Method and system for detecting upper and lower surface defects of 3D glass cover plate

By calculating the image availability index and adjusting the detection strategy, the problem of inaccurate image processing in 3D glass cover inspection was solved, ensuring the accuracy of the inspection results and product quality.

CN119044211APending Publication Date: 2024-11-29ANYANG NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411229980.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

In existing technologies for 3D glass cover defect detection, the lack of monitoring of image availability during image processing leads to inaccurate detection results and makes it impossible to flexibly adjust the manual inspection status, which may result in product quality issues.

Method used

By determining the importance level of the 3D glass cover, a strategy combining image detection and manual inspection is adopted to calculate the image availability index. The detection strategy is adjusted according to the index to ensure the accuracy of the detection results, including the calculation of object coverage, illumination qualification index, information entropy index, and exposure time index.

Benefits of technology

It enables usability monitoring of image detection results, flexibly adjusts manual detection status, reduces the probability of errors in the final detection results, and improves the reliability of product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119044211A_ABST
    Figure CN119044211A_ABST
Patent Text Reader

Abstract

The invention discloses a method and system for detecting upper and lower surface defects of a 3D glass cover plate, and particularly relates to the technical field of glass cover defect detection.For each target 3D glass cover plate to be detected, a corresponding first defect detection strategy and defect standard are determined firstly, and then an image available index of an image set corresponding to image detection defects is calculated; whether the image detection result is accurate or not is judged according to the image availability index, and when the image detection result is not accurate, a second defect detection strategy of the corresponding target 3D glass cover plate is determined according to the corresponding image availability index and the importance level to detect the defects of the upper surface and the lower surface of the target 3D glass cover plate. According to the method, the corresponding image availability can be monitored during image defect detection of each target 3D glass cover plate, and the subsequent manual detection state can be flexibly adjusted according to the actual image detection result, so that the probability that the final detection result has an error can be relatively small, and the probability that the quality of a factory product has a problem can be reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of glass cover defect detection, and in particular to a method and system for detecting defects on the upper and lower surfaces of a 3D glass cover. Background Art

[0002] 3D glass cover is a glass protective layer used in electronic devices, usually with a three-dimensional curved shape to adapt to the design requirements of the device; in order to ensure the quality and performance of 3D glass cover, improve the overall quality of the final product and user experience, generally before the 3D glass cover is put into use, the upper and lower surfaces of the 3D glass cover are inspected for defects, and only after it is determined to meet the standards, it is put into use, thereby ensuring the quality of the 3D glass cover leaving the factory;

[0003] The standards for 3D glass cover plates with different uses are also different. For example, for high-end electronic devices such as smartphones and tablets, the inspection standards are usually very strict; while for applications such as car displays and home appliance panels, the inspection standards may be relatively low. Generally speaking, for 3D glass cover plates with relatively low inspection standards, when inspecting defects on their upper and lower surfaces, the degree of defects can be directly determined by obtaining images of the upper and lower surfaces of the 3D glass cover plates and performing image analysis. However, for 3D glass cover plates with relatively high inspection standards, after determining the defects of the upper and lower surfaces through image analysis, manual secondary judgment is often required. In this way, by combining image processing and manual inspection, defects on the upper and lower surfaces of 3D glass cover plates can be efficiently and accurately inspected, saving human resources and time costs, making the entire inspection process more flexible and efficient.

[0004] However, in the above-mentioned technology, when defect detection is generally performed on the upper and lower surfaces of the 3D glass cover plate through image processing, it is often assumed that all the acquired images are available, and there is a lack of monitoring of image availability. If the acquired images of the upper and lower surfaces of the 3D glass cover plate are unreliable and image processing analysis is still performed to obtain the defect status of the upper and lower surfaces of the 3D glass cover plate, the detection result may be inaccurate. At the same time, the subsequent manual inspection status cannot be flexibly adjusted according to the actual image detection results, which may cause errors in the final detection results and lead to problems in the quality of the products leaving the factory. Summary of the invention

[0005] The purpose of the present invention is to solve the problem and to propose a method and system for detecting defects on the upper and lower surfaces of a 3D glass cover.

[0006] In a first aspect of the present invention, a method for detecting defects on the upper and lower surfaces of a 3D glass cover is first proposed, the method comprising:

[0007] For each target 3D glass cover plate to be inspected, before the inspection begins, the corresponding importance level is determined, and a first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image inspection, image inspection first and then manual inspection, and the manual inspection also includes the number of manual inspections;

[0008] For each target 3D glass cover plate, an image set corresponding to defects to be detected is obtained, and features are extracted for each image in the image set to obtain image feature data, an image availability index is obtained based on the image feature data, and whether the image detection result is accurate is determined based on the image availability index;

[0009] For each target 3D glass cover plate, when the image detection result is inaccurate, a second defect detection strategy for the corresponding target 3D glass cover plate is determined according to the corresponding image availability index and importance level;

[0010] The defect states of the upper and lower surfaces of the corresponding 3D glass cover are detected according to the second defect detection strategy, and whether the defect is qualified is determined according to the corresponding defect standard.

[0011] Optionally, the image feature data includes object coverage, illumination qualification index, information entropy index and exposure time index, and the object coverage is obtained by:

[0012] Obtain the completeness of the upper and lower surfaces of the target 3D glass cover plate covered in each image in the image set corresponding to the image detection defect, and obtain the object coverage rate of the target 3D glass cover plate image. The calculation formula is:

[0013]

[0014] Where Wsd is the object coverage, Wt x represents the xth image in the image set, n is the total number of images in the image set, and Wh is the number of images in the image set whose upper and lower surfaces of the target 3D glass cover plate covered are not 100% complete.

[0015] Optionally, the method for obtaining the illumination qualification index is:

[0016] The illumination qualification index includes the illumination uniformity index and the illumination brightness qualification index; the illumination uniformity index is obtained by:

[0017] Convert each image in the image set into a grayscale image, calculate the mean of the grayscale values ​​of all pixels in the grayscale image, and calculate the standard deviation of the image based on the mean. The inverse of the standard deviation is used as the illumination uniformity index. The calculation formula can be:

[0018]

[0019] Where σ is the standard deviation, M is the number of rows in the image, N is the number of columns in the image, and I gray (x,y) represents the gray value of pixel (x,y), μ is the mean;

[0020] The method for obtaining the light brightness qualified index is:

[0021] Each image in the image set is converted into a grayscale image, and the light brightness qualification index is obtained according to the grayscale values ​​of all pixels in the grayscale image and the preset brightness threshold. The calculation formula is:

[0022]

[0023] In the formula, Esx is the qualified index of light brightness, Yr b represents the gray value of the xth pixel in the image, f is the total number of pixel gray values ​​in the image, and Ye is the preset brightness threshold;

[0024] The illumination qualified index of each image is obtained according to the illumination uniformity index and the illumination brightness qualified index, and the average of the illumination qualified indexes of the images in the image set is calculated as the illumination qualified index RTW of the target 3D glass cover image.

[0025] Optionally, the method for obtaining the information entropy index is:

[0026] For each image in the image set, wavelet transform is performed through Daubechies wavelet function to decompose it into different frequency sub-bands;

[0027] For each frequency subband, calculate the grayscale histogram subband (i), Histogram represents the grayscale histogram, subband represents a frequency subband obtained after wavelet decomposition, and i represents the grayscale level;

[0028] Calculate the information entropy of each subband, the calculation formula is:

[0029]

[0030] In the formula, H subband represents the information entropy of the k-th subband, (P subband (i))·log2(P subband (i)) represents the contribution of each gray level i to the subband entropy value;

[0031] Calculate the information entropy index of the entire image. The calculation formula is:

[0032]

[0033] In the formula, Hyd is the information entropy index of the entire image, and u is the total number of sub-bands;

[0034] The information entropy index mean of the images in the image set is calculated as the information entropy index of the target 3D glass cover image.

[0035] Optionally, the exposure time index is obtained by:

[0036] Get the exposure time of each image in the image set, and calculate the exposure time index based on the preset maximum exposure time. The calculation formula is: Where RFg is the exposure time index, JR t represents the exposure time of the tth image in the image set, Jq is the maximum exposure time threshold, and n is the total number of images;

[0037] The image availability index obtained based on the image feature data includes:

[0038] The image availability index is obtained according to the object coverage, illumination qualification index, information entropy index and exposure time index. The calculation formula is:

[0039]

[0040] Where LTR is the image availability index, Wsd, PTW, Hye and RFg are the object coverage rate, illumination qualification index, information entropy index and exposure time index respectively, b1, b2, b3, b4 are the proportional coefficients of object coverage rate, illumination qualification index, information entropy index and exposure time index respectively, and b1, b2, b3, b4 are all greater than 0.

[0041] Optionally, judging whether the image detection result is accurate according to the image availability index includes:

[0042] Compare the image usability index with a preset image usability index threshold, and if the image usability index is greater than the preset image usability index threshold, the corresponding analyzed image detection result of the upper and lower surface defects of the target 3D glass cover plate is recorded as inaccurate;

[0043] If the image availability index is not greater than the preset image availability index threshold, the corresponding analyzed image detection results of the upper and lower surface defects of the target 3D glass cover plate are recorded as accurate, and the upper and lower surface defect detection of the corresponding target 3D glass cover plate is still performed according to the first defect detection strategy.

[0044] Optionally, determining a second defect detection strategy of a corresponding target 3D glass cover plate according to a corresponding image availability index and an importance level includes:

[0045] The importance level is converted into the corresponding importance value, and the adjustment index is obtained by combining the image availability index. The calculation formula is: TF = c1 × LTR + c2 × Scv, where TF is the adjustment index, LTR and Scv are the image availability index and importance value, respectively, c1 and c2 are the proportional coefficients of the image availability index and importance value, and c1 and c2 are both greater than 0;

[0046] For a target 3D glass cover plate whose first defect detection strategy is image detection, the second defect detection strategy is only manual detection, and the number of manual detections Td is: Td=RE×(1+TF), where RE is the preset minimum number of manual detections;

[0047] For the target 3D glass cover plate whose first defect detection strategy is image detection first and then manual detection, the second defect detection strategy is manual detection only, and the number of manual detections Tr is: Tr=HE×1+TF), where HE is the number of manual detections in the first defect detection strategy.

[0048] In a second aspect of the present invention, a system for detecting defects on the upper and lower surfaces of a 3D glass cover is provided, the system comprising:

[0049] The first defect detection strategy module: for each target 3D glass cover plate to be detected, before the detection is started, the corresponding importance level is determined, and the first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image detection, image detection first and then manual detection, and the manual detection also includes the number of manual detections;

[0050] Image availability module: For each target 3D glass cover plate, obtain the image set corresponding to the defects to be detected, extract features from each image in the image set, obtain image feature data, obtain the image availability index based on the image feature data, and judge whether the image detection result is accurate based on the image availability index;

[0051] Second defect detection strategy module: for each target 3D glass cover plate, when the image detection result is inaccurate, determining the second defect detection strategy of the corresponding target 3D glass cover plate according to the corresponding image availability index and importance level;

[0052] Judgment module: detects the defect status of the upper and lower surfaces of the corresponding 3D glass cover according to the second defect detection strategy, and determines whether it is qualified according to the corresponding defect standard.

[0053] Beneficial effects of the present invention:

[0054] The present invention proposes a method and system for detecting defects on the upper and lower surfaces of a 3D glass cover plate. For each target 3D glass cover plate to be detected, before the detection begins, a corresponding first defect detection strategy and defect standard are first determined, and then an image availability index of an image set corresponding to the image detection defect is calculated, and whether the image detection result is accurate is judged according to the image availability index. When the image detection result is inaccurate, a second defect detection strategy for the corresponding target 3D glass cover plate is determined according to the corresponding image availability index and importance level to detect defects on the upper and lower surfaces of the target 3D glass cover plate. In this way, the corresponding image availability during image detection defects of each target 3D glass cover plate can be monitored, and the subsequent manual detection state can be flexibly adjusted according to the actual image detection result, so that the probability of errors in the final detection result is relatively small, and the probability of quality problems of products leaving the factory is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described below in conjunction with the accompanying drawings.

[0056] Figure 1 A flow chart of a method for detecting defects on the upper and lower surfaces of a 3D glass cover provided by the present invention;

[0057] Figure 2 A framework diagram of a system for detecting defects on the upper and lower surfaces of a 3D glass cover provided by the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not 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.

[0059] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0060] The embodiment of the present invention provides a method and system for detecting defects on the upper and lower surfaces of a 3D glass cover. Figure 1 , Figure 1 A flow chart of a method for detecting defects on the upper and lower surfaces of a 3D glass cover provided by an embodiment of the present invention, the method comprising:

[0061] For each target 3D glass cover plate to be inspected, before the inspection begins, the corresponding importance level is determined, and the first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image inspection, image inspection first and then manual inspection, and the manual inspection also includes the number of manual inspections;

[0062] For each target 3D glass cover plate, an image set corresponding to defects to be detected is obtained, and features are extracted for each image in the image set to obtain image feature data, an image availability index is obtained based on the image feature data, and whether the image detection result is accurate is determined based on the image availability index;

[0063] For each target 3D glass cover plate, when the image detection result is inaccurate, a second defect detection strategy for the corresponding target 3D glass cover plate is determined according to the corresponding image availability index and importance level;

[0064] The defect states of the upper and lower surfaces of the corresponding 3D glass cover are detected according to the second defect detection strategy, and whether the defect is qualified is determined according to the corresponding defect standard.

[0065] Based on a method for detecting defects on the upper and lower surfaces of a 3D glass cover plate provided in an embodiment of the present invention, through the above-mentioned method, the corresponding image availability during image detection of defects of each target 3D glass cover plate can be monitored, and the subsequent manual inspection state can be flexibly adjusted according to the actual image detection results, so that the probability of errors in the final inspection results is relatively small, and the probability of quality problems of products leaving the factory is reduced.

[0066] It should be noted that, for each target 3D glass cover plate to be inspected, the corresponding importance level is determined before the inspection begins. The importance level is set by professionals based on the actual purpose of the target 3D glass cover plate and is not specifically limited. For example, assuming that the 3D glass cover plates used in high-end electronic devices such as smart phones and tablets usually have very strict inspection standards, so the importance level is relatively high; while the 3D glass cover plates used in applications such as vehicle-mounted display screens and home appliance panels may have relatively low inspection standards and the importance level may be average or low; the specific setting is based on the professional personnel and is not limited. In addition, for 3D glass cover plates with different importance levels, the corresponding first defect detection strategy is also different. For example, for 3D glass cover plates with a higher importance level, image inspection is generally performed first and then manual inspection, and the number of manual inspections is relatively more; for those with a general importance level, image inspection is generally performed first and then manual inspection, but the number of manual inspections is relatively less; for those with a lower importance level, image inspection is generally performed only. In this way, different defect detection strategies can be formulated according to actual situations to achieve efficient and rapid detection.

[0067] In one embodiment, the image feature data includes object coverage, illumination qualification index, information entropy index and exposure time index. The object coverage is obtained by:

[0068] Obtain the completeness of the upper and lower surfaces of the target 3D glass cover plate covered in each image in the image set corresponding to the image detection defect, and obtain the object coverage rate of the target 3D glass cover plate image. The calculation formula is:

[0069]

[0070] Where Wsd is the object coverage, Wt x represents the xth image in the image set, n is the total number of images in the image set, and Wh is the number of images in the image set whose upper and lower surfaces of the target 3D glass cover plate covered are not 100% complete.

[0071] It should be noted that the object coverage rate refers to the percentage of images in which the integrity of the upper and lower surfaces of the target 3D glass cover plate covered in all images corresponding to the image detection defects is not 100%. The larger the percentage, the more unusable the acquired image is. If image processing and analysis are still performed to obtain the defect status of the upper and lower surfaces of the 3D glass cover plate, the detection results may be inaccurate. The reason is that the integrity of the upper and lower surfaces of the target 3D glass cover plate covered in the image is not 100%, which means that some areas may not be fully captured by the camera. These uncaptured areas may contain important details or potential defects, which will result in the loss of such information in the subsequent image analysis and processing process, making it impossible to fully and accurately detect all defects on the glass cover plate, making the reliability and accuracy of the entire image detection process lower, which may eventually cause the detection results to deviate from the actual situation and increase the quality control risk in production.

[0072] In one implementation, the above method can more accurately calculate the object coverage, and the calculated object coverage is more consistent with the actual image situation.

[0073] In one embodiment, the method for obtaining the illumination qualification index is:

[0074] The qualified lighting index includes the lighting uniformity index and the lighting brightness qualified index; the method for obtaining the lighting uniformity index is:

[0075] Convert each image in the image set into a grayscale image, calculate the mean of the grayscale values ​​of all pixels in the grayscale image, and calculate the standard deviation of the image based on the mean. The inverse of the standard deviation is used as the illumination uniformity index. The calculation formula can be:

[0076]

[0077] Where σ is the standard deviation, M is the number of rows in the image, N is the number of columns in the image, and I gray (x,y) represents the gray value of pixel (x,y), μ is the mean;

[0078] The method for obtaining the qualified index of light brightness is as follows:

[0079] Each image in the image set is converted into a grayscale image, and the light brightness qualification index is obtained according to the grayscale values ​​of all pixels in the grayscale image and the preset brightness threshold. The calculation formula is:

[0080]

[0081] In the formula, Esx is the qualified index of light brightness, Yr b represents the gray value of the xth pixel in the image, f is the total number of pixel gray values ​​in the image, and Ye is the preset brightness threshold;

[0082] The illumination qualified index of each image is obtained according to the illumination uniformity index and the illumination brightness qualified index, and the average of the illumination qualified indexes of the images in the image set is calculated as the illumination qualified index RTW of the target 3D glass cover image.

[0083] It should be noted that the preset brightness threshold is set by professionals according to actual conditions and will not be limited or elaborated on in detail.

[0084] It should be noted that the present invention can use the following formula to calculate the illumination qualification index of each image, and the calculation expression is:

[0085]

[0086] In the formula, RTG is the qualified index of illumination, Qas and Esx are the qualified index of illumination uniformity and illumination brightness respectively, a1 and a2 are the proportional coefficients of illumination uniformity and illumination brightness respectively, and a1 and a2 are both greater than 0;

[0087] It should be noted that the illumination qualification index refers to the illumination uniformity and illumination brightness qualification of the image set corresponding to the image detection defect. If the illumination uniformity and illumination brightness qualification are smaller, it means that the acquired image is more unusable. If image processing and analysis are still performed to obtain the defect status of the upper and lower surfaces of the 3D glass cover, the detection result may be inaccurate. The reason is that when the illumination uniformity and illumination brightness qualification are low, there may be too dark or too bright areas in the image, causing some details to be hidden or blurred. These areas of uneven illumination may mask the defects on the upper and lower surfaces of the 3D glass cover, making it impossible for the detection algorithm to accurately identify potential problems in these areas, further reducing the accuracy of image analysis. Unqualified lighting conditions will directly lead to the loss or distortion of key information in the image, which will cause the subsequent defect detection results to deviate from the actual situation and increase the risk of misjudgment.

[0088] In one embodiment, the method for obtaining the information entropy index is:

[0089] For each image in the image set, wavelet transform is performed through Daubechies wavelet function to decompose it into different frequency sub-bands;

[0090] For each frequency subband, calculate the grayscale histogram subband (i), Histogram represents the grayscale histogram, subband represents a frequency subband obtained after wavelet decomposition, and i represents the grayscale level;

[0091] Calculate the information entropy of each subband, the calculation formula is:

[0092]

[0093] In the formula, H subband represents the information entropy of the kth subband, P subband (i)·log2(P subband (i)) represents the contribution of each gray level i to the subband entropy value;

[0094] Calculate the information entropy index of the entire image. The calculation formula is:

[0095]

[0096] In the formula, Hyd is the information entropy index of the entire image, and u is the total number of sub-bands;

[0097] The information entropy index mean of the images in the image set is calculated as the information entropy index Hye of the target 3D glass cover image.

[0098] It should be noted that the information entropy index refers to the complexity of the information about the upper surface of the 3D glass cover that can be extracted from the image, also known as the image information volume or image complexity, which reflects the richness of the details and information contained in the image. If the information entropy index is smaller, it means that the acquired image is more unusable. If image processing analysis is still performed to obtain the defect status of the upper and lower surfaces of the 3D glass cover, the detection result may be inaccurate. The reason is that a small information entropy index means that the amount of information in the image is low, reflecting the lack of details in the image or the texture and structure in the area are relatively simple. In this case, the image may have a large uniform area or noise, making it difficult to distinguish subtle defects or complex structural changes on the surface of the 3D glass cover. If the image information volume is insufficient, even if image processing analysis is performed, key surface features or flaws may not be captured, resulting in reduced accuracy of defect detection.

[0099] In one implementation, the benefit of using the Daubechies wavelet function to perform wavelet transform decomposition into different frequency sub-bands is that the Daubechies wavelet function has good time-frequency localization characteristics and can capture image details and edge information at different scales. Through wavelet transform, the image is decomposed into multiple frequency sub-bands, each of which represents the characteristics of the image within a specific frequency range, which helps to analyze subtle structures and texture changes in the image. The Daubechies wavelet function is particularly good at processing non-stationary signals, so it performs well in extracting complex details on the surface of 3D glass cover plates, and can effectively improve the calculation accuracy of the information entropy index, ensure accurate evaluation of the amount of image information, and thus improve the reliability of defect detection.

[0100] In one embodiment, the exposure time index is obtained by:

[0101] Get the exposure time of each image in the image set, and calculate the exposure time index based on the preset maximum exposure time. The calculation formula is: Where RFg is the exposure time index, JR t represents the exposure time of the tth image in the image set, Jq is the maximum exposure time threshold, and n is the total number of images;

[0102] The image availability index obtained based on the image feature data includes:

[0103] The image availability index is obtained according to the object coverage, illumination qualification index, information entropy index and exposure time index. The calculation formula is:

[0104]

[0105] Where LTR is the image availability index, Wsd, RTW, Hye and RFg are the object coverage rate, lighting qualification index, information entropy index and exposure time index respectively, b1, b2, b3, b4 are the proportional coefficients of object coverage rate, lighting qualification index, information entropy index and exposure time index respectively, and b1, b2, b3, b4 are all greater than 0.

[0106] It should be noted that the exposure time of each image may be obtained by other means using the storage records of the shooting device, which is not specifically limited. In addition, the preset maximum exposure time is set by professionals based on actual conditions and is not specifically limited.

[0107] It should be noted that the exposure time index refers to the difference between the exposure time of each image in the image set when it is taken and the preset maximum exposure time. If the difference is larger, it means that the acquired image is more unusable. Image processing and analysis are still performed to obtain the defect status of the upper and lower surfaces of the 3D glass cover, and the detection results may be inaccurate. The reason is that too long exposure time will cause overexposure in certain areas of the image, which will make the brightness of the image too high and cause the loss of detail information. Overexposed images may have large areas of bright spots, covering up the subtle defects and texture changes on the surface of the 3D glass cover, which in turn affects the detection algorithm's accurate recognition of these key areas. In addition, too long exposure time may also cause the image contrast to decrease and the edges to blur, further reducing the clarity and information content of the image, so that the information entropy index cannot reflect the actual surface complexity. These factors may lead to large deviations in the results of defect detection, making the detection results inaccurate and increasing the risk of misjudgment.

[0108] In one embodiment, judging whether the image detection result is accurate according to the image availability index includes:

[0109] Compare the image usability index with a preset image usability index threshold, and if the image usability index is greater than the preset image usability index threshold, the corresponding analyzed image detection result of the upper and lower surface defects of the target 3D glass cover plate is recorded as inaccurate;

[0110] If the image availability index is not greater than the preset image availability index threshold, the corresponding analyzed image detection results of the upper and lower surface defects of the target 3D glass cover plate are recorded as accurate, and the upper and lower surface defect detection of the corresponding target 3D glass cover plate is still performed according to the first defect detection strategy.

[0111] It should be noted that the preset image usability index threshold is set by professionals according to actual conditions and is not specifically limited.

[0112] In one embodiment, determining the second defect detection strategy of the corresponding target 3D glass cover plate according to the corresponding image availability index and importance level includes:

[0113] The importance level is converted into the corresponding importance value, and the adjustment index is obtained by combining the image availability index. The calculation formula is: TF = c1 × LTR + c2 × Scv, where TF is the adjustment index, LTR and Scv are the image availability index and importance value, respectively, c1 and c2 are the proportional coefficients of the image availability index and importance value, and c1 and c2 are both greater than 0;

[0114] For a target 3D glass cover plate whose first defect detection strategy is image detection, the second defect detection strategy is only manual detection, and the number of manual detections Td is: Td=RE×(1+TF), where RE is the preset minimum number of manual detections;

[0115] For the target 3D glass cover plate whose first defect detection strategy is image detection first and then manual detection, the second defect detection strategy is manual detection only, and the number of manual detections Tr is: Tr=HE×(1+TF), where HE is the number of manual detections in the first defect detection strategy.

[0116] It should be noted that manual inspection can identify complex defects that image inspection cannot accurately detect, especially those small or complex defects that are difficult to identify through algorithms. The human eye has sensitivity to details and intuitive judgment of anomalies, which is more reliable than image inspection in some cases.

[0117] In one implementation, through the above method, the subsequent manual inspection state can be flexibly adjusted according to the actual image detection results, which can reduce the probability of errors in the final detection results and reduce the probability of quality problems in factory products.

[0118] Based on the same inventive concept, the present invention also provides a system for detecting defects on the upper and lower surfaces of a 3D glass cover. Figure 2 , Figure 2 A framework diagram of a system for detecting defects on the upper and lower surfaces of a 3D glass cover provided by an embodiment of the present invention, the system comprising:

[0119] First defect detection strategy module: for each target 3D glass cover plate to be inspected, before the inspection begins, the corresponding importance level is determined, and the first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image inspection, image inspection first and then manual inspection, and the manual inspection also includes the number of manual inspections;

[0120] Image availability module: For each target 3D glass cover plate, obtain the image set corresponding to the defects to be detected, extract features from each image in the image set, obtain image feature data, obtain the image availability index based on the image feature data, and judge whether the image detection result is accurate based on the image availability index;

[0121] Second defect detection strategy module: for each target 3D glass cover plate, when the image detection result is inaccurate, determining the second defect detection strategy of the corresponding target 3D glass cover plate according to the corresponding image availability index and importance level;

[0122] Judgment module: detects the defect status of the upper and lower surfaces of the corresponding 3D glass cover according to the second defect detection strategy, and determines whether it is qualified according to the corresponding defect standard.

[0123] Based on a detection system for upper and lower surface defects of a 3D glass cover plate provided by an embodiment of the present invention, the above method can monitor the corresponding image availability when image detection defects are performed on each target 3D glass cover plate, and flexibly adjust the subsequent manual inspection state according to the actual image detection results, so that the probability of errors in the final inspection results is relatively small, and the probability of quality problems of products leaving the factory is reduced.

[0124] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for detecting defects on the upper and lower surfaces of a 3D glass cover, characterized in that: The following steps are involved: For each target 3D glass cover plate to be inspected, before the inspection begins, the corresponding importance level is determined, and a first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image inspection, image inspection first and then manual inspection, and the manual inspection also includes the number of manual inspections; For each target 3D glass cover plate, an image set corresponding to defects to be detected is obtained, and features are extracted for each image in the image set to obtain image feature data, an image availability index is obtained based on the image feature data, and whether the image detection result is accurate is determined based on the image availability index; For each target 3D glass cover plate, when the image detection result is inaccurate, a second defect detection strategy for the corresponding target 3D glass cover plate is determined according to the corresponding image availability index and importance level; The defect states of the upper and lower surfaces of the corresponding 3D glass cover are detected according to the second defect detection strategy, and whether the defect is qualified is determined according to the corresponding defect standard.

2. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 1, characterized in that: The image feature data includes object coverage, illumination qualification index, information entropy index and exposure time index. The object coverage is obtained by: Obtain the completeness of the upper and lower surfaces of the target 3D glass cover plate covered in each image in the image set corresponding to the image detection defect, and obtain the object coverage rate of the target 3D glass cover plate image. The calculation formula is: Where Wsd is the object coverage, Wt x represents the xth image in the image set, n is the total number of images in the image set, and Wh is the number of images in the image set whose upper and lower surfaces of the target 3D glass cover plate covered are not 100% complete.

3. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 2, characterized in that: The method for obtaining the illumination qualification index is: The illumination qualification index includes the illumination uniformity index and the illumination brightness qualification index; the illumination uniformity index is obtained by: Convert each image in the image set into a grayscale image, calculate the mean of the grayscale values ​​of all pixels in the grayscale image, and calculate the standard deviation of the image based on the mean. The inverse of the standard deviation is used as the illumination uniformity index. The calculation formula can be: Where σ is the standard deviation, M is the number of rows in the image, N is the number of columns in the image, and I gray (x,y) represents the gray value of pixel (x,y), μ is the mean; The method for obtaining the light brightness qualified index is: Each image in the image set is converted into a grayscale image, and the light brightness qualification index is obtained according to the grayscale values ​​of all pixels in the grayscale image and the preset brightness threshold. The calculation formula is: In the formula, Esx is the qualified index of light brightness, Yr b represents the gray value of the xth pixel in the image, f is the total number of pixel gray values ​​in the image, and Ye is the preset brightness threshold; The illumination qualified index of each image is obtained according to the illumination uniformity index and the illumination brightness qualified index, and the average of the illumination qualified indexes of the images in the image set is calculated as the illumination qualified index RTW of the target 3D glass cover image.

4. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 2, characterized in that: The method for obtaining the information entropy index is: For each image in the image set, wavelet transform is performed through Daubechies wavelet function to decompose it into different frequency sub-bands; For each frequency subband, calculate the grayscale histogram subband (i), Histogram represents the grayscale histogram, subband represents a frequency subband obtained after wavelet decomposition, and i represents the grayscale level; Calculate the information entropy of each subband, the calculation formula is: In the formula, H subband represents the information entropy of the kth subband, P subband (i)·log2(P subband (i)) represents the contribution of each gray level i to the subband entropy value; Calculate the information entropy index of the entire image. The calculation formula is: In the formula, Hyd is the information entropy index of the entire image, and u is the total number of sub-bands; The information entropy index mean of the images in the image set is calculated as the information entropy index Hye of the target 3D glass cover image.

5. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 2, characterized in that: The method for obtaining the exposure time index is: Get the exposure time of each image in the image set, and calculate the exposure time index based on the preset maximum exposure time. The calculation formula is: Where RFg is the exposure time index, JR t represents the exposure time of the tth image in the image set, Jq is the maximum exposure time threshold, and n is the total number of images; The image availability index obtained based on the image feature data includes: The image availability index is obtained according to the object coverage, illumination qualification index, information entropy index and exposure time index. The calculation formula is: Where LTR is the image availability index, Wsd, RTW, Hye and RFg are the object coverage rate, lighting qualification index, information entropy index and exposure time index respectively, b1, b2, b3, b4 are the proportional coefficients of object coverage rate, lighting qualification index, information entropy index and exposure time index respectively, and b1, b2, b3, b4 are all greater than 0.

6. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 1, characterized in that: Judging whether the image detection results are accurate based on the image availability index includes: Compare the image usability index with a preset image usability index threshold, and if the image usability index is greater than the preset image usability index threshold, the corresponding analyzed image detection result of the upper and lower surface defects of the target 3D glass cover plate is recorded as inaccurate; If the image availability index is not greater than the preset image availability index threshold, the corresponding analyzed image detection results of the upper and lower surface defects of the target 3D glass cover plate are recorded as accurate, and the upper and lower surface defect detection of the corresponding target 3D glass cover plate is still performed according to the first defect detection strategy.

7. The method for detecting defects on the upper and lower surfaces of a 3D glass cover according to claim 1, characterized in that: The second defect detection strategy for determining the corresponding target 3D glass cover plate according to the corresponding image availability index and importance level includes: The importance level is converted into the corresponding importance value, and the adjustment index is obtained by combining the image availability index. The calculation formula is: TF = c1 × LTR + c2 × Scv, where TF is the adjustment index, LTR and Scv are the image availability index and importance value, respectively, c1 and c2 are the proportional coefficients of the image availability index and importance value, and c1 and c2 are both greater than 0; For the target 3D glass cover plate whose first defect detection strategy is image detection, the second defect detection strategy is only manual detection, and the number of manual detections Td is: Td=RE×((1+TF), where RE is the preset minimum number of manual detections; For the target 3D glass cover plate whose first defect detection strategy is image detection first and then manual detection, the second defect detection strategy is manual detection only, and the number of manual detections Tr is: Tr=HE×(1+TF), where HE is the number of manual detections in the first defect detection strategy.

8. A system for detecting defects on the upper and lower surfaces of a 3D glass cover, characterized in that: The system comprises: The first defect detection strategy module: for each target 3D glass cover plate to be detected, before the detection is started, the corresponding importance level is determined, and the first defect detection strategy and defect standard of the target 3D glass cover plate are determined according to the importance level; the first defect detection strategy includes image detection, image detection first and then manual detection, and the manual detection also includes the number of manual detections; Image availability module: For each target 3D glass cover plate, obtain the image set corresponding to the defects to be detected, extract features from each image in the image set, obtain image feature data, obtain the image availability index based on the image feature data, and judge whether the image detection result is accurate based on the image availability index; Second defect detection strategy module: for each target 3D glass cover plate, when the image detection result is inaccurate, determining the second defect detection strategy of the corresponding target 3D glass cover plate according to the corresponding image availability index and importance level; Judgment module: detects the defect status of the upper and lower surfaces of the corresponding 3D glass cover according to the second defect detection strategy, and determines whether it is qualified according to the corresponding defect standard.