A method, device, equipment and storage medium for repairing a flexible panel

Abstract

The present disclosure relates to a method, device, computer equipment and storage medium for repairing a flexible panel, the method comprising: obtaining first image information of a flexible panel to be repaired; determining a first classification label of the flexible panel to be repaired according to a deep learning model and the first image information; determining working parameters of a repair device according to the first classification label, and repairing the flexible panel to be repaired according to the working parameters. In this way, the repair rate and efficiency of the flexible panel can be improved, thereby improving the production yield and production efficiency of the flexible panel.

Description

Technical Field

[0001] The present disclosure relates to the field of display panel production, and in particular to a repair method, device, equipment and storage medium for flexible panels. Background Art

[0002] Although flexible panels have been mass-produced at present, a large number of defective flexible panels such as Sandy mura and black masses are difficult to improve in a short period of time and need to be repaired by high-temperature baking. In the related art, whether to use high-temperature baking to repair flexible panels is still at the level of manual experience, that is, whether to repair, the repair effect, and the repair process all require manual testing. Alternatively, some production lines use a fixed 8-hour baking time to repair all flexible panels to be repaired, which is obviously not conducive to the repair rate and repair efficiency of flexible panels, thereby affecting the generation yield and production efficiency of flexible panels. Based on this, how to improve the production yield of flexible panels and improve the production efficiency of flexible panels has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The purpose of the present disclosure is to provide a method, device, equipment and storage medium for repairing a flexible panel, aiming to improve the production yield of the flexible panel and improve the production efficiency of the flexible panel.

[0004] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0005] According to a first aspect of an embodiment of the present disclosure, a method for repairing a flexible panel is provided, comprising:

[0006] Acquire first image information of the flexible panel to be repaired;

[0007] Determine a first classification label of the flexible panel to be repaired according to the deep learning model and the first image information;

[0008] The working parameters of the repair equipment are determined according to the first classification label, and the flexible panel to be repaired is repaired according to the working parameters.

[0009] Optionally, a deep learning model is determined according to a predetermined number of samples of the flexible panel to be repaired, which includes:

[0010] Dividing the predetermined number of samples of the flexible panel to be repaired into training samples and verification samples according to a preset ratio;

[0011] Training the training samples to obtain a first model;

[0012] Inputting the verification sample into the first model to obtain a first prediction accuracy rate of the first model;

[0013] Determine the deep learning model according to the first prediction accuracy of the first model.

[0014] Optionally, the training the training samples to obtain a first model includes:

[0015] Acquire second image information of the flexible panel in the training sample;

[0016] Acquire third image information of the flexible panel in the training sample obtained after the flexible panel in the training sample is repaired using an nth working parameter, where n is greater than or equal to 1;

[0017] Determining whether the third image information meets a restoration standard;

[0018] If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard;

[0019] Mapping the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample;

[0020] In this way, second image information and second classification labels of all flexible panels in the training sample are obtained to form a second image information set and a second classification label set;

[0021] The second image information set and the second classification label set are used as training sample data.

[0022] Optionally, the training the training samples to obtain the first model further comprises:

[0023] If the repair standard is not met, the test flexible panel is repaired using the n+1th working parameter until the test flexible panel is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair using the mth working parameter meets the repair standard;

[0024] If the fifth image does not meet the restoration standard, a preset classification label is used as a second classification label for the flexible panel in the training sample; wherein n+1 is less than m.

[0025] Optionally, different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

[0026] Optionally, determining the deep learning model according to the first prediction accuracy of the first model includes:

[0027] If the first prediction accuracy is less than the preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy;

[0028] The first model corresponding to the first prediction accuracy being greater than or equal to the preset accuracy is used as the deep learning model.

[0029] Optionally, determining the deep learning model according to the first prediction accuracy of the first model includes:

[0030] If the first prediction accuracy is less than the preset accuracy, determining a second prediction accuracy of an alternative second model;

[0031] The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

[0032] Optionally, inputting the verification sample into the first model to obtain a first prediction accuracy of the first model includes:

[0033] Obtaining a third classification label set of the verification sample;

[0034] The verification sample is input into the first model to obtain a set of prediction results of the sample to be verified;

[0035] The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

[0036] Optionally, repairing the flexible panel to be repaired according to the working parameters includes:

[0037] If there are multiple repair devices, according to the first target working parameter corresponding to the first classification tag, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair;

[0038] or

[0039] If there is a repair device, the working parameters of the repair device are adjusted to the second target working parameters corresponding to the first classification label, and the flexible panel to be repaired is repaired with the second target working parameters.

[0040] According to a second aspect of an embodiment of the present disclosure, a repair device for a flexible panel is provided, comprising:

[0041] A first acquisition module, used to acquire first image information of the flexible panel to be repaired;

[0042] A determination module, configured to determine a first classification label of the flexible panel to be repaired according to a deep learning model and the first image information;

[0043] A repair module is used to determine working parameters of a repair device according to the first classification label, and repair the flexible panel to be repaired according to the working parameters.

[0044] Optionally, the device further comprises:

[0045] The second determination module is used to determine the deep learning model according to a predetermined number of flexible panel samples to be repaired.

[0046] The second determining module further includes:

[0047] A division submodule, used for dividing a predetermined number of samples of the flexible panel to be tested into training samples and samples to be tested according to a preset ratio;

[0048] A first processing submodule, used for training the training samples to obtain a first model;

[0049] A second processing submodule, configured to input the verification sample into the first model to obtain a first prediction accuracy rate of the first model;

[0050] The third processing submodule is used to determine the deep learning model according to the first prediction accuracy of the first model.

[0051] Optionally, the first processing submodule is further used to:

[0052] Acquire second image information of the flexible panel in the training sample;

[0053] Acquire third image information of the flexible panel in the training sample obtained after the flexible panel in the training sample is repaired using an nth working parameter, where n is greater than or equal to 1;

[0054] Determining whether the third image information meets a restoration standard;

[0055] If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard;

[0056] Mapping the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample;

[0057] In this way, second image information and second classification labels of all flexible panels in the training sample are obtained to form a second image information set and a second classification label set;

[0058] The second image information set and the second classification label set are used as training sample data.

[0059] Optionally, the first processing submodule is further used to:

[0060] If the repair standard is not met, the test flexible panel is repaired using the n+1th working parameter until the test flexible panel is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair using the mth working parameter meets the repair standard;

[0061] If the fifth image does not meet the restoration standard, a preset classification label is used as a second classification label for the flexible panel in the training sample, wherein n+1 is less than m.

[0062] Optionally, different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

[0063] Optionally, the third processing submodule is further used to:

[0064] If the first prediction accuracy is less than the preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy;

[0065] The first model corresponding to the first prediction accuracy being greater than or equal to the preset accuracy is used as the deep learning model.

[0066] Optionally, the third processing submodule is further used to:

[0067] If the first prediction accuracy rate is less than the preset accuracy rate, determining a second prediction accuracy rate of an alternative second model;

[0068] The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

[0069] Optionally, the second processing submodule is further used to:

[0070] Obtaining a third classification label set of the verification sample;

[0071] Inputting the verification sample into the first model to obtain a set of prediction results of the verification sample;

[0072] The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

[0073] Optionally, the repair module is further used to:

[0074] If there are multiple repair devices, according to the first target working parameter corresponding to the first classification label, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair;

[0075] or

[0076] If there is a repair device, the working parameter of the repair device is adjusted to the second target working parameter according to the second working parameter corresponding to the first classification label, and the flexible panel to be repaired is repaired with the second target working parameter.

[0077] According to a third aspect of an embodiment of the present disclosure, a computer device is provided, including:

[0078] A processor and a memory for storing a computer program that can be run on the processor, wherein the processor is used to implement any of the above-mentioned steps of the method for repairing the flexible panel when running the computer program.

[0079] According to a fourth aspect of an embodiment of the present disclosure, there is provided a non-temporary computer-readable storage medium on which a computer program is stored, and the program is executed by a processor to implement any of the above-mentioned steps of the method for repairing a flexible panel.

[0080] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0081] In the embodiment of the present disclosure, the first image information of the flexible panel to be repaired is obtained; and the first classification label of the panel to be repaired is determined according to the deep learning model and the first image information; the working parameters of the repair equipment are determined according to the determined first classification label, and the flexible panel to be repaired is repaired according to the working parameters. That is to say, the embodiment of the present disclosure improves the intelligence in the repair process because the flexible panel to be repaired is intelligently classified by the deep learning model. In this way, since the flexible panel to be repaired can be intelligently classified without the need for classification by manual experience, the phenomenon of classification errors caused by manual participation can be reduced, and the requirements for operator experience can be reduced, thereby improving the operability of the repair process in the production process. In addition, since the flexible panel to be repaired is intelligently classified according to the deep learning model, and the working parameters of the repair equipment are determined based on the classification label, and the flexible panel to be repaired is repaired based on the working parameters, it is not necessary to use a unified repair scheme for all the flexible panels to be repaired, thereby reducing the phenomenon of unsuccessful repair or inadequate repair caused by the use of a unified repair scheme. In this way, the repair rate of the flexible panel is improved, and the production yield of the flexible panel is improved. At the same time, there is no need to manually perform repeated repair tests on the panel to be repaired to repair the flexible panel to be repaired. This improves the repair efficiency of the flexible panel and further improves the production efficiency of the flexible panel.

[0082] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0084] Figure 1 is a flow chart of a method for repairing a flexible panel according to an exemplary embodiment;

[0085] Figure 2 is another flow chart of a method for repairing a flexible panel according to an exemplary embodiment;

[0086] Figure 3 is a scene diagram of a method for repairing a flexible panel according to an exemplary embodiment;

[0087] Figure 4 is a functional module diagram of a flexible panel repair device according to an exemplary embodiment;

[0088] Figure 5 The invention is a block diagram of a flexible panel repairing device according to an exemplary embodiment. DETAILED DESCRIPTION

[0089] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0090] It is understandable that after the flexible panel is manufactured, some of the flexible panels with defects will need to be repaired. For example, by high-temperature baking, the optical defects such as black masses caused by light, Sandy mura, excessive brightness or darkness caused by static electricity, and uneven color or brightness distribution of the flexible panel can be repaired. In the related art, the defects of the flexible panel are judged by the human eye, that is, the type of defect and the severity of the defect are judged by the human eye, and then the defect is judged according to manual experience whether it is the above-mentioned repairable defect, and the working parameters of the repair equipment are required to repair the defect, such as the baking time of the baking oven. Based on this, due to the limited observation ability of the human eye, it is obvious that the type of defect and the severity of the defect judged manually are not accurate. Repairing the defect based on this inaccurate judgment result is obviously beneficial to the repair rate of the flexible panel, thereby affecting the production yield of the entire production process. In addition, if all defects are repaired using the working parameters of the maximum degree of defect that can be repaired, it is obviously not conducive to the repair efficiency, thereby affecting the production efficiency of the entire production process.

[0091] Based on this, the embodiments of the present disclosure provide a method for repairing a flexible panel, which is applied to the field of manufacturing of flexible panels, and in particular, to a repair system corresponding to the repair process of flexible panels. Exemplarily, the method for repairing a flexible panel is applied in a production workstation of a flexible panel, which includes a computer device integrated with a computing processing module, which can be used to process image information, and output the processing result of the image information, and can intelligently repair the flexible panel according to the processing result of the image information. The method for repairing a flexible panel provided in the embodiments of the present disclosure aims to improve the repair rate and efficiency of the flexible panel, thereby improving the production yield and production efficiency.

[0092] Figure 1 is a flow chart of a method for repairing a flexible panel according to an exemplary embodiment. Figure 1 As shown, the method comprises the following steps:

[0093] Step 101: Acquire first image information of a flexible panel to be repaired.

[0094] It should be added that after the production of the flexible panel is completed, optical defects such as Sandy mura and / or black masses will still appear, and these defects need to enter the next production process, that is, the repair process for repair. And these defects can be presented on the surface of the flexible panel. Here, the purpose of obtaining the image information of the flexible panel to be repaired is to determine the defect characteristics of the flexible panel to be repaired based on the image information, where the defect characteristics include at least one of the following: defect type and defect area. Here, the repair working parameters of the repair equipment required for different defect types are different, and the repair working parameters of the repair equipment required for different defect areas are also different. Among them, the defect type can reflect what kind of defect appears on the flexible panel, and the defect area can reflect the severity of the defect of the flexible panel. In this way, in this embodiment, the defect characteristics of the flexible panel to be repaired are determined by the image information of the flexible panel to be repaired, which is more delicate and more accurate than the human eye judging the defect characteristics of the flexible panel to be repaired.

[0095] In some embodiments, obtaining first image information of the flexible panel to be repaired may include:

[0096] The first image information of the flexible panel to be repaired is obtained through the image acquisition module installed on the repair device itself.

[0097] It is understandable that industrial cameras have high image stability, high transmission capacity, high anti-interference ability, etc., and are usually installed on machine assembly lines. Based on this, in some embodiments, the image acquisition module may include: an industrial camera. In this way, the first image information of the flexible panel acquired can have higher stability, which can meet the requirements for the captured image information in industrial production.

[0098] Furthermore, the industrial camera may be a CCD camera or a CMOS camera, without any limitation herein.

[0099] In some other embodiments, obtaining the first image information of the flexible panel to be repaired may also include:

[0100] The first image information of the flexible panel to be repaired is obtained from a device other than the repairing device.

[0101] Here, obtaining the first image information of the flexible panel to be repaired from a device other than the repair device may include: obtaining the image information of all flexible panels from the production equipment corresponding to the previous production process of the repair as the first image information of the flexible panel to be repaired. In this embodiment, by obtaining the image information of all flexible panels from the production equipment corresponding to the previous production process of the repair as the first image information of the flexible panel to be repaired, the omission of defective flexible panels due to manual selection can be reduced. Moreover, by placing the collection of the first image information in the previous production process of the repair, this embodiment can make the entire production and repair process more intelligent, reduce the manual participation from production to repair, and reduce labor costs.

[0102] It should be added that the production equipment corresponding to the previous production process being repaired may be assembly equipment in the assembly process of the flexible panel.

[0103] Of course, in other embodiments, obtaining the first image information of the flexible panel to be repaired from a device other than the repair device may also include: obtaining the first image information of the flexible panel to be repaired from a terminal device connected to a network with the repair device.

[0104] The terminal devices here may include: mobile terminal devices or fixed terminal devices, wherein the mobile terminal devices may include: any one of mobile phones, tablet computers, laptop computers and wearable devices; the fixed terminal devices may include: any one of desktop computers, all-in-one computers and servers. The server may be an independent server or a server cluster composed of multiple servers, such as the cloud.

[0105] Here, the networking mode between the terminal and the repair device may include but is not limited to: local area network connection, or connection through a base station, etc. Among them, the local area network connection may include but is not limited to: Bluetooth, wifi or Zigbbe connection.

[0106] In some embodiments, the step of acquiring first image information of the flexible panel to be repaired from a terminal device connected to a network with the repair device includes:

[0107] The processed first image information of the flexible panel to be repaired is obtained from a terminal device connected to a network with the repair device. Here, the processed first image information may include: grayscale processing of the original image of the flexible panel to be repaired, etc., so that the defect features contained in the first image information are more obvious, and some unnecessary image information is reduced from being input into the repair device, thereby reducing the amount of information processing of the repair device and improving the information processing efficiency.

[0108] In this embodiment, by acquiring the first image information of the flexible panel to be repaired through a terminal device connected to a network with the repair device, there is no need to install an additional image acquisition module in the production line, thus saving the cost of modifying the production line and improving the versatility of the production line.

[0109] Step 102: Determine a first classification label of the flexible panel to be repaired based on the deep learning model and the first image information.

[0110] Here, the deep learning model refers to the inherent rules and representation levels of the learning sample data, and the interpretation of the information obtained in the learning process, such as text, images, and sounds. The deep learning model can identify the defect features of the flexible panel in the image information, and obtain the first classification label of the flexible panel corresponding to the image information based on the defect features.

[0111] Here, the first classification label can be represented by characters, for example, letters, numbers, or a combination of letters and numbers. In short, different first classification labels correspond to different characteristics of the flexible panel to be repaired. For example, the characteristics of the flexible panel to be repaired corresponding to the first classification label p2.0 are different from those of the flexible panel to be repaired represented by the first classification label p4.0.

[0112] Here, the characteristics of the flexible panel to be repaired refer to the defect characteristics of the panel to be repaired. Different first classification labels correspond to different defect characteristics of the flexible panel to be repaired. The defect characteristics here may include, but are not limited to, the type of defect and the area of ​​the defect. It can be understood that the working parameters of the repair equipment for different types of defects are different for different types of defects. Similarly, the severity of the corresponding defects is different for different defect areas, and the working parameters of the repair equipment for defects of different severity are also different. Therefore, the working parameters of the repair equipment required for the flexible panels to be repaired corresponding to different first classification labels are different. Based on this, the first classification label can be used to indicate the working parameters set by the repair equipment to repair the flexible panel to be repaired, so that the working parameters of the repair equipment can be determined according to the first classification label.

[0113] Taking the working parameter as working time as an example, the flexible panel P2.0 to be repaired may indicate that the repair time required for this type of flexible panel to be repaired is 2 hours, while the flexible panel P4.0 to be repaired may indicate that the repair time required for this type of flexible panel to be repaired is 4 hours.

[0114] Exemplarily, determining the first classification label of the flexible panel to be repaired according to the deep learning model and the image information includes:

[0115] The image information is input into a deep learning model to determine a first classification label of the flexible panel to be repaired.

[0116] In fact, the deep learning model here can identify the input first image information and determine the first classification label of the flexible panel to be repaired corresponding to the input first image information.

[0117] Step 103: Determine working parameters of the repair device according to the first classification label, and repair the flexible panel to be repaired according to the working parameters.

[0118] Here, since the first classification label can indicate the working parameters of the repairing device for repairing the flexible panel to be repaired, the working parameters of the repairing device can be determined according to the first classification label, and then the flexible panel to be repaired can be repaired according to the working parameters.

[0119] Here, the working parameters may include but are not limited to: working time and / or working temperature, etc.

[0120] In the disclosed embodiment, by inputting the first image information of the panel to be repaired into the deep learning model, the flexible panel to be repaired corresponding to the first image information can be intelligently classified based on the deep learning model, and there is no need for classification based on manual experience, which can reduce the phenomenon of inaccurate classification caused by manual participation, thereby reducing the problem of low repair rate caused by inaccurate manual judgment, and improving the production yield of the flexible panel. In addition, since there is no need for manual experience classification, the requirements for operator experience can be reduced, and the operability of the repair process in the production process is improved. In addition, since the flexible panel to be repaired can be intelligently classified according to the deep learning model, and the working parameters of the repair equipment are determined based on the first classification label, and the flexible panel to be repaired is repaired according to the working parameters, it is not necessary to collect a unified repair plan for all the flexible panels to be repaired, so that the phenomenon of unsuccessful repair or inadequate repair caused by repairing with a unified repair plan can be reduced, which can also improve the repair rate of the flexible panel, thereby improving the production yield of the flexible panel. At the same time, it is not necessary to repair the flexible panel to be repaired by manually performing repeated repair tests on the panel to be repaired, so that the repair efficiency of the flexible panel is improved, thereby improving the production efficiency of the flexible panel.

[0121] Taking the above-mentioned defects that can be repaired by high-temperature baking as an example, in the related art, when high-temperature baking is used to repair the flexible panel, whether the repair can be performed, the repair effect and the working parameters of the repair equipment during the repair need to be tested. For example, the repair effect is tested every two hours. If the repair is successful, the flexible panel is taken out of the baking oven. If it is not successful, the baking is continued. This is obviously not conducive to the repair efficiency of the flexible panel. In the related art, a fixed-time high-temperature baking method is also used to repair the flexible panel to be repaired. For example, 8 hours are uniformly used for repair. When 8 hours are uniformly used for repair, some flexible panels may only need 2 hours to be repaired. The remaining 6 hours not only occupy the baking oven, making the baking oven unable to be fully utilized, but also are not conducive to the repair efficiency of the flexible panel; in addition, 8 hours are uniformly used for repair, and some flexible panels may not be repaired in place or insufficiently repaired, resulting in a low repair rate, which leads to low yield of flexible panel production.

[0122] Based on this, in this embodiment, by acquiring the first image information of the flexible panel to be repaired; and determining the first classification label of the flexible panel to be repaired according to the deep learning model and the image information; then determining the working parameters of the repair equipment according to the first classification label, and repairing the flexible panel to be repaired according to the working parameters, the phenomenon of low repair efficiency caused by manual testing can be reduced, and at the same time, the phenomenon of low resource utilization, low repair efficiency and low repair rate of the repair equipment caused by the repair of uniform working hours can also be reduced. Therefore, by acquiring the image information of the repaired flexible panel; determining the first classification label of the flexible panel to be repaired according to the deep learning model and the image information; and repairing the flexible panel to be repaired according to the first classification label, not only the production yield of the flexible panel can be improved, but also the production efficiency of the flexible panel can be improved.

[0123] In some exemplary embodiments, the method further comprises:

[0124] The deep learning model is determined based on a predetermined number of flexible panel samples to be repaired.

[0125] The deep learning model here is determined by a predetermined number of samples of the flexible panel to be repaired. It is understandable that for general neural network models, such as Alexnet, Googlenet, ZFNet or ResNet, when used in different scenarios, it is necessary to learn based on samples provided by a specific scenario to obtain a deep learning model for the scenario. In this embodiment, a predetermined number of samples of the flexible panel to be repaired are used to learn the general neural network model to obtain a deep learning model that can predict the first classification label of the flexible panel to be repaired.

[0126] Here, the predetermined number refers to the number of samples required for the deep learning model training. It is understandable that different deep learning models require different numbers of samples. In some embodiments, the predetermined number can be determined by the general neural network model used.

[0127] In some embodiments, the predetermined number of samples of the flexible panel to be repaired may include: flexible panels to be repaired with multiple defect characteristics. In order to make the samples of the flexible panel to be repaired more referenceable, the predetermined number of samples of the flexible panel to be repaired may also include: an equal number of flexible panels to be repaired with the same defect characteristics.

[0128] In the disclosed embodiment, the flexible panels to be repaired with various defect characteristics are used as samples of the flexible panels to be repaired, thereby making the samples more universal and referenceable.

[0129] In some other embodiments, see Figure 2 , Figure 2 is another flow chart of a method for repairing a flexible panel according to an exemplary embodiment. Figure 2 As shown, the method of determining a deep learning model according to a predetermined number of samples of the flexible panel to be repaired includes:

[0130] Step 201: Divide a predetermined number of samples of the flexible panel to be repaired into training samples and verification samples according to a preset ratio;

[0131] Step 202: training the training samples to obtain a first model;

[0132] Step 203: inputting the verification sample into the first model to obtain a first prediction accuracy rate of the first model;

[0133] Step 204: Determine the deep learning model according to the first prediction accuracy of the first model.

[0134] Here, the preset ratio can also be determined according to the ratio requirements of the training samples and verification samples for the neural network model currently used. Different neural network models have different requirements for the ratio of training samples to verification samples. For example, for a neural network model that requires more learning, the ratio of training samples to verification samples will be higher than that of a neural network model that requires more verification. In some embodiments, a predetermined number of the flexible panel samples to be repaired are divided into samples to be trained and samples to be verified at a ratio of 75% to 25%, that is, 75% of the flexible panel samples to be repaired are used as samples to be trained, and 25% of the flexible panel samples to be repaired are used as samples to be verified.

[0135] Here, the training of the samples to be trained to obtain the first model may include: using the neural network model to train the samples to be trained to obtain the first model.

[0136] In some embodiments, training the training samples to obtain the first model includes:

[0137] Acquire second image information of the flexible panel in the training sample;

[0138] Acquire third image information of the flexible panel in the training sample obtained after the flexible panel in the training sample is repaired using an nth working parameter, where n is greater than or equal to 1;

[0139] Determining whether the third image information meets a restoration standard;

[0140] If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard;

[0141] Mapping the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample;

[0142] In this way, second image information and second classification labels of all flexible panels in the training sample are obtained to form a second image information set and a second classification label set;

[0143] The second image information set and the second classification label set are used as training sample data.

[0144] Here, the second classification label has the same classification rule as the first classification label. The second classification label can be used to map and repair the working parameters corresponding to the flexible panel in the training sample.

[0145] It should be noted that the first category label is a category label in the second category label set.

[0146] In this embodiment, the flexible panel in the training sample is repaired using multiple working parameters in the current production environment, and then the working parameters that can accurately meet the repair standard are determined to map the second classification label of the flexible panel in the training sample. In this way, the accuracy of the second classification label division in the training sample can be high, so that the accuracy of the training sample is high, and then the deep learning model can reduce the learning of invalid data when training based on the training sample, improve the learning and training efficiency and accuracy of the learning and training, and improve the overall performance of the deep learning model.

[0147] In some other embodiments, training the training samples to obtain the first model further comprises:

[0148] If the repair standard is not met, the test flexible panel is repaired using the n+1th working parameter until the test flexible panel is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair using the mth working parameter meets the repair standard;

[0149] If the fifth image does not meet the restoration standard, a preset classification label is used as a second classification label for the flexible panel in the training sample; wherein n+1 is less than m.

[0150] Here, m can be the number of working parameters that can be adjusted or effectively repaired by the repair equipment. For example, taking the above-mentioned baking oven as an example, taking the working parameter as the working time as an example, the working time that the repair equipment can effectively repair is usually no more than 10 hours. If it is still unable to be repaired after more than 10 hours, it indicates that the flexible panel is an unrepairable flexible panel. At this time, if this type of flexible panel is repaired, it will obviously waste repair equipment resources, etc., which is not conducive to improving production efficiency. Usually, when repairing, 2 hours will be used as a detection cycle, so five detection cycles can be used in 10 hours to detect the repair effect. Here m can be taken as 5. That is to say, if the fifth working parameter, that is, the working parameter for 10 hours, still cannot repair the flexible panel, this type of flexible panel can be marked with a preset classification label.

[0151] In this embodiment, by setting a maximum value of the number of detection times, it is possible to reduce invalid testing processes and improve the efficiency of extracting training samples.

[0152] As another optional embodiment, different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

[0153] In this embodiment, different flexible panels in the training samples have different defect characteristics, which means that the same defect characteristic can be represented by one training sample, so that the number of training samples is small and comprehensive, so that the deep learning model requires less training data and has a faster training speed when training based on the training samples.

[0154] Exemplarily, taking the working parameter as working time, if the flexible panel sample to be repaired meets the predetermined standard after 2 hours of repair, the flexible panel sample to be repaired is marked as P2.0; if the flexible panel sample to be repaired meets the predetermined standard after 4 hours of repair, the flexible panel sample to be repaired is marked as P4.0; if the flexible panel sample to be repaired meets the predetermined standard after 8 hours of repair, the flexible panel sample to be repaired is marked as P8.0; if the flexible panel sample to be repaired meets the predetermined standard after 10 hours of repair, the flexible panel sample to be repaired is marked as P10.0; if the flexible panel sample to be repaired still does not meet the predetermined standard after 10 hours of repair, the flexible panel sample to be repaired is marked as NG.

[0155] Of course, in some other embodiments, the method further comprises: obtaining a predetermined number of samples of the flexible panel to be repaired from a server. Here, the server stores samples of the flexible panel to be repaired uploaded by other production lines.

[0156] This embodiment obtains a predetermined number of samples of the flexible panel to be repaired from the cloud server, so that the obtained sample data can be more comprehensive and obtained faster without the need for self-testing.

[0157] In step 203, inputting the verification sample into the first model to obtain a first prediction accuracy of the first model includes:

[0158] Obtaining a third classification label set of the verification sample;

[0159] Inputting the verification sample into the first model to obtain a set of prediction results of the verification sample;

[0160] The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

[0161] Here, the prediction result refers to the prediction result of the third classification label of the verification sample. For example, if the third classification label of the verification sample is P2.0, then if the prediction result of the verification sample is P3.0, it indicates that the prediction result does not match the third classification label.

[0162] Here, determining the first prediction accuracy of the first model according to the prediction result set and the third classification label set includes:

[0163] According to the matching conditions between the plurality of prediction result sets and the corresponding third classification label sets, respectively, determining the proportion of the prediction results matching the corresponding third classification labels;

[0164] According to the proportion, the first prediction accuracy of the first model is determined.

[0165] It can be understood that, the higher the proportion of the prediction results matching the corresponding third classification labels, the higher the first prediction accuracy of the first model.

[0166] In this way, the first prediction accuracy of the first model can be obtained by verifying the verification sample, so that it can be determined based on the first prediction accuracy whether to use the first model as a deep learning model for intelligent repair of the subsequent flexible panel to be repaired.

[0167] In some embodiments, the step 204, i.e., determining the deep learning model according to the first prediction accuracy of the first model, may include:

[0168] If the first prediction accuracy is less than the preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy;

[0169] The first model corresponding to which the model parameters have been adjusted and the first prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

[0170] Here, the preset accuracy rate may be the accuracy rate specified in the production standard of the flexible panel. Different manufacturers have different production standards, and thus the preset accuracy rates may be different.

[0171] In the embodiment of the present disclosure, by adjusting the model parameters of the first model so that the first prediction accuracy of the first model is greater than or equal to the preset accuracy, the existing first model can be continuously optimized so that the first model can be used as a deep learning model for determining the first classification label of the flexible panel to be repaired.

[0172] In some other embodiments, the step 204, i.e., determining the deep learning model according to the first prediction accuracy of the first model, may further include:

[0173] If the first prediction accuracy is less than the preset accuracy, determining a second prediction accuracy of an alternative second model;

[0174] The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

[0175] Here, the first model and the second model belong to different neural network models.

[0176] In the disclosed embodiment, by replacing the neural network model, a second model with a second prediction accuracy greater than or equal to the preset accuracy is used as the deep learning model, so that the first model does not need to be repeatedly trained and verified, thereby improving the acquisition efficiency of the deep learning model. At the same time, the adjustment of the model parameters of the first model can be reduced, and the callback steps of the model parameters when the deep learning model is used in other scenarios are reduced, thereby improving the versatility of the deep learning model.

[0177] In some other embodiments, repairing the panel to be repaired according to the first classification label may include:

[0178] If there are multiple repair devices, according to the first target working parameter corresponding to the first classification label, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair;

[0179] or

[0180] If there is a repair device, the working parameters of the repair device are adjusted to the second target working parameters corresponding to the first classification label, and the flexible panel to be repaired is repaired with the second target working parameters.

[0181] The first target working parameter and the second target working parameter here are both working parameters of the repairing device for repairing the flexible panel to be repaired.

[0182] In the case where there are multiple repair devices in the production environment, the repair devices have pre-configured working parameters. The disclosed embodiment can place the flexible panel to be repaired directly into the repair device with the first target working parameter corresponding to the first classification label of the flexible panel to be repaired according to the first classification label of the flexible panel to be repaired. Still taking the example of working parameters including working time, the working time set for repair device A is 2 hours, the working time set for repair device B is 4 hours, and the working time set for repair device C is 8 hours. The first classification label of the current flexible panel to be repaired is P2.0. Then, according to the first classification label, the P2.0 flexible panel to be repaired can be directly placed into the repair device A. In some embodiments, in order to make the repair system more intelligent, placing the flexible panel to be repaired into the repair device with the first target working parameter corresponding to the first classification label includes:

[0183] The flexible panel to be repaired is directly delivered to the repair device with the first target working parameter corresponding to the first classification label through the conveyor belt. In this way, the production process can be fully automated.

[0184] In the case where there is only one repair device in the production environment, the working parameters of the repair device are adjustable. The disclosed embodiment can adjust the working parameters of the repair device to the second working parameters corresponding to the first classification label of the flexible panel to be repaired. Still taking the example that the working parameters include working time, the first classification label of the current flexible panel to be repaired is P2.0, then according to the first classification label, the working parameters of the repair device can be adjusted to 2 hours. In this way, the working parameters of the repair device can be set more accurately, thereby reducing the waste of resources of the repair device, improving the repair efficiency of the repair device, and thus improving production efficiency.

[0185] In some other embodiments, repairing the flexible panel to be repaired according to the first classification label may also include:

[0186] If the first classification label indicates that the flexible panel to be repaired is a scrapped panel, the flexible panel to be repaired is removed from the repair equipment, or the flexible panel to be repaired is placed in a scrap pool.

[0187] In this way, by intelligently determining the first classification label of the flexible panel to be repaired, and removing the scrapped panel in time when the first classification label indicates that the flexible panel to be repaired is a scrapped panel, repair equipment resources can be specially allocated to improve the repair efficiency of the repair equipment, thereby improving production efficiency.

[0188] Furthermore, the present disclosure also provides a specific implementation to further understand the repair method of the flexible panel provided by the embodiment of the present disclosure.

[0189] Here, high-temperature baking is used as an example to repair optical defects such as black spots caused by light, Sandy mura, excessive brightness or darkness caused by static electricity, and uneven color or brightness distribution of the flexible panel. The repair equipment here is a high-temperature baking oven as an example.

[0190] See also Figure 3 , Figure 3 is a scene diagram of a method for repairing a flexible panel according to an exemplary embodiment. Figure 3 As shown, image information of the flexible panel to be repaired is collected by a CCD camera, and the first classification label of the flexible panel to be repaired corresponding to the image information is determined according to the deep learning model; according to the first classification label, the flexible panel to be repaired is placed in a baking oven with corresponding working parameters, or is placed in a scrap pool.

[0191] Here, first, a CCD camera is used to capture images of the optically poor flexible panel samples to be repaired in mass production. One image is captured for each flexible panel sample to be repaired, and the flexible panel samples to be repaired are placed in a baking oven for baking. Then, the flexible panel samples to be repaired are marked with the repair results. Here, considering the impact of production efficiency, repair yield, and the characteristics of the flexible panel samples to be repaired, the maximum working time set during the test is 10 hours. The flexible panel to be repaired that can be repaired to the standard specification in 2 hours of working time is marked as P2.0, 4 hours is marked as P4.0, 6 hours is marked as P6.0, 8 hours is marked as P8.0, 10 hours is marked as P10.0, and the flexible panel that does not meet the standard specification in 10 hours is marked as NG.

[0192] 75% of the above-mentioned flexible panel samples to be repaired are used as samples to be trained, and 25% are used as samples to be verified. The samples to be trained are used to obtain a first model; the samples to be verified are input into the first model to obtain a first prediction accuracy of the first model. If the first prediction accuracy does not reach the preset accuracy, the first model is optimized. Here, optimizing the first model may include: adjusting the model parameters of the first model so that the first prediction accuracy of the first model reaches the preset accuracy. In other embodiments, if the first prediction accuracy does not reach the preset accuracy, a second model whose accuracy reaches the preset accuracy is replaced.

[0193] Finally, the model with the preset accuracy is used as the deep learning model and input into the production workstation. The image of each flexible panel to be repaired is directly input into the deep learning model for automatic judgment. If it is a scrapped panel, it is directly scrapped. If it is not scrapped, the flexible panel to be repaired is placed in the repair equipment with the corresponding working parameters for repair according to the first classification label of the corresponding working time.

[0194] The disclosed embodiment intelligently classifies the flexible panels to be repaired through a deep learning model, and repairs the flexible panels to be repaired based on the classification results, thereby improving the repair efficiency and repair rate of the flexible panels to be repaired, while improving the production yield, increasing the mass production capacity, maximizing the use of equipment and resources, and reducing costs.

[0195] Figure 4 FIG. 1 is a block diagram of a repair device for a flexible panel according to an exemplary embodiment. Figure 4 The device comprises: a first acquisition module 41, a determination module 42 and a repair module 43; wherein,

[0196] The first acquisition module 41 is used to acquire first image information of the flexible panel to be repaired;

[0197] The first determination module 42 is used to determine a first classification label of the flexible panel to be repaired according to the deep learning model and the first image information;

[0198] The repair module 42 is used to determine the working parameters of the repair equipment according to the first classification label, and repair the flexible panel to be repaired according to the working parameters.

[0199] As an optional embodiment, the device further includes:

[0200] A second determination module is used to determine the deep learning model according to a predetermined number of samples of the flexible panel to be repaired;

[0201] The second determining module further includes:

[0202] A division submodule, used for dividing a predetermined number of samples of the flexible panel to be used into training samples and verification samples according to a preset ratio;

[0203] A first processing submodule, used for training the training samples to obtain a first model;

[0204] A second processing submodule, configured to input the verification sample into the first model to obtain a first prediction accuracy rate of the first model;

[0205] The third processing submodule is used to determine the deep learning model according to the first prediction accuracy of the first model.

[0206] As an optional embodiment, the first processing submodule is further used to:

[0207] Acquire second image information of the flexible panel in the training sample;

[0208] Acquire third image information of the flexible panel in the training sample obtained after the flexible panel in the training sample is repaired using an nth working parameter, where n is greater than or equal to 1;

[0209] Determining whether the third image information meets a restoration standard;

[0210] If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard;

[0211] Mapping the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample;

[0212] In this way, second image information and second classification labels of all flexible panels in the training sample are obtained to form a second image information set and a second classification label set;

[0213] The second image information set and the second classification label set are used as training sample data.

[0214] As an optional embodiment, the first processing submodule is further used to:

[0215] If the repair standard is not met, the test flexible panel is repaired using the n+1th working parameter until the test flexible panel is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair using the mth working parameter meets the repair standard;

[0216] If the fifth image does not meet the restoration standard, a preset classification label is used as a second classification label for the flexible panel in the training sample, wherein n+1 is less than m.

[0217] Different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

[0218] As an optional embodiment, the third processing submodule is further used for:

[0219] If the first prediction accuracy is less than the preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy;

[0220] The first model corresponding to the first prediction accuracy being greater than or equal to the preset accuracy is used as the deep learning model.

[0221] As an optional embodiment, the third processing submodule is further used for:

[0222] If the first prediction accuracy is less than the preset accuracy, determining a second prediction accuracy of an alternative second model;

[0223] The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

[0224] As an optional embodiment, the second processing submodule is further used to:

[0225] Obtaining a third classification label set of the verification sample;

[0226] Inputting the verification sample into the first model to obtain a set of prediction results of the verification sample;

[0227] The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

[0228] As an optional embodiment, the repair module 43 is further used for:

[0229] If there are multiple repair devices, according to the first target working parameter corresponding to the first classification label, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair;

[0230] or

[0231] If there is a repair device, the working parameters of the repair device are adjusted to the second target working parameters corresponding to the first classification label, and the flexible panel to be repaired is repaired with the second target working parameters.

[0232] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0233] Figure 5 is a block diagram of a repair device 500 according to an exemplary embodiment. For example, the repair device 500 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0234] Reference Figure 5 The repair device 500 may include one or more of the following components: a processing component 502 , a memory 504 , a power component 506 , a multimedia component 508 , an audio component 510 , an input / output (I / O) interface 512 , a sensor component 514 , and a communication component 516 .

[0235] The processing component 502 generally controls the overall operation of the repair device 500, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 502 may include one or more processors 520 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 502 may include one or more modules to facilitate the interaction between the processing component 502 and other components. For example, the processing component 502 may include a multimedia module to facilitate the interaction between the multimedia component 508 and the processing component 502.

[0236] The memory 504 is configured to store various types of data to support operations on the repair device 500. Examples of such data include instructions for any application or method operating on the repair device 500, contact data, phone book data, messages, pictures, videos, etc. The memory 504 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0237] The power component 506 provides power to the various components of the repair device 500. The power component 506 can include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the repair device 500.

[0238] The multimedia component 508 includes a screen providing an output interface between the repair device 500 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 508 includes a front camera and / or a rear camera. When the repair device 500 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0239] The audio component 510 is configured to output and / or input audio signals. For example, the audio component 510 includes a microphone (MIC), and when the repair device 500 is in an operation mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 504 or sent via the communication component 516. In some embodiments, the audio component 510 also includes a speaker for outputting audio signals.

[0240] I / O interface 512 provides an interface between processing component 502 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0241] Sensor assembly 514 includes one or more sensors for providing status assessment of various aspects for repair device 500. For example, sensor assembly 514 can detect the opening / closing state of repair device 500, the relative positioning of components, such as the display and keypad of repair device 500, sensor assembly 514 can also detect the position change of a component of repair device 500 or terminal 500, the presence or absence of user contact with terminal 500, the position or acceleration / deceleration of repair device 500 and the temperature change of repair device 500. Sensor assembly 514 can include proximity sensor, configured to detect the presence of nearby objects when there is no physical contact. Sensor assembly 514 can also include optical sensor, such as CMOS or CCD image sensor, for use in imaging applications. In some embodiments, this sensor assembly 514 can also include acceleration sensor, gyroscope sensor, magnetic sensor, pressure sensor or temperature sensor.

[0242] The communication component 516 is configured to facilitate wired or wireless communication between the repair device 500 and other devices. The terminal 500 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 516 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 516 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0243] In an exemplary embodiment, the repair device 500 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-mentioned methods.

[0244] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 504 including instructions, and the instructions can be executed by the processor 520 of the repair device 500 to complete the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0245] A non-temporary computer-readable storage medium, when instructions in the storage medium are executed by a processor of a terminal, enables the terminal to execute the flexible panel repair method described in the above embodiments.

[0246] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in this disclosure. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0247] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for repairing a flexible panel, characterized in that: include: Acquire first image information of the flexible panel to be repaired; Determine a first classification label of the flexible panel to be repaired according to the deep learning model and the first image information; Determining working parameters of a repair device according to the first classification label, and repairing the flexible panel to be repaired according to the working parameters; Determining the deep learning model according to a predetermined number of samples of the flexible panel to be repaired includes: dividing the predetermined number of samples of the flexible panel to be repaired into training samples and verification samples according to a preset ratio; training the training samples to obtain a first model; inputting the verification samples into the first model to obtain a first prediction accuracy of the first model; and determining the deep learning model according to the first prediction accuracy of the first model; Among them, the training of the training samples to obtain the first model includes: obtaining the second image information of the flexible panel in the training sample; obtaining the third image information of the flexible panel in the training sample after repairing the flexible panel in the training sample through the nth working parameter, wherein n is greater than or equal to 1; determining whether the third image information meets the repair standard; if it does not meet the repair standard, repairing the flexible panel in the training sample using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard; mapping the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample; thereby obtaining the second image information and second classification labels of all flexible panels in the training sample to form a second image information set and a second classification label set; and using the second image information set and the second classification label set as training sample data.

2. The repair method according to claim 1, characterized in that: The training of the training samples to obtain a first model further comprises: If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the flexible panel in the training sample is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair with the mth working parameter meets the repair standard; If the fifth image does not meet the restoration standard, a preset classification label is used as a second classification label for the flexible panel in the training sample; wherein n+1 is less than m.

3. The repair method according to claim 1 or 2, characterized in that: Different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

4. The repair method according to claim 1, characterized in that: The step of determining the deep learning model according to the first prediction accuracy of the first model includes: If the first prediction accuracy is less than a preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy; The first model corresponding to the first prediction accuracy being greater than or equal to the preset accuracy is used as the deep learning model.

5. The repair method according to claim 1, characterized in that: The step of determining the deep learning model according to the first prediction accuracy of the first model includes: If the first prediction accuracy is less than a preset accuracy, determining a second prediction accuracy of an alternative second model; The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

6. The repair method according to claim 1, characterized in that: Inputting the verification sample into the first model to obtain a first prediction accuracy rate of the first model includes: Obtaining a third classification label set of the verification sample; Inputting the verification sample into the first model to obtain a set of prediction results of the verification sample; The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

7. The method according to claim 1, characterized in that The repairing of the flexible panel to be repaired according to the working parameters includes: If there are multiple repair devices, according to the first target working parameter corresponding to the first classification tag, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair; or, If there is a repair device, according to the second target working parameter corresponding to the first classification label, the working parameter of the repair device is adjusted to the second target working parameter, and the flexible panel to be repaired is repaired with the second target working parameter; or, If the first classification label indicates that the flexible panel to be repaired is a scrapped panel, the flexible panel to be repaired is moved out of the repair equipment, or the flexible panel to be repaired is placed in a scrap pool.

8. A repair device for a flexible panel, characterized in that: include: A first acquisition module, used to acquire first image information of the flexible panel to be repaired; A first determination module, configured to determine a first classification label of the flexible panel to be repaired according to a deep learning model and the first image information; A repair module, configured to determine working parameters of a repair device according to the first classification label, and repair the flexible panel to be repaired according to the working parameters; A second determination module is used to determine the deep learning model according to a predetermined number of samples of the flexible panel to be repaired; The second determination module further includes: a division submodule, used to divide a predetermined number of samples of the flexible panel to be repaired into training samples and verification samples according to a preset ratio; a first processing submodule, used to train the training samples to obtain a first model; a second processing submodule, used to input the verification samples into the first model to obtain a first prediction accuracy rate of the first model; a third processing submodule, used to determine the deep learning model according to the first prediction accuracy rate of the first model; The first processing submodule is also used to: obtain second image information of the flexible panel in the training sample; obtain third image information of the flexible panel in the training sample after repairing the flexible panel in the training sample using the nth working parameter, where n is greater than or equal to 1; determine whether the third image information meets the repair standard; if it does not meet the repair standard, repair the flexible panel in the training sample using the n+1th working parameter until the fourth image information of the flexible panel in the training sample meets the repair standard; map the working parameters corresponding to the repair standard to the second classification label of the flexible panel in the training sample; thereby obtain the second image information and second classification labels of all flexible panels in the training sample to form a second image information set and a second classification label set; and use the second image information set and the second classification label set as training sample data.

9. The repair device according to claim 8, wherein the first processing submodule is further configured to: If the repair standard is not met, the flexible panel in the training sample is repaired using the n+1th working parameter until the flexible panel in the training sample is repaired using the mth working parameter, and then it is determined whether the fifth image information obtained after the repair with the mth working parameter meets the repair standard; If the fifth image does not meet the restoration standard, the preset classification label is used as the second classification label of the flexible panel in the training sample, wherein: n+1 is less than m.

10. The repair device according to claim 8 or 9, characterized in that: Different flexible panels in the training samples have different defect characteristics, wherein the defect characteristics include at least one of the following: defect type and defect area.

11. The repair device according to claim 8, characterized in that: The third processing submodule is further used for: If the first prediction accuracy is less than a preset accuracy, adjusting the model parameters of the first model until the first prediction accuracy of the first model is greater than or equal to the preset accuracy; The first model corresponding to the first prediction accuracy being greater than or equal to the preset accuracy is used as the deep learning model.

12. The repair device according to claim 8, characterized in that: The third processing submodule is further used for: If the first prediction accuracy is less than a preset accuracy, determining a second prediction accuracy of an alternative second model; The second model whose second prediction accuracy is greater than or equal to the preset accuracy is used as the deep learning model.

13. The device according to claim 8, characterized in that The second processing submodule is further used for: Obtaining a third classification label set of the verification sample; Inputting the verification sample into the first model to obtain a set of prediction results of the verification sample; The first prediction accuracy of the first model is determined according to the prediction result set and the third classification label set.

14. The device according to claim 8, characterized in that The repair module is further used for: If there are multiple repair devices, according to the first target working parameter corresponding to the first classification label, the flexible panel to be repaired is placed in the repair device set to the first target working parameter for repair; or, If there is a repair device, according to the second target working parameter corresponding to the first classification label, the working parameter of the repair device is adjusted to the second target working parameter, and the flexible panel to be repaired is repaired with the second target working parameter; or, If the first classification label indicates that the flexible panel to be repaired is a scrapped panel, the flexible panel to be repaired is moved out of the repair equipment, or the flexible panel to be repaired is placed in a scrap pool.

15. A repair device for a flexible panel, characterized in that: include: A processor and a memory for storing a computer program that can be run on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when running the computer program.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images