Online defect detection system for float glass production

Abstract

The invention discloses an on-line defect detection system for float glass production, relates to the technical field of glass defect detection, and solves the technical problem that the quality of a finished product does not reach the standard due to the fact that tiny defects or inclusions in a glass plate are not considered in the prior art. The method comprises the following steps: irradiating a glass plate by laser and collecting to obtain a light propagation image, preprocessing the light propagation image to obtain a gray-scale light image, taking a mode of the light intensity of pixel points in the gray-scale light image as standard light intensity, dividing abnormal pixel points based on the standard light intensity, and obtaining an abnormal feature map in the gray-scale light image based on the abnormal pixel points. Extracting light reaction features in the abnormal feature map; and matching the light reaction characteristics with standard geometric characteristics to obtain a defect shape, obtaining the pixel type and the pixel defect intensity of the defect shape based on the number of abnormal pixel points in the defect shape and the light intensity, and matching the defect shape, the pixel type and the defect pixel intensity with standard defect data to obtain corresponding defect data.

Description

Technical Field

[0001] The invention belongs to the field of glass defect detection, in particular to an online defect detection system for float glass production. Background Art

[0002] Float glass production is a manufacturing process widely used in the fields of construction, automobiles, electronics, etc. In the float glass production process, the glass liquid undergoes processes such as melting, forming, cooling, etc., and finally forms a flat glass plate. However, during the production process, due to various reasons (such as impure raw materials, equipment failure, improper process control, etc.), various defects such as scratches, bubbles, stains, cracks, etc. may appear on the glass plate. These defects not only affect the appearance quality of the product, but may also affect its performance, so strict detection and control are required.

[0003] The invention patent with application number CN2018104639436 discloses a method for detecting surface flatness defects of 3D curved glass based on line scanning laser. The method collects three-dimensional point cloud data through a visual imaging system, and constructs a three-dimensional model of the glass plate through three-dimensional reconstruction. The glass image is collected and grayscaled to obtain a first grayscale image and a second grayscale image, and matching evaluation is performed based on the first grayscale image and the second grayscale image. The method collects three-dimensional data of the glass plate through a visual presentation system. There may be tiny defects or internal inclusions on the glass plate that cannot be collected by the visual presentation system. If the tiny defects and internal inclusions are ignored, it will to a certain extent lead to the final product quality not meeting the standards, affecting the user experience.

[0004] The present invention provides an online defect detection system for float glass production to solve the above technical problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an online defect detection system for float glass production, which is used to solve the technical problem in the prior art that the quality of the finished product does not meet the standard due to failure to consider the presence of tiny defects or inclusions in the glass plate.

[0006] To achieve the above-mentioned object, the first aspect of the present invention provides an online defect detection system for float glass production, comprising: a laser scanning module, an image analysis module, a defect matching module and a database;

[0007] Laser scanning module: used to transport the glass plate and scan the entire glass plate through the laser scanning system;

[0008] Image analysis module: used to capture and generate a light propagation image of the glass plate; extract light reaction features in the image based on the light propagation image;

[0009] Defect matching module: used to obtain corresponding defect data based on light reaction feature matching.

[0010] Preferably, the extracting of light reaction features in an image based on a light propagation image comprises:

[0011] Extract light propagation images;

[0012] The light propagation image is denoised and converted into a grayscale image to obtain a grayscale light image, the grayscale value of each pixel in the grayscale light image is counted, and the grayscale value of the pixel is marked as light intensity;

[0013] The mode of the pixel light intensity is used as the standard light intensity; the abnormal pixel is obtained based on the standard light intensity and the pixel light intensity of each pixel;

[0014] Acquire an abnormal feature map in the grayscale light image based on abnormal pixels; extract light reaction features in the abnormal feature map.

[0015] Preferably, the step of obtaining abnormal pixels based on the standard light intensity and the pixel light intensity of each pixel includes:

[0016] A1: Extract standard light intensity and pixel light intensity of each pixel;

[0017] A2: Determine whether the light intensity of the pixel is equal to the standard light intensity; if yes, mark the corresponding pixel as a normal pixel; if no, jump to A3;

[0018] A3: Determine whether the light intensity of the pixel is greater than the standard light intensity; if yes, mark the corresponding pixel as a strong light pixel; if no, mark the corresponding pixel as a weak light pixel;

[0019] A4: Combine strong light pixels with weak light pixels to obtain abnormal pixels.

[0020] Preferably, the step of acquiring an abnormal feature map in a grayscale image based on abnormal pixels includes:

[0021] Extract abnormal pixel points in the grayscale light image; mark the light intensity of weak light pixel points in the abnormal pixel points as weak light intensity, and mark the light intensity of strong light pixel points in the abnormal pixel points as strong light intensity;

[0022] Set a geometric distance; combine other weak-light pixel points within the geometric distance around the weak-light pixel point to obtain a weak-light feature; combine all weak-light features to obtain a weak-light feature map;

[0023] Combining other strong light pixel points within the geometric distance around the strong light pixel point to obtain a strong light feature; combining all the strong light features to obtain a strong light feature map;

[0024] The weak light feature map is combined with the strong light feature map to obtain the abnormal feature map.

[0025] Preferably, obtaining corresponding defect data according to light reaction feature matching includes:

[0026] Extracting light reaction features;

[0027] Match the light reaction feature with the standard geometric feature in the database to obtain the defect shape of the light reaction feature; count the light intensity of the pixel points in the defect shape to obtain the defect pixel intensity and the pixel type of the corresponding defect shape;

[0028] The defect shape, pixel type and defect pixel intensity are matched with the standard defect data in the database to obtain the corresponding defect data.

[0029] Preferably, the statistically analyzing the light intensity of pixels within the defect shape to obtain the defect pixel intensity and the pixel type corresponding to the defect shape includes:

[0030] B1: Set the defective pixel ratio; determine whether the proportion of weak-light pixels in the defect shape is greater than or equal to the defective pixel ratio; if yes, mark the pixel type of the corresponding defect shape as a weak-light defect; if no, jump to B2;

[0031] B2: Determine whether the proportion of strong light pixels in the defect shape is greater than or equal to the defect pixel ratio; if yes, mark the pixel type of the corresponding defect shape as strong light defect; if no, mark the pixel type of the corresponding defect shape as mixed defect;

[0032] B3: Calculate the average value of the weak light intensities corresponding to all weak light pixels in the defect shape where the pixel type is a weak light defect to obtain the weak light pixel intensity; calculate the average value of the strong light intensities corresponding to all strong light pixels in the defect shape where the pixel type is a strong light defect to obtain the strong light pixel intensity;

[0033] B4: Count the number of strong light pixels and weak light pixels in the defect shape whose pixel type is mixed defect; mark the average value of the strong light intensity of the strong light pixels in the mixed defect as the mixed strong light intensity, and mark the average value of the weak light intensity of the weak light pixels in the mixed defect as the mixed weak light intensity; divide the sum of the product of the mixed strong light intensity and the number of strong light pixels and the product of the mixed weak light intensity and the number of weak light pixels by the sum of the number of strong light pixels and the number of weak light pixels to obtain the defect pixel intensity;

[0034] B5: Combine the weak light pixel intensity, strong light pixel intensity and mixed pixel intensity to obtain the defective pixel intensity.

[0035] Preferably, the step of matching the defect shape, pixel type and defect pixel intensity with standard defect data in a database to obtain corresponding defect data includes:

[0036] S1: extracting pixel types and defect pixel intensities of all defect shapes in the grayscale light image; extracting standard defect data from a database; the standard defect data includes standard pixel types and defect pixel intensity ranges corresponding to standard defect shapes caused by various defects on various glass plates that cause light propagation phenomena;

[0037] S2: Match the defect shape with the standard defect shape to obtain the defect light propagation phenomenon corresponding to the defect shape;

[0038] S3: Determine whether the pixel type of the defect shape meets the standard pixel type corresponding to the defect light propagation phenomenon; if yes, jump to S4; if no, mark the corresponding defect shape as an unrecorded defect;

[0039] S4: matching the defect pixel intensity of the defect shape with the defect pixel intensity range, marking the defect position of the defect shape matching the weak light pixel intensity as the back side of the glass plate, and marking the defect position of the defect shape matching the strong light pixel intensity as the front side of the glass plate; marking the back side of the glass plate and the front side of the glass plate as defect positions;

[0040] S5: Combine the defect type corresponding to the defect shape with the defect position to obtain defect data.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention irradiates a glass plate with a laser, and uses a sensor to collect a light propagation image generated by the light propagation phenomenon of the glass plate, removes noise from the light propagation image, and then performs grayscale processing to convert it into a grayscale light image. By using the grayscale value of each pixel in the grayscale light image and marking the grayscale value of the pixel as the light intensity of the pixel, the contrast of the image can be improved, and the defects can be made more obvious, thereby improving the sensitivity of detection; the mode of the light intensity of the pixel in the grayscale light image is used as the standard light intensity, and the pixel points are divided into normal pixel points and abnormal pixel points based on the standard light intensity, and the abnormal feature map in the grayscale light image is obtained based on the abnormal pixel points, and the light reaction features in the abnormal feature map are extracted. , ensuring the consistency of the benchmark and reducing misjudgments and missed detections; matching the light reaction features with the standard geometric features in the database to obtain the defect shape; comparing the proportion of abnormal pixels in the defect shape with the preset defect pixel ratio to obtain the pixel type of the defect shape, and calculating the defect light pixel intensity of the defect shape based on the light intensity of the abnormal pixels, matching the defect shape, pixel type and defect pixel intensity with the standard defect data in the database to obtain the corresponding defect data, which can more accurately identify the shape and type of the defect, conduct a comprehensive analysis of the defects from multiple dimensions, improve the accuracy and reliability of detection, quickly identify and classify defects, and improve the detection speed.

[0043] 2. The present invention is provided with a geometric distance, and the same abnormal pixel points within the geometric distance around the abnormal pixel point are combined to obtain abnormal features, and the abnormal features are combined to obtain an abnormal feature map, which can more finely extract the features around the abnormal pixel point, avoid misjudging isolated noise points as defects, and improve the accuracy of detection; extract the light reaction features in the abnormal feature map, and match them with the standard geometric features to obtain the defect shape, count the number of pixels in the defect shape to obtain the pixel type of the defect shape, and calculate the defect pixel intensity based on the pixel type, which is helpful to compare the defect conditions between different samples, improve the detection accuracy, and can more comprehensively detect the defects on the surface of the material and improve the detection coverage. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 The following is a flowchart of an embodiment of the present invention.

[0046] Figure 2 It is a system composition diagram of the present invention.

[0047] Figure 3 The figure is a complete flow chart of obtaining abnormal pixels in one embodiment of the present invention. DETAILED DESCRIPTION

[0048] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1-Figure 3 , the first aspect of the present invention provides an online defect detection system for float glass production, including: a laser scanning module, an image analysis module, a defect matching module and a database;

[0050] Laser scanning module: used to transport the glass plate and scan the entire glass plate through the laser scanning system.

[0051] Exemplarily, the glass sheet is slowly conveyed through the laser scanning area by a conveyor belt, and the glass sheet is scanned by the laser scanning system.

[0052] Image analysis module: used to capture and generate image data of the glass plate; extract light reaction features in the image based on the image data.

[0053] Exemplarily, a light propagation image generated after laser irradiation of a glass plate is collected by a linear array CCD camera, the light propagation image being a light image of reflected light, refracted light, scattered light and other light beams generated by the light propagation phenomenon occurring after the laser irradiates the glass plate and collected by the linear array CCD camera; noise removal is performed on the light propagation image, and grayscale processing is performed on the light propagation image to convert the light propagation image into a grayscale image, the converted grayscale image is marked as a grayscale light image, the grayscale value of each pixel in the grayscale light image is counted, and the grayscale value of the pixel is marked as the light intensity of the pixel; the mode of the pixel light intensity in the grayscale light image is counted to obtain a mode of the pixel light intensity of 128, and 128 is used as the standard light intensity, so that the pixel light intensity of normal pixels in the light propagation image of the glass plate can be identified as a control, and the pixel position corresponding to the defect on the glass plate in the light propagation image of the glass plate can be accurately identified.

[0054] Determine whether the light intensity of each pixel is equal to the standard light intensity, and mark the pixel with the same light intensity as the standard light intensity as a normal pixel, indicating that the pixel is a pixel obtained by the normal light reflection phenomenon after the laser is irradiated on the glass plate; if the light intensity of each pixel is not equal to the standard light intensity, it means that the pixel is an image generated after the laser is irradiated on the glass plate. Due to defects such as burrs and bubbles on the glass plate, the propagation path or propagation angle of the laser changes, resulting in the pixel points of the final light image being different from the normal pixels; the pixel points whose light intensity is greater than the standard light intensity are marked as strong light pixels, and the pixel points whose light intensity is less than the standard light intensity are marked as weak light pixels. The strong light pixels and the weak light pixels are combined to obtain abnormal pixels.

[0055] In another preferred embodiment, a pixel threshold can also be set. If the difference between the light intensity of the pixel point in the grayscale image and the standard light intensity is smaller than the pixel threshold, it can still be judged as the standard light intensity; or in the grayscale image, the standard light intensity is within a certain standard range rather than a fixed value, and the pixel point whose light intensity is within the standard range is a normal pixel point.

[0056] The light intensity of the weak light pixel is marked as weak light intensity, and the light intensity of the strong light pixel is marked as strong light intensity. In this embodiment, the geometric distance is set to the distance of 5 pixels. In the grayscale image, although two adjacent abnormal pixels may not be closely connected, they can still be classified as the same type of abnormal pixels. The reason is that the defects on the glass plate, such as burrs, are irregular, so the adjacent pixels may not be the same, and if the distance between two batches of more abnormal pixels is too large, they may be two different abnormal pixel sets caused by two defects. All other weak light pixels within the geometric distance around the weak light pixel are combined to obtain the weak light feature. The weak light feature is an abnormal pixel set, which is manifested as different areas. Small circles or other geometric figures, combine all weak-light features to obtain a weak-light feature map. The size of the obtained weak-light feature map is the same as that of the light image, but the figures in the image are all weak-light features; combine all other strong-light pixel points within the geometric distance around the strong-light pixel point to obtain the strong-light feature, and combine all the strong-light features to obtain the strong-light feature map; combine the weak-light feature map with the strong-light feature map to obtain the abnormal feature map; combine the weak-light features in the weak-light feature map with the strong-light features in the strong-light feature map to obtain the light reaction features. The shape corresponding to the defect on the glass plate can be accurately obtained through the weak-light features. Different defects have different shapes, so the specific defect type can be identified through the weak-light features to provide guidance for subsequent operations on the defects.

[0057] Defect matching module: used to obtain corresponding defect data based on light reaction feature matching.

[0058] Exemplarily, the light reaction feature is matched with the standard geometric feature in the database to obtain the defect shape of the light reaction feature; specifically, it may be circles of different sizes, irregular patterns or dots; in this embodiment, the defect pixel ratio is set to 90%. Because of the geometric distance, the pixels in the defect shape may not all be weak light pixels or all strong light pixels. Therefore, if the proportion of weak light pixels to all pixels in the defect shape is greater than or equal to a certain value, it can be regarded that most of the pixels in the defect shape are weak light pixels, and the pixel type of the defect shape is marked as a weak light defect. This value is the defect pixel ratio, and the defect pixel ratio can be set according to the value specifically determined as a defect on the glass plate; if the proportion of strong light pixels in the defect shape is greater than or equal to the defect pixel ratio, the pixel type of the defect shape is marked as a strong light defect. If the proportion of strong light pixels and the proportion of weak light pixels in the defect shape are both less than the defect pixel ratio, it means that the defect shape cannot be considered as a certain type of abnormal pixel, so the pixel type of the corresponding defect shape is marked as a mixed defect.

[0059] The weak light pixel intensity is obtained by taking the average value of the weak light intensities corresponding to all weak light pixels in the defect shape whose pixel type is a weak light defect; the strong light pixel intensity is obtained by taking the average value of the strong light intensities corresponding to all strong light pixels in the defect shape whose pixel type is a strong light defect; the number of strong light pixels and the number of weak light pixels in the defect shape whose pixel type is a mixed defect are counted, and the average value of the strong light intensities of the strong light pixels in the mixed defect is marked as the mixed strong light intensity, and the average value of the weak light intensities of the weak light pixels in the mixed defect is marked as the mixed weak light intensity; the defect pixel intensity is obtained by dividing the sum of the product of the mixed strong light intensity and the number of strong light pixels and the product of the mixed weak light intensity and the number of weak light pixels by the sum of the number of strong light pixels and the number of weak light pixels; the defect pixel intensity is obtained by combining the weak light pixel intensity, the strong light pixel intensity and the mixed pixel intensity; the defect is evaluated from multiple dimensions in combination with the pixel type and the light intensity, so as to reduce misjudgment and missed detection and improve the accuracy of detection.

[0060] The standard defect shape, standard pixel type and corresponding defect pixel intensity range corresponding to the laser light propagation scene when different defects occur in the glass plate are extracted from the database; the defect shape is matched with the standard defect shape, and the light propagation phenomenon corresponding to the standard defect shape with the most similar defect shape is marked as the defect light propagation phenomenon, and it is judged whether the pixel type of the defect shape conforms to the standard pixel type corresponding to the defect light propagation phenomenon. In this embodiment, the defect shape is an aperture, and the pixel type corresponding to the aperture in the database should be a strong light defect. It is judged that the pixel type corresponding to the aperture is a strong light defect, then the defect pixel intensity of the strong light defect is matched with the defect pixel intensity range. The defect pixel intensity range includes the strong light defect pixel intensity range and the weak light defect pixel intensity range. If the matching result is If it meets the strong light defect pixel intensity range, it means that the defect position of the defect is located on one side of the glass plate irradiated by the laser. If the matching result is that it meets the weak light defect pixel intensity range, it means that the defect position of the defect is located on the other side of the glass plate irradiated by the laser. If neither of them meets the requirements, it means that the corresponding defect may not be recorded in the database. However, if the defect light propagation phenomenon is that there are inclusions or cracks inside the glass plate, the defect pixel intensity range has only one range. If it does not meet the range, it means that the corresponding defect is not recorded in the database, and the corresponding defect is marked as an unrecorded defect. If an unrecorded defect occurs, the staff is required to manually supplement the record to improve the database. The defect type corresponding to the defect shape and the defect position are combined to obtain defect data, and the defect data is uploaded to the management interface for reporting.

[0061] Working principle of the present invention:

[0062] The present invention collects a light propagation image generated by laser irradiation to a glass plate through a sensor, performs grayscale processing on the light propagation image after denoising to obtain a grayscale light image, counts the grayscale value of each pixel in the grayscale light image, uses the grayscale value as the light intensity of the pixel, takes the mode of the light intensity as the standard light intensity, divides the pixel points into abnormal pixel points and normal pixel points based on the standard light intensity, obtains an abnormal feature map based on the abnormal pixel points and a preset geometric distance; extracts light reaction features in the abnormal feature map, matches the light reaction features with the standard geometric features to obtain a defect shape; obtains a pixel type and defect pixel intensity of the defect shape based on the number and light intensity of the abnormal pixels in the defect shape, matches the defect shape, pixel type and defect pixel intensity with standard defect data in a database to obtain corresponding defect data.

[0063] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An online defect detection system for float glass production, characterized in that: include: Laser scanning module, image analysis module, defect matching module and database; Laser scanning module: used to transport the glass plate and scan the entire glass plate through the laser scanning system; Image analysis module: used to capture and generate a light propagation image of the glass plate; extract light reaction features in the image based on the light propagation image; Defect matching module: used to obtain corresponding defect data based on light reaction feature matching.

2. The online defect detection system for float glass production according to claim 1, characterized in that: The step of extracting light reaction features in an image based on a light propagation image comprises: Extract light propagation images; The light propagation image is denoised and converted into a grayscale image to obtain a grayscale light image, the grayscale value of each pixel in the grayscale light image is counted, and the grayscale value of the pixel is marked as light intensity; The mode of the pixel light intensity is used as the standard light intensity; the abnormal pixel is obtained based on the standard light intensity and the pixel light intensity of each pixel; Acquire an abnormal feature map in the grayscale light image based on abnormal pixels; extract light reaction features in the abnormal feature map.

3. The online defect detection system for float glass production according to claim 2, characterized in that: The method of obtaining abnormal pixels based on the standard light intensity and the pixel light intensity of each pixel includes: A1: Extract standard light intensity and pixel light intensity of each pixel; A2: Determine whether the light intensity of the pixel is equal to the standard light intensity; if yes, mark the corresponding pixel as a normal pixel; if no, jump to A3; A3: Determine whether the light intensity of the pixel is greater than the standard light intensity; if yes, mark the corresponding pixel as a strong light pixel; if no, mark the corresponding pixel as a weak light pixel; A4: Combine strong light pixels with weak light pixels to obtain abnormal pixels.

4. The online defect detection system for float glass production according to claim 2, characterized in that: The method of acquiring an abnormal feature map in a grayscale image based on abnormal pixels includes: Extract abnormal pixel points in the grayscale light image; mark the light intensity of weak light pixel points in the abnormal pixel points as weak light intensity, and mark the light intensity of strong light pixel points in the abnormal pixel points as strong light intensity; Set a geometric distance; combine other weak-light pixel points within the geometric distance around the weak-light pixel point to obtain a weak-light feature; combine all weak-light features to obtain a weak-light feature map; Combining other strong light pixel points within the geometric distance around the strong light pixel point to obtain a strong light feature; combining all the strong light features to obtain a strong light feature map; The weak light feature map is combined with the strong light feature map to obtain the abnormal feature map.

5. The online defect detection system for float glass production according to claim 1, characterized in that: The obtaining corresponding defect data according to the light reaction feature matching includes: Extracting light reaction features; Match the light reaction feature with the standard geometric feature in the database to obtain the defect shape of the light reaction feature; count the light intensity of the pixel points in the defect shape to obtain the defect pixel intensity and the pixel type of the corresponding defect shape; The defect shape, pixel type and defect pixel intensity are matched with the standard defect data in the database to obtain the corresponding defect data.

6. The online defect detection system for float glass production according to claim 5, characterized in that: The method of obtaining the defect pixel intensity and the pixel type of the corresponding defect shape by counting the pixel light intensity within the defect shape includes: B1: Set the defective pixel ratio; determine whether the proportion of weak-light pixels in the defect shape is greater than or equal to the defective pixel ratio; if yes, mark the pixel type of the corresponding defect shape as a weak-light defect; if no, jump to B2; B2: Determine whether the proportion of strong light pixels in the defect shape is greater than or equal to the defect pixel ratio; if yes, mark the pixel type of the corresponding defect shape as strong light defect; if no, mark the pixel type of the corresponding defect shape as mixed defect; B3: Calculate the average value of the weak light intensities corresponding to all weak light pixels in the defect shape where the pixel type is a weak light defect to obtain the weak light pixel intensity; calculate the average value of the strong light intensities corresponding to all strong light pixels in the defect shape where the pixel type is a strong light defect to obtain the strong light pixel intensity; B4: Count the number of strong light pixels and weak light pixels in the defect shape whose pixel type is mixed defect; mark the average value of the strong light intensity of the strong light pixels in the mixed defect as the mixed strong light intensity, and mark the average value of the weak light intensity of the weak light pixels in the mixed defect as the mixed weak light intensity; divide the sum of the product of the mixed strong light intensity and the number of strong light pixels and the product of the mixed weak light intensity and the number of weak light pixels by the sum of the number of strong light pixels and the number of weak light pixels to obtain the defect pixel intensity; B5: Combine the weak light pixel intensity, strong light pixel intensity and mixed pixel intensity to obtain the defective pixel intensity.

7. The online defect detection system for float glass production according to claim 5, characterized in that: The step of matching the defect shape, pixel type and defect pixel intensity with standard defect data in a database to obtain corresponding defect data includes: S1: extracting pixel types and defect pixel intensities of all defect shapes in the grayscale light image; extracting standard defect data from a database; the standard defect data includes standard pixel types and defect pixel intensity ranges corresponding to standard defect shapes caused by various defects on various glass plates that cause light propagation phenomena; S2: Match the defect shape with the standard defect shape to obtain the defect light propagation phenomenon corresponding to the defect shape; S3: Determine whether the pixel type of the defect shape meets the standard pixel type corresponding to the defect light propagation phenomenon; if yes, jump to S4; if no, mark the corresponding defect shape as an unrecorded defect; S4: matching the defect pixel intensity of the defect shape with the defect pixel intensity range, marking the defect position of the defect shape matching the weak light pixel intensity as the back side of the glass plate, and marking the defect position of the defect shape matching the strong light pixel intensity as the front side of the glass plate; marking the back side of the glass plate and the front side of the glass plate as defect positions; S5: Combine the defect type corresponding to the defect shape with the defect position to obtain defect data.

Images