Defect detection method and device and electronic equipment

Abstract

The invention relates to a defect detection method and device and electronic equipment. The method comprises the following steps: acquiring an initial defect detection result obtained by performing defect detection on a plurality of display panels by detection equipment; the initial defect detection result comprises feature data of initial defects of a plurality of display panels; inputting the feature data of each initial defect into a preset defect identification model to obtain feature data of each real defect output by the preset defect identification model; wherein the preset defect identification model determines the real defect from the initial defects according to the feature data of the initial defects; and taking the feature data of each real defect as a final defect detection result. By adopting the method, the defect detection accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of display technology, and in particular to a defect detection method, device and electronic equipment. Background Art

[0002] The production process of display panels is long and complex, and there are hundreds of types of defects in the production process. Usually, the display panel must be defect-detected before the next step. According to the results of the defect detection, unqualified products are eliminated and repairable defects are repaired. Therefore, the accuracy of defect detection is crucial.

[0003] In the related art, an automatic inspection machine is used to batch detect a large number of defects to obtain defect detection results. Usually, there is over-detection in the defect detection results output by the automatic inspection machine. Therefore, the accuracy of the defect detection results output by the automatic inspection machine is low. Summary of the invention

[0004] Based on this, it is necessary to provide a defect detection method, device and electronic device that can improve the accuracy of defect detection in response to the above technical problems.

[0005] In a first aspect, the present application provides a defect detection method. The method comprises:

[0006] Acquire initial defect detection results obtained by performing defect detection on multiple display panels by a detection device; the initial defect detection results include feature data of initial defects of the multiple display panels;

[0007] Inputting the characteristic data of each of the initial defects into a preset defect recognition model to obtain the characteristic data of each of the real defects output by the preset defect recognition model; wherein the preset defect recognition model determines the real defect from each of the initial defects according to the characteristic data of each of the initial defects;

[0008] The characteristic data of each real defect is used as the final defect detection result.

[0009] In one of the embodiments, a preset defect recognition model determines the real defect from each of the initial defects based on the feature data of each of the initial defects, including: determining the target feature data of each of the initial defects from the feature data of each of the initial defects; determining whether each of the initial defects is a real defect based on the target feature data of each of the initial defects; wherein each of the target feature data is associated with each other.

[0010] In one of the embodiments, the process of constructing a preset defect recognition model includes: obtaining a sample data set; training an initial defect recognition model based on the sample data set to obtain a candidate defect recognition model; evaluating the accuracy of the candidate defect recognition model based on a preset function; when the evaluation result indicates that the accuracy of the candidate defect recognition model meets the preset accuracy condition, the candidate defect recognition model is used as the preset defect recognition model.

[0011] In one embodiment, obtaining a sample data set includes:

[0012] Obtain sample defect detection results output by a detection device for defect detection of multiple sample display panels; the sample defect detection results include sample feature data of multiple sample defects of the sample display panel; the sample feature data include data values ​​corresponding to multiple candidate features; according to the data values ​​corresponding to each candidate feature of each sample defect, screen each candidate feature to obtain a first target feature of each sample defect, and obtain a data value corresponding to a second target feature of each sample defect, wherein the data value corresponding to the second target feature is used to characterize the position distribution of the sample defect in the defect detection process; determine the sample data set according to the data value corresponding to the first target feature and the data value corresponding to the second target feature of each sample defect.

[0013] In one embodiment, the sample feature data includes a defect label of the sample defect; the defect label is used to characterize whether the sample defect is a real defect; each candidate feature is screened according to the data value corresponding to each candidate feature of each sample defect to obtain a first target feature of each sample defect, including: determining a plurality of first correlation features from each candidate feature according to the defect label of each sample defect; determining a plurality of second correlation features from other candidate features in each candidate feature except each first correlation feature; and using the first correlation feature and the second correlation feature as the first target feature.

[0014] In one of the embodiments, based on the defect label of each of the sample defects, a plurality of first correlation features are determined from each of the candidate features, including: for each first sample defect whose defect label is a first label, based on the data values ​​corresponding to each of the candidate features of each of the first sample defects, determining the first data distribution corresponding to each of the candidate features; for each second sample defect whose defect label is a second label, based on the data values ​​corresponding to each of the candidate features of each of the second sample defects, determining the second data distribution corresponding to each of the candidate features; for each of the candidate features, performing a difference comparison between the first data distribution and the second data distribution corresponding to the candidate feature to obtain a comparison result; and based on the comparison result of each of the candidate features, determining a plurality of first correlation features from each of the candidate features.

[0015] In one of the embodiments, a plurality of second correlation features are determined from other candidate features among the candidate features except for each of the first correlation features, including: calculating the correlation between each of the other candidate features based on the data values ​​of each of the other candidate features of each of the sample defects to obtain a plurality of correlation calculation values; determining a target correlation calculation value greater than a preset correlation threshold from each of the correlation calculation values; and using other candidate features to which the data values ​​used to calculate the target correlation calculation value belong as the second correlation features.

[0016] In one embodiment, the initial defect recognition model is constructed based on a random forest algorithm, and the initial defect recognition model includes multiple random classification tree models.

[0017] In a second aspect, the present application also provides a defect detection device. The device comprises:

[0018] An acquisition module, used to acquire an initial defect detection result obtained by a detection device performing defect detection on a plurality of display panels; the initial defect detection result includes characteristic data of initial defects of the plurality of display panels;

[0019] A determination module, used for inputting the characteristic data of each of the initial defects into a preset defect recognition model to obtain the characteristic data of each of the real defects output by the preset defect recognition model; wherein the preset defect recognition model determines the real defect from each of the initial defects according to the characteristic data of each of the initial defects;

[0020] The result module is used to use the characteristic data of each real defect as the final defect detection result.

[0021] In one embodiment, the determination module is specifically used to: determine the target feature data of each initial defect from the feature data of each initial defect; determine whether each initial defect is a real defect based on the target feature data of each initial defect; wherein each target feature data is associated with each other.

[0022] In one of the embodiments, the device also includes a construction module, which is used to: obtain a sample data set; train an initial defect recognition model based on the sample data set to obtain a candidate defect recognition model; evaluate the accuracy of the candidate defect recognition model based on a preset function; when the evaluation result indicates that the accuracy of the candidate defect recognition model meets the preset accuracy condition, use the candidate defect recognition model as the preset defect recognition model.

[0023] In one embodiment, the building block is specifically configured to:

[0024] Obtain sample defect detection results output by a detection device for defect detection of multiple sample display panels; the sample defect detection results include sample feature data of multiple sample defects of the sample display panel; the sample feature data include data values ​​corresponding to multiple candidate features; according to the data values ​​corresponding to each candidate feature of each sample defect, screen each candidate feature to obtain a first target feature of each sample defect, and obtain a data value corresponding to a second target feature of each sample defect, wherein the data value corresponding to the second target feature is used to characterize the position distribution of the sample defect in the defect detection process; determine the sample data set according to the data value corresponding to the first target feature and the data value corresponding to the second target feature of each sample defect.

[0025] In one embodiment, the sample feature data includes a defect label of the sample defect; the defect label is used to characterize whether the sample defect is a real defect; a construction module is specifically used to: determine a plurality of first correlation features from each candidate feature according to the defect label of each sample defect; determine a plurality of second correlation features from other candidate features among each candidate feature except each first correlation feature; and use the first correlation feature and the second correlation feature as the first target feature.

[0026] In one embodiment, the construction module is specifically used to: for each first sample defect whose defect label is a first label, determine the first data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each first sample defect; for each second sample defect whose defect label is a second label, determine the second data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each second sample defect; for each candidate feature, perform a difference comparison between the first data distribution and the second data distribution corresponding to the candidate feature to obtain a comparison result; and determine multiple first correlation features from each candidate feature according to the comparison result of each candidate feature.

[0027] In one embodiment, the construction module is specifically used to: calculate the correlation between each other candidate feature based on the data value of each other candidate feature of each sample defect to obtain multiple correlation calculation values; determine a target correlation calculation value greater than a preset correlation threshold from each correlation calculation value; and use the other candidate feature to which the data value used to calculate the target correlation calculation value belongs as the second correlation feature.

[0028] In one embodiment, the initial defect recognition model is constructed based on a random forest algorithm, and the initial defect recognition model includes a plurality of random classification tree models.

[0029] In a third aspect, the present application further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in any one of the first aspects are implemented.

[0030] In the above-mentioned defect detection method, device and electronic device, an initial defect detection result including characteristic data of initial defects of multiple display panels is obtained by acquiring the detection device for performing defect detection on multiple display panels; the characteristic data of each initial defect is input into a preset defect recognition model to obtain characteristic data of each real defect output by the preset defect recognition model; and the characteristic data of each real defect is used as the final defect detection result. The preset defect recognition model determines the real defect from each initial defect based on the characteristic data of each initial defect. The detection device is, for example, an automatic inspection machine. In this way, based on the initial defect detection result output by the automatic inspection machine, the preset defect recognition model is used to perform screening and identification again, and the real defect is determined from the initial defects. The final defect detection result obtained has higher accuracy, thereby improving the accuracy of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 is a schematic flow chart of a defect detection method in one embodiment;

[0033] Figure 2 A schematic diagram of a process for constructing a preset defect recognition model in an embodiment;

[0034] Figure 3 A schematic diagram of a process for obtaining a sample data set in one embodiment;

[0035] Figure 4 A schematic diagram of a process for determining a first target feature in one embodiment;

[0036] Figure 5 A schematic diagram of a process for determining a first correlation feature in one embodiment;

[0037] Figure 6 A schematic diagram showing a comparison between a first data distribution and a second data distribution of a first candidate feature in an embodiment;

[0038] Figure 7 A schematic diagram of a process for determining a second correlation feature in one embodiment;

[0039] Figure 8 A schematic diagram of the arrangement of a display panel in an embodiment;

[0040] Fig. 9 A schematic diagram of a process for constructing a preset defect recognition model in an embodiment;

[0041] Fig.10 A schematic diagram of a process for identifying whether each initial defect is a real defect in one embodiment;

[0042] Fig.11 is a structural block diagram of a defect detection device in one embodiment;

[0043] Fig.12 FIG. 4 is a diagram showing the internal structure of an electronic device in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It should be understood that many specific details are set forth in the following description to facilitate a full understanding of the present application, but the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0046] For electronic products, especially display panels, the production process is long, with many sites and complex processes. There are hundreds of types of defects generated during the production process. Usually, the display panel must be defect-detected before the next step. Based on the results of the defect detection, unqualified ones are eliminated and repairable defects are repaired. If the defects of the display panel are not detected and flow into the next link, it will cause batch scrapping, resulting in greater economic losses.

[0047] As the PPI (Pixels Per Inch) of display panels continues to increase, inspection equipment often over-inspects defects in order to avoid missing them. Over-inspection means, for example, reporting product process window fluctuations that are not defects as defects. A high over-inspection rate (number of over-inspected defects / total number of defects) will affect the judgment of the status of the online process, and a high over-inspection rate will also increase the number of ineffective repair defects.

[0048] Traditional automatic inspection machines (AT) and electrical properties testers (EPM) detect a large number of defects, and the pass rate fluctuates randomly and may be abnormally high. In related technologies, when the pass rate is too high, people manually judge whether the defects detected by the inspection equipment are true to reduce the pass rate, but this method is inefficient.

[0049] Therefore, there is an urgent need to solve the problem of how to efficiently classify the massive defects detected by the testing equipment before reporting the test results, avoid excessively high pass rate, save testing resources, and improve the accuracy of the test results of the testing equipment.

[0050] In view of this, an embodiment of the present application provides a defect detection method, which can be applied to a detection device so that after obtaining preliminary defect detection results, the detection device can further screen the detection results, efficiently utilize detection resources, and output more accurate defect detection results.

[0051] It should be noted that the defect detection method provided in the embodiment of the present application may be executed by a defect detection device, which may be implemented as part or all of an electronic device through software, hardware, or a combination of software and hardware. The electronic device may be a detection device for detecting defects in a display panel, such as the above-mentioned automatic inspection machine (AT). Optionally, the electronic device may also be other types of devices such as a personal computer, a laptop computer, a tablet computer, and a server. In the following method embodiments, the execution subject is described as a detection device.

[0052] In one embodiment, Figure 1 As shown, a defect detection method is provided, comprising the following steps:

[0053] Step 101, obtaining initial defect detection results obtained by performing defect detection on multiple display panels by a detection device. The initial defect detection results include feature data of initial defects of the multiple display panels.

[0054] The detection device can perform defect detection on multiple display panels at the same time. Optionally, after each display panel is arranged in an array on the glass substrate, the detection device is used to perform defect detection on each display panel.

[0055] The initial defect refers to the defects detected by the detection device on multiple display panels, each of which is regarded as an initial defect. The initial defect detection result includes characteristic data of each initial defect.

[0056] Optionally, the characteristic data of the initial defect includes, for example, the defect serial number, the serial number of the display panel to which the initial defect belongs, the position of the initial defect in the display panel to which it belongs, the serial number of the data line connected to the pixel unit corresponding to the initial defect, etc. The characteristic data of the initial defect contains a large amount of data, which is not fully illustrated here.

[0057] It can be understood that some of the initial defects may be defects in the same display panel.

[0058] Step 102, input the characteristic data of each initial defect into a preset defect recognition model to obtain the characteristic data of each real defect output by the preset defect recognition model. The preset defect recognition model determines the real defect from each initial defect according to the characteristic data of each initial defect.

[0059] Traditional detection equipment outputs the characteristic data of the above-mentioned initial defects as defect detection results, which have a high over-detection rate, that is, some of the initial defects are real defects and others are false defects, so the accuracy of the defect detection results is low.

[0060] Since the characteristic data of each initial defect can characterize whether the initial defect is a real defect to a certain extent, in an embodiment of the present application, the characteristic data of each initial defect is input into a preset defect recognition model, and each initial defect is identified using the preset defect recognition model to determine the real defect. In other words, based on the characteristic data of each initial defect, each initial defect is screened, false defects are removed, and real defects are output to improve the accuracy of the detection results.

[0061] Among them, the preset defect recognition model is a pre-trained model used to identify whether each initial defect is a real defect. The preset defect recognition model can be preset in the detection equipment. In this way, after obtaining the characteristic data of the initial defect, the detection equipment does not output the characteristic data of the initial defect first, but inputs the characteristic data of the initial defect into the preset defect recognition model, and uses the preset defect recognition model to perform identification and screening to obtain the characteristic data of each real defect output by the preset defect recognition model. In this way, the detection equipment can output the characteristic data of each real defect as the final defect detection result.

[0062] Optionally, as mentioned above, the execution subject of the method can also be other devices, such as a personal computer, etc. For example, if the execution subject is a personal computer, the preset defect recognition model can be preset in the personal computer, and the detection device outputs the characteristic data of each initial defect after defect detection. The personal computer obtains the characteristic data of each initial defect, and then uses the preset defect recognition model to identify it to obtain the characteristic data of each real defect.

[0063] Step 103: taking the characteristic data of each real defect as the final defect detection result.

[0064] The above-mentioned defect detection method obtains the initial defect detection result including the characteristic data of the initial defects of the multiple display panels obtained by the detection equipment performing defect detection on the multiple display panels; inputs the characteristic data of each of the initial defects into a preset defect recognition model to obtain the characteristic data of each of the real defects output by the preset defect recognition model; and uses the characteristic data of each of the real defects as the final defect detection result. The preset defect recognition model determines the real defect from each of the initial defects based on the characteristic data of each of the initial defects. The detection equipment is, for example, an automatic inspection machine. In this way, based on the initial defect detection result output by the automatic inspection machine, the preset defect recognition model is used to perform screening and identification again, and the real defects are determined from the initial defects. The final defect detection result obtained has a higher accuracy, thereby improving the accuracy of defect detection.

[0065] The construction process of the preset defect model is described below.

[0066] In one embodiment, Figure 2 The figure shows a flow chart of building a preset defect recognition model provided by an embodiment of the present application. The process of building the preset defect recognition model includes:

[0067] Step 201: Obtain a sample data set.

[0068] The sample data set may include feature data of many sample defects, and the feature data of the sample defects may be referred to as sample feature data.

[0069] Optionally, each sample defect may be an initial defect of each display panel detected by a detection device when performing defect detection on some display panels in the past.

[0070] It can be understood that among the sample defects, some sample defects are real defects and some sample defects are false defects.

[0071] Step 202: train the initial defect recognition model according to the sample data set to obtain a candidate defect recognition model.

[0072] A training data set and a test data set may be established from the sample data set. Optionally, the ratio of the training data set to the test data set is 3:7.

[0073] The training data set can be used to train the initial defect recognition model, and the test data set can be used to test the accuracy of the trained initial defect recognition model.

[0074] The initial defect recognition model is implemented based on a deep learning algorithm. Optionally, the initial defect recognition model is built based on a random forest algorithm. The initial defect recognition model includes multiple random classification tree models. In this way, the generated logic can be visualized and explained. During the training process, the generated random classification tree logic can be integrated and imported back to the detection device, that is, a new random forest model with efficient classification logic is generated to guide the inspection filtering.

[0075] Optionally, the number of random classification tree models is 10. In the embodiment of the present application, compared with the 45% accuracy and recall rate of the prior art production line, the detection accuracy of the detection equipment can reach 81% and the recall rate can reach 77% by using the generated 10 random classification tree models for detection.

[0076] Step 203: Evaluate the accuracy of the candidate defect recognition model based on a preset function.

[0077] The preset function may be an evaluation function for evaluating the accuracy of the candidate defect recognition model. Exemplarily, the score function in the sklearn library may be used as the evaluation function. The score function is used to evaluate the accuracy of the model. When the input of the candidate defect recognition model is the feature data and labels of the test set, the score function returns the accuracy of the candidate defect recognition model, and the best score is 1. Optionally, when the value returned by the score function is 1 or the difference with 1 is less than a preset difference threshold, it is determined that the evaluation result is that the accuracy of the candidate defect recognition model meets the preset accuracy condition.

[0078] Step 204 , when the evaluation result indicates that the accuracy of the candidate defect recognition model meets the preset accuracy condition, the candidate defect recognition model is used as the preset defect recognition model.

[0079] In the embodiment of the present application, the deep learning algorithm based on the random forest algorithm is trained by extracting the feature data of the historical output of the automatic detection equipment, and an efficient random classification tree model is produced, and a preset defect recognition model including multiple random classification tree models is obtained. In this way, the preset defect recognition model is input into the detection equipment, making it easier for the detection equipment to recognize and run the preset defect recognition model, and then using the preset defect recognition model to improve the accuracy of the defect detection results finally output by the detection equipment.

[0080] In one embodiment, Figure 3 A schematic diagram of a process for obtaining a sample data set provided in an embodiment of the present application is shown. Obtaining a sample data set includes:

[0081] Step 301, obtaining sample defect detection results output by a detection device for performing defect detection on multiple sample display panels. The sample defect detection results include sample feature data of multiple sample defects of the sample display panel; the sample feature data includes data values ​​corresponding to multiple candidate features.

[0082] Optionally, the detection device may perform defect detection on multiple sample display panels of different batches for multiple times, and combine the results of all defect detections to obtain the sample defect detection result.

[0083] The sample defect detection result includes a plurality of sample feature data, each sample feature data corresponds to a different sample defect, and each sample defect may belong to a different sample display panel.

[0084] For each sample feature data, the sample feature data includes data values ​​corresponding to multiple candidate features, wherein the candidate features are features of sample defects. For example, there are 200 candidate features in total, and the sample feature data of each sample defect includes data values ​​corresponding to the 200 candidate features.

[0085] Step 302, based on the data values ​​corresponding to the candidate features of each sample defect, each candidate feature is screened to obtain the first target feature of each sample defect, and the data value corresponding to the second target feature of each sample defect is obtained. The data value corresponding to the second target feature is used to characterize the position distribution of the sample defect during the defect detection process.

[0086] Since there are a large number of candidate features, and only some of them are helpful in determining whether a sample defect is a real defect, it is necessary to screen all candidate features to select the first target feature and the second target feature.

[0087] Among them, according to the data values ​​corresponding to each candidate feature of each sample defect, it can be determined whether the candidate features are associated with each other, and the associated candidate features can be determined as the first target features, and the data values ​​of the first target features can be determined. Secondly, there are some other features that are helpful in judging whether the sample defects are real defects, but these features are not candidate features, so it is necessary to additionally determine these second target features and obtain the data values ​​of these second target features corresponding to each sample defect.

[0088] Step 303 : determining a sample data set according to the data value corresponding to the first target feature and the data value corresponding to the second target feature of each sample defect.

[0089] The data value corresponding to the first target feature and the data value corresponding to the second target feature of each sample defect may together constitute a sample data set.

[0090] In an embodiment of the present application, the first target feature is determined by screening the candidate features, and the second target feature is determined. The sum of the number of the first target features and the second target features is less than the number of candidate features, which is equivalent to dimensionality reduction of the candidate features. The first target feature and the second target feature help to determine whether the sample defect is a real defect. In this way, the training process of the preset defect recognition model can be made more efficient and accurate.

[0091] In one embodiment, Figure 4 A schematic diagram of a process for determining a first target feature provided by an embodiment of the present application is shown. According to the data values ​​corresponding to each candidate feature of each sample defect, each candidate feature is screened to obtain the first target feature of each sample defect, including:

[0092] Step 401 : determining a plurality of first correlation features from candidate features according to the defect labels of the sample defects.

[0093] Among them, the sample feature data includes the defect label of the sample defect; the defect label is used to characterize whether the sample defect is a real defect.

[0094] Optionally, each sample defect can be manually analyzed in advance to determine whether each sample defect is a real defect, thereby determining a defect label for each sample defect. The defect label for each sample defect is input into the detection device, so that the detection device can obtain the defect label for each sample defect.

[0095] In one embodiment, Figure 5 A schematic diagram of a process for determining a first correlation feature provided by an embodiment of the present application is shown. According to the defect labels of each sample defect, multiple first correlation features are determined from each candidate feature, including:

[0096] Step 501 : for each first sample defect whose defect label is the first label, determine the first data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each first sample defect.

[0097] Optionally, the defect label may include a first label and a second label. For example, the first label may indicate that the sample defect is a real defect, and the second label may indicate that the sample defect is a false defect.

[0098] Optionally, the defect label may include a first label, a second label and a third label. For example, the first label may indicate that the sample defect is a real defect, the second label may indicate that the sample defect is a false defect, and the third label may indicate that it is impossible to confirm whether the sample defect is a real defect.

[0099] For ease of understanding, the embodiment of the present application is described by taking the example that a defective label includes a first label and a second label.

[0100] Among them, the sample defect whose defect label is the first label can be used as the first sample defect. Here, the data value of each candidate feature of each first sample defect can be obtained for statistics to form a first data distribution corresponding to each candidate feature. Optionally, the first data distribution corresponding to each candidate feature can be a bar distribution graph, and of course it can also be other forms of distribution graphs, which are not specifically limited here. For example, for the first candidate feature, the data values ​​of the first candidate features of all first sample defects are selected for statistics to form a first bar distribution graph of the first candidate feature as the first data distribution.

[0101] Step 502 : for each second sample defect whose defect label is the second label, determine the second data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each second sample defect.

[0102] The sample defect whose defect label is the second label can be used as the second sample defect. Here, the data value of each candidate feature of each second sample defect can be obtained for statistics to form a second data distribution corresponding to each candidate feature. Optionally, the second data distribution corresponding to each candidate feature can be a bar distribution graph, and of course it can also be a distribution graph in other forms, which is not specifically limited here. For ease of comparison, the statistical form of the first data distribution and the second data distribution is consistent. For example, for the above-mentioned first candidate feature, the data values ​​of the first candidate feature of all second sample defects are selected for statistics to form a second bar distribution graph of the first candidate feature as the second data distribution. Exemplary, such as Figure 6 A schematic diagram showing a comparison of a first data distribution and a second data distribution of a first candidate feature.

[0103] Optionally, each pixel unit in the display panel is connected to a different gate line and data line, and works under the action of the signals output by the gate line and the data line. Each gate line and data line has a serial number identifier. The same sample defect may correspond to multiple pixel units, that is, multiple pixel units located together have faults, forming a certain sample defect. Each candidate feature may include the serial number identifier of the starting gate line connected to the pixel unit corresponding to the sample defect, the serial number identifier of the ending gate line connected to the pixel unit corresponding to the sample defect, the serial number identifier of the starting data line connected to the pixel unit corresponding to the sample defect, and the serial number identifier of the ending data line connected to the pixel unit corresponding to the sample defect. Among them, the starting gate line is the gate line with the earliest serial number among the gate lines connected to the multiple pixel units corresponding to the sample defect. Similarly, the ending gate line is the gate line with the latest serial number among the gate lines connected to the multiple pixel units corresponding to the sample defect. The definition of the data line is similar and will not be repeated.

[0104] Since each gate line and data line has a serial number identifier and the serial number identifier is continuous, for example, from 1 to 150. For candidate features with similar data values ​​having continuous characteristics, when counting the first data distribution and the second data distribution, a segmented statistical method, such as 0-55, 56-82, 83-110, ..., can be adopted for data processing, and then the discrete data is divided into statistical distributions within the data segments for subsequent difference comparison.

[0105] Step 503: for each candidate feature, a first data distribution and a second data distribution corresponding to the candidate feature are compared to obtain a comparison result.

[0106] Step 504: Determine a plurality of first correlation features from the candidate features according to the comparison results of the candidate features.

[0107] Among them, for each candidate feature, if the first data distribution and the second data distribution corresponding to the candidate feature are significantly different, it means that the candidate feature can help determine whether the sample defect is a real defect, and then the candidate feature can be determined as the first correlation feature.

[0108] For example, refer to Figure 6 , the first data distribution of the first candidate feature and the second data distribution have obvious differences, and the first candidate feature can be used as the first correlation feature. The difference comparison method is not specifically limited here, as long as it can be determined that the first data distribution and the second data distribution have obvious differences.

[0109] Step 402: Determine a plurality of second correlation features from among the candidate features other than the first correlation features.

[0110] As mentioned above, each first correlation feature is determined from each candidate feature. The remaining candidate features except each first correlation feature among all candidate features can be used as other candidate features, and multiple second correlation features can be further determined from all other candidate features.

[0111] Optionally, the correlation between other candidate features may be calculated based on the data values ​​corresponding to other candidate features of each sample defect, and a plurality of second correlation features may be determined according to the correlation calculation results.

[0112] Please refer to Figure 7 , which shows a schematic diagram of a process for determining a second correlation feature provided by an embodiment of the present application. Determining a plurality of second correlation features from other candidate features in each candidate feature except each first correlation feature includes:

[0113] Step 701 , calculating the correlation between each other candidate feature according to the data value of each other candidate feature of each sample defect, and obtaining a plurality of correlation calculation values.

[0114] The correlations between other candidate features may be calculated pairwise.

[0115] Optionally, the correlation between every two other candidate features is calculated based on a preset correlation formula.

[0116] For example, any two other candidate features can be taken as a group, and the correlation between the two other candidate features in each group is calculated based on a preset correlation formula. Optionally, the preset correlation formula is a formula for calculating a simple correlation coefficient, that is, a formula for a Pearson correlation coefficient or a linear correlation coefficient formula, and the definition formula is:

[0117]

[0118] r is the calculated correlation value, and the absolute value of r is between 0 and 1. Generally, the closer r is to 1, the stronger the correlation between X and Y. Conversely, the closer r is to 0, the weaker the correlation between X and Y. i Y is all the data values ​​corresponding to the first other candidate feature in a set of other candidate features. i is all the data values ​​corresponding to the second other candidate feature in a set of other candidate features. n is the number of data values, that is, the number of sample defects.

[0119] Step 702: Determine a target correlation calculation value that is greater than a preset correlation threshold from among the correlation calculation values.

[0120] The preset correlation threshold may be predetermined. For example, Table 1 shows the corresponding relationship between the absolute value of r and the correlation degree. The preset correlation threshold may be determined according to Table 1. For example, the preset correlation threshold is 0.6.

[0121] Table 1

[0122]

[0123] Step 703: other candidate features to which the data values ​​used to calculate the target correlation calculation value belong are used as second correlation features.

[0124] For example, if the correlation calculation value calculated based on the data values ​​corresponding to two other candidate features of a group is greater than a preset correlation threshold, then the two other candidate features of the group can be used as the second correlation features.

[0125] Step 403: Use the first correlation feature and the second correlation feature as the first target feature.

[0126] In the embodiment of the present application, based on the association between the candidate features, the first target feature can be determined efficiently and accurately, and feature dimensionality reduction can be achieved, so as to help improve the efficiency and accuracy of subsequent model training.

[0127] In one embodiment, obtaining the data value corresponding to the second target feature of each sample defect includes:

[0128] Determine the specification data of the display panel to which each sample defect belongs during the defect detection process; and determine the data value corresponding to the second target feature of each sample defect based on the specification data corresponding to each sample defect and a preset relationship conversion table.

[0129] The specification data of the display panel corresponding to each sample defect is generated during the detection process. For example, the specification data of the display panel corresponding to a sample defect is L2E3428B5032ET. The specification data represents the position distribution of the display panel during the detection process, and therefore can also represent the position distribution of the sample defect during the detection process.

[0130] like Figure 8 A schematic diagram of the arrangement of multiple display panels is shown, and AA, AB, EU, etc. are display panels. During the defect detection process, multiple display panels are arranged on a glass substrate, and the glass substrate may include a G1 region and a G2 region. Taking the specification data L2E3428B5032ET as an example, the third-to-last digit is 2, which indicates that the display panel is located in the G2 region. The second-to-last digit is E. According to the preset relationship conversion table, E indicates that the display panel is in the 4th row; the first-to-last digit is T. According to the preset relationship conversion table, T indicates that the display panel is in the 17th column. Among them, the preset relationship conversion table includes a plurality of correspondences between letters and numbers, exemplarily, as shown in Table 2:

[0131] Table 2

[0132]

[0133] The second target feature may include, for example, the area of ​​the display panel in the glass substrate to which the sample defect belongs during the defect detection process, the row of the display panel, and the column of the display panel. Therefore, based on the above specification data, the data value corresponding to each second target feature of the sample defect may be determined.

[0134] In the embodiment of the present application, the basic electrical data generated by the detection equipment is used to obtain grayscale data with high repeatability and accuracy that can be quantified in the device dimension, which is different from the data generated by the classification and calculation of the image, and the first-hand device identification data is obtained. In this way, by extracting the data values ​​of the first target feature and the second target feature, the constructed deep learning algorithm is trained by model simulation, so that the preset defect recognition model with more scientific classification logic and better classification effect can be obtained by multiple optimization fitting. Subsequently, the detection equipment can be used to directly screen real defects, save labor costs, and improve detection efficiency. Here, since the detection chamber of the detection equipment is a high-level dust-free environment, it will not affect the product yield. Therefore, the preset defect recognition model is mounted on the detection equipment, which saves costs and is highly effective, and reduces the detection cycle of display panel defect detection.

[0135] Based on the above description, the construction process of the preset defect recognition model is as follows: Fig. 9 shown.

[0136] In one embodiment, a preset defect recognition model determines real defects from each initial defect based on feature data of each initial defect, including: determining target feature data of each initial defect from the feature data of each initial defect; determining whether each initial defect is a real defect based on the target feature data of each initial defect; wherein each target feature data is associated with each other.

[0137] The feature data of each initial defect includes the data value of each candidate feature of each initial defect, and the target feature data is the data value corresponding to the first target feature of each initial defect and the data value corresponding to the second target feature of each initial defect.

[0138] The preset defect recognition model identifies the data value corresponding to the first target feature of each initial defect and the data value corresponding to the second target feature of each initial defect based on the random forest algorithm, thereby determining whether each initial defect is a real defect. The feature data of each real defect is used as the final defect detection result.

[0139] Optionally, the final defect detection result may include data values ​​of each candidate feature of each real defect.

[0140] Based on the above description, the process of identifying whether each initial defect is a real defect is as follows: Fig.10 shown.

[0141] In the embodiment of the present application, the electrical data generated by the defect detection by the detection equipment is used as the basic data. Through data processing, a large amount of feature data is reduced in dimension and optimized, and more effective feature data is extracted. Then, a deep learning algorithm is used for model training, and finally a deep learning model is obtained. The deep learning algorithm is implemented based on the random forest algorithm, so that the model logic can be visualized through visualization means such as grphviz, so as to achieve the effects of more scientific inspection classification of detection equipment, more efficient classification effect, and less manpower investment.

[0142] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0143] Based on the same inventive concept, the embodiment of the present application also provides a defect detection device for implementing the defect detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more defect detection device embodiments provided below can refer to the limitations on the defect detection method above, and will not be repeated here.

[0144] In one embodiment, Fig.11 As shown, a defect detection device is provided, and the defect detection device 1100 includes: an acquisition module 1101, a determination module 1102 and a result module 1103, wherein:

[0145] An acquisition module 1101 is used to acquire initial defect detection results obtained by a detection device performing defect detection on multiple display panels; the initial defect detection results include feature data of initial defects of the multiple display panels;

[0146] The determination module 1102 is used to input the characteristic data of each initial defect into a preset defect recognition model to obtain the characteristic data of each real defect output by the preset defect recognition model; wherein the preset defect recognition model determines the real defect from each initial defect according to the characteristic data of each initial defect;

[0147] The result module 1103 is used to use the characteristic data of each real defect as the final defect detection result.

[0148] In one embodiment, the determination module 1102 is specifically used to: determine target feature data of each initial defect from the feature data of each initial defect; determine whether each initial defect is a real defect based on the target feature data of each initial defect; wherein each target feature data is associated with each other.

[0149] In one embodiment, the device also includes a construction module, which is used to: obtain a sample data set; train an initial defect recognition model based on the sample data set to obtain a candidate defect recognition model; evaluate the accuracy of the candidate defect recognition model based on a preset function; when the evaluation result indicates that the accuracy of the candidate defect recognition model meets the preset accuracy condition, use the candidate defect recognition model as the preset defect recognition model.

[0150] In one embodiment, the building module is specifically used to:

[0151] Obtain sample defect detection results output by a detection device for defect detection of multiple sample display panels; the sample defect detection results include sample feature data of multiple sample defects of the sample display panel; the sample feature data include data values ​​corresponding to multiple candidate features; according to the data values ​​corresponding to each candidate feature of each sample defect, screen each candidate feature to obtain a first target feature of each sample defect, and obtain a data value corresponding to a second target feature of each sample defect, wherein the data value corresponding to the second target feature is used to characterize the position distribution of the sample defect in the defect detection process; determine a sample data set according to the data value corresponding to the first target feature and the data value corresponding to the second target feature of each sample defect.

[0152] In one embodiment, the sample feature data includes a defect label of the sample defect; the defect label is used to characterize whether the sample defect is a real defect; a construction module is specifically used to: determine a plurality of first correlation features from each candidate feature according to the defect label of each sample defect; determine a plurality of second correlation features from other candidate features in each candidate feature except each first correlation feature; and use the first correlation feature and the second correlation feature as the first target feature.

[0153] In one embodiment, a construction module is specifically used to: for each first sample defect whose defect label is a first label, determine the first data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each first sample defect; for each second sample defect whose defect label is a second label, determine the second data distribution corresponding to each candidate feature according to the data value corresponding to each candidate feature of each second sample defect; for each candidate feature, perform a difference comparison between the first data distribution corresponding to the candidate feature and the second data distribution to obtain a comparison result; and determine multiple first correlation features from each candidate feature according to the comparison result of each candidate feature.

[0154] In one embodiment, a construction module is specifically used to: calculate the correlation between each other candidate feature based on the data value of each other candidate feature of each sample defect to obtain multiple correlation calculation values; determine a target correlation calculation value that is greater than a preset correlation threshold from each correlation calculation value; and use other candidate features to which the data value used to calculate the target correlation calculation value belongs as the second correlation feature.

[0155] In one embodiment, the initial defect recognition model is constructed based on a random forest algorithm, and the initial defect recognition model includes a plurality of random classification tree models.

[0156] Each module in the above defect detection device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0157] In one embodiment, an electronic device is provided. The electronic device may be a server, and its internal structure diagram may be as follows: Fig.12 As shown. The electronic device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the electronic device is used to store defect detection data. The input / output interface of the electronic device is used to exchange information between the processor and an external device. The communication interface of the electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a defect detection method is implemented.

[0158] Those skilled in the art will understand that Fig.12 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0159] In one embodiment, an electronic device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0160] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0161] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0162] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0163] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0164] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A defect detection method, characterized in that: The method comprises: Acquire initial defect detection results obtained by performing defect detection on multiple display panels by a detection device; the initial defect detection results include feature data of initial defects of the multiple display panels; Inputting the characteristic data of each of the initial defects into a preset defect recognition model to obtain the characteristic data of each of the real defects output by the preset defect recognition model; wherein the preset defect recognition model determines the real defect from each of the initial defects according to the characteristic data of each of the initial defects; The characteristic data of each of the real defects is used as the final defect detection result.

2. The method according to claim 1, characterized in that The preset defect recognition model determines the real defect from each of the initial defects according to the characteristic data of each of the initial defects, including: Determining target feature data of each of the initial defects from the feature data of each of the initial defects; Determining whether each of the initial defects is a real defect according to the target feature data of each of the initial defects; Among them, each of the target feature data is associated with each other.

3. The method according to claim 1, characterized in that The construction process of the preset defect recognition model includes: Get a sample dataset; According to the sample data set, an initial defect recognition model is trained to obtain a candidate defect recognition model; Based on a preset function, evaluating the accuracy of the candidate defect recognition model; When the evaluation result indicates that the accuracy of the candidate defect recognition model meets the preset accuracy condition, the candidate defect recognition model is used as the preset defect recognition model.

4. The method according to claim 3, characterized in that The obtaining of the sample data set comprises: Acquire sample defect detection results output by a detection device for performing defect detection on a plurality of sample display panels; the sample defect detection results include sample feature data of a plurality of sample defects of the sample display panels; the sample feature data include data values ​​corresponding to a plurality of candidate features; According to the data values ​​corresponding to the candidate features of the sample defects, the candidate features are screened to obtain the first target features of the sample defects, and the data values ​​corresponding to the second target features of the sample defects are obtained, wherein the data values ​​corresponding to the second target features are used to characterize the position distribution of the sample defects in the defect detection process; The sample data set is determined according to the data value corresponding to the first target feature and the data value corresponding to the second target feature of each of the sample defects.

5. The method according to claim 4, characterized in that The sample feature data includes a defect label of the sample defect; the defect label is used to indicate whether the sample defect is a real defect; and the candidate features are screened according to the data values ​​corresponding to the candidate features of the sample defects to obtain the first target features of the sample defects, including: Determining a plurality of first correlation features from the candidate features according to the defect labels of the sample defects; Determine a plurality of second correlation features from other candidate features among the candidate features except the first correlation features; The first correlation feature and the second correlation feature are used as the first target feature.

6. The method according to claim 5, characterized in that The determining of a plurality of first correlation features from the candidate features according to the defect labels of the sample defects includes: For each first sample defect whose defect label is the first label, determining, according to the data value corresponding to each candidate feature of each first sample defect, a first data distribution corresponding to each candidate feature; For each second sample defect whose defect label is the second label, determining, according to the data value corresponding to each candidate feature of each second sample defect, a second data distribution corresponding to each candidate feature; For each of the candidate features, performing a difference comparison between the first data distribution and the second data distribution corresponding to the candidate feature to obtain a comparison result; According to the comparison results of the candidate features, a plurality of first correlation features are determined from the candidate features.

7. The method according to claim 5, characterized in that The determining of a plurality of second correlation features from other candidate features among the candidate features except the first correlation features comprises: Calculating the correlation between the other candidate features according to the data values ​​of the other candidate features of the sample defects to obtain a plurality of correlation calculation values; Determine a target correlation calculation value greater than a preset correlation threshold value from each of the correlation calculation values; Other candidate features to which the data values ​​used to calculate the target correlation calculation value belong are used as the second correlation features.

8. The method according to claim 3, characterized in that The initial defect recognition model is constructed based on a random forest algorithm, and the initial defect recognition model includes a plurality of random classification tree models.

9. A defect detection device, characterized in that: The device comprises: An acquisition module, used to acquire initial defect detection results obtained by a detection device performing defect detection on a plurality of display panels; the initial defect detection results include characteristic data of initial defects of the plurality of display panels; A determination module, used for inputting the characteristic data of each of the initial defects into a preset defect recognition model to obtain the characteristic data of each of the real defects output by the preset defect recognition model; wherein the preset defect recognition model determines the real defect from each of the initial defects according to the characteristic data of each of the initial defects; The result module is used to use the characteristic data of each real defect as the final defect detection result.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images