Screen defect detection method and device, equipment and storage medium

Abstract

The invention discloses a screen defect detection method and device, equipment and a storage medium, and relates to the technical field of information processing, and the screen defect detection method comprises the steps: obtaining a screen detection image of a target screen; the screen detection image is input to a target defect detection student model, a screen defect detection result output by the target defect detection student model is obtained, and the target defect detection student model is obtained through iterative training according to a target defect detection teacher model. Screen defect detection is realized by using the target defect detection student model, so that the target defect detection student model can be applied to edge equipment with relatively low operational capability, and the processing result of the target defect detection student model is close to that of the target defect detection student model, so that the accuracy of screen detection is improved, and the cost of equipment upgrading is saved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a screen defect detection method, device, equipment and storage medium. Background Art

[0002] In the field of high-precision screen defect detection, for some unknown defects and minor defects, there are still problems of low detection accuracy such as false detection and missed detection. At present, the application of deep learning models in the field of defect detection has greatly improved the detection accuracy of the detection system. Although the application of deep learning models has developed rapidly, it faces a problem: the better the performance of the model, the higher the requirements for the equipment, and the higher the cost of equipment expenditure for enterprises.

[0003] When companies want to save costs on testing equipment, traditional deep learning models cannot fully utilize the computing power of the model in testing equipment with low computing power, resulting in slow operation, low efficiency, and even the possibility of failure to operate normally.

[0004] Therefore, how to enable the model to be applied on low-cost devices while maintaining the model's ability to detect screen defects is a problem that needs to be solved urgently. Summary of the invention

[0005] The main purpose of this application is to provide a screen defect detection method, device, equipment and storage medium, aiming to enable the model to be used in edge devices with low computing power, and its processing results are close to traditional models, thereby improving the accuracy of screen detection and saving the cost of equipment upgrades.

[0006] To achieve the above objectives, the present application proposes a screen defect detection method, the method comprising:

[0007] Acquire a screen detection image of the target screen;

[0008] The screen detection image is input into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on a target defect detection teacher model.

[0009] In one embodiment, before inputting the screen detection image into the target defect detection student model to obtain the screen defect detection result output by the target defect detection student model, the method further includes:

[0010] Acquire several groups of screen image data;

[0011] Inputting each of the screen image data and the sample labels corresponding to each of the screen image data into one or more initial defect detection teacher models for iterative training to obtain a plurality of pending defect detection teacher models;

[0012] Based on each of the pending defect detection teacher models, the target defect detection teacher model is determined.

[0013] In one embodiment, determining the target defect detection teacher model based on each of the pending defect detection teacher models includes:

[0014] Inputting a plurality of sets of screen image data into each of the pending defect detection teacher models to obtain a predicted value output by each of the pending defect detection teacher models;

[0015] Obtaining a standard data label, and calculating a cross entropy value between the predicted value and the standard data label;

[0016] Based on the cross entropy value, a target defect detection teacher model is determined.

[0017] In one embodiment, the target defect detection teacher model includes a feature layer;

[0018] The training process of the target defect detection student model includes:

[0019] Obtain soft labels and screen image data of the target defect detection teacher model;

[0020] Based on a preset loss function, the initial defect detection student model is connected to the feature layer of the target defect detection teacher model, so that the initial defect detection student model calculates the feature gap corresponding to the feature layer of the target defect detection teacher model;

[0021] The soft label, the feature information and the screen image data are input into an initial defect detection student model for iterative training to obtain a target defect detection student model.

[0022] In one embodiment, obtaining a soft label of a target defect detection teacher model includes:

[0023] Inputting several groups of screen image data into a target defect detection teacher model to obtain several prediction results output by the target defect detection teacher model;

[0024] Based on each of the prediction results, a soft label corresponding to the target defect detection teacher model is generated.

[0025] In one embodiment, after the screen detection image is input into the target defect detection student model and the screen defect detection result output by the target defect detection student model is obtained, the method further includes:

[0026] Based on the screen defect detection result, generating a screen defect detection report;

[0027] Pushing the screen defect detection report to the inspector so that the inspector can repair the target screen;

[0028] Acquiring screen defect detection data of the target screen after repair;

[0029] Based on the screen defect detection data, a screen repair effect is determined.

[0030] In one embodiment, after determining the screen repair effect based on the screen defect detection data, the method further includes:

[0031] Determining a repair effect level of the screen repair effect;

[0032] If the repair effect level meets the preset effect level, the screen repair method of the inspector is obtained;

[0033] The screen repair method is associated and combined with a screen defect detection report corresponding to the screen repair effect to generate a screen repair record, and the screen repair record is stored in a preset database.

[0034] In addition, to achieve the above-mentioned purpose, the present application also proposes a screen defect detection device, the screen defect detection device comprising:

[0035] An acquisition module, used for acquiring a screen detection image of a target screen;

[0036] A detection module is used to input the screen detection image into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on a target defect detection teacher model.

[0037] In addition, to achieve the above-mentioned purpose, the present application also proposes a screen defect detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the screen defect detection method described above.

[0038] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the screen defect detection method described above are implemented.

[0039] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the screen defect detection method described above are implemented.

[0040] The present application provides a screen defect detection method, apparatus, device and storage medium. The screen defect detection method obtains a screen detection image of a target screen, and then inputs the screen detection image into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model. The target defect detection student model is obtained by iterative training based on a target defect detection teacher model, so that the model can be used in edge devices with lower computing power, and its processing results are close to those of traditional models, thereby improving the accuracy of screen detection and saving the cost of equipment upgrades. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0042] 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, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0043] Figure 1 A schematic diagram of a process flow provided for Embodiment 1 of the screen defect detection method of the present application;

[0044] Figure 2 A schematic diagram of a flow chart provided for Embodiment 2 of the screen defect detection method of the present application;

[0045] Figure 3 A schematic diagram of a flow chart provided for Embodiment 3 of the screen defect detection method of the present application;

[0046] Figure 4 An example diagram of the process implementation provided for the screen defect detection method of this application;

[0047] Figure 5 This is a schematic diagram of the module structure of the screen defect detection device according to an embodiment of the present application;

[0048] Figure 6 A schematic diagram of the device structure of the hardware operating environment involved in the screen defect detection method in the embodiment of the present application;

[0049] Figure 7 A structural example diagram of a teacher model provided for the screen defect detection method of this application.

[0050] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0052] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0053] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a big data service platform, a screen defect detection system, etc. The following takes the screen defect detection system as an example to illustrate this embodiment and the following embodiments.

[0054] Based on this, the present application embodiment provides a screen defect detection method, referring to Figure 1 , Figure 1 A flow chart of the screen defect detection method according to the present invention.

[0055] In this embodiment, the screen defect detection method includes steps S11 to S13:

[0056] Step S11, obtaining a screen detection image of the target screen;

[0057] It should be noted that the target screen refers to the screen that needs to be inspected for screen defects, including multiple types of screens such as LCD screens, which are not limited here and can be set according to actual conditions.

[0058] It should be further explained that the screen detection image refers to an image that reflects the screen quality of the target screen, wherein the screen detection image may be one or more screen image data taken under different brightness, different environments and different angles, so as to expand the diversity and comprehensiveness of the data and improve the accuracy of screen defect detection. No limitation is made here. For example, the number of screen detection images and the shooting angle can be determined by professionals to determine how to cover the maximum range of screen defect detection with the minimum number of images and angles, etc., and can be set according to actual conditions.

[0059] Specifically, the image of the target screen may be captured by an industrial camera, which is not limited here.

[0060] Step S12, inputting the screen detection image into the target defect detection student model to obtain the screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on the target defect detection teacher model.

[0061] It should be noted that the target defect detection student model refers to a student model obtained after knowledge distillation and iterative training of multiple target defect detection teacher models, wherein the target defect detection student model can select a smaller, simpler network for model training, so as to replicate the model performance of the teacher model as much as possible through training, etc., so that the student model can be used in edge devices with lower computing power. At the same time, its processing results are close to the processing results output by the teacher model, thereby effectively saving the cost of equipment upgrades.

[0062] It should be further explained that the screen defect detection results refer to the defect detection results for the target screen, including defect type (such as light leakage, scratches, bright spots, dark spots, stray spots, and polarizer impurity points, etc.), defect location, defect size, number of defects, defect severity, defect distribution, detection time, detection equipment information, screen model, production batch, etc., without limitation here.

[0063] Furthermore, the target defect detection teacher model refers to a teacher model that meets the knowledge distillation standard after iterative training and model evaluation. The target defect detection teacher model can use a large, complex and excellent convolutional neural network, which can be referred to Figure 7 , Figure 7 A structural example diagram of a teacher model provided for the screen defect detection method of this application, which highlights the contrast of point defects and enhances the detection capability of small targets through methods such as precise search for small targets, small models and lightweight deployment, and low-quality / low-contrast image enhancement. At the same time, it realizes the positioning and detection of tiny point defect targets such as bright spots, dark spots, and polarizer impurity points, thereby achieving the requirements of high precision and high speed in screen image data defect detection.

[0064] In addition, the soft label refers to the detection result output by the target defect detection teacher model. The screen image data refers to the training samples used for model training, including normal screen image data and abnormal screen image data, etc., which are not limited here.

[0065] This embodiment obtains a screen detection image of the target screen, and then inputs the screen detection image into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on the target defect detection teacher model, so that the model can be used in edge devices with lower computing power, and its processing results are close to those of traditional models, thereby improving the accuracy of screen detection and saving the cost of equipment upgrades.

[0066] Based on this, the present application embodiment provides a screen defect detection method, referring to Figure 2 , Figure 2 A flow chart of the second embodiment of the screen defect detection method of the present application.

[0067] In a feasible implementation manner, before inputting the screen detection image into the target defect detection student model to obtain the screen defect detection result output by the target defect detection student model, the method further includes:

[0068] Step S21, obtaining several groups of screen image data;

[0069] It should be noted that the screen image data is a data sample used for model training, wherein the teacher model needs to be trained using screen image data with real object labels and complete spatial information, such as object categories and position information of classified defects in screen inspection, so that the teacher model can accurately understand and express 3D scenes under natural language instructions by learning real object relationships and spatial layouts.

[0070] Step S22, inputting each of the screen image data and the sample labels corresponding to each of the screen image data into one or more initial defect detection teacher models for iterative training to obtain a plurality of pending defect detection teacher models;

[0071] It should be noted that the pending defect detection teacher model refers to a teacher model that has completed model training but has not undergone knowledge distillation condition evaluation.

[0072] Specifically, data preprocessing, such as data cleaning and normalization, is performed on each of the screen image data and the sample labels corresponding to each of the screen image data, so that each of the screen image data and the sample labels corresponding to each of the screen image data are input into several initial defect detection teacher models to obtain the predicted values ​​output by each of the initial defect detection teacher models, and then based on the predicted values ​​and the sample labels, the model loss value is calculated using the loss function. In this embodiment, the loss function can be set according to actual needs and is not specifically limited here. After the model loss value is calculated, the training process ends, and the error back propagation algorithm is used to update the model parameters in several initial defect detection teacher models before the next training. During the training process, it is determined whether the updated initial defect detection teacher models meet the preset training end conditions. If so, the updated initial defect detection teacher models are used as the target defect detection teacher models. If not, the model training continues, wherein the preset training end conditions include loss convergence and reaching the maximum number of iterations threshold.

[0073] Step S23, determining the target defect detection teacher model based on each of the pending defect detection teacher models.

[0074] It should be noted that in deep neural network training, even models trained by the same network will have differences when facing the same data. Therefore, in the knowledge distillation of multiple teacher models, in order to make the teacher model participating in knowledge transfer each time the relatively best model, it is necessary to identify multiple teacher models for different data. In this embodiment, the cross entropy between the predicted value of the teacher model and the standard data is used as the best quantitative indicator for identifying the quality of the model. Therefore, in this embodiment, the cross entropy between the predicted output of the teacher model for the current input data and the standard data label is used as a quantitative indicator of whether this teacher model can transfer sufficient positive knowledge to the student model.

[0075] Specifically, several groups of screen image data are input into each of the pending defect detection teacher models to obtain the predicted values ​​output by each of the pending defect detection teacher models, and then the standard data labels are obtained, and the cross-entropy values ​​between the predicted values ​​and the standard data labels are calculated, so as to determine the target defect detection teacher model based on the cross-entropy values.

[0076] This embodiment obtains several groups of screen image data, and then inputs each of the screen image data and the sample labels corresponding to each of the screen image data into one or more initial defect detection teacher models for iterative training to obtain several pending defect detection teacher models, thereby determining the target defect detection teacher model based on each of the pending defect detection teacher models, thereby improving the screen detection accuracy, and by training multiple teacher models to enhance the robustness and adaptability of the subsequent knowledge distillation process, thereby achieving continuous model optimization and reducing the risk of detection errors.

[0077] Based on this, the present application embodiment provides a screen defect detection method, referring to Figure 3 , Figure 3 A flow chart of the screen defect detection method according to the present invention.

[0078] In a feasible implementation manner, determining the target defect detection teacher model based on each of the pending defect detection teacher models includes:

[0079] Step S31, inputting a plurality of groups of screen image data into each of the pending defect detection teacher models to obtain a predicted value output by each of the pending defect detection teacher models;

[0080] It should be noted that the predicted value refers to the screen defect prediction result of the screen image data output by the teacher model.

[0081] Step S32, obtaining a standard data label, and calculating a cross entropy value between the predicted value and the standard data label;

[0082] It should be noted that the standard data labels refer to the real labeled data used to train and evaluate the model, wherein the standard data labels are known and correct, and they indicate which category each sample should be classified into. For example, in classification problems, the standard data labels are usually integers representing different categories. For example, in image classification tasks, each image has a corresponding exact category label.

[0083] It should be further explained that the cross-entropy value (Cross-Entropy Loss) is a loss function commonly used in classification problems, which is used to measure the difference between the model output probability distribution and the true label probability distribution.

[0084] Specifically, the standard data label is obtained, and the cross entropy value between the predicted value and the standard data label is calculated, wherein the calculation can be performed using a cross entropy calculation formula, and the cross entropy calculation formula is:

[0085] L i n (t, T) = -∑Ti log(t i )

[0086] Among them, i represents the category of the data, represents the nth teacher model; T represents the standard data label, and represents the predicted probability of the current teacher model for the i-th category label.

[0087] Thus, the cross entropy value is calculated by the cross entropy calculation formula.

[0088] Step S33, determining a target defect detection teacher model based on the cross entropy value.

[0089] Specifically, according to the properties of the cross-entropy function, the smaller the cross-entropy between the standard data label and the predicted value of the current teacher model, the closer the prediction of this teacher model for this category is to the standard data, and the more correct the knowledge mastered by this teacher model is. Therefore, the cross-entropy values ​​are sorted in reverse order to determine the target defect detection teacher model according to the preset sorting. For example, the top five pending defect detection teacher models are used as the target defect detection teacher models.

[0090] This embodiment obtains the predicted values ​​output by each of the pending defect detection teacher models by inputting several groups of screen image data into the pending defect detection teacher models, and then obtains the standard data labels, and calculates the cross-entropy values ​​between the predicted values ​​and the standard data labels, thereby determining the target defect detection teacher model based on the cross-entropy value, thereby improving the model's generalization ability for data, and achieving more accurate and objective model evaluation, thereby effectively improving the robustness and stability of screen defect detection, and then improving the accuracy and precision of screen defect detection. At the same time, in the knowledge distillation of multiple teacher models, the teacher model participating in knowledge transfer each time is kept as a relatively optimal model.

[0091] In a feasible implementation manner, the target defect detection teacher model includes a feature layer; and the training process of the target defect detection student model includes:

[0092] Step S41, obtaining the soft label and screen image data of the target defect detection teacher model;

[0093] It should be noted that the soft labels are used to deal with the uncertainty in classification tasks. In traditional hard labels, each sample is explicitly assigned to a category. For example, in image classification tasks, a picture is either a cat or a dog, and the label is certain. In contrast, soft labels do not simply assign samples to a single category, but assign a probability distribution to each sample, indicating the possibility that the sample belongs to each category. In this embodiment, the soft labels represent the knowledge about screen defect detection obtained by the target defect detection teacher model after model training and model prediction.

[0094] Specifically, several groups of screen image data are input into the target defect detection teacher model to obtain several prediction results output by the target defect detection teacher model, and then based on each of the prediction results, a soft label corresponding to the target defect detection teacher model is generated, and the screen image data is obtained.

[0095] Step S42, based on a preset loss function, connecting the initial defect detection student model with the feature layer of the target defect detection teacher model, so that the initial defect detection student model calculates the feature gap corresponding to the feature layer of the target defect detection teacher model;

[0096] It should be noted that in order to further improve the accuracy of the student model, the information of the feature layer of the target defect detection teacher model needs to be transmitted to the student model so that the information after feature extraction can be directly output to the student model for learning, thereby effectively improving the accuracy of model prediction.

[0097] It should be further explained that the preset loss function refers to a pre-set loss function that connects the feature layer of the teacher model with the student model using a mean square error. In this embodiment, a connection loss function can be used for connection. The connection loss function is specifically as follows:

[0098] L hin =∑ω KD k ||F tk -R(F s )|| 2 2

[0099] Among them, L hin is the loss function; k refers to the kth teacher model; F represents the feature layer; t is the target defect detection teacher model; s is the initial defect detection student model; R is the dimension adjustment function added to make the dimension of the feature layer connecting the teacher model and the student model the same.

[0100] Therefore, according to the properties of this loss function, the feature layer of the student model during training can be brought closer to the feature layer of the teacher model group, thereby achieving the effect of receiving feature layer information.

[0101] Specifically, according to a preset loss function, such as a connection loss function, the initial defect detection student model is connected to the feature layer of the target defect detection teacher model, so that the initial defect detection student model can calculate the feature gap corresponding to the feature layer of the target defect detection teacher model, wherein the feature gap refers to the distance between the features of the student model at a specific intermediate layer and the features of the corresponding layer of the teacher model (such as the L1 or L2 norm). There is no restriction on the calculation method of the feature gap, which can be set according to actual conditions.

[0102] In addition, the target defect detection teacher model also includes an aggregation layer. In this embodiment, the preliminary classification probability of each teacher model passes through the aggregation layer after passing through the softmax function with a temperature coefficient added, and the hidden knowledge extracted from the soft labels of multiple output teacher models is averaged to ensure that the knowledge of the soft labels passed to the student network can represent all the teacher model levels of the teacher model group.

[0103] Step S43, inputting the soft label, the feature information and the screen image data into the initial defect detection student model for iterative training to obtain a target defect detection student model.

[0104] Specifically, the soft labels, the feature information, and the screen image data are preprocessed, such as data cleaning and normalization, so that the soft labels, the feature information, the screen image data, and the sample labels corresponding to the soft labels, the feature information, and the screen image data are input into several initial defect detection student models to obtain the predicted values ​​output by the initial defect detection student models, and then based on the predicted values ​​and the sample labels, the model loss value is calculated using the loss function. In this embodiment, the loss function can be set according to actual needs and is not specifically limited here. After the model loss value is calculated, the training process ends, and the error back propagation algorithm and gradient update are used to update the model parameters in the initial defect detection student model, and then the next training is carried out, wherein the student model adjusts the parameters through back propagation so that the feature space structure extracted by the student model in the middle layer replicates the feature space of the teacher model as much as possible. At the same time, the student model should not only match the real data label, but also try to make the feature representation of its internal layer close to the feature representation of the teacher model under the same input.

[0105] During the training process, it is determined whether the updated initial defect detection student model meets the preset training end conditions. If so, the updated initial defect detection student model is used as the target defect detection student model. If not, the model training continues, wherein the preset training end conditions include loss convergence and reaching the maximum number of iterations threshold, etc.

[0106] Furthermore, since the prediction results generated by the teacher model for the same input data reflect high-level relational reasoning and spatial understanding, the student model needs to try to imitate the behavior of the teacher model. For example, during training, it not only minimizes its own prediction error for unlabeled data, but also adjusts its own learning goals according to the soft targets provided by the teacher model, that is, tries to make its output close to the output of the teacher model.

[0107] This embodiment obtains the soft label and screen image data of the target defect detection teacher model, and then connects the initial defect detection student model with the feature layer of the target defect detection teacher model based on a preset loss function, so that the initial defect detection student model can calculate the feature gap corresponding to the feature layer of the target defect detection teacher model, thereby inputting the soft label, the feature information and the screen image data into the initial defect detection student model for iterative training to obtain the target defect detection student model, and then making the feature layer of the student model during training closer to the feature layer of the teacher model group, so as to realize the distillation of the learning capabilities of multiple teacher models into the student model. The student model has the average learning capabilities of multiple teacher models, so that the prediction results output by the student model are consistent with the prediction results output by the teacher model, so as to improve the accuracy of the student model prediction. At the same time, through the characteristics of lightweight and other characteristics of the student model, it can also be used in edge devices with low computing power to save the cost of equipment upgrade.

[0108] In a feasible implementation manner, the step of obtaining a soft label of a target defect detection teacher model includes:

[0109] Step S51, inputting several groups of screen image data into a target defect detection teacher model to obtain several prediction results output by the target defect detection teacher model;

[0110] It should be noted that the prediction result refers to the defect detection prediction result about the screen image data.

[0111] Step S52: generating a soft label corresponding to the target defect detection teacher model based on each of the prediction results.

[0112] Specifically, in one embodiment, a preliminary classification probability is determined based on each of the prediction results, and then the preliminary classification probability is passed through a softmax function with a temperature coefficient added and then passed through an aggregation layer, and the hidden knowledge extracted from the softened labels corresponding to multiple teacher models in the target defect detection teacher model is averaged to generate a soft label corresponding to the target defect detection teacher model, wherein the adjustment of the temperature parameter can make the probability distribution smoother or more prominent, and the specific adjustment depends on the requirements of the soft label, which is not limited here.

[0113] Additionally, in another embodiment, the soft label may be generated as a weighted average or other combination of the output probabilities of multiple teacher models, thereby integrating the opinions of multiple models to provide more robust and comprehensive soft labels and improve detection accuracy.

[0114] This embodiment obtains several prediction results output by the target defect detection teacher model by inputting several groups of screen image data into the target defect detection teacher model, and then generates soft labels corresponding to the target defect detection teacher model based on each of the prediction results, thereby ensuring that the knowledge of the softened labels passed to the student network can represent all teacher model levels of the teacher model group, so as to pass more information and background knowledge to the student model, so that the student model can better understand and process data, thereby improving the student model performance and accelerating the training process of the student model. At the same time, because the teacher model conveys comprehensive judgments on different situations, it provides more comprehensive training signals, so that the student model can better cope with various changes and noises.

[0115] In a feasible implementation manner, after inputting the screen detection image into the target defect detection student model and obtaining the screen defect detection result output by the target defect detection student model, the method further includes:

[0116] Step S61, generating a screen defect detection report based on the screen defect detection result;

[0117] It should be noted that the screen defect detection report refers to a detection report generated based on the screen defect detection results, wherein the screen defect detection report may include the production information, detection information, detection results and analysis of the detection results of the target screen, so that the inspectors can quickly understand the detection results of the target screen and clarify the repair direction through the screen defect detection report, and there is no limitation here.

[0118] Specifically, the screen defect detection result may be detected and analyzed according to a detection result analysis tool, and then the analysis result may be associated and combined with the generation information of the target screen, etc., to generate a screen defect detection report.

[0119] Step S62, pushing the screen defect detection report to the detection personnel, so that the detection personnel can repair the target screen;

[0120] Specifically, the screen defect detection report is pushed to the inspector so that the inspector can repair the target screen, wherein the pushing method includes website, APP, mini-program and email, etc., which are not limited here, so that the inspector can view the screen defect detection report through various devices, evaluate the screen defect detection report, and then repair the target screen.

[0121] Step S63, obtaining screen defect detection data of the target screen after repair;

[0122] It should be noted that the screen defect detection data refers to relevant data used to detect and identify screen defects, including screen image data (such as photographic images of the screen taken under different lighting conditions, different angles and different environments, or screenshots of the screen area, etc.), defect locations in the image, defect type attributes (such as scratches, bright spots, oil spots, etc.), and screen defect size, shape, color and other information. There is no limitation here and it can be set according to actual conditions to more accurately judge the defect situation of the target screen.

[0123] Specifically, the screen defect detection data of the target screen after repair can be obtained through screen detection equipment, and screen testing software can be used to perform a comprehensive screen test, so as to detect various defects on the screen and the type, location and severity of the defects, such as dead spots, bright spots, color spots, lines, color, grayscale, brightness, contrast, etc. At the same time, the performance and potential defects of the screen under different conditions can be detected, thereby improving the diversity and comprehensiveness of screen defect detection data. There is no limitation here and it can be set according to actual conditions.

[0124] Step S64: determining the screen repair effect based on the screen defect detection data.

[0125] It should be noted that the screen repair effect refers to the degree to which the screen display quality is improved after the defects or damage on the screen are processed through specific technologies or methods, including the repair effect level, repair defect type and repair method, etc.

[0126] Specifically, the screen defect detection data is compared with the screen defect detection data in the screen defect detection report before repair, so as to pay attention to whether the previous problems (such as dead spots, bright spots) exist after repair, and whether the repair effect meets the expected standards (such as color accuracy and whether grayscale performance is improved). Then, the digitized screen detection data is quantitatively analyzed. For example, the changes in the number of dead spots and bright spots are calculated to evaluate the degree of improvement in color errors or grayscale non-uniformity. Then, the comparison results and analysis results are analyzed and integrated to determine the screen repair effect. At the same time, by analyzing the data before and after the repair, the repair process and strategy can be continuously optimized to improve the effectiveness of future repair work.

[0127] This embodiment generates a screen defect detection report based on the screen defect detection result, and pushes the screen defect detection report to the inspector so that the inspector can repair the target screen and then obtain the screen defect detection data of the target screen after the repair, so as to determine the screen repair effect based on the screen defect detection data, thereby improving the efficiency of the repair work and the repair quality, reducing the screen return rate, and saving enterprise expenditure costs.

[0128] In a feasible implementation manner, after determining the screen repair effect based on the screen defect detection data, the method further includes:

[0129] Step S71, determining a repair effect level of the screen repair effect;

[0130] It should be noted that the repair effect level refers to the comparative evaluation between the repaired screen and the original state or ideal state. The repair effect level can be set according to different standards and requirements. For example, the repair effect level includes perfect repair (screen defects are completely repaired, no repair marks can be detected, and the display effect is exactly the same as a new screen or an undamaged screen), good repair (screen defects have been significantly improved. Although some repair marks may still be detected, it will not have a negative impact on the user experience), general repair (screen defects have improved, but the repair marks are still obvious, which may have a certain impact on the user experience), poor repair (screen defects are only partially repaired, the repair marks are obvious, and the user experience is greatly affected), unrepaired, etc. The repair effect level can be set according to actual needs and is not limited here.

[0131] Specifically, the repair effect level of the screen repair effect may be determined by evaluation using a detection tool or data feedback from a detection person, and is not limited here.

[0132] Step S72, if the repair effect level meets the preset effect level, obtaining the screen repair method of the inspector;

[0133] It should be noted that the preset effect level refers to a pre-set repair effect level that meets the requirements. In addition, the screen repair method refers to the improvement behavior and method made by the inspection personnel on the screen defects, such as adjusting the screen parameters, etc., which is not limited here and can be set according to the actual situation.

[0134] Specifically, if the repair effect level meets the preset effect level, the screen repair method of the inspector is obtained.

[0135] Step S73, associating and combining the screen repair method with the screen defect detection report corresponding to the screen repair effect, generating a screen repair record, and storing the screen repair record in a preset database.

[0136] It should be noted that the screen repair record is used to record screen defects and effective methods for solving the screen defects, so as to optimize the subsequent screen defect repair process and improve the repair efficiency. In addition, the preset database is used to store screen repair records of various screen defects, so that the inspection personnel or the inspection system can quickly obtain effective repair methods for screen defects, further improving the repair efficiency.

[0137] This embodiment determines the repair effect level of the screen repair effect, and if the repair effect level meets the preset effect level, obtains the screen repair method of the inspector, thereby associating and combining the screen repair method with the screen defect detection report corresponding to the screen repair effect, generates a screen repair record, and stores the screen repair record in a preset database, thereby improving the repair quality and efficiency, and at the same time facilitating the inheritance and sharing of screen repair knowledge, thereby effectively reducing the amount of screen damage and improving the satisfaction of enterprises and users.

[0138] For example, to help understand the implementation process of the screen defect detection method, please refer to Figure 4 , Figure 4 An example diagram of the process implementation provided for the screen defect detection method of this application.

[0139] It should be noted that the examples in the figures are only used to understand the present application and do not constitute a limitation on the screen defect detection method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0140] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0141] This application also provides a screen defect detection device, please refer to Figure 5 , the screen defect detection device comprises:

[0142] An acquisition module 51 is used to acquire a screen detection image of a target screen;

[0143] The detection module 52 is used to input the screen detection image into the target defect detection student model to obtain the screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on the target defect detection teacher model.

[0144] The screen defect detection device is also used for:

[0145] Acquire several groups of screen image data;

[0146] Inputting each of the screen image data and the sample labels corresponding to each of the screen image data into one or more initial defect detection teacher models for iterative training to obtain a plurality of pending defect detection teacher models;

[0147] Based on each of the pending defect detection teacher models, the target defect detection teacher model is determined.

[0148] The screen defect detection device is also used for:

[0149] Inputting a plurality of sets of screen image data into each of the pending defect detection teacher models to obtain a predicted value output by each of the pending defect detection teacher models;

[0150] Obtaining a standard data label, and calculating a cross entropy value between the predicted value and the standard data label;

[0151] Based on the cross entropy value, a target defect detection teacher model is determined.

[0152] The screen defect detection device is also used for:

[0153] Obtain soft labels and screen image data of the target defect detection teacher model;

[0154] Based on a preset loss function, the initial defect detection student model is connected to the feature layer of the target defect detection teacher model, so that the initial defect detection student model calculates the feature gap corresponding to the feature layer of the target defect detection teacher model;

[0155] The soft label, the feature information and the screen image data are input into an initial defect detection student model for iterative training to obtain a target defect detection student model.

[0156] The screen defect detection device is also used for:

[0157] Inputting several groups of screen image data into a target defect detection teacher model to obtain several prediction results output by the target defect detection teacher model;

[0158] Based on each of the prediction results, a soft label corresponding to the target defect detection teacher model is generated.

[0159] The screen defect detection device is also used for:

[0160] Based on the screen defect detection result, generating a screen defect detection report;

[0161] Pushing the screen defect detection report to the inspector so that the inspector can repair the target screen;

[0162] Acquiring screen defect detection data of the target screen after repair;

[0163] Based on the screen defect detection data, a screen repair effect is determined.

[0164] The screen defect detection device is also used for:

[0165] Determining a repair effect level of the screen repair effect;

[0166] If the repair effect level meets the preset effect level, the screen repair method of the inspector is obtained;

[0167] The screen repair method is associated and combined with a screen defect detection report corresponding to the screen repair effect to generate a screen repair record, and the screen repair record is stored in a preset database.

[0168] The screen defect detection device provided by the present application adopts the screen defect detection method in the above embodiment, which can solve the technical problems in the background technology. Compared with the prior art, the beneficial effects of the screen defect detection device provided by the present application are the same as the beneficial effects of the screen defect detection method provided by the above embodiment, and other technical features in the screen defect detection device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0169] The present application provides a screen defect detection device, which includes: at least one processor; and a memory that is communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the screen defect detection method in the above-mentioned embodiment one.

[0170] Reference below Figure 6, which shows a schematic diagram of the structure of a screen defect detection device suitable for implementing an embodiment of the present application. The screen defect detection device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The screen defect detection device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0171] like Figure 6 As shown, the screen defect detection device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the screen defect detection device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the screen defect detection device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a screen defect detection device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0172] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0173] The screen defect detection device provided by the present application adopts the screen defect detection method in the above embodiment, which can solve the technical problems in the background technology. Compared with the prior art, the beneficial effects of the screen defect detection device provided by the present application are the same as the beneficial effects of the screen defect detection method provided by the above embodiment, and the other technical features in the screen defect detection device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0174] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0175] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0176] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the screen defect detection method in the above-mentioned embodiment.

[0177] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0178] The computer-readable storage medium may be included in the screen defect detection device; or may exist independently without being assembled into the screen defect detection device.

[0179] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the screen defect detection device, the screen defect detection device:

[0180] Acquire a screen detection image of the target screen;

[0181] The screen detection image is input into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on a target defect detection teacher model.

[0182] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0183] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0184] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0185] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned screen defect detection method, and can solve the technical problems in the background technology. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the screen defect detection method provided in the above-mentioned embodiment, and will not be repeated here.

[0186] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the above-mentioned screen defect detection method when executed by a processor.

[0187] The computer program product provided in this application can solve the technical problems in the background technology. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiment of this application are the same as the beneficial effects of the screen defect detection method provided in the above embodiment, which will not be repeated here.

[0188] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A screen defect detection method, characterized in that: include: Acquire a screen detection image of the target screen; The screen detection image is input into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on a target defect detection teacher model.

2. The screen defect detection method according to claim 1, characterized in that: Before inputting the screen detection image into the target defect detection student model to obtain the screen defect detection result output by the target defect detection student model, the method further includes: Acquire several groups of screen image data; Inputting each of the screen image data and the sample labels corresponding to each of the screen image data into one or more initial defect detection teacher models for iterative training to obtain a plurality of pending defect detection teacher models; Based on each of the pending defect detection teacher models, the target defect detection teacher model is determined.

3. The screen defect detection method according to claim 2, characterized in that: The step of determining the target defect detection teacher model based on each of the pending defect detection teacher models includes: Inputting a plurality of sets of screen image data into each of the pending defect detection teacher models to obtain a predicted value output by each of the pending defect detection teacher models; Obtaining a standard data label, and calculating a cross entropy value between the predicted value and the standard data label; Based on the cross entropy value, a target defect detection teacher model is determined.

4. The screen defect detection method according to claim 1, characterized in that: The target defect detection teacher model includes a feature layer; The training process of the target defect detection student model includes: Obtain soft labels and screen image data of the target defect detection teacher model; Based on a preset loss function, the initial defect detection student model is connected to the feature layer of the target defect detection teacher model, so that the initial defect detection student model calculates the feature gap corresponding to the feature layer of the target defect detection teacher model; The soft label, the feature information and the screen image data are input into an initial defect detection student model for iterative training to obtain a target defect detection student model.

5. The screen defect detection method according to claim 4, characterized in that: The step of obtaining a soft label of a target defect detection teacher model includes: Inputting several groups of screen image data into a target defect detection teacher model to obtain several prediction results output by the target defect detection teacher model; Based on each of the prediction results, a soft label corresponding to the target defect detection teacher model is generated.

6. The screen defect detection method according to claim 1, characterized in that: After inputting the screen detection image into the target defect detection student model and obtaining the screen defect detection result output by the target defect detection student model, the method further includes: Based on the screen defect detection result, generating a screen defect detection report; Pushing the screen defect detection report to the inspector so that the inspector can repair the target screen; Acquiring screen defect detection data of the target screen after repair; Based on the screen defect detection data, a screen repair effect is determined.

7. The screen defect detection method according to claim 6, characterized in that: After determining the screen repair effect based on the screen defect detection data, the method further includes: Determining a repair effect level of the screen repair effect; If the repair effect level meets the preset effect level, the screen repair method of the inspector is obtained; The screen repair method is associated and combined with a screen defect detection report corresponding to the screen repair effect to generate a screen repair record, and the screen repair record is stored in a preset database.

8. A screen defect detection device, characterized in that: include: An acquisition module, used for acquiring a screen detection image of a target screen; A detection module is used to input the screen detection image into a target defect detection student model to obtain a screen defect detection result output by the target defect detection student model, wherein the target defect detection student model is obtained by iterative training based on a target defect detection teacher model.

9. A screen defect detection device, characterized in that: The screen defect detection device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the screen defect detection method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the screen defect detection method according to any one of claims 1 to 7 are implemented.

Images