based on multiple imaging surfaces proposed in this invention. Figure 5 This is a partial structural diagram of a glass defect layering detection device based on multiple imaging planes proposed in this invention, in which a motor is added to the rotating wheel; Figure 6 This is a schematic diagram of the coupling assembly in a glass defect layering detection device based on multiple imaging planes proposed in this invention.
[0019] In the diagram: 1. Conveying module; 2. Scanning module; 3. Fusion analysis module; 4. Front linear scan camera; 5. Side area scan camera; 6. Side edge area scan camera; 7. Rotary wheel; 8. Sensor 1; 9. Sensor 2; 10. Sensor 3; 11. Sensor 4; 12. Motor; 13. Reducer; 14. Rotating shaft; 15. Coupling; 16. Limiting ring. Detailed Implementation
[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0021] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0022] Example 1: Reference Figure 1 and Figure 2 A glass defect layering detection device based on multiple imaging planes includes: a conveying module 1, a scanning module 2, and a fusion analysis module 3. The scanning module 2 is mounted on the conveying module 1, and the fusion analysis module 3 is connected to the scanning module 2 via a wiring harness. The scanning module 2 has a front linear scan camera 4, a side area scan camera 5, and a side area scan camera 6. The front linear scan camera 4 and the side area scan camera 5 are mounted on the conveying module 1 and are used to acquire front and side images of the glass. The side area scan camera 6 is mounted on both sides of the conveying module 1 and is used to acquire side images of the glass.
[0023] This invention, through the combination of a front linear array camera 4, a side array camera 5, and a side-mounted array camera 6, can acquire images of the detected glass from the front, side, and side layers, providing multi-angle support for the judgment of glass defects, avoiding missed acquisition blind spots, and preventing incomplete defect statistics; and can also provide an image basis for defect fusion analysis.
[0024] The fusion analysis module 3 performs defect detection on images from different imaging surfaces, spatially correlates the defects detected on each imaging surface, and combines spatially correlated defects with the imaging surface priority and credibility verification mechanism for fusion judgment, outputting a unified glass defect detection result. This eliminates the duplicate identification and misjudgment of defects caused by multi-imaging surface detection, and improves the accuracy and comprehensiveness of glass defect detection.
[0025] The conveying module 1 includes an infeed conveying section, a scanning section, and an outfeed conveying section. The scanning module 2 is located in the scanning section of the conveying module 1. In use, the glass to be inspected is placed in the infeed conveying section. Each imaging component on the conveying module 1 is equipped with a corresponding sensing device in its vicinity. The sensing device is connected to the adjacent imaging component via a wiring harness. When the sensing device senses that the glass is approaching, it triggers the corresponding imaging component to continuously acquire images. When the sensing device senses that the glass is leaving, the corresponding imaging component stops acquiring images. When the glass has completely left the scanning section, the scanning module 2 transmits the acquired images to the fusion analysis module 3. The fusion analysis module 3 performs fusion analysis on the detection results of different imaging surfaces.
[0026] The method for spatial correlation discrimination of defects detected on each imaging plane is as follows: Step 1: Load the reference data of the glass being tested and identify the actual dimensions of the glass: length l0, width b0, height h0; Reference materials can be derived from standard design data or constructed through reverse engineering via image acquisition.
[0027] Step 2: On the images of different imaging surfaces, establish a three-dimensional rectangular coordinate system with the same corner of the glass as the origin, the side containing the length as the x-axis, the side containing the width as the y-axis, and the side containing the height as the z-axis. Step 3: Calculate the coordinates of the defect center point based on its position in the image, and then register the defect center point coordinates. The method for registering the coordinates of the defect center point in step three is as follows: identify the length l1, width b1, and height h1 of the glass in the image, and calculate the registration coefficients γ for the length, width, and height respectively. l γ b γ h : γ l =l0 / l1;γ b =b0 / b1; γ h =h0 / h1; For the coordinates (x, y) of the i-th defect in the image i ,y i ,z i The registered coordinates are (γ) l x i ,γ b y i ,γ h z i ).
[0028] Step 4: Calculate the distance between the center points of defects on two different imaging surfaces. If the distance between the center points of defects on two different imaging surfaces is less than the preset distance threshold δ (the preset distance threshold δ can be preset according to the size of the glass and the detection requirements, and there is no fixed value), it means that the two defects are spatially related; otherwise, the two defects are not spatially related.
[0029] When determining whether there is a spatial correlation between defects on different imaging surfaces, the spatial location (γ) of the defects on two different imaging surfaces is considered. l1 x1,γ b1 y1,γ h1 z1) and (γ) l2 x2,γ b2 y2,γ h2 z2), where γ l1 and γ l2 γ is the registration coefficient for the lengths of the two different imaging planes of the glass being tested. b1 and γ b2 γ is the registration factor for the width of the two different imaging planes of the tested glass. h1 and γ h2 The registration coefficient is high on two different imaging planes of the tested glass, and the distance between the center points of these two defects is:
[0030] Compare d with the pre-input preset distance threshold δ. If d is less than δ, it means that the two defects from different imaging surfaces are spatially related and are the same actual defect; otherwise, the two defects are not spatially related and are two independent defects.
[0031] By identifying spatial correlations of defects, the same actual defect can be distinguished from independent defects. The multi-faceted detection results of the same actual defect can be grouped into a fusion unit, while independent defects without spatial correlation can be retained as independent units. This avoids duplicate statistics and misjudgments of the same actual defect, reduces the burden of defect detection and the duplication of detection results, and improves the accuracy and comprehensiveness of glass defect layer detection.
[0032] Example 2: refer to Figure 3 Based on Example 1, the fusion analysis method of the fusion analysis module further includes: Step 1: Label the detection results from different imaging surfaces; Step II: Determine whether the defects detected by different imaging surfaces are spatially correlated; Step 3: For defects with spatial correlation, perform priority determination based on the imaging surface type; Step IV: Introduce a credibility verification mechanism during the priority determination process; Step V: Finally, output the fused defect results.
[0033] By utilizing the design of the fusion analysis method, spatial correlation determination can be made for the detection results of different imaging surfaces, and the more reliable detection results can be obtained through comparative analysis, which can reduce the false alarm rate and avoid duplicate statistics.
[0034] In step I, the detection results from different imaging surfaces are labeled with the type of imaging surface, defect category, severity level, score, spatial location of the defect, and area range.
[0035] In glass images, there is a significant change in brightness at the defective area compared to the non-defective area. The fusion analysis module 3 can extract the contour of the defect based on the brightness change in the image, and then compare the contour with the defect data in the network or the library for detection, and label the detection results from different imaging surfaces.
[0036] The types of imaging surfaces include front, side, and edge; Defect categories include scratches, pitting, bubbles, gaps, chipped edges, internal streaks, etc. The severity level is divided into at least three levels, and the classification is based on the number of pre-entered defect thresholds. The number of defect thresholds is set according to the type and size of the glass and the inspection requirements.
[0037] In this embodiment, a first defect threshold and a second defect threshold (the first and second defect thresholds can be preset according to the size of the glass and the detection requirements, and there are no fixed values) are pre-inputted and divided as follows: Defects smaller than the first defect threshold are considered minor or severe. Defects with sizes between the first and second defect thresholds are considered generally severe. A defect size larger than the second defect threshold is considered extremely serious. The scoring is determined based on the defect type, severity level, and size. First, the defect type, severity level, and size are pre-inputted and corresponding scoring criteria are established, with a scoring range of 0-10 points. Then, the weights for defect type, severity level, and size are set, with defect type having a weight of 0.4, severity level having a weight of 0.4, and size having a weight of 0.2. The scoring criteria and weights can be adjusted according to the type and size of the glass and the testing requirements. Finally, each individual score is multiplied by its corresponding weight and summed to obtain the comprehensive score for the defect. The comprehensive score is rounded to one decimal place for subsequent severity level classification.
[0038] The registered defect coordinates are (γ) l x i ,γ b y i ,γ hz i ), which is the spatial location of the defect.
[0039] For defect types with vector characteristics, such as scratches and internal striations, identify the spatial location of the start and end points (γ). l x ia ,γ b y ia ,γ h z ia ) and (γ l x ib ,γ b y ib ,γ h z ib (γ) l x ia ,γ b y ia ,γ h z ia ) and (γ l x ib ,γ b y ib ,γ h z ib This can reflect the area where the defect is located.
[0040] By designing step I, the detected defects are sorted out, and each defect can be assigned feature labels such as imaging surface affiliation, defect type, location, and parameters, avoiding defect data chaos and inability to trace the source; at the same time, step I can also provide data support for subsequent judgments on whether defects have spatial correlation and credibility.
[0041] Example 3: Based on Example 2, the priority determination method in step III is further as follows: When the defect is a surface defect, the front imaging detection result is used first; when the defect is an edge defect or close to the glass boundary, the side imaging or side edge imaging detection result is used first; when both imaging surfaces give a valid judgment, the defect category and score are combined for a comprehensive judgment.
[0042] Surface defects in glass include scratches, bubbles, and pitting, while edge defects include chipping and gaps. Through the design of step III, the priority of different types of defects is determined based on their degree of visibility on different imaging surfaces, making the defect detection results more accurate.
[0043] The credibility verification mechanism in step IV includes: Are the defect categories consistent or have a reasonable mapping relationship? Are there any significant conflicts in the defect severity levels? Is the defect area range consistent?
[0044] Among the three conditions above, the more spatially related the two defects are, the more accurate and reliable the fusion analysis results of these two defects will be. If there is a significant conflict, any of the following detection strategies can be used: The results of high-priority imaging surface detection shall prevail; Reduce the severity level of the defect and mark it as a candidate for re-inspection; Output multiple test results for manual review.
[0045] By designing step IV, a credibility verification mechanism is used to fuse and determine defects with spatial correlation. By comparing the detection results of the same actual defect from different imaging surfaces, erroneous judgments caused by single detection anomalies can be avoided, the rationality and reliability of defect priority determination can be improved, and the accuracy of detection results can be further increased.
[0046] Step V outputs the fused defect results, including the fused defect category and severity level, defect imaging surface source identifier, and fusion judgment confidence mark.
[0047] Through the design of step V, the entire fusion analysis is output. For the same actual defect, high-priority, high-reliability imaging surface data is used as the core, and effective supplementary features from other imaging surfaces are integrated to eliminate duplicate detection results and form a unified defect feature description. Moreover, the fused results can retain core information such as the spatial location, type, and parameter characteristics of the defect, providing a basis for subsequent glass screening and defect tracing, and improving the overall accuracy and efficiency of glass defect detection.
[0048] Example 4: Reference Figure 4 The conveying module 1 is equipped with a rotating wheel 7 and a first sensor group and a second sensor group mounted on the rotating wheel 7, wherein... The first sensor group is located in the left half of the rotating wheel 7, including sensor 8 and sensor 9. Sensor 8 is located at the front end of the rotating wheel 7, and sensor 9 is located at the rear end of the rotating wheel 7. The second sensor group is symmetrically arranged in the right half of the rotating wheel 7, including sensor 3 10 and sensor 4 11. Sensor 3 10 is located at the front end of the rotating wheel 7, and sensor 4 11 is located at the rear end of the rotating wheel 7. By setting up the first and second sensor groups, the horizontal status of the glass on the conveying module 1 can be detected. Whether the glass is horizontal affects the judgment of the original size of the glass, the type of defect, and the range of the defect area. Therefore, the glass must be in a horizontal state before entering the scanning module 2.
[0049] Sensors 8 and 9 are set in parallel. If the time difference between the glass detected by sensors 8 and 9 is less than the preset maximum time difference value, it means that the left edge of the glass is parallel to the rotating wheel 7 and is in a horizontal state. If the time difference detected by sensors 8 and 9 is not less than the preset maximum time difference value, it means that the front edge of the glass is not parallel to the rotating wheel 7 and is in a skewed state. At this time, the state of the glass needs to be adjusted. Sensors 10 and 11 are also set in parallel and can detect the horizontal state of the right edge of the glass. Sensor 1 (8) and Sensor 3 (10) are set in parallel. If Sensor 1 (8) and Sensor 3 (10) detect a time difference between the glass and the ...
[0050] Sensors 8, 9, and 10 are infrared ranging sensors. Since the lines containing sensors 8 and 9 intersect at a single point (sensor 8), a plane is defined by these two intersecting lines. Sensors 8, 9, and 10 form this plane, which is parallel to the surface of the conveying module 1. If the distance difference between sensors 8, 9, and 10 and the glass is less than a preset distance threshold, it can be determined whether the glass is skewed in the vertical direction. If the distance difference is not less than the preset distance threshold, it indicates that the detected glass is not parallel to the surface of the conveying module 1. This will affect the detection results of glass defects and may even cause the glass to fall and break. Simultaneously, sensors 8, 9, and 10 form a triangle. The stability of the triangle ensures the stability of the plane containing sensors 8, 9, and 10, allowing for real-time detection of the glass's state. When the detected glass is not parallel to the surface of the conveying module 1, an alarm is issued promptly, reminding the operator to adjust the glass's condition. When one of the sensors 8, 9, and 10 is damaged, sensor 11 can participate in the detection of the glass condition.
[0051] It should be noted that the "preset maximum time difference" and "preset spacing threshold" can be preset according to the glass conveying speed, size and detection requirements, and there are no fixed values.
[0052] By setting up the first sensor group and the second sensor group, the horizontal state of the glass can be detected simultaneously in the conveying direction and in the direction perpendicular to the conveying direction, thus avoiding the influence of glass tilt on the defect detection results.
[0053] Example 5: Reference Figure 5 and Figure 6 Based on embodiment 4, the difference is that the conveying module 1 is also equipped with a motor 12, the motor 12 is connected to the rotating shaft 14 through the reducer 13, and the rotating shaft 14 is connected to the rotating wheel 7 through the coupling assembly.
[0054] By connecting the motor 12 to the rotating wheel 7, the sensing directions of the first sensor group and the second sensor group can be adjusted according to the shape and size of the glass, making the horizontal state detection applicable to glass of various shapes and sizes, thus increasing the universality of the invention; and by achieving dynamic detection through the back-and-forth rolling of the rotating wheel 7, the horizontal state detection can be verified by sensing different positions of the glass, further confirming whether the glass is horizontal.
[0055] The coupling assembly includes a coupling (15) and a retaining ring (16). The retaining ring (16) is a ring with a slot and is located at the head of the coupling (15). The coupling (15) has a corresponding retaining block, which is fitted into the slot of the retaining ring (16).
[0056] The limiting ring 16 can control the rotation of the rotating wheel 7 within a certain angle, preventing the first and second sensor groups from contacting the glass surface during rotation, thus avoiding scratches or even damage to the glass. Furthermore, the limiting ring 16 prevents the first and second sensor groups from being subjected to rigid impacts due to the instantaneous change in rotation direction when the rotating wheel 7 changes direction, which could cause damage to the first and second sensor groups over time. Simultaneously, it also prevents the rigid impacts from altering the orientation of the first and second sensor groups, thus avoiding errors in detecting the horizontal state of the glass.
[0057] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A glass defect delamination detection device based on multiple imaging planes, comprising: The system comprises a transport module (1), a scanning module (2), and a fusion analysis module (3). The scanning module (2) is mounted on the transport module (1). The fusion analysis module (3) is connected to the scanning module (2) via a wiring harness. The scanning module (2) has a front linear array camera (4), a side array camera (5) and a side array camera (6). The front linear array camera (4) and the side array camera (5) are mounted on the conveying module (1) and are used to collect front and side images of the glass. The side array camera (6) is mounted on both sides of the conveying module (1) and is used to collect side images of the glass. The fusion analysis module (3) can perform defect detection on images of different imaging surfaces, perform spatial correlation discrimination on defects detected on each imaging surface, and combine spatially correlated defects with imaging surface priority and credibility verification mechanism for fusion judgment, outputting unified glass defect detection results, thereby eliminating repeated identification and misjudgment of defects caused by multi-imaging surface detection, and improving the accuracy and comprehensiveness of glass defect detection. The method for spatial correlation discrimination of defects detected on each imaging plane is as follows: Step 1: Load the reference data of the glass to be tested and identify the actual dimensions of the glass, namely length l0, width b0, and height h0; Step 2: On the images of different imaging surfaces, establish a three-dimensional rectangular coordinate system with the same corner of the glass as the origin, the side containing the length as the x-axis, the side containing the width as the y-axis, and the side containing the height as the z-axis. Step 3: Calculate the coordinates of the defect center point based on its position in the image, and then register the defect center point coordinates. Step 4: Calculate the distance between the center points of defects on two different imaging surfaces. If the distance between the center points of defects on two different imaging surfaces is less than the preset distance threshold δ, it indicates that the two defects are spatially related; otherwise, the two defects are not spatially related.
2. The glass defect layering detection device based on multiple imaging planes according to claim 1, characterized in that, The method for registering the coordinates of the defect center point in step three is as follows: identify the length l1, width b1, and height h1 of the glass in the image, and calculate the registration coefficients γ for the length, width, and height respectively. l γ b γ h : c l =l0 / l1;γ b =b0 / b1;γ h =h0 / h1; For the coordinates (x, y) of the i-th defect in the image i ,y i ,z i The registered coordinates are (γ) l x i ,γ b y i ,γ h z i ).
3. The glass defect layering detection device based on multiple imaging planes according to claim 1, characterized in that, The method for calculating the distance between the center points of defects on two different imaging planes in step four is as follows: For the spatial location of the defect on two different imaging planes (γ) l1 x1,γ b1 y1,γ h1 z1) and (γ) l2 x2,γ b2 y2,γ h2 z2), where γ l1 and γ l2 γ is the registration coefficient for the lengths of the two different imaging planes of the glass being tested. b1 and γ b2 γ is the registration factor for the width of the two different imaging planes of the tested glass. h1 and γ h2 The registration coefficient is high on two different imaging planes of the tested glass, and the distance between the center points of these two defects is: 。 4. The glass defect layering detection device based on multiple imaging planes according to claim 1, characterized in that, The fusion analysis method includes: Step 1: Label the detection results from different imaging surfaces; Step II: Determine whether the defects detected by different imaging surfaces are spatially correlated; Step 3: For defects with spatial correlation, perform priority determination based on the imaging surface type; Step IV: Introduce a credibility verification mechanism during the priority determination process; Step V: Finally, output the fused defect results.
5. The glass defect layering detection device based on multiple imaging planes according to claim 4, characterized in that, In step I, the detection results from different imaging surfaces are labeled with the type of imaging surface, defect category, severity level, score, spatial location of the defect, and area range.
6. The glass defect layering detection device based on multiple imaging planes according to claim 4, characterized in that, The credibility verification mechanism in step IV includes: whether the defect categories are consistent or have a reasonable mapping relationship; whether there is a significant conflict in the defect severity level; and whether the defect area range is consistent.
7. The glass defect layering detection device based on multiple imaging planes according to claim 4, characterized in that, The output of the fused defect results in step V includes the fused defect category and severity level, the source identifier of the imaging plane to which the defect belongs, and the confidence mark of the fusion judgment.
8. The glass defect layering detection device based on multiple imaging planes according to claim 1, characterized in that, The conveying module (1) is equipped with a rotating wheel (7) and a first sensor group and a second sensor group disposed on the rotating wheel (7), wherein, The first sensor group is located in the left half of the rotating wheel (7), including sensor one (8) and sensor two (9). Sensor one (8) is located at the front end of the rotating wheel (7), and sensor two (9) is located at the rear end of the rotating wheel (7). The second sensor group is symmetrically arranged in the right half of the rotating wheel (7), including sensor three (10) and sensor four (11). Sensor three (10) is located at the front end of the rotating wheel (7), and sensor four (11) is located at the rear end of the rotating wheel (7). By setting up the first sensor group and the second sensor group, the horizontal state of the glass on the conveying module (1) can be detected.
9. A glass defect layering detection device based on multiple imaging planes according to claim 8, characterized in that, The conveying module (1) is also equipped with a motor (12), which is connected to the rotating shaft (14) through a reducer (13), and the rotating shaft (14) is connected to the rotating wheel (7) through a coupling assembly; The coupling assembly includes a coupling (15) and a retaining ring (16). The retaining ring (16) is a ring with a slot and is located at the head of the coupling (15). The coupling (15) has a corresponding retaining block, which is fitted into the slot of the retaining ring (16).