Defect classification method and computer-readable storage medium

Abstract

This application relates to a defect classification method, apparatus, computer device, readable storage medium, and program product. The method includes: driving a target display unit in a target display panel based on multiple first driving voltages, and acquiring first optical detection data corresponding to each of the multiple first driving voltages; inputting the multiple first driving voltages and the first optical detection data corresponding to each of the multiple first driving voltages into a pre-trained multilayer perceptron to predict at least one second driving voltage and second optical detection data corresponding to the second driving voltage; and determining the defect classification result of the target display unit based on the multiple first driving voltages, the first optical detection data corresponding to the multiple first driving voltages, the at least one second driving voltage, and the second optical detection data corresponding to the second driving voltage. This method can quickly and accurately classify whether defects are compensable.

Description

Technical Field

[0001] This application relates to the field of display technology, and in particular to a defect classification method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] With the development of display technology, the demand for and application scope of display panels are constantly expanding, and they are widely used in mobile phones, televisions, automotive displays, wearable devices, tablets, laptops, commercial displays and other fields.

[0003] In the manufacturing process of display panels such as Liquid Crystal Display (LCD) and Organic Light-Emitting Diode (OLED), Mura is a common visual defect that manifests as areas on the screen with uneven brightness, color, or contrast (e.g., stripes, patches, or localized color shifts), affecting the display effect.

[0004] Currently, in the production process of display panels, the determination of whether Mura (damage) is compensable mainly relies on human experience and traditional algorithms, which suffers from a high misjudgment rate (for example, misjudging compensable brightness deviations as irreparable physical damage). Therefore, there is an urgent need for a defect classification method that can quickly and accurately determine whether Mura is compensable. Summary of the Invention

[0005] Therefore, it is necessary to provide a defect classification method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can quickly and accurately classify whether a defect is compensable, in order to address the aforementioned technical problems.

[0006] Firstly, this application provides a defect classification method, including:

[0007] The target display unit in the target display panel is driven by multiple first driving voltages, and the first optical detection data corresponding to the multiple first driving voltages are obtained respectively;

[0008] Multiple first driving voltages and the first optical detection data corresponding to the multiple first driving voltages are input into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage. The pre-trained multilayer perceptron is trained using a loss function that includes physical constraint terms.

[0009] Based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage, and second optical detection data corresponding to the second driving voltage, the defect classification result of the target display unit is determined, and the defect classification result includes compensable / uncompensable.

[0010] In one embodiment, the training method of the multilayer perceptron includes:

[0011] A multilayer perceptron is trained using a loss function that includes a first constraint term and / or a second constraint term, based on multiple sample data points. Each sample data point includes a sample driving voltage and corresponding sample optical detection data. The first constraint term is used to constrain the smoothness of the predicted data point sequence output by the multilayer perceptron, which consists of the predicted driving voltage and the corresponding predicted optical detection data. The second constraint term is used to constrain the monotonicity of the predicted optical detection data as the predicted driving voltage changes in the predicted data point sequence.

[0012] In one embodiment, the loss function further includes a third constraint term, which is used to minimize the prediction error of the multilayer perceptron for the sample data points.

[0013] In one embodiment, based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage, and second optical detection data corresponding to the second driving voltage, the defect classification result of the target display unit is determined as follows:

[0014] Based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage, and second optical detection data corresponding to the second driving voltage, a luminous efficiency curve corresponding to the target display unit is formed. The luminous efficiency curve is used to represent the functional relationship between the optical detection data and the predicted driving voltage.

[0015] Based on the luminous efficiency curve, the defect classification results of the target display unit are determined.

[0016] In one embodiment, before inputting the plurality of first driving voltages and the first optical detection data corresponding to the plurality of first driving voltages into the pre-trained multilayer perceptron, the following steps are included:

[0017] The first optical detection data is normalized based on the corresponding expected optical detection data.

[0018] And / or, the first driving voltage is normalized based on the threshold voltage and anchor voltage of the target display panel.

[0019] In one embodiment, before driving the target display unit in the target display panel based on a plurality of first driving voltages and acquiring first optical detection data corresponding to the plurality of first driving voltages, the process includes:

[0020] Drive the target display unit in the target display panel based on the initial driving voltage, and acquire the current optical detection data of the target display unit;

[0021] Based on the deviation between the current optical detection data and the expected optical detection data corresponding to the initial driving voltage, the initial driving voltage is calibrated to obtain the corresponding first driving voltage.

[0022] Secondly, this application also provides a defect classification device, comprising:

[0023] The data acquisition module is used to drive the target display unit in the target display panel based on multiple first driving voltages, and acquire the first optical detection data corresponding to the multiple first driving voltages respectively;

[0024] The data generation module is used to input multiple first driving voltages and the first optical detection data corresponding to the multiple first driving voltages into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage. The pre-trained multilayer perceptron is trained using a loss function that includes physical constraint terms.

[0025] The defect classification module is used to determine the defect classification result of the target display unit based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage and second optical detection data corresponding to the second driving voltage. The defect classification result includes compensateable / uncompensable.

[0026] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the above-mentioned embodiments.

[0027] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0028] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the above claims.

[0029] The aforementioned defect classification methods, devices, computer equipment, computer-readable storage media, and computer program products sample the actual driving voltage and optical response (optical detection data) of the display unit and input them into a pre-trained multilayer perceptron to predict the optical response of the display unit under other driving voltages. This enables data point completion between a small number of actual sampling points. Furthermore, the pre-trained multilayer perceptron is trained using a loss function that includes physical constraints, ensuring that the predicted output conforms to the basic physical laws of display devices. Therefore, classifying whether display unit defects are compensable based on the actual driving voltage and its optical response, as well as the predicted driving voltage and its optical response, allows for comprehensive judgment based on more complete and physically consistent luminescence characteristic data, improving the efficiency and accuracy of classification. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a diagram illustrating the application environment of a defect classification method in one embodiment.

[0032] Figure 2 This is a flowchart illustrating a defect classification method in one embodiment;

[0033] Figure 3 This is a flowchart illustrating step S206 in one embodiment;

[0034] Figure 4 This is a flowchart illustrating the defect classification method in another embodiment;

[0035] Figure 5 This is a flowchart illustrating the defect classification method in yet another embodiment;

[0036] Figure 6 This is a structural block diagram of a defect classification device in one embodiment;

[0037] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0039] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0040] Traditional algorithms rely solely on optical detection data, making it difficult to accurately fit the luminous efficiency of display units (e.g., OLEDs). This results in a high misjudgment rate of whether defects (e.g., muras) can be compensated, or failure to compensate (e.g., overcompensation leading to the generation of new muras outside the original mura area).

[0041] Based on this, embodiments of this application provide a defect classification method that can be applied to, for example... Figure 1 In the application environment shown, the defect classification system 10 includes a signal generator 11, an optical inspection device 12, and a computer device 13. The signal generator 11 can output pattern signals corresponding to multiple first driving voltages to the target display panel 00. The optical inspection device 12 can acquire first optical inspection data corresponding to the multiple first driving voltages. The optical inspection device 12 can include a dot-matrix colorimeter probe 121 and an area array imaging colorimeter 122. The area array imaging colorimeter 122 can be an Automated Optical Inspection (AOI) camera 122 with colorimetric measurement capabilities. The computer device 13 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and IoT devices.

[0042] In one exemplary embodiment, such as Figure 2 As shown, a defect classification method is provided, which can be applied to... Figure 1 Taking computer device 13 as an example, the explanation includes the following steps S202 to S206. Wherein:

[0043] S202, drive the target display unit in the target display panel based on multiple first driving voltages, and acquire the first optical detection data corresponding to the multiple first driving voltages respectively.

[0044] In this implementation, multiple first driving voltages can be used to drive target display units in the target display panel to display different grayscale levels. Each first driving voltage can be used to simultaneously drive one or more target display units. A target display unit can be any display unit in the target display panel (e.g., a pixel or pixel area). In one possible implementation, all display units in the target display panel can be used as target display units, and the target display panel can be driven to display different grayscale images based on the multiple first driving voltages to obtain optical detection data of each target display unit within a preset brightness range.

[0045] For example, please continue to refer to Figure 1 The computer device 13 can drive the target display unit in the target display panel through the signal generator 11 based on multiple first driving voltages. Figure 1 (Not shown in the image) The optical detection device 12 can simultaneously acquire optical detection data of the target display unit to obtain first optical detection data corresponding to multiple first driving voltages. Here, the first optical detection data includes, but is not limited to, brightness data, chromaticity data, or contrast data.

[0046] In one possible implementation, optical detection data of the target display panel can be acquired using a high-resolution imaging colorimeter to obtain an XYZ image, where the Y component represents luminance, and the X and Z components, combined with the Y component, can be used to calculate chromaticity. For each XYZ image, each pixel location (h, w) has a set of XYZ data. For each grayscale m, there corresponds to a first driving voltage. Furthermore, for each target display unit in each target display panel i, a multimodal dataset can be obtained. Each sampled data point can be represented as That is, including the first driving voltage And the resulting first optical detection data XYZ. Here, This can correspond to the voltage combination of each color sub-pixel (e.g., red, green, and blue sub-pixels) within the target display unit. .

[0047] S204, multiple first driving voltages and the first optical detection data corresponding to the multiple first driving voltages are input into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage. The pre-trained multilayer perceptron is trained using a loss function that includes physical constraint terms.

[0048] For example, multiple first driving voltages and the first optical detection data corresponding to each of the multiple first driving voltages can be normalized and fused before being input into a pre-trained multilayer perceptron. In one possible implementation, the fusion processing can be based on a first formula, which includes:

[0049] (1)

[0050] The loss function, which includes a physical constraint term, enables the multilayer perceptron to learn an input-output mapping that conforms to the physical laws governing the illumination of display panels (including normal and defective display panels) during training. The input to the pre-trained multilayer perceptron can be M sampled data points corresponding to at least one target display unit. Here, M is the number of gray levels sampled. The output of the pre-trained multilayer perceptron can be N predicted data points. Here, N is the number of predicted data points (e.g., 5). Each predicted data point can be represented as... That is, including the second driving voltage And predicting the resulting second optical detection data In one possible implementation, for each target display unit, its corresponding predicted data point can be an interpolation point between corresponding sampled data points. For example, the second driving voltage can be an interpolation between corresponding first driving voltages.

[0051] S206, based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages respectively, at least one second driving voltage and second optical detection data corresponding to the second driving voltage, determine the defect classification result of the target display unit, the defect classification result including compensable / uncompensable.

[0052] Understandably, for each target display unit, the computer device 13 can obtain a denser set of data points (e.g., expanding from containing M sampled data points to containing M sampled data points and N predicted data points). Furthermore, the defect classification result of the target display unit can be determined based on the characteristics of this set of data points. Here, the defect can be a mura.

[0053] In the aforementioned defect classification method, the actual driving voltage and optical response (optical detection data) of the display unit are sampled and input into a pre-trained multilayer perceptron to predict the optical response of the display unit under other driving voltages. This method can complete the data points between a small number of actual sampling points. Furthermore, the pre-trained multilayer perceptron is trained using a loss function that includes physical constraints, ensuring that the predicted output conforms to the basic physical laws of the display device. Therefore, classifying whether the defects of the display unit are compensable based on the actual driving voltage and its optical response, as well as the predicted driving voltage and its optical response, allows for a comprehensive judgment based on more complete and physically consistent luminous characteristic data, thus improving the efficiency and accuracy of the classification.

[0054] In some embodiments, the training method of the above-mentioned multilayer perceptron includes:

[0055] A multilayer perceptron is trained using a loss function that includes a first constraint term and / or a second constraint term, based on multiple sample data points. Each sample data point includes a sample driving voltage and corresponding sample optical detection data. The first constraint term is used to constrain the smoothness of the predicted data point sequence output by the multilayer perceptron, which consists of the predicted driving voltage and the corresponding predicted optical detection data. The second constraint term is used to constrain the monotonicity of the predicted optical detection data as the predicted driving voltage changes in the predicted data point sequence.

[0056] The predicted data point sequence can refer to a set of ordered data points output by a multilayer perceptron, which are arranged according to the magnitude of the predicted driving voltage.

[0057] For example, multiple normalized sample data points can be used as input to a multilayer perceptron. The learning objective of the multilayer perceptron can be to predict interpolation points (i.e., predicted data points) located between these known sample data points. During training, the sum of squares or absolute values ​​of the differences between predicted optical detection data (e.g., brightness values) of adjacent predicted data points (adjacent on the voltage axis) can be calculated as a smoothing loss to penalize sudden jumps in the predicted data point sequence, thus smoothing the output curve. The sum of the differences among all adjacent predicted points where the predicted brightness value decreases as the driving voltage increases can be calculated as a trend loss to penalize portions of the predicted data point sequence that violate the physical law that brightness increases monotonically with voltage.

[0058] In one embodiment, the loss function further includes a third constraint term, which is used to minimize the prediction error of the multilayer perceptron for the sample data points.

[0059] For example, during training, the deviation between the multilayer perceptron's predicted optical response to a sample driving voltage and the actual sample optical detection data acquired at that voltage can be calculated as the sampling point loss. For instance, the mean square error can be used to calculate this deviation, driving the multilayer perceptron's predicted output value for a known driving voltage to approximate its corresponding true measurement value. In one possible implementation, the above loss function can be determined based on a second formula, which includes:

[0060] (2)

[0061] in, The loss function is defined as the sampling point loss, which is the third constraint term. As the first constraint term, This is the second constraint term. , These are adjustable positive weighting coefficients. In one possible implementation, the sampling point loss can be determined based on a third formula, which includes:

[0062] (3)

[0063] in, For sampling point loss, For real data points, These are the predicted data points output by the multilayer perceptron.

[0064] In this embodiment, by using the smoothing loss constraint model training, the luminous response sequence generated by the model can be smooth and continuous, avoiding non-physical abrupt jumps or oscillations at interpolation points, thus more realistically reflecting the actual luminous characteristics of the display panel; by using the trend loss constraint model training, it can be ensured that the predicted brightness value monotonically does not decrease with the driving voltage, which conforms to actual physical laws; by using the sampling point loss constraint model training, it can be ensured that the model maintains a high-precision fit to limited known measurement data, reducing the deviation from the true data due to excessive pursuit of smoothness or trend.

[0065] In one embodiment, such as Figure 3 As shown, step S206 above may include:

[0066] S2061, based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages respectively, at least one second driving voltage and second optical detection data corresponding to the second driving voltage, a luminous efficiency curve corresponding to the target display unit is formed. The luminous efficiency curve is used to represent the functional relationship between the optical detection data and the predicted driving voltage.

[0067] S2062, Based on the luminous efficiency curve, determine the defect classification result of the target display unit.

[0068] For example, for each target display unit, the computer device 13 can merge all its known sampling data points (first driving voltage and first optical detection data) with the interpolated data points predicted by the multilayer perceptron (second driving voltage and second optical detection data) to form a denser set of data points. Then, a luminous efficiency curve characterizing the light emission characteristics of the target display unit can be formed by curve fitting (e.g., polynomial fitting, spline interpolation, etc.) or by directly connecting these ordered data points. It is understood that, since the multilayer perceptron is trained using a loss function that includes physical constraints, the luminous efficiency curve conforms to the physical laws of light emission from the display panel. Furthermore, the defect classification result can be determined based on the shape and characteristics of the luminous efficiency curve. For example, when the luminous efficiency curve shows an overall translation or regular deformation, it can be determined that the target display unit is compensable (i.e., there is no irreversible physical damage). When the luminous efficiency curve shows a broken or non-monotonic distortion, it can be determined that the target display unit is uncompensable (i.e., it cannot be compensated by the driver IC).

[0069] In this embodiment, by using a denser set of data points generated based on model prediction to fit the luminous efficiency curve, the fitted luminous efficiency curve can more accurately and realistically reflect the response characteristics of the target display unit at different gray levels.

[0070] In one embodiment, such as Figure 4 As shown, the above defect classification method may further include steps S2031 and / or S2032. Wherein:

[0071] S2031, the first optical detection data is normalized based on the corresponding expected optical detection data.

[0072] Here, the expected optical detection data refers to the ideal optical data value that the target display unit should achieve when a specific first driving voltage is applied. For optical detection data represented in the XYZ color space, the expected value of the luminance component Y can be used as the normalization reference. In one possible implementation, the normalized first optical detection data can be expressed as:

[0073] (4)

[0074] S2032, the first driving voltage is normalized based on the threshold voltage and anchor voltage of the target display panel.

[0075] The threshold voltage can refer to the minimum driving voltage required to make the target display unit start emitting light. The anchor voltage can be a selected reference voltage value, such as the maximum allowable driving voltage of the target display unit, the typical driving voltage at a certain standard gray level (e.g., the highest gray level), or a fixed system reference voltage. The threshold voltage can be used to offset and calibrate the first driving voltage. The anchor voltage can be used to scale the offset first driving voltage. In one possible implementation, the normalized first driving voltage can be expressed as:

[0076] (5)

[0077] In this embodiment, by normalizing the optical detection data input to the model, the influence of absolute brightness values ​​is eliminated. Without changing the color coordinates, the optical data can be converted to a scale based on ideal brightness, thereby directly highlighting the differences in relative luminous efficiency between different display units. By normalizing the driving voltage input to the model, the difference in driving start point caused by process fluctuations in different units is eliminated by using threshold voltage, and the range of driving voltage is normalized by using anchor voltage. This allows the model to focus more on learning the essential relationship between voltage change and optical response, effectively improving the model's generalization ability and adaptability to display panels of different specifications.

[0078] Furthermore, the above S204 may include:

[0079] S2041, the normalized first driving voltages are fused with the corresponding normalized first optical detection data to obtain multiple multimodal sampling vectors.

[0080] For example, the sampled M first driving voltages and corresponding first optical detection data can be normalized, and the m-th first driving voltage after normalization can be denoted as... The m-th first optical detection data after normalization can be denoted as... The fusion process can employ either direct concatenation or weighted fusion. When using direct concatenation along dimensions, a 6-dimensional multimodal sampling vector can be formed. .

[0081] Multiple normalized first driving voltages can be directly spliced ​​or weighted and fused with their corresponding normalized first optical detection data to obtain multiple multimodal sampling vectors.

[0082] S2043, input multiple multimodal sampling vectors into a pre-trained multilayer perceptron to predict at least one multimodal prediction vector; wherein, each multimodal prediction vector includes a second driving voltage and second optical detection data corresponding to the second driving voltage.

[0083] For example, a multilayer perceptron can output a set of N multimodal prediction vectors. N represents the number of predicted data points, and the second driving voltage can be the interpolated voltage between the first driving voltage and the second driving voltage. Each multimodal prediction vector can be denoted as... .

[0084] In this way, by fusing multimodal data, the electrical characteristics of the driving voltage and the optical characteristics of the optical detection data can be integrated into a unified multimodal feature vector, avoiding the information loss problem caused by a single feature input. This enables the multilayer perceptron to learn the essential mapping relationship between electrical and optical features, rather than independent single feature rules. Furthermore, the normalized multimodal data eliminates dimensional differences, avoiding model training bias caused by the different numerical scales of the driving voltage and optical data. This improves the prediction accuracy and stability of the multilayer perceptron for optical response under interpolated voltage, and adapts to the nonlinear light emission characteristics of the target display unit in the display panel.

[0085] In one embodiment, such as Figure 5 As shown, the above defect classification method may further include steps S2011 and S2012. Wherein:

[0086] S2011, drive the target display unit in the target display panel based on the initial driving voltage, and acquire the current optical detection data of the target display unit.

[0087] S2012, based on the deviation between the current optical detection data and the expected optical detection data corresponding to the initial driving voltage, the initial driving voltage is calibrated to obtain the corresponding first driving voltage.

[0088] For example, please continue to refer to Figure 1 The signal generator 11 and the dot colorimeter probe 121 can form a closed-loop feedback system. The computer device 13 can control the signal generator 11 to output an initial driving voltage signal corresponding to a specific grayscale to the target display panel 00 to drive it to display the corresponding image. Simultaneously, the dot colorimeter probe 121 measures the central area (or a specific calibration point) of the target display panel 00 to obtain current optical detection data (e.g., current brightness value). The computer device 13 can calculate the deviation between this current brightness value and a preset desired brightness value at this specific grayscale. If the absolute value of the deviation exceeds a preset tolerance threshold, the computer device 13 can fine-tune the voltage value output by the signal generator 11 through the feedback system and drive and measure again, iterating this process until the deviation between the current brightness value and the desired brightness value falls within the preset tolerance threshold. It is understood that this process can be performed separately for multiple different grayscales to obtain a set of calibrated first driving voltages that are optically aligned with the desired optical detection data.

[0089] In this embodiment, by performing optical calibration of the driving voltage before acquiring multimodal data, the multilayer perceptron model can be provided with more unified and purer input data, thereby further improving the accuracy of classifying whether defects are compensable.

[0090] In an exemplary embodiment, S204 may include: inputting a set of sampling data points corresponding to a plurality of target display units in the target display panel into a pre-trained multilayer perceptron to predict at least one predicted data point corresponding to the plurality of target display units; wherein each sampling data point in the set of sampling data points includes a first driving voltage and first optical detection data corresponding to the first driving voltage, and each predicted data point includes a second driving voltage and second optical detection data corresponding to the second driving voltage.

[0091] For example, the input to the pre-trained multilayer perceptron can be a sampled data point. Here, M is the number of sampled gray levels, which is the number of target display units (e.g., pixels) in the target display panel. The output of the pre-trained multilayer perceptron can be a predicted data point. In one possible implementation, the target display panel can be divided into multiple target display units by pixel blocks, the sampled data point set of each target display unit can be extracted, a batched input matrix can be constructed, this matrix can be input to the pre-trained multilayer perceptron, and the predicted data points of all target display units can be output simultaneously to achieve parallel prediction.

[0092] In an exemplary embodiment, S206 may include: in the front-end process stage, determining the defect classification result of the target display unit based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage and second optical detection data corresponding to the second driving voltage.

[0093] In this way, by determining the defect classification results of the target display unit in the pre-open cell (Pre-OC) stage of display panel manufacturing, material waste in the subsequent process stages can be reduced, and overall production efficiency and yield can be improved.

[0094] In summary, the above defect classification method, by sampling the actual driving voltage and optical response (optical detection data) of the display unit and inputting them into a pre-trained multilayer perceptron to predict the optical response of the display unit under other driving voltages, can achieve data point completion between a small number of actual sampling points. Furthermore, the pre-trained multilayer perceptron is trained using a loss function that includes physical constraints, ensuring that the predicted output conforms to the basic physical laws of display devices. Therefore, classifying whether display unit defects are compensable based on the actual driving voltage and its optical response, as well as the predicted driving voltage and its optical response, enables a comprehensive judgment based on more complete and physically consistent luminous characteristic data, thus improving the efficiency and accuracy of classification.

[0095] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0096] Based on the same inventive concept, this application also provides a defect classification apparatus for implementing the defect classification method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more defect classification apparatus embodiments provided below can be found in the limitations of the defect classification method described above, and will not be repeated here.

[0097] In one exemplary embodiment, such as Figure 6 As shown, a defect classification device 300 is provided, including: a data acquisition module 301, a data generation module 302, and a defect classification module 303, wherein:

[0098] The data acquisition module 301 is used to drive the target display unit in the target display panel based on multiple first driving voltages, and acquire the first optical detection data corresponding to the multiple first driving voltages respectively.

[0099] The data generation module 302 is used to input multiple first driving voltages and the first optical detection data corresponding to the multiple first driving voltages into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage. The pre-trained multilayer perceptron is trained using a loss function that includes physical constraint terms.

[0100] The defect classification module 303 is used to determine the defect classification result of the target display unit based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage and second optical detection data corresponding to the second driving voltage. The defect classification result includes compensateable / uncompensable.

[0101] In one embodiment, the defect classification device 300 further includes a model training module, used for:

[0102] A multilayer perceptron is trained using a loss function that includes a first constraint term and / or a second constraint term, based on multiple sample data points. Each sample data point includes a sample driving voltage and corresponding sample optical detection data. The first constraint term is used to constrain the smoothness of the predicted data point sequence output by the multilayer perceptron, which consists of the predicted driving voltage and the corresponding predicted optical detection data. The second constraint term is used to constrain the monotonicity of the predicted optical detection data as the predicted driving voltage changes in the predicted data point sequence.

[0103] In one embodiment, the loss function further includes a third constraint term, which is used to minimize the prediction error of the multilayer perceptron for the sample data points.

[0104] In one embodiment, the defect classification module 303 includes:

[0105] The curve generation submodule is used to form a luminous efficiency curve corresponding to the target display unit based on multiple first driving voltages, first optical detection data corresponding to the multiple first driving voltages, at least one second driving voltage, and second optical detection data corresponding to the second driving voltage. The luminous efficiency curve is used to represent the functional relationship between the optical detection data and the predicted driving voltage.

[0106] The defect classification submodule is used to determine the defect classification result of the target display unit based on the luminous efficiency curve.

[0107] In one embodiment, the defect classification device 300 further includes a normalization processing module, used for:

[0108] The first optical detection data is normalized based on the corresponding expected optical detection data.

[0109] And / or, the first driving voltage is normalized based on the threshold voltage and anchor voltage of the target display panel.

[0110] In one embodiment, the defect classification device 300 further includes a voltage calibration module for:

[0111] Drive the target display unit in the target display panel based on the initial driving voltage, and acquire the current optical detection data of the target display unit;

[0112] Based on the deviation between the current optical detection data and the expected optical detection data corresponding to the initial driving voltage, the initial driving voltage is calibrated to obtain the corresponding first driving voltage.

[0113] Each module in the aforementioned defect classification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0114] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a defect classification method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0115] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0116] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0117] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0118] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0119] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0121] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A defect classification method, characterized in that, The method includes: The target display unit in the target display panel is driven by multiple first driving voltages, and the first optical detection data corresponding to the multiple first driving voltages are obtained respectively; The plurality of first driving voltages and the first optical detection data corresponding to the plurality of first driving voltages are input into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage. The pre-trained multilayer perceptron is trained using a loss function that includes physical constraint terms. Based on the plurality of first driving voltages, the first optical detection data corresponding to the plurality of first driving voltages, at least one second driving voltage, and the second optical detection data corresponding to the second driving voltage, the defect classification result of the target display unit is determined, and the defect classification result includes compensable / uncompensable.

2. The method according to claim 1, characterized in that, The training methods for the multilayer perceptron include: A multilayer perceptron is trained based on multiple sample data points using a loss function that includes a first constraint term and / or a second constraint term. Each of the multiple sample data points includes a sample driving voltage and corresponding sample optical detection data. The first constraint term is used to constrain the smoothness of the predicted data point sequence output by the multilayer perceptron, which consists of the predicted driving voltage and the corresponding predicted optical detection data. The second constraint term is used to constrain the monotonicity of the predicted optical detection data as the predicted driving voltage changes in the predicted data point sequence.

3. The method according to claim 2, characterized in that, The loss function also includes a third constraint term, which is used to minimize the prediction error of the multilayer perceptron for the sample data points.

4. The method according to claim 1, characterized in that, The method of determining the defect classification result of the target display unit based on the plurality of first driving voltages, the first optical detection data corresponding to the plurality of first driving voltages, at least one second driving voltage, and the second optical detection data corresponding to the second driving voltage includes: Based on the plurality of first driving voltages, the first optical detection data corresponding to the plurality of first driving voltages, at least one second driving voltage, and the second optical detection data corresponding to the second driving voltage, a luminous efficiency curve corresponding to the target display unit is formed. The luminous efficiency curve is used to represent the functional relationship between the optical detection data and the predicted driving voltage. Based on the luminous efficiency curve, the defect classification result of the target display unit is determined.

5. The method according to claim 1, characterized in that, Before inputting the plurality of first driving voltages and the first optical detection data corresponding to the plurality of first driving voltages into the pre-trained multilayer perceptron, the following steps are included: The first optical detection data is normalized based on the corresponding expected optical detection data. And / or, the first driving voltage is normalized based on the threshold voltage and anchor voltage of the target display panel.

6. The method according to claim 5, characterized in that, The step of inputting the plurality of first driving voltages and the first optical detection data corresponding to the plurality of first driving voltages into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage includes: The normalized first driving voltages are fused with the corresponding normalized first optical detection data to obtain multiple multimodal sampling vectors. The plurality of multimodal sampling vectors are input into a pre-trained multilayer perceptron to predict at least one multimodal prediction vector; wherein each multimodal prediction vector includes a second driving voltage and second optical detection data corresponding to the second driving voltage.

7. The method according to claim 1, characterized in that, Before driving the target display unit in the target display panel based on multiple first driving voltages and acquiring the first optical detection data corresponding to the multiple first driving voltages, the process includes: Drive the target display unit in the target display panel based on the initial driving voltage, and acquire the current optical detection data of the target display unit; Based on the deviation between the current optical detection data and the expected optical detection data corresponding to the initial driving voltage, the initial driving voltage is calibrated to obtain the corresponding first driving voltage.

8. The method according to claim 1, characterized in that, The step of inputting the plurality of first driving voltages and the first optical detection data corresponding to the plurality of first driving voltages into a pre-trained multilayer perceptron to predict at least one second driving voltage and the second optical detection data corresponding to the second driving voltage includes: The sampling data point sets corresponding to the multiple target display units in the target display panel are jointly input into a pre-trained multilayer perceptron to predict at least one prediction data point corresponding to each of the multiple target display units; wherein, each sampling data point in the sampling data point set includes a first driving voltage and first optical detection data corresponding to the first driving voltage, and each prediction data point includes a second driving voltage and second optical detection data corresponding to the second driving voltage.

9. The method according to claim 1, characterized in that, The method of determining the defect classification result of the target display unit based on the plurality of first driving voltages, the first optical detection data corresponding to the plurality of first driving voltages, at least one second driving voltage, and the second optical detection data corresponding to the second driving voltage includes: In the front-end process, based on the plurality of first driving voltages, the first optical detection data corresponding to the plurality of first driving voltages, at least one second driving voltage and the second optical detection data corresponding to the second driving voltage, the defect classification result of the target display unit is determined.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 9.

Images