Silicon-based OLED brightness compensation method based on artificial intelligence

Abstract

The invention provides a silicon-based OLED brightness compensation method based on artificial intelligence. The method comprises the following steps: S1, obtaining a display image of a silicon-based OLED device; s2, converting the display image into a gray matrix, and expanding an environment temperature scalar into an environment temperature matrix with the same row and column number as the gray matrix; s3, splicing the two matrixes into a dual-channel input tensor input neural network model; s4, outputting a temperature distribution matrix through the neural network model, and obtaining a predicted temperature value of each pixel point according to the distribution matrix; s5, according to the temperature distribution matrix, obtaining a brightness compensation value of each pixel point by using a table look-up method for each pixel point; and S6, performing brightness compensation on the silicon-based OLED based on the brightness compensation value, generating a compensated display image, and outputting the compensated display image to the silicon-based OLED for display. The technical problem that compensation is not accurate enough due to the fact that the coupling relation between image gray level distribution and a temperature field is not fully considered in existing silicon-based OLED brightness compensation and pixel-level temperature-brightness accurate matching cannot be achieved is solved.

Description

Technical Field

[0001] This invention relates to the field of OLED display technology, and more particularly to a silicon-based OLED brightness compensation method based on artificial intelligence. Background Technology

[0002] Silicon-based OLEDs enable miniaturized displays with high resolution, high contrast, low power consumption, and low cost, making them a core component for next-generation mobile terminals such as augmented reality (AR) and virtual reality (VR). However, temperature changes can cause fluctuations in the brightness of silicon-based OLEDs when displaying the same grayscale, leading to unstable or even distorted display effects. The impact of temperature on pixel brightness stems not only from the environment but is also closely related to the content of the displayed image. Therefore, achieving pixel-level temperature sensing and brightness compensation has become a key challenge in improving the performance of silicon-based OLED displays.

[0003] Currently, temperature-brightness compensation technology for silicon-based OLEDs uses temperature sensors to detect temperature and then queries a pre-stored table showing the relationship between temperature and compensated brightness to obtain the common cathode compensation voltage, thus compensating for brightness drift caused by temperature changes. However, existing technologies fail to fully consider the coupling relationship between image content (such as grayscale distribution) and the temperature field, and typically rely only on measured images or average grayscale for modeling, making it impossible to achieve pixel-level precise temperature-brightness matching. Therefore, there is a problem of inaccurate brightness compensation. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based silicon-based OLED brightness compensation method, which solves the technical problem that existing silicon-based OLED brightness compensation technologies do not fully consider the coupling relationship between image grayscale distribution and temperature field, and cannot achieve pixel-level temperature-brightness precise matching, resulting in inaccurate compensation.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] This invention provides an artificial intelligence-based method for brightness compensation of silicon-based OLEDs, comprising the following steps: S1, acquiring a display image of a silicon-based OLED device; S2, converting the display image into a grayscale matrix, and expanding the ambient temperature scalar to an ambient temperature matrix with the same number of rows and columns as the grayscale matrix; S3, concatenating the grayscale matrix and the ambient temperature matrix into a dual-channel input tensor, and inputting it into a neural network model; S4, outputting a temperature distribution matrix through the neural network model, and obtaining a predicted temperature value for each pixel based on the temperature distribution matrix; S5, obtaining a pixel-by-pixel brightness compensation value for each pixel using a lookup table method based on the temperature distribution matrix; S6, performing point-to-point brightness compensation on the silicon-based OLED based on the brightness compensation value, generating a compensated display image, and outputting it to the silicon-based OLED display.

[0007] Furthermore, the conversion of the displayed image into a grayscale matrix in step S2 is obtained through the following steps: the grayscale matrix obtained by performing grayscale processing on the displayed image using an algorithm in MATLAB.

[0008] Furthermore, the structure of the neural network model described in step S3 includes: a dual-channel input layer, comprising an image grayscale matrix input layer and an ambient temperature matrix input layer of the same size, used to receive a dual-channel input tensor of size H×W×2; a two-stage encoder structure, each stage containing two convolutional layers with a kernel size of 3×3 and an activation function of ReLU, and extracting the input features through max pooling; and a decoding structure, which uses transposed convolution to upsample the extracted features to restore the original size, concatenates the upsampled feature map with the corresponding feature map from the encoding stage, and then further integrates the information through two convolutional layers, finally outputting a temperature distribution matrix of size 100×100 for a single channel.

[0009] Furthermore, the training set of the neural network model is established through the following steps: S01, establishing a multilayer temperature distribution simulation model of silicon-based OLED in finite element analysis software, defining the thermal conductivity and boundary conditions of each layer of material; S02, performing thermal distribution simulation of silicon-based OLED displaying images of different grayscale at different ambient temperatures, generating corresponding temperature distribution simulation results; S03, measuring the average surface temperature data of silicon-based OLED under different ambient temperatures and grayscale images using a high and low temperature test chamber; S04, calibrating the simulation results by adjusting the parameters of the thermal simulation model, so that the error between it and the measured data is less than a set threshold; S05, pairing the temperature distribution matrix generated by the calibrated thermal simulation model with the corresponding grayscale matrix and ambient temperature matrix to form a training sample group.

[0010] Furthermore, in step S01, a multilayer temperature distribution simulation model of silicon-based OLED is established, including but not limited to glass substrate, OCA, color filter, POL, encapsulation layer, transparent cathode, OLED layer, anode, metal interconnect layer and silicon substrate.

[0011] Furthermore, the specific steps for measuring the average surface temperature in step S03 include: bonding multiple thermocouples at equal intervals on the silicon-based OLED surface; dividing the ambient temperature into N values ​​and the grayscale into M values; collecting the thermocouple outputs for each combination of ambient temperature and grayscale values, and calculating the average temperature value.

[0012] Furthermore, in step S04, calibrating the simulation results by adjusting the parameters of the thermal simulation model includes adjusting the thermal conductivity of each layer of material and the convective heat transfer coefficient and thermal radiation coefficient of the surface in the simulation model.

[0013] Compared with the prior art, the present invention has at least the following beneficial effects:

[0014] This application achieves pixel-level temperature distribution prediction and pixel-by-pixel brightness compensation by concatenating the image grayscale matrix and the ambient temperature matrix into a dual-channel input tensor and inputting it into a neural network model. Combined with a training set generation mechanism that integrates finite element thermal simulation and measured data calibration, this approach overcomes the shortcomings of traditional techniques that do not fully consider the coupling relationship between image grayscale distribution and temperature field and cannot achieve pixel-level precise matching. This effectively improves the brightness compensation accuracy of silicon-based OLEDs, ensuring brightness consistency of the same grayscale in different display scenarios. At the same time, the sample generation mechanism that combines simulation and measurement enhances the model's generalization ability and the accuracy of temperature field prediction. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the silicon-based OLED brightness compensation method based on artificial intelligence according to the present invention.

[0017] Figure 2 This is a schematic diagram of the neural network model (outputting the temperature distribution matrix) of the present invention.

[0018] Figure 3 A flowchart illustrating the process of establishing a training set for training neural network models in this invention;

[0019] Figure 4 This is a visualization diagram of a set of training samples for this invention (actually a 100×100 numerical matrix). Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to specific embodiments.

[0022] like Figure 1 As shown, this invention provides a silicon-based OLED brightness compensation method based on artificial intelligence, the specific steps of which include:

[0023] S1. Obtain the display image of the silicon-based OLED device;

[0024] S2. Display the image and then process the grayscale matrix in MATLAB using an algorithm to expand the ambient temperature scalar into an ambient temperature matrix with the same number of rows and columns as the grayscale matrix.

[0025] S3. Concatenate the grayscale matrix and the ambient temperature matrix into a dual-channel input tensor, and input them into the neural network model.

[0026] S4. Obtain the temperature of each pixel based on the temperature distribution matrix output by the neural network model;

[0027] S5. Silicon-based OLEDs use a lookup table method to obtain the brightness compensation value for each pixel by generating a temperature distribution matrix through a neural network model for each pixel.

[0028] S6. The silicon-based OLED uses the compensated image generated after point-to-point brightness compensation as a new display image and sends it to the silicon-based OLED for display.

[0029] like Figure 2 As shown, the neural network model of this invention includes the following:

[0030] 2.1 A 100×100 display image is processed by an algorithm in MATLAB to obtain a grayscale matrix with 100 rows and 100 columns, which is used as the input of one channel of the network. The temperature scalar is expanded into a 100×100 temperature matrix, and concatenated with the grayscale matrix to form a dual-channel input tensor of size 100×100×2.

[0031] 2.2 The model front end is set with a two-stage encoder structure. Each stage contains two convolutional layers with a kernel size of 3×3 and an activation function of ReLU. The input features are compressed and extracted through max pooling.

[0032] 2.3 Decoding Structure: In the decoding stage, transposed convolution is used to upsample the extracted features to restore their original size. The upsampled feature map is then concatenated with the corresponding feature map from the encoding stage to achieve the fusion of low-level details and high-level semantic information. Subsequently, two convolutional layers are used to further integrate the information, ultimately outputting a temperature distribution matrix with a single-channel size of 100×100.

[0033] Preferably, the network uses the mean squared error loss function (MSE) as the optimization objective during training, uses the adaptive moment estimation Adam optimizer for parameter updates, and sets the initial learning rate to 0.001.

[0034] It is worth noting that, unlike existing technologies that use average grayscale or single-channel image input for brightness compensation, this invention introduces a structured dual-channel fusion mechanism, which can effectively extract the coupling characteristics of spatial features and temperature influencing factors, thereby improving the accuracy and generalization of the model in pixel-level temperature prediction.

[0035] Furthermore, the neural network model of this invention is not limited to a specific structural form and can be a Convolutional Neural Network (CNN), a Residual Network (ResNet), or other deep learning structures suitable for image regression prediction tasks. The network can contain several convolutional layers, pooling layers, and fully connected layers. Hyperparameters such as kernel size, number of layers, activation function, optimizer, and learning rate can be flexibly adjusted according to the specific display scenario, computing resources, and model accuracy requirements. The core of this invention lies in using the image grayscale matrix and the ambient temperature matrix as input to output pixel-level temperature distribution results, and is not limited to a specific network configuration.

[0036] like Figure 3-4 As shown, in this invention, to improve the diversity and physical accuracy of the neural network model training set, a training set generation mechanism combining finite element thermal simulation and measured data calibration is proposed. The specific steps include:

[0037] 1.1 A temperature distribution simulation model of silicon-based OLED is established in the steady-state thermal analysis module of the finite element analysis software. Preferably, in this embodiment, the model is based on a 1.03-inch silicon-based OLED.

[0038] 1.1.1 Establish a multilayer model of silicon-based OLED, including but not limited to glass substrate, OCA, color filter, POL, encapsulation layer, transparent cathode, OLED layer, anode, metal interconnect layer and silicon substrate;

[0039] 1.1.2 Define the thermal conductivity of each material layer and the boundary conditions of the model, including the convective heat transfer coefficient and the thermal radiation coefficient; the specific parameters are shown in Table 1.

[0040] Table 1

[0041]

[0042]

[0043] 1.2 Thermal distribution simulations were performed on silicon-based OLEDs displaying images of different grayscale levels under different ambient temperatures, and the corresponding temperature distribution simulation results T were obtained. i，j .

[0044] 1.3 A silicon-based OLED temperature distribution measurement platform was built in a high and low temperature test chamber.

[0045] 1.4 Obtaining the average temperature value T' of different grayscale images displayed on silicon-based OLEDs at different ambient temperatures i，j .

[0046] 1.4.1 Three K-type thermocouples are bonded at equal intervals from near to far on the surface of the silicon-based OLED;

[0047] 1.4.2 Set the temperature range of the high and low temperature test chamber to 0℃~50℃, divide it into 6 parts, and denot them as T. i T i (i = 1 to 6) represent 0℃, 10℃, 20℃, 30℃, 40℃, and 50℃ respectively. The temperature of the high and low temperature test chamber is divided into N parts, T i (i = 1 to N); Collect screen surface temperature under different grayscale values ​​of display images at different ambient temperatures to obtain some high-precision measurement data;

[0048] 1.4.3 Preferably, the grayscale of the displayed image is divided into 8 parts from 0 to 255, denoted as G. j (j=1~8),G j Given values ​​of 0, 50, 75, 100, 128, 150, 200, and 255, the grayscale of the displayed image will be divided into M parts, and the image G will be... j (j=1~M) are fed into a silicon-based OLED;

[0049] 1.4.4 Read the outputs of the three thermocouple thermometers, calculate the average temperature, and denot it as T'. i，j .

[0050] 1.5 The simulation results are calibrated by adjusting parameters such as the thermal conductivity of each layer of material in the simulation model, the convective heat transfer coefficient of the surface, and the thermal radiation coefficient, so that the simulation temperature T is adjusted accordingly. i，j and the actual test temperature T' i，j The difference between them is less than the set value ε. In this embodiment, ε = 1℃. The obtained average temperature data and simulation data, as well as the difference between them, are shown in Table 2:

[0051] Table 2

[0052]

[0053] 1.6 The pixel-level temperature distribution matrix generated in batches from the parameter-tuned thermal simulation model is paired with its corresponding input image grayscale matrix and ambient temperature matrix to form a training sample group, which is then input into the neural network model for training.

[0054] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A silicon-based OLED brightness compensation method based on artificial intelligence, characterized in that, Includes the following steps: S1. Obtain the display image of the silicon-based OLED device; S2. Convert the displayed image into a grayscale matrix, and expand the ambient temperature scalar into an ambient temperature matrix with the same number of rows and columns as the grayscale matrix; S3. Concatenate the grayscale matrix and the ambient temperature matrix into a dual-channel input tensor and input it into the neural network model; S4. Output a temperature distribution matrix through the neural network model, and obtain the predicted temperature value of each pixel based on the temperature distribution matrix; S5. Based on the temperature distribution matrix, obtain the brightness compensation value for each pixel using a lookup table method. S6. Perform point-to-point brightness compensation on the silicon-based OLED based on the brightness compensation value, generate the compensated display image, and output it to the silicon-based OLED display.

2. The silicon-based OLED brightness compensation method based on artificial intelligence according to claim 1, characterized in that, The conversion of the displayed image into a grayscale matrix in step S2 is obtained through the following steps: This will display the grayscale matrix obtained after the image has been processed into grayscale using an algorithm in MATLAB.

3. The silicon-based OLED brightness compensation method based on artificial intelligence according to claim 1, characterized in that, The structure of the neural network model described in step S3 includes: The dual-channel input layer includes an image grayscale matrix input layer and an ambient temperature matrix input layer of the same size, which is used to receive a dual-channel input tensor of size H×W×2; The two-stage encoder structure consists of two convolutional layers with a kernel size of 3×3 and an activation function of ReLU, and the input features are compressed and extracted through max pooling. The decoding structure uses transposed convolution to upsample the extracted features to restore the original size. The upsampled feature map is then concatenated with the corresponding feature map from the encoding stage. Subsequently, two layers of convolution are used to further integrate the information, and finally, a temperature distribution matrix with a single channel size of 100×100 is output.

4. The silicon-based OLED brightness compensation method based on artificial intelligence according to any one of claims 1-3, characterized in that, The training set for the neural network model is established through the following steps: S01. Establish a multilayer temperature distribution simulation model of silicon-based OLED in finite element analysis software, and define the thermal conductivity and boundary conditions of each layer of material. S02. Perform thermal distribution simulation of silicon-based OLED displaying images of different grayscale levels under different ambient temperatures, and generate corresponding temperature distribution simulation results; S03. The average surface temperature data of silicon-based OLED under different ambient temperatures and grayscale images were measured by using a high and low temperature test chamber. S04. The simulation results are calibrated by adjusting the parameters of the thermal simulation model so that the error between them and the measured data is less than a set threshold. S05. Pair the temperature distribution matrix generated by the calibrated thermal simulation model with the corresponding grayscale matrix and ambient temperature matrix to form a training sample group.

5. The silicon-based OLED brightness compensation method based on artificial intelligence according to claim 4, characterized in that, In step S01, a multilayer temperature distribution simulation model of silicon-based OLED is established, including but not limited to glass substrate, OCA, color filter, POL, encapsulation layer, transparent cathode, OLED layer, anode, metal interconnect layer and silicon substrate.

6. The silicon-based OLED brightness compensation method based on artificial intelligence according to claim 4, characterized in that, The specific steps for measuring the average surface temperature in step S03 include: Multiple thermocouples are bonded at equal intervals on the surface of a silicon-based OLED. The ambient temperature is divided into N values, and the grayscale is divided into M values; Collect the thermocouple output for each combination of ambient temperature and grayscale value, and calculate the average temperature value.

7. The silicon-based OLED brightness compensation method based on artificial intelligence according to claim 4, characterized in that, In step S04, calibrating the simulation results by adjusting the parameters of the thermal simulation model includes adjusting the thermal conductivity of each layer of material and the convective heat transfer coefficient and thermal radiation coefficient of the surface in the simulation model.

Images