Intelligent dimming COB display screen system and control method

Abstract

The application relates to the technical field of display control, and discloses a COB display screen system and a control method for intelligent dimming, the COB display screen system for intelligent dimming comprises a light field sensor array module, which generates a light field distribution by collecting ambient light in real time; an super-pixel segmentation module, which extracts super-pixel features according to a display image and light field information; a model prediction controller module, which models based on the super-pixel features and generates an optimal brightness control signal; a reinforcement learning engine module, which dynamically optimizes feature generation and control parameters according to an operating state; and a distributed muLED driving circuit module, which drives a muLED light-emitting unit according to the control signal. The application realizes accurate perception of display ambient light and fine regional division of display content by introducing the light field sensor array module and the super-pixel segmentation module for ambient light perception, and improves the environmental adaptability and local contrast of the display screen under various illumination conditions.

Description

Technical Field

[0001] This invention relates to the field of display control technology, specifically to an intelligent dimming COB display screen system and control method. Background Technology

[0002] With the popularization of advanced packaging technologies such as COB, high-quality μLED displays have been widely used in various professional and civilian scenarios. However, the viewing comfort and energy consumption under different lighting conditions are receiving increasing attention. An ideal display system should be able to intelligently sense changes in ambient light and the displayed image, and adjust its brightness in real time, so as to not only provide viewers with a consistent best visual experience, but also save energy to the maximum extent.

[0003] In the existing technology, there are some preliminary intelligent dimming solutions. These solutions usually integrate photosensitive elements into the display device to detect the overall intensity of ambient light. When the external light changes, the control unit inside the system will adjust the overall brightness of the backlight or light-emitting unit of the display screen in a unified manner according to the preset logic algorithm. Compared with traditional display devices with constant brightness, this automatic adjustment mechanism can improve the viewing comfort to a certain extent and achieve basic energy-saving effect. It is a beneficial attempt for display technology to move towards intelligence.

[0004] Despite this, existing technologies still exhibit several inherent flaws in practical applications. First, their single-point or few-point illumination detection methods cannot accurately capture the complex and uneven light field distribution in real-world environments. When the display faces complex scenes such as alternating sunlight from a window and indoor lighting, global brightness adjustments based on partial data can lead to localized overbrightness or underbrightness, affecting image depth and detail. Second, existing technologies largely rely on reactive feedback control strategies, which are essentially passive responses to errors and lack the ability to predict future system states. Therefore, when illumination or screen content changes rapidly, they are prone to causing delays, overshoots, and even frequent jitter in brightness adjustment, damaging visual stability and causing unnecessary energy consumption. More importantly, the control parameters of these systems are difficult to change once set, making them unable to adapt to changing display content, environmental spectra, and long-term dynamic factors such as hardware aging. Their control performance gradually deteriorates over time and with changes in usage scenarios, lacking self-optimization and continuous adaptation capabilities. Therefore, those skilled in the art propose an intelligent dimming COB display system and control method to address these issues. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent dimming COB display system and control method, which solves the problems of poor display effect and high energy consumption caused by inaccurate environmental perception, rigid control strategies, and lack of adaptive capabilities in existing technologies.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent dimming COB display system, comprising:

[0007] The light field sensor array module is used to collect ambient light intensity data in real time through multi-point distributed ambient light sensors, and generate ambient light field distribution information based on the collected data;

[0008] The superpixel segmentation module is used to receive the display image to be displayed and the ambient light field distribution information, and to segment the display image to generate superpixel feature data.

[0009] The model prediction controller module is used to establish a system state space model based on the superpixel feature data and generate an optimal brightness control signal according to a preset control objective function.

[0010] The reinforcement learning engine module is used to dynamically adjust the generation method of the superpixel feature data and the generation parameters of the optimal brightness control signal based on the superpixel feature data and the running state information generated during the generation of the optimal brightness control signal.

[0011] A distributed μLED driving circuit module is used to receive the optimal brightness control signal and drive the μLED light-emitting units in the COB display screen according to the optimal brightness control signal.

[0012] Preferably, the light field sensor array module includes:

[0013] The acquisition unit is used to acquire ambient light intensity values ​​at different spatial locations through multiple ambient light sensors deployed around the display screen;

[0014] The processing unit is used to interpolate and fit the ambient light intensity value to generate a two-dimensional ambient light field data matrix.

[0015] The filtering unit is used to perform spatial smoothing filtering on the ambient light field data matrix to eliminate local abrupt errors and form ambient light field distribution information.

[0016] Preferably, the superpixel segmentation module includes:

[0017] An image preprocessing unit is used to perform color space transformation and edge enhancement processing on the image to be displayed;

[0018] The feature extraction unit is used to combine the preprocessed display image with the ambient light field distribution information to generate a pixel-level color-position-light triplet feature vector.

[0019] The segmentation execution unit is used to cluster the feature vectors using a clustering-based algorithm, output multiple superpixel regions, and generate the superpixel feature data.

[0020] Preferably, the feature extraction unit constructs pixel-level triplet feature vectors through the following steps:

[0021] The image to be displayed is converted into a perceptually uniform CIELAB color space, and the color components of each pixel are extracted.

[0022] Obtain the two-dimensional spatial coordinates of each pixel in the image and perform normalization processing;

[0023] From the ambient light field distribution information, the local ambient light intensity value corresponding to each pixel position is extracted.

[0024] Preferably, the model prediction controller module includes:

[0025] The state modeling unit is used to construct the system state vector and control input vector based on the superpixel feature data, and to establish a linear state space model.

[0026] The objective function construction unit is used to define the control objective function, which includes the deviation term between the target brightness and the actual brightness, the control input variation term, and the brightness uniformity constraint term.

[0027] The optimal solution unit is used to solve the control objective function based on the rolling optimization strategy and output the corresponding optimal brightness control signal.

[0028] Preferably, solving the control objective function based on the rolling optimization strategy includes the following steps:

[0029] Within each sampling period, the system response trajectory for multiple future control periods is predicted starting from the latest state vector;

[0030] A constrained optimization problem is constructed based on minimizing the control objective function, and a set of control sequences is obtained by solving the optimization problem in each cycle;

[0031] The first control input in the control sequence is output as the optimal brightness control signal for the current cycle, and then the cycle begins.

[0032] Preferably, the reinforcement learning engine module includes:

[0033] The state perception unit is used to acquire the convergence error, control input change rate and system stability index during the generation process of the superpixel feature data and the optimal brightness control signal, as operating state information.

[0034] The strategy evaluation unit is used to calculate the strategy value function based on the running status information and evaluate the control performance under the current parameter settings.

[0035] The policy update unit is used to dynamically update the extraction method of the superpixel feature data and the weight parameters of the brightness control signal based on the policy gradient method.

[0036] Preferably, the strategy value function is a weighted linear combination of system state error, control cost, and response stability, including:

[0037] The brightness error term is used to represent the root mean square error between the current output brightness of the display screen and the reference brightness.

[0038] The control energy consumption term is used to represent the total power consumption estimate caused by changes in all control inputs;

[0039] The response jitter term is a statistical indicator used to represent the brightness fluctuation rate within a continuous control period.

[0040] Preferably, the distributed μLED driving circuit module includes:

[0041] A brightness mapping unit is used to calculate the target luminous intensity required by each luminous unit based on the optimal brightness control signal.

[0042] A drive signal generation unit is used to map the target luminous intensity into the duty cycle parameter of a pulse width modulation signal and generate a corresponding PWM drive signal.

[0043] A distributed drive execution unit is used to apply the PWM drive signal to the corresponding μLED light-emitting unit in the COB display screen through multiple independent control channels.

[0044] A method for controlling a COB display screen with intelligent dimming is also provided, including the following steps:

[0045] S1. Collect ambient light intensity data, obtain light information at different locations through multi-point distributed ambient light sensors, and construct ambient light field distribution information;

[0046] S2. Receive the display image to be displayed and the ambient light field distribution information, preprocess the display image and extract features, and generate superpixel feature data by combining the light field information;

[0047] S3. Construct a state space model based on the superpixel feature data, and solve for the optimal brightness control signal in the current period based on the preset control objective function;

[0048] S4. Based on the superpixel feature data and the running status information during the optimal brightness control signal generation process, adjust the feature extraction method and the optimal brightness control signal generation parameters;

[0049] S5. Calculate the target luminous intensity of each luminous unit according to the optimal brightness control signal, generate a PWM signal and drive the corresponding μLED luminous unit in the COB display screen.

[0050] This invention provides an intelligent dimming COB display system and control method. It has the following beneficial effects:

[0051] 1. By introducing a light field sensor array module and an ambient light sensing superpixel segmentation module, this invention achieves accurate perception of display ambient light and fine-grained division of display content. Compared with existing technologies that rely on only a single or a few sensors for global brightness adjustment, this invention solves the problem of its inability to cope with complex and uneven lighting environments, and improves the environmental adaptability and local contrast of the display screen under various lighting conditions.

[0052] 2. This invention employs a model predictive controller and designs a multi-objective optimization function that includes brightness error, control energy consumption, and superpixel stability, achieving smooth and efficient dynamic control of display brightness. Compared with the reactive PID control or simple threshold adjustment methods commonly used in the prior art, this invention fundamentally solves the problems of frequent brightness jumps, flickering, and control overshoot that are easily caused by traditional methods, thereby ensuring the stability of the display effect and effectively reducing the overall power consumption of the system.

[0053] 3. This invention integrates a reinforcement learning engine module, which learns by combining a reward function that integrates power consumption, image fidelity, and adjustment stability, and continuously optimizes the weight parameters of the superpixel segmentation and MPC modules. This enables the system control strategy to be adaptive and self-evolving, achieving optimal long-term performance. Compared with the control systems in the prior art, which typically use fixed, offline calibrated parameters, this invention solves the inherent shortcomings of existing solutions in terms of performance degradation and poor adaptability when facing variable display content and long-term environmental changes. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the system architecture of the present invention;

[0055] Figure 2 This is a schematic diagram of the optical field sensor array module architecture of the present invention;

[0056] Figure 3 This is a schematic diagram of the superpixel segmentation module architecture of the present invention;

[0057] Figure 4 This is a schematic diagram of the model prediction controller module architecture of the present invention;

[0058] Figure 5 This is a schematic diagram of the reinforcement learning engine module architecture of the present invention;

[0059] Figure 6 This is a schematic diagram of the distributed μLED driving circuit module architecture of the present invention;

[0060] Figure 7 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0061] The technical solutions in 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, and 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.

[0062] Please see the appendix Figure 1 -Appendix Figure 6 This invention provides an intelligent dimming COB display system, comprising:

[0063] The light field sensor array module is used to collect ambient light intensity data in real time through multi-point distributed ambient light sensors, and generate ambient light field distribution information based on the collected data;

[0064] Specifically, in this embodiment, the light field sensor array module is used to collect ambient light intensity data in real time and generate ambient light field distribution information based on the collected data, providing basic data support for the brightness adjustment of the display screen. This module captures light intensity from different directions through multiple ambient light sensors distributed around the display screen and converts this data into high-precision ambient light field distribution data, providing accurate input for subsequent brightness adjustment control.

[0065] In the intelligent dimming COB display system of this invention, the core task of the light field sensor array module is to collect ambient light intensity data through the ambient light sensor array to construct a complete ambient light field distribution information. This data not only affects the brightness adjustment of the display screen, but also the stability and energy efficiency of the overall display effect.

[0066] First, the acquisition unit obtains ambient light intensity data at different spatial locations using multiple ambient light sensors deployed around the display screen. The light field sensor array is arranged in a multi-point distribution around the display screen, covering light information at different locations in the environment through multi-dimensional acquisition. These sensor points can perceive changes in ambient light intensity in real time, ensuring comprehensive monitoring of external lighting conditions.

[0067] In some embodiments, the acquisition unit can adjust the sensor's acquisition frequency and sensitivity in real time to adapt to different ambient light changes. For example, at night or under strong light conditions, the sensor's acquisition frequency can be increased to capture more lighting details and improve the acquisition accuracy of ambient light information.

[0068] In this embodiment, the processing unit performs interpolation fitting on the collected ambient light intensity values ​​to generate a two-dimensional ambient light field data matrix. The collected light intensity data typically contains spatial intervals or measurement errors, thus requiring interpolation. Generally, interpolation methods based on local smoothing, such as quadratic or cubic interpolation, can be used to ensure that the generated data matrix accurately reflects the distribution of ambient light.

[0069] For example, by interpolating the light intensity data, the processing unit can generate a two-dimensional light field matrix, where each element represents the light intensity at a specific location in the environment. This matrix provides the foundational information for subsequent brightness compensation and image processing.

[0070] Alternatively, the filtering unit performs spatial smoothing filtering on the generated ambient light field data matrix to eliminate local abrupt errors and ultimately form ambient light field distribution information. Generally, ambient light data may be affected by local noise, sensor errors, or reflected light, resulting in significant data fluctuations. Filtering effectively removes these errors, generating smooth ambient light field distribution data.

[0071] Specifically, the filtering unit uses Gaussian filtering or other smoothing filtering algorithms to process the ambient light field data matrix. Gaussian filtering is typically calculated based on the following formula:

[0072]

[0073] Among them: B comp I is the brightness compensation matrix; env This is the raw ambient light intensity data; K gauss Γ is the Gaussian convolution kernel; Γ is the perceptron coefficient matrix; This represents the element-wise multiplication operation of a matrix.

[0074] In practice, Gaussian filtering effectively smooths ambient light data, reducing abrupt changes in light intensity and making subsequent brightness adjustments smoother and more stable. Through this processing, the resulting ambient light field distribution information accurately reflects the spatial distribution characteristics of ambient light, thus providing reliable data support for subsequent image processing and brightness control.

[0075] In this embodiment, the ambient light data collected by the ambient light sensor array, after the above processing, can be converted into light field data that meets display requirements. In some embodiments, the light field data can be further optimized using the spherical harmonic expansion method. Specifically, through spherical harmonic expansion, the captured ambient light data can be represented as a linear combination of spherical harmonic basis functions. The formula is as follows:

[0076]

[0077] Where: L env (θ,φ) represents the ambient light field data; Y lm (θ,φ) are the spherical harmonic basis functions; c lm θ is the expansion coefficient; θ and φ are the directional angles of the light field, respectively.

[0078] In this approach, spherical harmonic unfolding can effectively capture the global distribution characteristics of the ambient light field, improve the accuracy of the ambient light field data, and provide higher quality data support for the brightness adjustment of the display screen.

[0079] The superpixel segmentation module is used to receive the display image to be displayed and the ambient light field distribution information, and to segment the display image to generate superpixel feature data.

[0080] Specifically, in this embodiment, the main task of the superpixel segmentation module is to receive the display image to be displayed and the ambient light field distribution information, perform image preprocessing and segmentation, and generate superpixel feature data. This module performs saliency analysis on the image and utilizes superpixel segmentation technology to decompose the display image into several regions with similar visual characteristics, thereby providing a foundation for subsequent brightness adjustment and image processing.

[0081] In the aforementioned technical solution, the light field sensor array module has already generated ambient light field distribution information, providing crucial data for subsequent image processing. Based on this, the superpixel segmentation module will continue to process the displayed image and, combined with ambient light information, output superpixel feature data, which provides data support for subsequent brightness adjustment control and optimization of the system.

[0082] In this embodiment, the image preprocessing unit first performs color space transformation and edge enhancement processing on the display image to be displayed. Normally, the display image is presented in the RGB color space, but to adapt to brightness adjustment requirements and better handle the influence of ambient light, the image usually needs to be converted to other suitable color spaces, such as CIELAB or HSV.

[0083] Specifically, the image preprocessing unit performs color space conversion on the image (e.g., from RGB to CIELAB) to separate the color and brightness information of each pixel, thus facilitating subsequent brightness control and color processing. Edge enhancement uses image processing techniques such as the Sobel operator to highlight important edge information in the image, enabling the subsequent segmentation process to more accurately identify and process different image regions.

[0084] Alternatively, the feature extraction unit further combines the preprocessed display image with ambient light field distribution information to generate a triplet feature vector for each pixel, containing color, spatial, and illumination information. This feature vector comprehensively represents the visual characteristics and ambient lighting effects of each pixel in the image.

[0085] Specifically, the feature extraction unit first extracts the color information of each pixel (such as the L, a, and b components in the CIELAB color space), and then normalizes the spatial information by combining it with the two-dimensional spatial coordinates (x, y coordinates) of each pixel in the image. In addition, the ambient light intensity value of the corresponding pixel (ambient light data provided by the light field sensor array module) is also included in the feature vector to enhance the performance of the pixel under different lighting conditions.

[0086] For example, the feature vector of a pixel can be represented as:

[0087]

[0088] Where: l i ,a i ,b i x is the CIELAB color component of the i-th pixel; i ,y i These are the spatial coordinates of the pixel; This represents the ambient light intensity value at that pixel location.

[0089] In this way, the feature extraction unit can accurately capture the color, position, and illumination information of each pixel, providing rich feature data for subsequent image segmentation and brightness adjustment.

[0090] The segmentation execution unit is responsible for inputting the extracted feature vectors into the superpixel segmentation algorithm. It uses a clustering-based algorithm to cluster pixels, outputting multiple superpixel regions and generating superpixel feature data. Generally, superpixel segmentation uses the SLIC (Simple Linear Iterative Clustering) algorithm. This algorithm defines seed points for superpixels on the image and, using these seed points as centers, it iteratively clusters the image into several small regions.

[0091] In this embodiment, the improved SLIC algorithm uses the following distance metric formula:

[0092]

[0093] Where: l i ,a i ,b i and l j ,a j ,b j These are the color components of the superpixel; x i ,x j β is the spatial location; s is the size of the superpixel; β and γ are weighting coefficients; and This represents the ambient light intensity value for the corresponding pixel.

[0094] This distance metric formula takes into account pixel color differences, spatial location differences, and ambient light intensity differences. By optimizing the distance metric, the superpixel segmentation algorithm can fully consider the influence of ambient light while ensuring image quality, thus avoiding display instability caused by changes in ambient light.

[0095] In practice, the cluster centers are updated according to the following formula to ensure the accuracy and stability of the superpixels:

[0096]

[0097] Where: w i The weights output by the MPC module; p i The feature vector for each pixel; C k J is the cluster center of the k-th superpixel; mpc λ represents the optimization objective of the MPC module; λ is the adjustment coefficient. S represents the update result of the k-th superpixel cluster center in the current iteration; k The set of image pixel indices belonging to the k-th cluster center.

[0098] In this way, the center of the superpixel cluster can be continuously updated in each iteration, ensuring that the image segmentation effect meets the requirements of brightness adjustment.

[0099] The model prediction controller module is used to establish a system state space model based on superpixel feature data and generate the optimal brightness control signal according to the preset control objective function.

[0100] Specifically, in this embodiment, the Model Predictive Controller (MPC) module is primarily responsible for optimizing the display screen's brightness adjustment strategy based on superpixel feature data and the aforementioned generated ambient light field distribution information. The MPC module establishes a state-space model to adjust the display screen's brightness in real time, ensuring the display effect remains stable and efficient under various ambient light changes. Combining the superpixel feature data, the MPC module optimizes the brightness output according to the control objective function, thereby achieving precise control of the COB display screen.

[0101] In the aforementioned technical solution, the light field sensor array module and the superpixel segmentation module provide ambient light intensity data and superpixel feature data of the image, respectively. This data provides accurate input information to the MPC module, enabling it to calculate the optimal brightness control signal based on different display requirements and lighting environments.

[0102] In this embodiment, the main task of the state modeling unit is to construct the state-space model of the system based on the superpixel feature data. Generally, the state-space model uses a discrete-time model to describe the dynamic behavior of the system, which includes state vectors, control inputs, and external disturbances.

[0103] Specifically, the state-space model takes the following form:

[0104] x k+1 =Ax k +Bu k +Ed k ;

[0105] Where: x k This is the system's state vector, containing the current brightness state of the display screen, the image content state, and the superpixel segmentation results; u k For control input, it represents the current brightness adjustment signal; d k External disturbances, such as changes in ambient light intensity, are represented by A and B, which are system matrices describing the relationship between the system state and the control input, respectively. The disturbance matrix E describes the impact of external disturbances on the system. k+1 This represents the state of the system at the next moment.

[0106] In some embodiments, the state modeling unit can continuously adjust matrices A, B, and E based on system feedback to adapt to dynamically changing display content and environmental conditions. This method ensures that the system can adaptively adjust under different conditions, maintaining the stability and accuracy of brightness adjustment.

[0107] Alternatively, the objective function building block is used to define the control objective function, which is then used to generate the optimal brightness control signal. The objective function typically includes multiple terms to comprehensively consider precise brightness adjustment, energy efficiency, and display stability.

[0108] Specifically, the control objective function can be expressed in the following form:

[0109]

[0110] Where: y k r represents the actual output brightness of the system. k Δu represents the desired brightness; Q is the weight matrix, representing the weights of the brightness error; Δu k R is the weight matrix for controlling the change in input; ρ is the superpixel stability constraint coefficient, which ensures the stability of the superpixel segmentation result in each control cycle. For the superpixel state at time k; ||·|| F N is the Flobenius norm, used to measure the change in superpixel state; p The prediction time domain length represents the number of time steps for predicting the system's future output; N c To control the time domain length, it represents the number of time steps for optimizing the control variables; u k The control input represents the current brightness adjustment signal.

[0111] Through this objective function, the MPC module can optimize the brightness control signal at each time step, minimizing brightness error, control input variation, and superpixel segmentation fluctuations, thereby ensuring smooth adjustment of display brightness and optimal energy efficiency.

[0112] In this embodiment, the function of the optimal solution unit is to solve for the optimal brightness control signal based on a preset objective function and a scrolling optimization strategy. Generally, the scrolling optimization strategy recalculates the optimization problem in each sampling period to adapt to real-time changes in display content and environmental conditions.

[0113] The core idea of ​​the rolling optimization strategy is to predict the system response for multiple future control cycles within each time step and optimize based on the predicted response. This process can typically be expressed in the following form:

[0114]

[0115] Among them: J mpc (u k The optimization objective for the current control cycle is ). This is the optimal brightness control signal.

[0116] Through rolling optimization, the MPC module can output the most suitable brightness control signal for the current system state in each time period, ensuring optimized display effect.

[0117] Specifically, the rolling optimization strategy includes the following steps:

[0118] Within each sampling period, the latest state vector x k Starting from this point, predict multiple future control cycles u k +1,u k +2,…,u k +N p The system's response trajectory. In this way, the system can predict future behavior at each point in time and formulate optimal strategies for subsequent control cycles.

[0119] A constrained optimization problem is constructed based on minimizing the control objective function. This optimization problem includes multiple terms such as brightness error, control input variation, and superpixel stability. Constraints may include maximum power consumption limits and display uniformity requirements.

[0120] At the end of each control cycle, an optimization problem is solved to obtain a set of control sequences, and these control signals are applied to the system. Within each cycle, the MPC module outputs the control signal from the first step as the optimal brightness control signal for the current cycle and proceeds to the next rolling cycle.

[0121] In this way, the rolling optimization strategy can continuously adjust and optimize the brightness control signal in different control cycles, thereby ensuring the stability and real-time performance of the display effect.

[0122] The reinforcement learning engine module is used to dynamically adjust the generation method of superpixel feature data and the generation parameters of the optimal brightness control signal based on the superpixel feature data and the running status information generated during the generation of the optimal brightness control signal.

[0123] Specifically, in this embodiment, the reinforcement learning engine module dynamically adjusts the generation method of the feature data and the generation parameters of the brightness control signal based on the superpixel feature data and the operating status information generated during the generation of the optimal brightness control signal. Through adaptive learning and optimization, the reinforcement learning engine can effectively adjust the control strategy of the MPC module under constantly changing display content and environmental conditions to ensure optimal system performance and stable display effects.

[0124] In the aforementioned modules, the superpixel segmentation module provides the superpixel feature data of the image, which the MPC module then uses for brightness control. Building upon this, the reinforcement learning engine module dynamically adjusts the MPC module's control strategy through a feedback mechanism to adapt to different display requirements and ambient light changes.

[0125] In this embodiment, the function of the state-aware unit is to acquire superpixel feature data and operational state information during the generation process of the optimal brightness control signal, and to provide this information to the strategy evaluation unit. Generally, the system's operational state information includes the convergence error of brightness adjustment, the rate of change of control input, and the stability indicators of the system response.

[0126] Specifically, the state-aware unit will extract the following information from the system:

[0127] Superpixel feature data includes information such as color, position, and illumination of each superpixel region.

[0128] The process of generating the optimal brightness control signal: The optimal brightness control signal generated in the MPC module can be compared with the actual brightness output to obtain the brightness error.

[0129] Operating status indicators include the rate of change of control inputs (e.g., the increment of control signals) and system stability (e.g., the smoothness and consistency of displayed content).

[0130] This state information provides feedback signals to the reinforcement learning engine, helping it to determine the effectiveness of the current control strategy and make dynamic adjustments to the strategy accordingly.

[0131] In some embodiments, the strategy evaluation unit calculates the strategy value function based on the collected operational status information and evaluates the control performance under the current parameter settings. The strategy value function is a key indicator for measuring the quality of a control strategy, and it can be used to evaluate the performance of the current strategy in a real-world environment.

[0132] Specifically, the policy value function V(s) t This can be represented as:

[0133]

[0134] Where: s t Let a be the state of the system at time t; t The action at time t (i.e., the optimal brightness control signal); r k (s t ,a t ) is based on strategy a at time t t The reward obtained; γ is the discount factor, used to represent the weight of long-term rewards.

[0135] Specifically, the strategy evaluation unit continuously calculates and updates the strategy value function to assess the effectiveness of the current control strategy. When the evaluation result is unsatisfactory, the strategy evaluation unit transmits feedback information to the strategy update unit for optimization.

[0136] In some embodiments, the policy value function can also be designed as a weighted linear combination that considers system state error, control cost, and response stability, specifically including the following:

[0137] Brightness error term‖y k -r k || 2 : Used to represent the mean square error between the current output brightness of the display and the reference brightness, aiming to minimize the error between the actual brightness and the desired brightness.

[0138] Control energy consumption item || Δu k || 2 : Used to represent the total power consumption estimate caused by all changes in control inputs. The control strategy needs to minimize unnecessary energy consumption.

[0139] Response jitter item A statistical indicator used to represent the brightness fluctuation rate within a continuous control period, ensuring smooth brightness changes and avoiding excessive fluctuations.

[0140] These terms together constitute a weighted combination of the strategy value function, ensuring that the system achieves the best balance between brightness error, energy efficiency, and stability.

[0141] In this embodiment, the policy update unit is responsible for dynamically updating the generation method of superpixel feature data and the weight parameters of the brightness control signal according to the policy gradient method. The policy gradient method is an optimization algorithm in reinforcement learning that calculates the policy gradient and adjusts the policy parameters according to the gradient to maximize the cumulative reward.

[0142] Specifically, the policy gradient method can be expressed as the following formula:

[0143]

[0144] Where: θ t Here, α is the policy parameter; α is the learning rate. The policy gradient represents the direction of change of the current policy parameter θ; θ t+1 Let J(θ) be the updated policy parameters at time t+1; J(θ) is the policy performance index function.

[0145] In this way, the strategy update unit gradually adjusts the superpixel feature data generation method and the brightness control signal generation parameters based on the feedback information of the current strategy, in order to adapt to the ever-changing display requirements.

[0146] Specifically, the policy update unit adjusts the weights of the superpixel feature extraction method and the brightness control signal based on the reward signal obtained from reinforcement learning and the policy evaluation results. For example, when the displayed image is too dark or too bright, the policy update unit may adjust the generation method of the brightness control signal or optimize the superpixel segmentation algorithm to ensure better brightness adjustment.

[0147] In some embodiments, the reward function r t It is used to measure the performance of the current strategy and adjust the strategy parameters based on the performance. The design of the reward function typically needs to consider the following aspects:

[0148] Brightness error: By calculating the difference between the actual brightness and the desired brightness, the reward function can encourage the system to reduce brightness error.

[0149] Energy efficiency: In the control strategy, the reward function should also consider power consumption and encourage the use of more efficient control inputs.

[0150] Display quality: The reward function can also be optimized based on the quality of the display, for example, by measuring the quality of the displayed image through the Structural Similarity Index (SSIM).

[0151] The reward function can be:

[0152] r t =-‖ΔP power ‖+α1·SSIM(I disp ,I orig )-β1·TV(S sp );

[0153] Where: ΔP power The display shows the change in power consumption, indicating the power change caused by control adjustments, and reflects the energy efficiency target; P power Indicates power consumption; I disp and I orig These represent the currently displayed image and the original image, respectively; SSIM is the structural similarity index; TV(S sp ) represents the total variation value of the superpixel; ‖·‖ represents the Euclidean norm, which measures the magnitude of power consumption variation; α1 represents the weighting coefficient of the SSIM similarity term; β1 represents the weighting coefficient of the superpixel smoothing term.

[0154] Through this reward design, the system is able to balance the relationship between brightness adjustment, energy efficiency, and display effect.

[0155] The distributed μLED driver circuit module is used to receive the optimal brightness control signal and drive the μLED light-emitting units in the COB display screen according to the optimal brightness control signal.

[0156] Specifically, in this embodiment, the distributed μLED driving circuit module receives the optimal brightness control signal and drives the light-emitting units in the COB display screen according to the signal. The main task of this module is to convert the optimal brightness control signal generated by the model predictive controller (MPC) module into a driving current signal suitable for driving each μLED light-emitting unit. Through precise current adjustment, the driving circuit can ensure that the brightness output of each light-emitting unit in the COB display screen reaches the predetermined target value.

[0157] In the aforementioned modules, the light field sensor array module and the superpixel segmentation module provide the system with ambient light field information and superpixel feature data of the image, while the MPC module generates the optimal brightness control signal through state modeling and rolling optimization strategies. Based on this input data, the distributed μLED driving circuit module performs specific current adjustment operations to drive each μLED emitting unit of the display screen, thereby precisely controlling the brightness and color temperature of the display screen.

[0158] In this embodiment, the function of the brightness mapping unit is to calculate the target luminous intensity required by each luminous unit based on the optimal brightness control signal, and generate the corresponding driving current signal. Generally, there is a certain correspondence between the target luminous intensity and the desired brightness, and the brightness mapping unit generates the current adjustment value based on this relationship.

[0159] Specifically, the mapping relationship between the target luminous intensity and the control signal can be expressed by the following formula:

[0160] I target =f(B target );

[0161] Among them: I target B represents the target luminous intensity. target is the desired brightness value; f(·) is the mapping function between brightness and current.

[0162] In some embodiments, the mapping function f(·) may be a linear relationship, but it may also be a non-linear relationship, such as an exponential or logarithmic function, depending on the characteristics of the μLED and the display requirements.

[0163] Through this mapping relationship, the brightness mapping unit can calculate the current requirement of each light-emitting unit based on the optimal brightness control signal, so as to achieve precise brightness adjustment.

[0164] In some embodiments, the function of the drive signal generation unit is to map the target luminous intensity to the duty cycle parameter of a pulse width modulation (PWM) signal and generate a corresponding PWM drive signal. The PWM signal controls the brightness output of the LED by adjusting the duty cycle. The adjustment of the duty cycle directly affects the average current of the LED, thereby achieving brightness regulation.

[0165] Specifically, the driving signal generation unit generates signals based on the target luminescence intensity I. target Calculate the duty cycle PWM duty The mapping relationship can be represented as:

[0166]

[0167] Among them: I max The maximum drive current for the LED; PWM duty For the corresponding PWM duty cycle; I target The target luminous intensity.

[0168] By adjusting the PWM duty cycle, the drive signal generation unit can precisely control the brightness of each μLED light-emitting unit.

[0169] In this embodiment, the distributed drive execution unit is responsible for applying PWM drive signals to the corresponding light-emitting units in the COB display screen through multiple independent control channels. The display screen is typically composed of multiple μLED light-emitting units, each of which requires an independent drive signal to ensure consistent brightness and color temperature across the entire display screen.

[0170] Typically, a distributed drive execution unit sends PWM signals to different μLED arrays through multiple parallel channels, and each channel can independently adjust the brightness of the connected light-emitting units. Specifically, a distributed drive execution unit may include multiple drive chips or circuit boards, with each chip responsible for a certain number of μLEDs to achieve precise zoned adjustment.

[0171] In this way, the distributed drive execution unit can ensure precise control of the brightness of each light-emitting unit in the display screen, thereby achieving the expected display effect and brightness uniformity.

[0172] As an alternative, the distributed μLED driver circuit module also needs to adjust the LED drive current to meet different brightness requirements. The relationship between drive current and brightness can usually be expressed by the following formula:

[0173]

[0174] Among them: I drive The driving current for each light-emitting unit; B target B represents the desired brightness. actual This refers to the actual brightness. Let K be the expected and actual brightness values ​​at the i-th historical moment, respectively; p K i K d Here are the proportional, integral, and derivative coefficients of the PID controller; n is the number of historical time steps used for the integral term. The rate of change of the target brightness over time.

[0175] This formula adjusts the drive current based on a PID control algorithm to reduce brightness errors and ensure accurate brightness adjustment.

[0176] The intelligent dimming COB display control method described below can be referred to in correspondence with the intelligent dimming COB display system described above.

[0177] Please see the appendix Figure 7 A method for controlling a COB display screen with intelligent dimming includes the following steps:

[0178] S1. Collect ambient light intensity data, obtain light information at different locations through multi-point distributed ambient light sensors, and construct ambient light field distribution information;

[0179] S2. Receive the display image to be displayed and the ambient light field distribution information, preprocess the display image and extract its features, and combine the light field information to generate superpixel feature data;

[0180] S3. Construct a state-space model based on superpixel feature data, and solve for the optimal brightness control signal in the current period based on the preset control objective function;

[0181] S4. Based on the superpixel feature data and the running status information during the generation of the optimal brightness control signal, adjust the feature extraction method and the optimal brightness control signal generation parameters;

[0182] S5. Calculate the target luminous intensity of each luminous unit based on the optimal brightness control signal, generate a PWM signal and drive the corresponding μLED luminous unit in the COB display screen.

[0183] The method in this embodiment can be used to execute the above system embodiment, and its principle and technical effect are similar, so it will not be described again here.

[0184] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart dimming COB display system, characterized in that, include: The light field sensor array module is used to collect ambient light intensity data in real time through multi-point distributed ambient light sensors, and generate ambient light field distribution information based on the collected data; The superpixel segmentation module is used to receive the display image to be displayed and the ambient light field distribution information, and to segment the display image to generate superpixel feature data. The superpixel segmentation module includes: An image preprocessing unit is used to perform color space transformation and edge enhancement processing on the image to be displayed; The feature extraction unit is used to combine the preprocessed display image with the ambient light field distribution information to generate a pixel-level color-position-ambient light intensity triplet feature vector. The segmentation execution unit is used to cluster the feature vectors using a clustering-based algorithm, output multiple superpixel regions, and generate the superpixel feature data. The feature extraction unit constructs pixel-level triplet feature vectors through the following steps: The image to be displayed is converted into a perceptually uniform CIELAB color space, and the color components of each pixel are extracted. Obtain the two-dimensional spatial coordinates of each pixel in the image and perform normalization processing; From the ambient light field distribution information, extract the local ambient light intensity value corresponding to each pixel position; The model prediction controller module is used to establish a system state space model based on the superpixel feature data and generate an optimal brightness control signal according to a preset control objective function. The reinforcement learning engine module is used to dynamically adjust the generation method of the superpixel feature data and the generation parameters of the optimal brightness control signal based on the superpixel feature data and the running state information generated during the generation of the optimal brightness control signal. The reinforcement learning engine module includes: The state perception unit is used to acquire the convergence error, control input change rate and system stability index during the generation process of the superpixel feature data and the optimal brightness control signal, as operating state information. The strategy evaluation unit is used to calculate the strategy value function based on the running status information and evaluate the control performance under the current parameter settings. The policy update unit is used to dynamically update the extraction method of the superpixel feature data and the weight parameters of the brightness control signal based on the policy gradient method. A distributed μLED driving circuit module is used to receive the optimal brightness control signal and drive the μLED light-emitting units in the COB display screen according to the optimal brightness control signal.

2. The intelligent dimming COB display system according to claim 1, characterized in that, The light field sensor array module includes: The acquisition unit is used to acquire ambient light intensity values ​​at different spatial locations through multiple ambient light sensors deployed around the display screen; The processing unit is used to interpolate and fit the ambient light intensity value to generate a two-dimensional ambient light field data matrix. The filtering unit is used to perform spatial smoothing filtering on the ambient light field data matrix to eliminate local abrupt errors and form ambient light field distribution information.

3. The intelligent dimming COB display system according to claim 1, characterized in that, The model prediction controller module includes: The state modeling unit is used to construct the system state vector and control input vector based on the superpixel feature data, and to establish a linear state space model. The objective function construction unit is used to define the control objective function, which includes the deviation term between the target brightness and the actual brightness, the control input variation term, and the brightness uniformity constraint term. The optimal solution unit is used to solve the control objective function based on the rolling optimization strategy and output the corresponding optimal brightness control signal.

4. The intelligent dimming COB display system according to claim 3, characterized in that, The solution of the control objective function based on the rolling optimization strategy includes the following steps: Within each sampling period, the system response trajectory for multiple future control periods is predicted starting from the latest state vector; A constrained optimization problem is constructed based on minimizing the control objective function, and a set of control sequences is obtained by solving the optimization problem in each cycle; The first control input in the control sequence is output as the optimal brightness control signal for the current cycle, and then the cycle begins.

5. The intelligent dimming COB display system according to claim 1, characterized in that, The strategy value function is a weighted linear combination of system state error, control cost, and response stability, including: The brightness error term is used to represent the root mean square error between the current output brightness of the display screen and the reference brightness. The control energy consumption term is used to represent the total power consumption estimate caused by changes in all control inputs; The response jitter term is a statistical indicator used to represent the brightness fluctuation rate within a continuous control period.

6. The intelligent dimming COB display system according to claim 1, characterized in that, The distributed μLED driving circuit module includes: A brightness mapping unit is used to calculate the target luminous intensity required by each luminous unit based on the optimal brightness control signal. A drive signal generation unit is used to map the target luminous intensity into the duty cycle parameter of a pulse width modulation signal and generate a corresponding PWM drive signal. A distributed drive execution unit is used to apply the PWM drive signal to the corresponding μLED light-emitting unit in the COB display screen through multiple independent control channels.

7. A method for controlling a COB display screen with intelligent dimming, applied to a COB display screen system with intelligent dimming as described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Collect ambient light intensity data, obtain light information at different locations through multi-point distributed ambient light sensors, and construct ambient light field distribution information; S2. Receive the display image to be displayed and the ambient light field distribution information, preprocess the display image and extract features, and generate superpixel feature data by combining the light field information; S3. Construct a state space model based on the superpixel feature data, and solve for the optimal brightness control signal in the current period based on the preset control objective function; S4. Based on the superpixel feature data and the running status information during the optimal brightness control signal generation process, adjust the feature extraction method and the optimal brightness control signal generation parameters; S5. Calculate the target luminous intensity of each luminous unit according to the optimal brightness control signal, generate a PWM signal and drive the corresponding μLED luminous unit in the COB display screen.

Images