Multi-scene compatible transparent LED screen display regulation and control system

Abstract

The invention relates to the technical field of intelligent display control, and provides a multi-scene compatible transparent LED screen display regulation and control system which comprises an ambient light intensity detection device, a display regulation and control platform and a transparent LED screen. The ambient light intensity detection device collects ambient light intensity data in real time and sends the ambient light intensity data to the display regulation and control platform; the display regulation and control platform comprises a data storage module, a display effect optimization module and a control module, the data storage module stores light intensity data, the display effect optimization module extracts latest data to construct a multi-dimensional environment light intensity data set, and after a trained display effect optimization model is input, a display parameter optimization result is generated and sent to the control module; and the control module adjusts the screen display effect according to the optimization parameters. And the transparent LED screen dynamically adjusts the display content according to the control instruction. According to the system, display parameters are optimized through real-time environment light intensity data, and multi-scene self-adaptive display regulation and control are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent display control, and more specifically, the present invention relates to a transparent LED screen display regulation system compatible with multiple scenarios. Background Art

[0002] With the continuous development of technology, as a new type of display technology, transparent LED screens have gradually been widely used in many fields such as advertising display, building decoration, and exhibition display. Transparent LED screens, with their unique transparent display effect, high contrast, and bright colors, can bring a brand-new visual experience to the audience. However, transparent LED screens face the problem of multi-scenario compatibility in practical applications. Under different ambient light intensity conditions, such as indoor natural light, indoor lighting, and outdoor strong light, the display effect of transparent LED screens will be greatly affected. For example, in a strong light environment, the display content on the screen may be difficult to see due to reflection or too low contrast; while in a weak light environment, the brightness of the screen may be too high, resulting in a dazzling display effect or increased energy consumption. Existing transparent LED screen display regulation technologies usually adopt simple brightness adjustment or fixed display parameter settings, and cannot dynamically adjust according to the real-time change of ambient light intensity, making it difficult to meet the display requirements in multiple scenarios.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: existing transparent LED screen display regulation technologies cannot real-time sense multi-dimensional information of ambient light intensity, such as light intensity, color temperature, contrast, etc., resulting in inaccurate adjustment of display parameters; lack of an effective display effect optimization model, unable to learn and predict based on historical ambient light intensity data, and it is difficult to achieve dynamic and intelligent optimization of display parameters; under different ambient light intensity conditions, the display effect of transparent LED screens cannot reach the best state, affecting the user experience, the service life, and energy efficiency of the screen. Summary of the Invention

[0004] The present invention provides a transparent LED screen display regulation system compatible with multiple scenarios, including:

[0005] An ambient light intensity detection device, configured to collect ambient light intensity data in real time and send the collected light intensity data to the display regulation platform;

[0006] The display control platform includes a data storage module, a display effect optimization module, and a control module; the data storage module is used to store the light intensity data; the display effect optimization module extracts the latest light intensity data from the storage module to construct a multi-dimensional ambient light intensity data set; inputs the constructed multi-dimensional ambient light intensity data set into the trained display effect optimization model to obtain the optimized result of the display parameters of the transparent LED screen, and sends the optimized result display parameters to the control module; the control module is used to adjust the display effect of the transparent LED screen according to the optimized result display parameters;

[0007] The transparent LED screen is used to dynamically adjust the display content according to the optimized result display parameters sent by the control module.

[0008] Further, the ambient light intensity detection device includes a light intensity sensor, a color temperature sensor, and a contrast detection device; the light intensity sensor collects the ambient light intensity value in real time, the color temperature sensor collects the ambient light color temperature value, and the contrast detection device collects the ambient light contrast value, and all data is uploaded to the display control platform.

[0009] Further, the optimized result of the display effect includes the brightness, contrast, and color temperature parameters of the transparent LED screen, and the control module sends different display adjustment instructions to the transparent LED screen according to the optimized result display parameters.

[0010] Further, the display effect optimization model is trained in the following way:

[0011] Obtain historical ambient light intensity data from the storage module, extract a multi-dimensional ambient light intensity data set from the historical data using a dynamic sampling window, and construct a training sample set;

[0012] Construct a pre-training model for predicting display parameters, and train the pre-training model based on the training sample set;

[0013] Based on the trained pre-training model, construct a prediction model for display effect optimization, and train the prediction model based on the training sample set to obtain the trained display effect optimization model.

[0014] Further, training the pre-training model based on the training sample set includes:

[0015] For each training sample, extract multi-dimensional ambient light intensity data from the training sample using the dynamic sampling window method;

[0016] Use the data of the first length within each dynamic sampling window as the source data, and the remaining data as the target data. One source data and the corresponding target data form a sub-training sample;

[0017] Train the pre-training model based on the sub-training sample.

[0018] Furthermore, the loss of the pre-training model is calculated using the following formula:

[0019]

[0020] where: ⊙ represents element-wise multiplication; represents the target data predicted by the pre-training model; Y represents the true target data; M represents the mask matrix corresponding to the target data; ∥·∥ 2 represents the square of the 2-norm of the matrix; M i,j represents the element at the i-th row and j-th column of the mask matrix; T represents the time step in the target data; C represents the number of variables of the multi-dimensional ambient light intensity data.

[0021] Furthermore, the pre-training model includes a first feature mapping layer, an encoding layer, and a prediction module;

[0022] The first feature mapping layer is used to perform feature mapping encoding on the input data to obtain the feature representation of the input data;

[0023] The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation;

[0024] The prediction module is used to predict the display parameters based on the high-level representation.

[0025] Furthermore, the size of the feature representation output by the first feature mapping layer is d emb , which is determined using the following formula:

[0026]

[0027] where: L represents the first length; C represents the number of variables of the multi-dimensional ambient light intensity data; d min represents the minimum length of the feature representation; d max represents the maximum length of the feature representation.

[0028] Furthermore, the prediction model includes: a second feature mapping layer, an encoding layer, and a classification module;

[0029] The second feature mapping layer is used to perform feature mapping encoding on the input data to obtain the feature representation of the input data;

[0030] The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation;

[0031] The classification module is used to predict the display parameters of the transparent LED screen based on the high-level representation;

[0032] wherein, the encoding layer is the encoding layer in the trained pre-training model.

[0033] Further, after constructing the training sample set based on the multi-dimensional ambient light intensity data, it further includes: for each sample, performing processing on the sample in the variable dimension by using brightness adaptive calibration:

[0034]

[0035] Where: B cal represents the calibrated brightness value; B raw represents the original brightness value; α represents the brightness calibration coefficient; E represents the current ambient light intensity; E ref represents the reference ambient light intensity.

[0036] According to the above embodiments of the present invention, it has at least the following beneficial effects: The transparent LED screen display control system can effectively improve the display effect of the transparent LED screen in multiple scenarios. By using the ambient light intensity detection device to collect ambient light intensity data in real time, including multi-dimensional information such as light intensity, color temperature, and contrast, the system can accurately perceive the ambient light conditions in different scenarios. Combining with the display effect optimization model in the display control platform, based on technical means such as dynamic sampling windows and pre-trained models, it can quickly and accurately generate the optimized results of the display parameters of the transparent LED screen, and then dynamically adjust parameters such as the brightness, contrast, and color temperature of the screen, so that the screen can present the best display effect in various lighting environments. Whether it is clearly visible in strong light or soft and comfortable in weak light, it can meet the needs of users and improve the user experience.

[0037] In addition, the system can also improve the service life and energy efficiency of the transparent LED screen. Through precise adjustment of the display parameters, it is possible to avoid excessive brightening or darkening of the screen in unsuitable ambient light, reduce the loss of the screen caused by long-term high-load operation, and thus extend the service life of the screen. At the same time, reasonable adjustment of the brightness and contrast can reduce the energy consumption of the screen, achieving energy conservation on the premise of ensuring the display effect. For places where transparent LED screens are widely used, such as commercial advertisements and public displays, it has significant economic and environmental benefits. Description of the Drawings

[0038] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become easily understandable. In the drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, where:

[0039] Figure 1 is a schematic structural diagram of a multi-scenario compatible transparent LED screen display control system provided by an embodiment of the present invention. Detailed Embodiments

[0040] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and then implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to fully convey the scope of the present invention to those skilled in the art.

[0041] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: all hardware, all software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0042] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0043] The following reference Figure 1 , Figure 1 is a schematic structural diagram of a multi-scenario compatible transparent LED screen display control system provided for an embodiment of the present invention. As Figure 1 shown, a multi-scenario compatible transparent LED screen display control system includes:

[0044] An ambient light intensity detection device 101, configured to collect ambient light intensity data in real time and send the collected light intensity data to the display control platform;

[0045] A display control platform 102, including a data storage module, a display effect optimization module, and a control module; the data storage module is configured to store the light intensity data; the display effect optimization module extracts the latest light intensity data from the storage module to construct a multi-dimensional ambient light intensity data set; inputs the constructed multi-dimensional ambient light intensity data set into a trained display effect optimization model to obtain an optimized result of the display parameters of the transparent LED screen, and sends the optimized result display parameters to the control module; the control module is configured to adjust the display effect of the transparent LED screen according to the optimized result display parameters;

[0046] A transparent LED screen 103, configured to dynamically adjust the display content according to the optimized result display parameters sent by the control module.

[0047] It should be noted that the present invention proposes a transparent LED screen display control system compatible with multiple scenarios, aiming to dynamically adjust the display effect of the transparent LED screen by real-time detecting the ambient light intensity to meet the display requirements in different scenarios. Among them, the ambient light intensity detection device refers to a device capable of real-time collecting ambient light intensity data, including a light intensity sensor, a color temperature sensor, a contrast detection device, etc., for sensing the lighting conditions of the surrounding environment. The display control platform is the core part of the system, including a data storage module, a display effect optimization module, and a control module, which are used to process the collected light intensity data and generate optimized display parameters, and then control the display effect of the transparent LED screen. The transparent LED screen is a display device that dynamically adjusts the display content according to the optimized result display parameters sent by the control module, and can change its own display parameters in real time according to the change of the ambient light intensity to achieve the best display effect.

[0048] Specifically, the light intensity sensor in the ambient light intensity detection device is used to real-time collect the ambient light intensity value, that is, the luminous flux received per unit area, usually measured in lux (lx); the color temperature sensor is used to collect the ambient light color temperature value, that is, the relative color temperature of the light source, usually measured in Kelvin (K). The higher the color temperature, the bluer the light, and the lower the color temperature, the redder the light; the contrast detection device is used to collect the ambient light contrast value, that is, the relative difference between the bright part and the dark part in the ambient light. The higher the contrast, the clearer the display effect. These data are uploaded to the display control platform. The data storage module in the display control platform is used to store the collected light intensity data for subsequent analysis and processing; the display effect optimization module extracts the latest light intensity data from the storage module to construct a multi-dimensional ambient light intensity data set, which contains data in multiple dimensions such as light intensity, color temperature, and contrast, and then inputs this data set into the trained display effect optimization model to obtain the optimized result of the display parameters of the transparent LED screen. These optimization results include parameters such as brightness, contrast, and color temperature; the control module sends different display adjustment instructions to the transparent LED screen according to the optimized result display parameters to adjust the display effect of the transparent LED screen. The transparent LED screen then dynamically adjusts the display content according to the optimized result display parameters sent by the control module to adapt to different ambient light intensity conditions.

[0049] Preferably, for the construction of the display effect optimization model, first obtain historical ambient light intensity data from the data storage module, extract a multi-dimensional ambient light intensity data set from the historical data using a dynamic sampling window, and construct a training sample set. The dynamic sampling window is a data sampling method that can dynamically adjust the sampling range according to the time series characteristics of the data to better capture the variation law of the ambient light intensity. Then construct a pre-training model for predicting display parameters and train the pre-training model based on the training sample set. The training process of the pre-training model includes extracting multi-dimensional ambient light intensity data for each training sample using the dynamic sampling window method, using the data of the first length within each dynamic sampling window as the source data, and the remaining data as the target data. A source data and the corresponding target data form a sub-training sample, and the pre-training model is trained based on the sub-training sample. The loss calculation of the pre-training model uses a specific formula, which calculates the difference between the target data predicted by the pre-training model and the real target data through operations such as element-wise multiplication and the square of the 2-norm of the matrix to optimize the parameters of the model. The pre-training model includes a first feature mapping layer, an encoding layer, and a prediction module. The first feature mapping layer is used to perform feature mapping encoding on the input data to obtain the feature representation of the input data, and the size of the output feature representation is determined by a specific formula that comprehensively considers factors such as the first length, the number of variables of the multi-dimensional ambient light intensity data, the minimum length and the maximum length of the feature representation to ensure the effectiveness and rationality of the feature representation; the encoding layer is used to extract deep features from the feature representation to obtain a high-level representation; the prediction module is used to predict the display parameters based on the high-level representation.

[0050] In some embodiments, the ambient light intensity detection device includes a light intensity sensor, a color temperature sensor, and a contrast detection device; the light intensity sensor collects the ambient light intensity value in real time, the color temperature sensor collects the ambient light color temperature value, the contrast detection device collects the ambient light contrast value, and all the data is uploaded to the display control platform.

[0051] It should be noted that the ambient light intensity detection device mentioned in the present invention includes a light intensity sensor, a color temperature sensor, and a contrast detection device, and these sensors work together to collect various parameters of the ambient light in real time. The light intensity sensor is used to measure the intensity of the ambient light, that is, the luminous flux received per unit area, usually measured in lux (lx). The color temperature sensor is used to measure the color temperature of the ambient light. The color temperature is a physical quantity representing the color of the light source, which describes the relative color temperature of the light source, and the unit is Kelvin (K). The contrast detection device is used to measure the contrast of the ambient light. The contrast refers to the relative difference between the bright part and the dark part in the environment. The ambient light with high contrast can make the display content clearer. These data are uploaded to the display control platform in real time so that the system can dynamically adjust the display effect of the transparent LED screen according to these data, thereby providing the best visual experience under different ambient light conditions.

[0052] Specifically, the light intensity sensor, color temperature sensor, and contrast detection device are the core components of the ambient light intensity detection device. The light intensity sensor can use components such as photodiodes to convert the light signal into an electrical signal through the photoelectric effect, thereby measuring the intensity of the ambient light. The color temperature sensor is usually based on thermocouples or other temperature-sensitive components and determines the color temperature by measuring the temperature of the light source. The contrast detection device can calculate the contrast by analyzing the brightness distribution of the ambient light. For example, the contrast value is determined by measuring the light intensity difference between the bright and dark parts of the ambient light. After the measurement results of these sensors are converted into digital signals, they are uploaded to the display control platform through a communication interface (such as serial communication or wireless communication). In the display control platform, this data is stored in the data storage module so that the subsequent display effect optimization module can use this data to adjust the display parameters of the transparent LED screen.

[0053] Preferably, to improve the accuracy and reliability of the ambient light intensity detection, the measurement process of the sensors can be optimized. For example, the light intensity sensor can be set to sample periodically, and the sampling period can be adjusted according to the actual application scenario, such as sampling once per second, to ensure that the changes in the ambient light intensity can be reflected in real time. The color temperature sensor can adopt a multi-point measurement method, measuring the color temperature at different positions and taking the average value to reduce the measurement error. The contrast detection device can combine image processing technology to calculate the contrast more accurately by analyzing the brightness distribution map of the ambient light. In addition, to further improve the adaptability of the system, the collected data can be preprocessed in the display control platform, such as removing noise through a filtering algorithm or comprehensively processing the data of multiple sensors through data fusion technology to obtain more accurate ambient light intensity information. These optimization measures can significantly improve the display effect of the transparent LED screen under different ambient light conditions, ensuring that the screen can provide a clear, comfortable, and energy-saving visual experience in various scenarios.

[0054] In some embodiments, the display effect optimization results include the brightness, contrast, and color temperature parameters of the transparent LED screen, and the control module sends different display adjustment instructions to the transparent LED screen according to the optimized result display parameters.

[0055] It should be noted that the display effect optimization results mentioned in the present invention include the brightness, contrast, and color temperature parameters of the transparent LED screen, and these parameters are calculated by the display effect optimization module in the display control platform. The control module sends different display adjustment instructions to the transparent LED screen according to these optimized result display parameters to dynamically adjust the display effect of the screen. Brightness refers to the intensity of the screen's light emission, usually measured in nits (cd / m 2)The unit is; contrast refers to the relative difference between the bright and dark parts of the content displayed on the screen. High contrast can make the displayed content clearer; color temperature describes the color tendency of the content displayed on the screen, usually in Kelvin (K). By dynamically adjusting these parameters, the transparent LED screen can provide the best visual experience under different ambient light intensity conditions.

[0056] Specifically, the brightness parameter of the transparent LED screen refers to the intensity of the screen's light emission, which can be achieved by adjusting the backlight intensity of the screen or the light-emitting power of the pixels. The contrast parameter refers to the relative difference between the bright and dark parts of the content displayed on the screen, which can be optimized by adjusting the backlight uniformity of the screen and the gray level of the pixels. The color temperature parameter describes the color tendency of the content displayed on the screen, which can be achieved by adjusting the RGB color balance of the screen. The control module sends different display adjustment instructions to the transparent LED screen according to the optimized result display parameters. These instructions can include specific brightness values, contrast values, and color temperature values. For example, the brightness value can be set to a specific nit value, the contrast value can be set to a percentage, and the color temperature value can be set to a Kelvin value. The specific settings of these parameters can be dynamically adjusted according to the data collected by the ambient light intensity detection device to ensure that the transparent LED screen can provide a clear, comfortable, and energy-saving visual experience under different ambient light intensity conditions.

[0057] Preferably, in order to achieve more accurate display effect optimization, the display effect optimization module in the display control platform can adopt advanced algorithms to calculate the brightness, contrast, and color temperature parameters. For example, the brightness parameter can be calculated by analyzing the light intensity data collected by the ambient light intensity detection device and combining with the preset brightness adjustment curve. The contrast parameter can be optimized by analyzing the ambient light contrast data and combining with the display characteristics of the screen. The color temperature parameter can be adjusted by analyzing the ambient light color temperature data and combining with the color balance model of the screen. In actual operation, the control module can generate specific control instructions according to the optimized result display parameters. For example, the backlight intensity of the screen can be adjusted by a PWM (pulse width modulation) signal, and the contrast and color temperature of the screen can be adjusted by digital signals. In addition, in order to further improve the adaptability of the system, the data collected in the display control platform can be preprocessed, such as removing noise through a filtering algorithm, or comprehensively processing the data of multiple sensors through data fusion technology to obtain more accurate ambient light intensity information. These optimization measures can significantly improve the display effect of the transparent LED screen under different ambient light conditions, ensuring that the screen can provide a clear, comfortable, and energy-saving visual experience in various scenarios.

[0058] In some embodiments, the display effect optimization model is trained in the following manner:

[0059] Obtain historical ambient light intensity data from the storage module, extract a multi-dimensional ambient light intensity data set from the historical data using a dynamic sampling window, and construct a training sample set;

[0060] Construct a pre-trained model for predicting display parameters, and train the pre-trained model based on the training sample set;

[0061] Based on the trained pre-trained model, construct a prediction model for display effect optimization, and train the prediction model based on the training sample set to obtain a trained display effect optimization model.

[0062] It should be noted that the display effect optimization model mentioned in the present invention constructs a training sample set by obtaining historical ambient light intensity data from a data storage module and using a dynamic sampling window to extract a multi-dimensional ambient light intensity data set from it. The pre-trained model is trained based on this training sample set, and the prediction model is constructed based on the trained pre-trained model for display effect optimization. This process involves data collection, processing, and model training, aiming to optimize the display parameters of the transparent LED screen by learning the patterns in the historical data to adapt to different ambient light intensity conditions.

[0063] Specifically, the historical ambient light intensity data refers to multi-dimensional data such as light intensity, color temperature, and contrast that were previously collected and stored in the data storage module. The dynamic sampling window is a data sampling method that can dynamically adjust the sampling range according to the time series characteristics of the data to better capture the changing patterns of the ambient light intensity. The training sample set is composed of these sampled data and is used to train the pre-trained model. The pre-trained model is a basic model that provides a basis for subsequent display effect optimization by learning the data features in the training sample set. The prediction model is a further optimized model based on the pre-trained model and is specifically used to predict the optimal display parameters of the transparent LED screen according to the current ambient light intensity data. These parameters include brightness, contrast, and color temperature, etc., and their settings directly affect the display effect of the screen.

[0064] Preferably, the construction process of the display effect optimization model can be further refined. First, when obtaining historical ambient light intensity data from the data storage module, it is necessary to ensure the integrity and accuracy of the data. These data are usually stored in the form of a time series, containing information such as light intensity, color temperature, and contrast at different time points. The size of the dynamic sampling window can be adjusted according to the actual application scenario. For example, for rapidly changing ambient light intensity, the sampling window can be set smaller to capture changes more promptly. When constructing the training sample set, the data can be preprocessed, such as removing noise and filling in missing values. The training process of the pre-trained model can adopt machine learning algorithms, such as neural networks, with input parameters including multi-dimensional data such as light intensity, color temperature, and contrast, and the output being the optimized display parameters. The prediction model can be further optimized based on this, for example, by adjusting the network structure or adding regularization terms to improve the generalization ability of the model. In practical applications, the training and optimization of the model need to be adjusted according to a large amount of experimental data to ensure that the model can accurately predict the display parameters, thereby achieving the best display effect of the transparent LED screen under different ambient light intensity conditions.

[0065] In some embodiments, training the pre-trained model based on the training sample set includes:

[0066] For each training sample, extract multi-dimensional ambient light intensity data from the training sample using the dynamic sampling window method;

[0067] Use the data of the first length within each dynamic sampling window as the source data, and the remaining data as the target data. One source data and the corresponding target data form a sub-training sample;

[0068] Train the pre-trained model based on the sub-training sample.

[0069] It should be noted that the training process of the pre-trained model mentioned in the present invention involves using the dynamic sampling window method to extract multi-dimensional ambient light intensity data from each training sample, and dividing the data into source data and target data to form sub-training samples. This method aims to simulate the light intensity changes in the real environment and let the model learn how to predict the subsequent display parameters (target data) based on the known ambient light intensity data (source data). In this way, the pre-trained model can better adapt to the display parameter optimization requirements under different ambient light intensity conditions and provide a more accurate prediction basis for subsequent display effect optimization.

[0070] Specifically, the dynamic sampling window method is a data processing technique used to extract representative sample segments from time series data. In the present invention, each training sample represents the ambient light intensity data over a period of time, which includes multiple dimensions such as light intensity, color temperature, and contrast. The source data is extracted from the first length of data before each dynamic sampling window, and this part of the data is used as the input of the model to simulate known ambient light intensity conditions. The target data is the data after the source data and is used as the output of the model, that is, the display parameters that the model needs to predict. The first length is a parameter representing the time span of the source data and can be adjusted according to the actual application scenario. For example, in a scenario where the ambient light intensity changes rapidly, the first length can be set shorter so that the model can respond to environmental changes more quickly. In this way, the sub-training samples can effectively train the pre-trained model to learn how to predict subsequent display parameters based on the known ambient light intensity data, thereby optimizing the display effect of the transparent LED screen.

[0071] Preferably, the training process of the pre-trained model can be further refined. When constructing the sub-training samples, the source data and the target data can be normalized to eliminate the dimensional differences between different dimensional data and improve the training efficiency of the model. For example, data such as light intensity, color temperature, and contrast can be normalized to the range of 0 to 1 respectively. In addition, to improve the generalization ability of the model, regularization techniques such as L2 regularization can be introduced during the training process to prevent the model from overfitting. In terms of the model structure, the pre-trained model can adopt a deep learning architecture such as a recurrent neural network (RNN) or its variant long short-term memory network (LSTM), and these network structures can effectively process time series data. The input parameter includes the pre-processed source data, and the output is the target data. During the training process, the model optimizes the parameters by minimizing the difference between the predicted value and the true value. For example, the mean squared error can be used as the loss function, and the model parameters can be updated through the backpropagation algorithm. Through these optimization measures, the pre-trained model can more accurately predict the display parameters, thereby achieving the best display effect of the transparent LED screen under different ambient light intensity conditions.

[0072] In some embodiments, the following formula is used to calculate the loss of the pre-trained model:

[0073]

[0074] where: ⊙ represents element-wise multiplication; represents the target data predicted by the pre-trained model; Y represents the true target data; M represents the mask matrix corresponding to the target data; ∥·∥ 2 represents the square of the 2-norm of the matrix; M i,jRepresents the element in the i-th row and j-th column of the mask matrix; T represents the time step in the target data; C represents the number of variables of the multi-dimensional ambient light intensity data.

[0075] It should be noted that the loss calculation method of the pre-trained model mentioned in the present invention is implemented through a specific formula. The core idea of this formula is to compare the difference between the target data predicted by the pre-trained model and the real target data, and combine the mask matrix to process the missing or irrelevant parts in the data. This loss calculation method can effectively evaluate the prediction performance of the model and guide the training process of the model, enabling it to better learn the characteristics of the ambient light intensity data, thereby improving the accuracy of display parameter prediction.

[0076] Specifically, several key concepts are involved in the loss calculation formula of the pre-trained model. First is the element-wise multiplication operation, which means multiplying the elements in the same position of two matrices. This operation is used to compare the differences between the predicted values and the real values element by element. Secondly, the mask matrix is a matrix with the same shape as the target data, used to mark which parts of the target data are valid (i.e., need to participate in the loss calculation) and which parts are invalid (such as missing data or irrelevant data). The mask value of the valid part is usually 1, and the mask value of the invalid part is 0. Through the mask matrix, those irrelevant or missing data points can be ignored, thereby improving the accuracy and robustness of the loss calculation. Finally, the square of the 2-norm of the matrix is a mathematical method for measuring the size of matrix elements. It calculates the overall error size of the matrix by taking the square root of the sum of the squares of all elements in the matrix and then squaring it again. This calculation method can comprehensively consider the errors of all elements, thereby providing a comprehensive performance evaluation index.

[0077] Preferably, the specific process of loss calculation can be further refined. In practical applications, the predicted target data and the true target data of the pre-trained model are first converted into matrices of the same shape. Then, the difference between the predicted value and the true value is multiplied by the mask matrix in an element-wise manner, so that only the elements with a mask value of 1 (i.e., valid data points) are taken into account. Next, the square of the 2-norm of the calculated difference matrix is obtained to get the final loss value. This loss value will be used in the optimization process of the model, and the parameters of the model are adjusted through the backpropagation algorithm to minimize the difference between the predicted value and the true value. In terms of model construction, the pre-trained model can adopt a deep learning architecture, such as a convolutional neural network (CNN) or a recurrent neural network (RNN), and the specific choice depends on the characteristics of the ambient light intensity data. For example, if the data has time series characteristics, then RNN or its variant, the long short-term memory network (LSTM), may be a better choice. The input parameters include the preprocessed multi-dimensional ambient light intensity data, and the output is the optimized display parameters. In this way, the pre-trained model can more effectively learn the characteristics of the ambient light intensity data, thereby improving the accuracy and robustness of display parameter prediction.

[0078] In some embodiments, the pre-trained model includes a first feature mapping layer, an encoding layer, and a prediction module;

[0079] The first feature mapping layer is used to perform feature mapping encoding on the input data to obtain the feature representation of the input data;

[0080] The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation;

[0081] The prediction module is used to predict the display parameters based on the high-level representation.

[0082] It should be noted that the pre-trained model mentioned in the present invention includes a first feature mapping layer, an encoding layer, and a prediction module. The role of the first feature mapping layer is to perform feature mapping encoding on the input multi-dimensional ambient light intensity data to obtain the feature representation of the input data. The encoding layer extracts deep features from the feature representation to obtain a high-level representation. The prediction module predicts the display parameters of the transparent LED screen based on the high-level representation. The design of this structure aims to gradually extract the key features in the data through hierarchical processing and finally generate optimized display parameters, so as to achieve the best display effect of the transparent LED screen under different ambient light intensity conditions.

[0083] Specifically, the first feature mapping layer is the input layer of the pre-trained model, which receives multi-dimensional ambient light intensity data extracted from the dynamic sampling window. This data includes dimensions such as light intensity, color temperature, and contrast. The process of feature mapping encoding is to convert the original data into a more easily processed feature representation. For example, the data is mapped to a new feature space through linear transformation or non-linear transformation. The encoding layer is the core part of the model, which is responsible for extracting deep features from the feature representation. These deep features can capture complex patterns and relationships in the data. The encoding layer can adopt various neural network structures, such as convolutional neural network (CNN) or recurrent neural network (RNN). The specific choice depends on the characteristics of the data. The prediction module is the output layer of the model. Based on the high-level representation extracted by the encoding layer, it predicts the display parameters of the transparent LED screen, such as brightness, contrast, and color temperature. The predicted values of these parameters will be sent to the control module to adjust the display effect of the transparent LED screen.

[0084] Preferably, the construction process of the pre-trained model can be further refined. In the first feature mapping layer, a multi-layer perceptron (MLP) or a convolutional layer can be used to implement feature mapping encoding. For example, if the input data is time series data, a one-dimensional convolutional layer can be used to extract local features. The encoding layer can adopt a long short-term memory network (LSTM) or a gated recurrent unit (GRU). These recurrent neural network structures can effectively handle long-term dependencies in time series data. The prediction module can adopt a fully connected layer, and its output is the display parameter of the transparent LED screen. During the model training process, the mean squared error can be used as the loss function, and the model parameters are optimized through the backpropagation algorithm. In addition, to improve the generalization ability of the model, regularization techniques such as Dropout or L2 regularization can be introduced during the training process. Through these optimization measures, the pre-trained model can more accurately predict the display parameters, thus achieving the best display effect of the transparent LED screen under different ambient light intensity conditions.

[0085] In some embodiments, the size of the feature representation output by the first feature mapping layer is d emb , which is determined by the following formula:

[0086]

[0087] where: L represents the first length; C represents the number of variables of the multi-dimensional ambient light intensity data; d min represents the minimum length of the feature representation; d max represents the maximum length of the feature representation.

[0088] It should be noted that the size of the feature representation output by the first feature mapping layer in the present invention is determined by a specific formula. This formula comprehensively considers factors such as the first length, the number of variables in the multi-dimensional ambient light intensity data, the minimum length and the maximum length of the feature representation, etc., to ensure the effectiveness and rationality of the feature representation. In this way, the size of the feature representation can be dynamically adjusted so that it can not only contain sufficient information but also not cause waste of computing resources due to being too large, thereby improving the efficiency and performance of the model.

[0089] Specifically, the first feature mapping layer is an important part of the pre-trained model, and its function is to perform feature mapping encoding on the input multi-dimensional ambient light intensity data to obtain the feature representation of the input data. The first length refers to the time step used to construct the source data in the dynamic sampling window, which determines the length of the time series information contained in the source data. The number of variables in the multi-dimensional ambient light intensity data refers to the number of different dimensions contained in the input data, such as light intensity, color temperature, contrast, etc. The minimum length and the maximum length of the feature representation are two constraint parameters used to limit the size range of the feature representation to ensure that the feature representation can not only contain sufficient information but also not cause waste of computing resources due to being too large. Through this formula, the size of the feature representation can be dynamically adjusted according to different input data and application scenarios, thereby improving the adaptability and efficiency of the model.

[0090] Preferably, the determination process of the size of the feature representation of the first feature mapping layer can be further refined. In practical applications, the first length can be adjusted according to the change frequency of the ambient light intensity data. For example, in a scenario where the ambient light intensity changes rapidly, the first length can be set shorter to respond to the environmental change faster. The number of variables in the multi-dimensional ambient light intensity data can be set according to the actual collected data dimensions. For example, if the collected data includes three dimensions of light intensity, color temperature, and contrast, the number of variables is 3. The minimum length and the maximum length of the feature representation can be adjusted according to the complexity of the model and the computing resources. For example, if the computing resources are limited, the maximum length can be appropriately reduced. In this way, it can be ensured that the size of the feature representation can meet the requirements of the model and will not cause waste of computing resources due to being too large. In addition, the first feature mapping layer can use a multi-layer perceptron (MLP) or a convolutional layer to implement feature mapping encoding, and the specific selection depends on the characteristics of the input data. For example, if the input data is time series data, a one-dimensional convolutional layer can be used to extract local features. Through these optimization measures, the first feature mapping layer can extract the features of the input data more effectively, thereby improving the performance of the pre-trained model.

[0091] In some embodiments, the prediction model includes: a second feature mapping layer, an encoding layer, and a classification module;

[0092] The second feature mapping layer is used to perform feature mapping encoding on the input data to obtain a feature representation of the input data;

[0093] The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation;

[0094] The classification module is used to predict the display parameters of the transparent LED screen based on the high-level representation;

[0095] Wherein, the encoding layer is the encoding layer in a trained pre-trained model.

[0096] It should be noted that the prediction model mentioned in the present invention includes a second feature mapping layer, an encoding layer, and a classification module. The role of the second feature mapping layer is to perform feature mapping encoding on the input data to obtain a feature representation of the input data. The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation. The classification module then predicts the display parameters of the transparent LED screen based on the high-level representation. Among them, the encoding layer is the encoding layer in a trained pre-trained model, which means that the encoding layer of the pre-trained model is utilized during the construction of the prediction model to achieve better feature extraction and display parameter prediction effects.

[0097] Specifically, the second feature mapping layer is the input layer of the prediction model, which receives multi-dimensional ambient light intensity data extracted from the dynamic sampling window. These data include dimensions such as light intensity, color temperature, and contrast. The process of feature mapping encoding is to convert the original data into a more easily processed feature representation, for example, by linearly transforming or non-linearly transforming the data into a new feature space. The encoding layer is the core part of the model, which is responsible for extracting deep features from the feature representation. These deep features can capture complex patterns and relationships in the data. The encoding layer uses the encoding layer in a trained pre-trained model, which means that the encoding layer has learned the feature representation of the ambient light intensity data through the pre-training process and can be directly used in the prediction model. The classification module is the output layer of the model, which predicts the display parameters of the transparent LED screen, such as brightness, contrast, and color temperature, based on the high-level representation extracted by the encoding layer. The predicted values of these parameters will be sent to the control module to adjust the display effect of the transparent LED screen.

[0098] Preferably, the construction process of the prediction model can be further refined. In the second feature mapping layer, a multi-layer perceptron (MLP) or a convolutional layer can be used to implement feature mapping encoding. For example, if the input data is time series data, a one-dimensional convolutional layer can be used to extract local features. The encoding layer adopts the encoding layer in the trained pre-trained model, which can be a long short-term memory network (LSTM) or a gated recurrent unit (GRU). These recurrent neural network structures can effectively handle the long-term dependencies in time series data. The classification module can adopt a fully connected layer, and its output is the display parameters of the transparent LED screen. During the model training process, the mean square error can be used as the loss function, and the model parameters can be optimized through the backpropagation algorithm. In addition, in order to improve the generalization ability of the model, regularization techniques such as Dropout or L2 regularization can be introduced during the training process. Through these optimization measures, the prediction model can more accurately predict the display parameters, so as to achieve the best display effect of the transparent LED screen under different ambient light intensity conditions.

[0099] In some embodiments, after constructing the training sample set based on the multi-dimensional ambient light intensity data, it further includes: for each sample, performing processing on the sample in the variable dimension by using brightness adaptive calibration:

[0100]

[0101] Where: B cal represents the calibrated brightness value; B raw represents the original brightness value; α represents the brightness calibration coefficient; E represents the current ambient light intensity; E ref represents the reference ambient light intensity.

[0102] It should be noted that the brightness adaptive calibration mentioned in the present invention is a method for processing multi-dimensional ambient light intensity data samples. Through this calibration, the brightness value of the sample in the variable dimension can be adjusted to make it more in line with the actual display requirements. This method takes into account the difference between the current ambient light intensity and the reference ambient light intensity, and is adjusted through the brightness calibration coefficient, so as to ensure the consistency and comfort of the display effect of the transparent LED screen under different ambient light intensity conditions. This calibration method can effectively solve the problem of display effect difference caused by the change of ambient light intensity, and improve the adaptability and user experience of the transparent LED screen.

[0103] Specifically, the brightness adaptive calibration involves several key parameters. First is the original brightness value, which is the uncalibrated brightness value directly collected from the ambient light intensity detection device and reflects the initial brightness state of the screen in the current environment. Second is the brightness calibration coefficient, which is an adjustment parameter used to control the intensity of brightness calibration. Its value can be set according to the actual application scenario. For example, in a scenario with a large change in ambient light intensity, the coefficient can be appropriately increased to enhance the calibration effect. Third is the current ambient light intensity, which is the real-time collected ambient light intensity value and reflects the current lighting conditions. Finally is the reference ambient light intensity, which is a preset standard value used as the calibration benchmark. By calculating the difference between the current ambient light intensity and the reference ambient light intensity and combining with the brightness calibration coefficient, the calibrated brightness value can be obtained, thus realizing the adaptive adjustment of brightness. The specific settings of these parameters need to be optimized according to the display characteristics and application scenarios of the transparent LED screen to ensure the best display effect after calibration.

[0104] Preferably, the specific implementation process of the brightness adaptive calibration can be further refined. In practical applications, the brightness calibration coefficient can be dynamically adjusted according to the change range of the ambient light intensity and the display characteristics of the transparent LED screen. For example, when the ambient light intensity changes from indoor natural light (about 100 - 1000 lux) to outdoor strong light (about 10000 - 100000 lux), the brightness calibration coefficient can be appropriately increased to ensure that the screen can still be clearly displayed under strong light. At the same time, the reference ambient light intensity can be set according to the application scenario. For example, in an indoor environment, the reference ambient light intensity can be set to 500 lux, while in an outdoor environment, it can be set to 10000 lux. During the calibration process, first calculate the difference between the current ambient light intensity and the reference ambient light intensity, and then adjust the original brightness value according to the brightness calibration coefficient to obtain the calibrated brightness value. This calibration method can effectively solve the display consistency problem of the transparent LED screen under different ambient light intensity conditions and improve the user experience. In addition, to further improve the calibration accuracy, the change rate of the ambient light intensity can be introduced during the calibration process. When the ambient light intensity changes rapidly, the calibration sensitivity can be appropriately increased to quickly adapt to the environmental changes.

[0105] The above embodiments of the present invention have the following beneficial effects: The system can collect multi-dimensional data such as ambient light intensity, color temperature, and contrast in real time, and construct a training sample set through a dynamic sampling window to provide accurate data support for display parameter optimization. The double-layer architecture based on the pre-trained model and the prediction model can efficiently process the ambient light intensity data, extract deep features through the feature mapping layer and the encoding layer, and finally output optimized parameters such as brightness, contrast, and color temperature, enabling the transparent LED screen to adapt to different lighting environments and significantly improving the display effect.

[0106] The system can combine historical data with real-time detection results, adopt the dynamic sampling window method and brightness adaptive calibration technology to ensure the accuracy and adaptability of display parameter optimization. By calculating the loss function of the pre-trained model and dynamically adjusting the dimension of feature representation, the training efficiency of the model can be optimized. At the same time, the classification module can predict the best display parameters based on high-level features, enabling the transparent LED screen to intelligently adjust the display content, enhancing visual comfort and multi-scene compatibility.

[0107] Furthermore, the storage medium of the embodiment of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0108] The above description is only some preferred embodiments of the present invention and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with technical features having similar functions (but not limited to) disclosed in the embodiments of the present invention.

Claims

1. A multi-scenario compatible transparent LED screen display control system, characterized in that: include: An ambient light intensity detection device is used to collect ambient light intensity data in real time and send the collected light intensity data to the display control platform; Display control platform, including data storage module, display effect optimization module and control module; The data storage module is used to store the light intensity data; The display effect optimization module extracts the latest light intensity data from the storage module to construct a multi-dimensional ambient light intensity data set; inputs the constructed multi-dimensional ambient light intensity data set into the trained display effect optimization model to obtain the display parameter optimization result of the transparent LED screen, and sends the optimization result display parameters to the control module; the control module is used to adjust the display effect of the transparent LED screen according to the optimization result display parameters; The transparent LED screen is used to dynamically adjust the display content according to the optimization result display parameters sent by the control module.

2. The multi-scenario compatible transparent LED screen display control system according to claim 1 is characterized in that: The ambient light intensity detection device includes a light intensity sensor, a color temperature sensor and a contrast detection device; the light intensity sensor collects the ambient light intensity value in real time, the color temperature sensor collects the ambient light color temperature value, and the contrast detection device collects the ambient light contrast value, and all data are uploaded to the display control platform.

3. The multi-scenario compatible transparent LED screen display control system according to claim 1 is characterized in that: The display effect optimization result includes brightness, contrast and color temperature parameters of the transparent LED screen, and the control module sends different display adjustment instructions to the transparent LED screen according to the optimization result display parameters.

4. The multi-scenario compatible transparent LED screen display control system according to claim 1 is characterized in that: The display effect optimization model is trained in the following way: Acquire historical ambient light intensity data from the storage module, extract a multidimensional ambient light intensity data set from the historical data using a dynamic sampling window, and construct a training sample set; Constructing a pre-trained model for predicting display parameters, and training the pre-trained model based on the training sample set; A prediction model for display effect optimization is constructed based on the trained pre-trained model, and the prediction model is trained based on the training sample set to obtain a trained display effect optimization model.

5. The multi-scenario compatible transparent LED screen display control system according to claim 4 is characterized in that: Training the pre-trained model based on the training sample set includes: For each training sample, a dynamic sampling window method is used to extract multi-dimensional ambient light intensity data from the training sample; The first length of data in each dynamic sampling window is used as source data, and the remaining data is used as target data. One source data and the corresponding target data constitute a sub-training sample. The pre-training model is trained based on the sub-training samples.

6. The multi-scenario compatible transparent LED screen display control system according to claim 4 is characterized in that: The loss of the pre-trained model is calculated using the following formula: Among them: ⊙ means multiplication by element position; represents the target data predicted by the pre-trained model; Y represents the actual target data; M represents the mask matrix corresponding to the target data; ∥·∥ 2 Represents the square of the 2-norm of the matrix; M i,j Represents the element in the i-th row and j-th column of the mask matrix; T represents the time step in the target data; C represents the number of variables of the multidimensional ambient light intensity data.

7. The multi-scenario compatible transparent LED screen display control system according to claim 4 is characterized in that: The pre-trained model includes a first feature mapping layer, a coding layer and a prediction module; The first feature mapping layer is used to perform feature mapping encoding on the input data to obtain a feature representation of the input data; The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation; The prediction module is used to predict display parameters based on the high-level representation.

8. The multi-scenario compatible transparent LED screen display control system according to claim 7 is characterized in that: The size of the feature representation output by the first feature mapping layer is d emb , determined using the following formula: Where: L represents the first length; C represents the number of variables of multidimensional ambient light intensity data; d min Indicates the minimum length of feature representation; d max Indicates the maximum length of the feature representation.

9. The multi-scenario compatible transparent LED screen display control system according to claim 7, characterized in that: The prediction model includes: a second feature mapping layer, a coding layer and a classification module; The second feature mapping layer is used to perform feature mapping encoding on the input data to obtain a feature representation of the input data; The encoding layer is used to extract deep features from the feature representation to obtain a high-level representation; The classification module is used to predict the display parameters of the transparent LED screen based on the high-level representation; The coding layer is a coding layer in a trained pre-trained model.

10. The multi-scenario compatible transparent LED screen display control system according to claim 1, characterized in that: After constructing the training sample set based on the multi-dimensional ambient light intensity data, the method further includes: for each sample, processing the sample in the variable dimension by using brightness adaptive calibration: Among them: B cal Indicates the brightness value after calibration; B raw represents the original brightness value; α represents the brightness calibration coefficient; E represents the current ambient light intensity; E ref Indicates the reference ambient light intensity.

Images