A method for self-adaptive regulation and control of brightness of outdoor display screen in strong light environment

Abstract

This invention discloses a method for adaptive brightness control of outdoor displays under strong light conditions, comprising three collaborative steps: first, distributing ambient light sensing nodes in each display module; second, using the Kriging interpolation algorithm to reconstruct a two-dimensional non-uniform ambient light field on the display surface, driving local dimming zones to achieve non-uniform brightness control; third, extracting semantic features of video stream content in real time, distinguishing between high-frequency information areas and low-frequency color areas, and applying differentiated directional brightness gains to each; and fourth, constructing a dynamic three-dimensional lookup table with ambient illuminance, driving current, and node temperature as three-dimensional independent variables, and performing feedforward color pre-correction before pixel data is transmitted to the driving chip. This invention solves the problems of existing single-point sensing schemes being unable to adapt to non-uniform illumination, dimming ignoring content semantics, and the inability to dynamically compensate for color drift under extreme brightness.

Description

Technical Field

[0001] This invention belongs to the field of display screen technology, specifically relating to a method for adaptive brightness control of outdoor LED displays under strong light conditions. Background Technology

[0002] Outdoor LED displays consist of pixel arrays composed of red, green, and blue LED chips. Grayscale adjustment and color mixing are achieved by controlling the driving current of each chip or the duty cycle of pulse width modulation. The brightness requirements of the display in outdoor environments change dynamically with the ambient lighting conditions, ranging from 1000 nits (cloudy conditions) to 10000 nits (strong sunlight conditions).

[0003] Existing outdoor displays generally adopt a global automatic brightness adjustment system based on a single-point ambient light sensor. This system has an ambient light sensor fixedly installed on the outer frame of the display screen. The illuminance value collected by the sensor is converted into a uniform brightness setting value for the entire screen through a preset piecewise linear mapping table. Then, a uniform pulse width modulation duty cycle adjustment command is broadcast to all LED driver chips to achieve synchronous brightness increase and decrease across the entire screen. The core features of the above solution are: a single collection point, uniform control granularity across the entire screen, fixed mapping logic, and complete oblivion to the displayed content.

[0004] The above solution has the following three shortcomings: First, insufficient spatial sampling leads to the failure of zoned brightness control. Single-point sensors can only reflect the local light intensity near their installation location and cannot perceive the spatial differences in illuminance between different areas of the same screen caused by factors such as building obstruction, tree shadows, and local reflections from glass curtain walls. When one area of ​​the screen is in shadow and another area is directly exposed to the sun, the difference in illuminance between the two areas can exceed an order of magnitude. Since the control system can only execute a uniform brightness command on the entire screen based on the single-point acquisition value, the shadowed areas generate unnecessary light pollution and energy waste due to excessive brightening, while the directly exposed areas are still not bright enough and the content is unreadable.

[0005] Second, ignoring the semantic features of the displayed content leads to a degradation in visual quality. Existing dimming solutions are completely unaware of the currently playing content. They use the same brightness gain strategy for areas containing high-frequency information such as text and edges as for areas containing low-frequency information such as gradient color blocks. When the full-screen brightness is forcibly increased to the limit, large areas of color blocks in the picture that do not need high brightness are also brightened by the same amount, resulting in a large amount of ineffective energy consumption and waste. At the same time, the luminous efficiency of R, G, and B LED chips decreases at different rates under excessive current driving, and the heating of the PN junction causes unequal shifts in the spectral peak wavelengths of different color LEDs, resulting in obvious color distortion in the color mixing results.

[0006] Third, the lack of a dynamic color compensation mechanism leads to the loss of color fidelity under extreme brightness conditions. The color calibration of existing systems is usually calibrated once at the factory under standard conditions and solidified into a static Gamma curve or a simple color matrix. Under actual strong light conditions, the driving current may be several times the rated value, the junction temperature of the device may be significantly exceeded by the rated range, the static compensation parameters will completely fail, and there is no mechanism to sense and correct the color drift that occurs in real time, resulting in serious distortion of the brand color of the advertising content. Summary of the Invention

[0007] The purpose of this invention is to provide a method for adaptive brightness control of outdoor displays in strong light environments, in order to solve the technical problems in the prior art where single-point sensors cannot adapt to non-uniform spatial illumination, dimming schemes ignore the semantics of display content leading to visual quality degradation, and color drift cannot be dynamically compensated under extreme brightness conditions.

[0008] This invention is implemented as follows: a method for adaptive brightness control of an outdoor display screen in strong light environments, comprising the following collaboratively executed steps: Ambient light field reconstruction and zone mapping: Ambient light sensor nodes are distributed in an interleaved manner on each display module of the outdoor LED display screen. The illuminance values ​​of each node are periodically collected and stabilized by median filtering. The Kriging space interpolation algorithm is used to reconstruct a two-dimensional non-uniform ambient light field matrix covering the entire screen based on the illuminance sampling values ​​of sparse sensor nodes. The physical coordinate mapping relationship between each cell of the light field matrix and the local dimming control zone is established. The brightness adjustment of each zone is independently driven by the light field illuminance estimation value corresponding to each dimming zone. Content semantic-driven differentiated brightness gain: Real-time parallel feature extraction is performed on pixel data in each dimming zone of the input video stream. The grayscale histogram statistical features and high-frequency energy index of the Sobel operator space are calculated for the zone. Based on the comparison between the high-frequency energy index and the preset judgment threshold, the zone is judged as a high-frequency information-dominated zone or a low-frequency color-dominated zone. A directional peak brightness gain is applied to edge pixels in the high-frequency information-dominated zone whose gradient amplitude exceeds the preset pixel-level threshold. A lower level of basic brightness gain is applied to the low-frequency color-dominated zone. The final brightness command of each zone is synthesized by combining the ambient light field illuminance estimate. Dynamic 3D lookup table color calibration: In the offline calibration stage, the color characteristics of representative LED modules under multiple operating conditions are measured using ambient illuminance, drive current, and node temperature as three-dimensional independent variables. Corresponding RGB pre-correction offset data is generated for each operating point and stored in the multi-dimensional lookup table library. In the online operation stage, the ambient illuminance, drive current, and node temperature parameters of each zone are collected in real time. Through trilinear interpolation, an effective 3D lookup table matching the current operating condition is synthesized in the multi-dimensional lookup table library. Before the pixel data is transmitted to the LED driver chip, the RGB digital values ​​of all pixels in each zone are pre-corrected and mapped forward. The pre-corrected RGB data after color compensation is output.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: By deploying distributed sensing nodes at the module level and introducing the Kriging spatial interpolation algorithm to reconstruct the two-dimensional ambient light field on the display surface, non-uniform partitioned brightness control with millimeter-level spatial resolution is achieved. This enables the brightness output of each area of ​​the screen to be precisely matched with its actual illumination intensity, eliminating the problem of local over-brightness or insufficient brightness caused by global dimming, and reducing ineffective light pollution and energy consumption.

[0010] By introducing a content semantically driven differentiated brightness gain strategy, targeted peak brightness enhancement is applied only to edge pixels in high-frequency information areas, avoiding ineffective brightening of low-frequency color block areas. This reduces overall system power consumption while ensuring content readability and extends device lifespan by reducing LED overdrive.

[0011] By constructing a dynamic three-dimensional lookup table color compensation mechanism that includes three-dimensional independent variables such as ambient illuminance, drive current, and node temperature, the RGB channel data is pre-corrected in real time under different extreme brightness conditions. This ensures that the display can maintain broadcast-level color gamut fidelity even at extreme brightness levels above 5000 nits, effectively solving the color distortion problem under high brightness conditions. Attached Figure Description

[0012] Figure 1 This is a diagram of the overall system architecture of the present invention; Figure 2 This is a flowchart of the Kriging light field reconstruction algorithm for module A of the present invention; Figure 3 This is a flowchart of the semantic analysis and partition brightness instruction synthesis process for module B of the present invention. Figure 4 This is a flowchart of the workflow of the adaptive 3D lookup table color calibration module C of the present invention. Figure 5 This is a sequence diagram showing the coordinated operation of the three modules of this invention. Detailed Implementation

[0013] To further understand the invention's content, features, and effects, the following embodiments are provided, and detailed descriptions are given in conjunction with the accompanying drawings.

[0014] The adaptive brightness control method for outdoor displays under strong light conditions proposed in this invention is based on a four-layer closed-loop system architecture consisting of a sensing layer, a computing layer, a control layer, and a display execution layer. (See [link to relevant documentation]). Figure 1 The complete control process consists of three core modules that operate in synergy: Module A is the spatial reconstruction and partitioning mapping of non-uniform ambient light field based on distributed multi-point sensing; Module B is the dynamic compensation of visual contrast driven by the fusion of ambient light field and display content semantics; and Module C is the adaptive three-dimensional lookup table color dynamic calibration under high brightness and high current thermal coupling conditions. The three modules operate in synergy, with a target response latency of less than 16.7 milliseconds, matching a 60Hz display refresh rate.

[0015] Module A: Spatial Reconstruction and Partition Mapping of Non-Uniform Ambient Light Field In the modular hardware architecture of outdoor LED displays, each display module is used as a deployment unit for sensor nodes. A low-power, high-precision digital ambient light sensor is soldered to one of the four corners of each display module according to the diagonal staggered principle. It is recommended to use a digital illuminance sensor with an I2C interface. The sensor collects the ambient light illuminance through the light-transmitting hole on the front bezel of the module. The measurement range is 0 to 120,000 lux, and the resolution is better than 1 lux. All sensor nodes are connected to the microcontroller on the receiving card in a daisy-chain topology via the I2C bus. The microcontroller polls all sensor nodes at a period of 50 milliseconds, collects and caches the real-time illuminance sampling values ​​of all sensor nodes on the screen, forming a sparse illuminance sampling matrix.

[0016] Median filtering is performed on the illuminance values ​​of each sensor node collected by the microcontroller: the historical observation values ​​of each sensor node for 5 consecutive collection cycles are retained, and the median is taken as the stable output value of the current cycle to eliminate the interference of transient noise spikes caused by rapid cloud movement, bird obstruction, etc. on subsequent interpolation calculations.

[0017] For the execution steps of the Kriging space interpolation algorithm, please refer to [link / reference]. Figure 2First, based on the Euclidean distance and illuminance difference between all sparse sampling point pairs, the experimental variogram values ​​are statistically calculated according to the distance interval to describe the statistical characteristics of illuminance differences between sampling point pairs at different spatial distances. Second, a spherical variogram model is used to perform least-squares fitting on the experimental variogram to determine three model parameters: nugget value, sill value, and range. The nugget value characterizes sensor measurement noise and microscopic spatial variation, the structural variance describes the magnitude of illuminance variation caused by spatial correlation, and the range characterizes the influence range of spatial correlation. Then, for each target dimming zone center point that needs to be estimated, a set of ordinary kriging equations is constructed, with the sum of all weight coefficients equal to 1 as the normalization factor. The optimal weighting coefficients of each sensor node relative to the target point are solved in the FPGA through LU matrix decomposition (to ensure the unbiasedness of the estimation). Finally, the stable illuminance observations of each sensor node are linearly weighted and summed using the obtained weighting coefficients to obtain the estimated ambient illuminance value of the target dimming zone center point. The above Kriging estimation process is performed sequentially on the center points of all dimming zones in the entire screen to finally form a two-dimensional ambient light field estimation matrix covering the entire screen. The matrix dimension is consistent with the local dimming zone grid. Each dimming zone is in 32×32 pixel units. For example, for a display screen with 48 modules wide × 27 modules high, the resolution of the light field estimation matrix is ​​48×27.

[0018] By mapping each cell of the two-dimensional ambient light field estimation matrix to the local dimming control zone of the LED display hardware, an index mapping relationship between the light field zone and the dimming zone is established, so that subsequent brightness control commands can be directly addressed and issued with the dimming zone as the smallest control granularity.

[0019] Module B: Content-Semantic Driven Dynamic Compensation for Visual Contrast This module performs real-time frame-by-frame and partition-by-partition semantic feature extraction of the input video stream within the FPGA. (See [link to FPGA documentation]). Figure 3 Using local dimming zones (32×32 pixels) as the analysis unit, the following two types of features are calculated in parallel for the pixel data in each zone.

[0020] Brightness distribution characteristics: Calculate the grayscale histogram of all pixels in the partition, extract the mean and standard deviation of the histogram to characterize the overall brightness level and grayscale dynamic range of the partition. The grayscale values ​​are obtained by converting RGB values ​​through the standard brightness weight formula, with the weights being: red channel 0.299, green channel 0.587, and blue channel 0.114.

[0021] Spatial high-frequency energy characteristics: Apply 3×3 Sobel convolution kernels in the horizontal and vertical directions to the grayscale image within the partition, calculate the edge gradient magnitude of each pixel (the square root of the sum of the squares of the horizontal and vertical gradient components), and then average the gradient magnitudes of all pixels within the partition to obtain the high-frequency energy index of the partition, which is used to quantify the density of high-frequency information such as text outlines and graphic boundaries within the partition.

[0022] For each dimming zone, the high-frequency energy index is compared with the preset judgment threshold. If the high-frequency energy index is not lower than the judgment threshold, the zone is judged as a high-frequency information-dominated zone, and contrast should be prioritized. If the high-frequency energy index is lower than the judgment threshold, the zone is judged as a low-frequency color-dominated zone, and color accuracy should be prioritized. The judgment threshold is set to 15 gray levels by default (based on the average gradient amplitude of the range of 0 to 255 gray levels). On-site calibration is supported through the debugging interface, and the recommended value range is 10 to 30 gray levels.

[0023] For a partition identified as a high-frequency information-dominated area, the minimum distinguishable contrast threshold is determined based on the estimated ambient illuminance value corresponding to that partition: when the estimated ambient illuminance value is higher than 10,000 lux, the minimum distinguishable contrast threshold is set to 7:1 (based on the outdoor reading contrast requirements of ISO 9241-303 standard); when the estimated ambient illuminance value is in the range of 1,000 lux to 10,000 lux, the minimum distinguishable contrast threshold is determined by linear interpolation between 3:1 and 7:1.

[0024] Based on the minimum discernible contrast threshold and the actual brightness ratio of foreground pixels to background pixels in the current partition content, the peak brightness gain coefficient required to meet the contrast requirements is calculated, with a value ranging from 1.0 to 4.0 (dimensionless). This peak gain coefficient only applies to edge pixels (high-frequency detail feature pixels) within the partition with a gradient amplitude of not less than 30 gray levels. The base brightness gain coefficient is applied to pixels in the low-frequency color-dominant area. Under strong light conditions, the base gain coefficient does not exceed 1.3 (dimensionless), thereby realizing a differentiated gain strategy of high-frequency directional enhancement and low-frequency conservative adjustment.

[0025] By combining the light field illuminance estimates from module A with the aforementioned zoning gain strategy, a final target brightness command is synthesized for each dimming zone. The target brightness command is represented by the percentage of the drive current gain of the local dimming control zone. It is obtained by nonlinearly mapping the light field illuminance value through a pre-calibrated piecewise cubic spline curve (covering an illuminance range of 0 to 120,000 lux, corresponding to a drive current gain range of 10% to 100%), multiplying it by the average gain coefficient of the corresponding zone, limiting it to the maximum current specification, and then outputting it to the local dimming drive unit of the control layer.

[0026] Module C: Adaptive 3D Lookup Table Color Dynamic Calibration This module addresses the problem of severe color shift in LED displays under extreme high-brightness conditions, which is difficult to compensate for in real time. It constructs a physical compensation model for color shift using ambient illuminance, drive current, and node temperature as three-dimensional independent variables. Based on this model, a dynamic three-dimensional lookup table is generated. (See [link to relevant documentation]). Figure 4 .

[0027] In the offline calibration phase, under a controlled laboratory environment, a spectroradiometer was used to systematically measure the color characteristics of representative LED module samples under various operating conditions. Measurement levels were divided according to three-dimensional independent variables: ambient illuminance was divided into 6 levels: 0 lux, 1000 lux, 5000 lux, 20000 lux, 60000 lux, and 120000 lux (simulated using artificial light); driving current was divided into 8 levels, corresponding to current values ​​at brightness levels of 10%, 20%, 30%, 40%, 60%, 80%, 90%, and 100%; and node temperature was divided into 5 levels. The temperature settings are 25 degrees Celsius, 40 degrees Celsius, 55 degrees Celsius, 70 degrees Celsius, and 85 degrees Celsius. The above three-dimensional division forms a total of 240 operating condition measurement points. The actual CIE chromaticity coordinates of the R, G, and B LEDs are measured at each operating condition point, compared with the standard target chromaticity coordinates, and the color deviation is calculated. The pre-correction offset to be applied to the R, G, and B digital input signals is calculated through inverse color space transformation. A corresponding 12-bit precision (4096 levels) three-dimensional lookup table pre-correction data block is generated for each of the 240 operating condition points and stored in the non-volatile flash memory of the receiver card to form a multi-dimensional lookup table library.

[0028] During online operation, three types of operating parameters are collected in real time: the estimated ambient illuminance value of the zone output from module A, the current actual driving current value from the current feedback register of the driver chip, and the estimated junction temperature value from the NTC thermistor of the LED module backplane. Based on the collected combination of the three operating parameters, trilinear interpolation is performed in the multidimensional lookup table library: when the operating parameter falls exactly on the calibration level node, the corresponding data block is directly called; when the operating parameter falls between two adjacent levels, the three-dimensional lookup table data of the adjacent levels are linearly weighted and interpolated to synthesize the effective three-dimensional lookup table of the current operating condition, so as to achieve continuous coverage of the operating parameters. Before the pixel data is transmitted to the LED driver chip, the FPGA performs feedforward pre-correction mapping on all the pixel RGB digital values ​​of the current zone through the synthesized effective three-dimensional lookup table, outputs the color-compensated pre-corrected RGB data, and then transmits it to the driver chip to execute the actual drive, forming a complete feedforward color calibration closed loop.

[0029] Three-module collaborative runtime sequence For the coordinated operation sequence of the three modules within each control cycle (16.7 milliseconds, corresponding to 60Hz), please refer to [link / reference]. Figure 5At the start of the blanking period of each video frame, module A completes the filtering of the sensor data and updates the light field matrix (when the sensor data update cycle is 50 milliseconds, module A updates the light field matrix every 3 frames, and the intermediate frames use the results of the previous cycle); module B processes the pixel data of each frame in real time in parallel in the FPGA pipeline, and completes the calculation of the partition gain coefficient and the synthesis of the brightness command for the current frame during the blanking period; after module B completes the brightness command, module C synchronously completes the three-dimensional lookup table index of the current working condition and the preloading of the feedforward pre-correction data; during the effective display period, module C performs real-time online correction of the pixel data stream and outputs it synchronously to the driver chip. The end-to-end delay of the entire control link is controlled within one frame cycle (16.7 milliseconds) to ensure real-time response to the video content.

[0030] In Module A, the matrix solution of the Kriging equations is implemented in parallel on the FPGA through LU decomposition. In Module B, Sobel convolution kernel operation and histogram statistics are both pipelined in the parallel computing unit of the FPGA to ensure that real-time requirements are met. In Module C, trilinear interpolation calculation is also completed in the FPGA in a hardware-accelerated manner to ensure that LUT synthesis and preloading are completed within the blanking period.

[0031] Alternative implementation methods In Module A, the radial basis function (RBF) interpolation method can be used to replace Kriging interpolation to reconstruct the two-dimensional ambient light field. The radial basis function interpolation uses the Gaussian function as the kernel function and the Euclidean distance between each sampling point as the independent variable. After solving the linear equation system to obtain the weight coefficients of each sampling point, spatial interpolation is performed. The implementation complexity of this alternative method is slightly lower than that of the Kriging method. It can be used as an engineering simplification for application scenarios where the sensor node distribution is relatively regular. However, the interpolation accuracy may be lower than that of the Kriging method in scenarios where the sensor nodes are sparse or unevenly distributed.

[0032] In module C, a lightweight fully connected neural network can be used to replace the three-dimensional lookup table mechanism based on multidimensional lookup tables. Ambient illuminance, drive current, node temperature, and RGB input values ​​are used as network inputs, and the corrected RGB output values ​​are used as network outputs. Supervised training is performed using multi-condition measured data during the factory calibration phase. During online operation, a fixed-point inference engine deployed in the FPGA performs real-time forward computation to output the pre-corrected value. This alternative solution has practical value in products equipped with high-performance FPGAs or dedicated AI inference acceleration chips.

Claims

1. A method for adaptive brightness control of an outdoor display screen under strong light conditions, characterized in that: The following steps are included in the coordinated execution: Ambient light field reconstruction and zone mapping steps: Ambient light sensor nodes are distributed in an interleaved manner on each display module of the outdoor LED display screen. The illuminance values ​​of each sensor node are periodically collected and stabilized by median filtering. The Kriging space interpolation algorithm is used to reconstruct a two-dimensional non-uniform ambient light field matrix covering the entire screen based on the illuminance sampling values ​​of sparse sensor nodes. The physical coordinate mapping relationship between each cell of the light field matrix and the local dimming control zone is established. The brightness adjustment of each zone is independently driven by the light field illuminance estimation value corresponding to each dimming zone. The content semantic-driven differential brightness gain step performs real-time parallel feature extraction on the pixel data in each dimming zone of the input video stream, calculates the grayscale histogram statistical features of the zone and the spatial high-frequency energy index based on the Sobel operator, and determines the zone as a high-frequency information-dominant zone or a low-frequency color-dominant zone based on the comparison result of the high-frequency energy index and the preset judgment threshold. Directional peak brightness gain is applied to edge pixels in the high-frequency information-dominant zone whose gradient amplitude exceeds the preset pixel-level threshold, and basic brightness gain is applied to the low-frequency color-dominant zone. The final brightness command of each zone is synthesized by combining the ambient light field illuminance estimation value. The dynamic 3D lookup table color calibration process involves measuring the color characteristics of representative LED modules under multiple operating conditions during the offline calibration phase, using ambient illuminance, drive current, and node temperature as three-dimensional independent variables. This generates corresponding RGB pre-correction offset data for each operating point and stores it in a multi-dimensional lookup table library. During the online operation phase, the ambient illuminance, drive current, and node temperature parameters of each zone are collected in real time. Through trilinear interpolation, an effective 3D lookup table matching the current operating condition is synthesized in the multi-dimensional lookup table library. Before the pixel data is transmitted to the LED driver chip, feedforward pre-correction mapping is performed on the RGB digital values ​​of all pixels in each zone, and the pre-corrected RGB data after color compensation is output.

2. The method according to claim 1, characterized in that: The ambient light sensing nodes are digital ambient light sensors with I2C interfaces, deployed diagonally at one of the four corners of each display module. The measurement range is 0 to 120,000 lux, and the resolution is better than 1 lux. All sensing nodes are connected to the microcontroller of the receiving card in a daisy-chain topology via the I2C bus. The microcontroller collects the real-time illuminance sampling values ​​of all sensing nodes on the screen with a polling period of 50 milliseconds.

3. The method of claim 1, wherein: The median filtering stabilization process is as follows: for each sensor node, the historical observation values ​​of 5 consecutive acquisition cycles are retained, and the median is taken as the stable output value of the current cycle. The Kriging spatial interpolation algorithm adopts a spherical variogram model and determines the three parameters of nugget value, sill value and range through least squares fitting. For the center point of each target dimming zone, a set of ordinary Kriging equations is constructed. The optimal weight coefficients of each sensor node for the target point are solved with the normalization constraint that the sum of each weight coefficient is equal to 1. The illuminance observation values ​​of each sensor node are linearly weighted and summed using the weight coefficients to obtain the illuminance estimate of the target point. The Kriging equations are solved in parallel in the FPGA through LU matrix decomposition. Each dimming zone is a 32×32 pixel unit.

4. The method of claim 1, wherein: The calculation method of the spatial high-frequency energy index is as follows: apply 3×3 Sobel convolution kernels in the horizontal and vertical directions to the grayscale image in the dimming zone, calculate the edge gradient magnitude of each pixel, and then take the average of the gradient magnitudes of all pixels in the zone. The grayscale value is obtained by weighted summation by assigning weights of 0.299, 0.587 and 0.114 to the three RGB channels respectively. The default value of the preset judgment threshold is 15 gray levels, and the value range is 10 to 30 gray levels. On-site calibration is supported through the debugging interface.

5. The method of claim 1, wherein: For a partition identified as a high-frequency information-dominant area, the minimum distinguishable contrast threshold is determined according to the following rules based on the estimated ambient illuminance value of the partition: when the estimated ambient illuminance value is higher than 10,000 lux, the minimum distinguishable contrast threshold is set to 7:1; when the estimated ambient illuminance value is in the range of 1,000 lux to 10,000 lux, the minimum distinguishable contrast threshold is determined by linear interpolation of the degree value between 3:1 and 7:

1. The directional peak brightness gain coefficient ranges from 1.0 to 4.0, the preset pixel-level threshold is fixed at 30 gray levels, and the basic brightness gain coefficient applied to the low-frequency color-dominant area does not exceed 1.3 under strong light conditions.

6. The method of claim 1, wherein: The offline calibration stage is divided into measurement levels according to three-dimensional independent variables: ambient illuminance is divided into 6 levels, namely 0 lux, 1000 lux, 5000 lux, 20000 lux, 60000 lux, and 120000 lux; driving current is divided into 8 levels, corresponding to current values ​​of 10%, 20%, 30%, 40%, 60%, 80%, 90%, and 100% of brightness levels; and node temperature is divided into 5 levels, namely 25 degrees Celsius, 40 degrees Celsius, 55 degrees Celsius, 70 degrees Celsius, and 85 degrees Celsius, forming a total of 240 operating condition measurement points. The actual CIE chromaticity coordinates of the R, G, and B LEDs at each operating condition point are measured using a spectroradiometer. After comparison with the standard target chromaticity coordinates, the RGB pre-correction offset of each operating condition point is calculated through inverse color space transformation, generating a 12-bit precision three-dimensional lookup table pre-correction data block, which is stored in the non-volatile flash memory of the receiving card to form a multi-dimensional lookup table library.

7. The method of claim 1, wherein: The trilinear interpolation is performed as follows: when the collected ternary operating condition parameter combination falls exactly on the calibration gear node, the corresponding three-dimensional lookup table data block is directly called; when the ternary operating condition parameter combination falls between adjacent gears, the three-dimensional lookup table data of the adjacent gears are linearly weighted and interpolated to synthesize an effective three-dimensional lookup table suitable for the current operating condition. The node temperature parameter is obtained in real time by measuring the NTC thermistor deployed on the back panel of the LED module, and the drive current parameter is read in real time by the current feedback register of the drive chip.

8. An outdoor LED display screen brightness adaptive control system for implementing the method of any one of claims 1 to 7, characterized in that: include: The sensing layer includes ambient light sensing nodes distributed in an interleaved manner on each display module of the display screen, and an I2C bus network for aggregating data from the sensing nodes. The computing layer includes the receiver card main control FPGA, which integrates a Kriging light field reconstruction unit, a content semantic analysis unit, and a three-dimensional lookup table color calibration unit. The three units process sensor layer data, input video stream data, and temperature data in parallel. The control layer, including the local dimming control zone matrix and the PWM current gain control module, receives the zone brightness instructions from the calculation layer and independently adjusts the drive current of each zone. The display execution layer includes LED module driver chips and R, G, B three-color LED chip arrays, which receive pixel data after color pre-correction to complete the actual light output; The computing layer receives drive current parameters from the current feedback register of the driver chip and temperature parameters from the NTC thermistor of the LED module backplane through the three-dimensional lookup table color calibration unit, forming a forward control link from the sensing layer to the computing layer to the control layer to the display execution layer, and a closed-loop feedback link from the temperature sensor to the three-dimensional lookup table color calibration unit.

Images