LED display screen correction method and system

Abstract

The invention discloses an LED display screen correction method which comprises the following steps: performing three-dimensional point cloud scanning on the surface of a display screen through an unmanned aerial vehicle group, and constructing a brightness-chromaticity-time sequence three-dimensional feature matrix; sequentially executing brightness balance compensation based on a particle swarm algorithm by adopting a multi-stage optimization strategy; performing color gamut dynamic matching based on deep learning; gray scale response correction driven by a time domain transfer function; temperature, current and light attenuation parameters are monitored in real time through the distributed optical fiber sensor network and fed back to the correction engine. According to the LED display screen correction method and system, the correction precision, efficiency and reliability are comprehensively improved through three-dimensional scanning of the unmanned aerial vehicle group, the multi-stage optimization engine, the intelligent driving system and the like. Compared with a traditional scheme, the brightness uniformity is improved by 7.4%, the color standard delta Eab is reduced by 71%, the energy consumption is reduced by 23.7%, and the technical bottleneck of curved screen correction is broken through.

Description

Technical Field

[0001] The present invention relates to the technical field of LED display, and in particular to a correction method and system for an LED display screen. Background Art

[0002] The existing LED display screen correction technology has the following defects:

[0003] The contradiction between discreteness and attenuation: the brightness error of the same batch of LED lamp beads when leaving the factory is as high as 20%-40%, the chromaticity deviation is 5nm, and the oxidation of silicone grease during use causes the average annual attenuation rate to exceed 15%.

[0004] Dynamic environmental interference: Traditional cabinet calibration relies on constant darkroom conditions, while on-site calibration is significantly affected by temperature / humidity fluctuations and splicing mechanical errors.

[0005] Insufficient real-time performance: The refresh rate of existing systems is generally lower than 60Hz, which cannot achieve nonlinear compensation during grayscale transitions. Summary of the Invention

[0006] In response to the above-mentioned deficiencies in the prior art, the present invention provides a method and system for correcting an LED display screen. The purpose is to provide a real-time dynamic correction system and method for an LED display screen that combines drone visual acquisition, machine vision algorithms, and multi-level parameter optimization strategies. The system is suitable for improving display quality in complex scenarios such as large outdoor screens and stage backgrounds.

[0007] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0008] The LED display screen calibration method includes the following steps:

[0009] S1: Use a swarm of drones to perform 3D point cloud scanning on the display surface and construct a 3D feature matrix of brightness, chromaticity, and timing.

[0010] S2: Use multi-level optimization strategy to execute in sequence:

[0011] Brightness balance compensation based on particle swarm algorithm;

[0012] Color gamut dynamic matching based on deep learning;

[0013] Grayscale response correction driven by time domain transfer function;

[0014] S3: Real-time monitoring of temperature, current, and optical attenuation parameters through a distributed fiber optic sensor network, and feedback to the correction engine.

[0015] Furthermore, the step S1 includes:

[0016] S11: Use TOF synchronous trigger mechanism to control the drone imaging unit;

[0017] S12: Correct lens distortion using LiDAR ranging data to generate a sub-pixel aligned display surface topology map;

[0018] S13: Extract the normalized brightness distribution under the Gamma2.2 curve and calculate the CIE1931XYZ space color gamut coverage.

[0019] Furthermore, the multi-level optimization strategy in step S2 includes:

[0020] S21: In the brightness balance compensation stage, the particle swarm algorithm parameters c1 = 1.5, c2 = 1.7 are set, and the inertia weight ω decreases linearly from 0.9 to 0.4;

[0021] S22: In the color gamut dynamic matching stage, a training set containing 100,000 color block samples is used, and the loss function is the sum of squared color differences ΔEab;

[0022] S23: In the grayscale response correction stage, the pre-emphasis time constant is calculated according to the formula t_rise=K×exp(-T / τ), where τ=2ms.

[0023] Furthermore, the feedback control in step S3 includes:

[0024] S31: Use the Kalman filter to predict the LED lamp life. The calculation formula is L_t = L_0 × (1-α)^t, where α = 0.15 / year;

[0025] S32: When the brightness attenuation of a single LED is detected to be more than 15%, the local recalibration process is automatically started;

[0026] S33: Dynamically adjust the PWM duty cycle according to the change of the ambient temperature, with the temperature coefficient γ = 0.5% / °C.

[0027] LED display calibration system, including:

[0028] UAV visual acquisition module, equipped with a dual-spectral imaging system and LiDAR rangefinder;

[0029] Edge computing nodes, integrated with FPGA chips and configured with YUV4:4:4 color space conversion units;

[0030] Multi-stage correction engine, including a three-stage cascade processing architecture of brightness compensation channel, color gamut mapping channel and timing compensation channel;

[0031] Intelligent drive unit, using GaN-based PWM driver.

[0032] Furthermore, the UAV visual acquisition module includes:

[0033] The spiral trajectory planning algorithm generates a 3D point cloud scanning path within the 0.5-3m airspace in front of the display screen;

[0034] Adaptive exposure control unit, according to the formula E_t=K×(L_amb+σ 2 / SNR_target) dynamically adjusts imaging parameters, where SNR_target ≥ 45dB;

[0035] The multispectral fusion unit superimposes visible light and near-infrared imaging data to generate a composite feature map.

[0036] Furthermore, the multi-stage correction engine includes:

[0037] In the brightness compensation channel, the particle swarm optimization algorithm is used to allocate the brightness weight of the LED module, and the iterative convergence threshold ε = 0.005;

[0038] The color gamut mapping channel deploys a deep residual network based on the ResNet-18 architecture. The input layer receives CIE1931XYZ colorimetric data and outputs a 3×3 colorimetric conversion matrix.

[0039] The timing compensation channel constructs the time domain transfer function H(s)=K / (Ts+1), performs pre-emphasis processing on the driving signal, and compensates for the grayscale response delay.

[0040] Furthermore, the intelligent driving unit includes:

[0041] Fourth-order current calibration circuit, achieving 0-20mA current closed-loop control through Hall effect sensor;

[0042] PAM4 encoding module, which encodes the correction data into a four-level pulse signal;

[0043] The temperature compensation submodule dynamically adjusts the drive voltage according to the PT1000 sensor data, with a compensation coefficient of β = -0.02V / °C.

[0044] Furthermore, the LED display screen calibration system also includes:

[0045] Embedded self-check module performs a quick correction verification every 24 hours and generates a CRC32 check code;

[0046] The detachable heat dissipation component controls the module temperature at ΔT≤5°C through the heat pipe structure;

[0047] The security encryption unit uses the AES-256 algorithm to perform end-to-end encryption transmission of calibration data.

[0048] The beneficial effects of the present invention are:

[0049] The LED display screen calibration method and system of this invention utilizes drone swarm 3D scanning, a multi-level optimization engine, and an intelligent drive system to comprehensively improve calibration accuracy, efficiency, and reliability. Compared to traditional solutions, it improves brightness uniformity by 7.4%, reduces color accuracy (ΔEab) by 71%, and reduces energy consumption by 23.7%. It also overcomes the technical bottleneck of curved screen calibration. All system indicators meet or exceed design requirements, providing a complete solution for LED display quality control in ultra-large and complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a diagram of the LED display screen correction system architecture of the present invention;

[0051] Figure 2 is a flow chart of the correction engine of the present invention;

[0052] Figure 3 This is the trajectory planning model diagram of the present invention.

[0053] Figure 4 Schematic diagram of the time domain response curve of the present invention; DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present invention are described clearly and completely below with reference to the accompanying drawings.

[0055] like Figures 1 to 4 As shown in the figure, the LED display screen calibration method and system are implemented based on a curved LED display screen in a sports stadium with a radius of 15m, a total area of ​​300㎡, and a resolution of 7680×4320. A complete "drone collaboration + multi-level optimization + intelligent drive" calibration system is constructed. The system hardware components are as follows:

[0056] UAV visual acquisition module

[0057] Deploy 4 six-rotor drones. The flight parameters are shown in Table 1. Each drone is equipped with:

[0058] Dual-spectral imaging system: visible light camera, Sony IMX585, resolution 8000 × 6000, F1.8 aperture, and near-infrared camera, wavelength 850 nm, resolution 4096 × 2160.

[0059] Velodyne VLP-32C LiDAR rangefinder, ranging accuracy ±2mm, scanning frequency 20Hz.

[0060] Onboard computing unit, NVIDIA Jetson AGX Orin, with a computing power of 275TOPS.

[0061] The drone swarm uses a swarm collaboration algorithm:

[0062] The flight height is adjustable from 0.8 to 2.5 meters, and the distance from the screen is dynamically maintained at 0.5 to 1.2 meters;

[0063] Speed ​​0.3m / s, overlapping rate of adjacent tracks ≥35%, guaranteed by TOF synchronous trigger mechanism; point cloud density reaches 500 points / cm 2 , generating a sub-pixel topological map.

[0064] parameter index Hovering accuracy ±3cm (RTK positioning) Battery life 45 minutes (dual battery hot swap) Wind resistance level Level 7 Communication Protocol 5GNR+Wi-Fi6E dual redundant links

[0065] Table 1

[0066] Edge computing nodes:

[0067] Adopting Xilinx Versal ACAP chip, AIEngine+FPGA architecture:

[0068] Implement YUV4:4:4 color space conversion.

[0069] The integrated non-uniformity correction (NUC) algorithm eliminates lens vignetting with a compensation rate of 99.2%. It also features a built-in AES-256 encryption module and a data transmission rate of up to 40Gbps.

[0070] Multi-level correction engine:

[0071] Build a three-stage processing pipeline:

[0072] Brightness compensation channel:

[0073] Input: 16-bit RAW image data;

[0074] Processing: Brightness weight allocation based on improved particle swarm algorithm;

[0075] Particle dimension = number of LED modules, in this embodiment N = 512;

[0076] Iteration number = 200, convergence threshold ε = 0.0045;

[0077] The inertia weight ω decreases linearly from 0.92 to 0.38;

[0078] Gamut mapping channel:

[0079] Input: CIE1931XYZ chromaticity data, ΔEab<1.3;

[0080] Processing: Deep learning model based on ResNet-18;

[0081] Training set: 120,000 sets of color patch samples, covering the Rec.709 / 2020 color gamut;

[0082] Output: 3×3 chroma conversion matrix, with float32 precision;

[0083] Timing compensation channel:

[0084] Input: grayscale response timing data, 0-255 steps;

[0085] Processing: Time domain transfer function pre-emphasis.

[0086] Transfer function: Response time after compensation ≤1.8ms;

[0087] Intelligent drive unit:

[0088] GaN-based PWM driver (model EPC2218):

[0089] Fourth-order current calibration circuit, output stability ±0.03%;

[0090] PAM4 coding transmission density is increased by 32% compared to traditional PWM.

[0091] Temperature compensation coefficient β = -0.021V / ℃, multi-point temperature measurement accuracy ±0.08℃.

[0092] Calibration process:

[0093] UAV dynamic calibration stage;

[0094] 3D point cloud construction:

[0095] Use the spiral trajectory planning algorithm to generate the UAV flight path:

[0096] r(t)=R max ×(1-e -kt ), θ(t)=ωt, z(t)=Z_0+v_zt.

[0097] Where R_max = 7.5m, k = 0.15, ω = 0.8rad / s, Z_0 = 0.8m, v_z = 0.05m / s. Lens distortion is corrected by LiDAR data (radial distortion coefficients k1 = 0.012, k2 = 0.0003). Multispectral data fusion:

[0098] Brightness features are extracted from visible light images and normalized using the Gamma2.2 curve;

[0099] Near-infrared imaging detects the oxidation area of ​​silicone grease, and marks the area with a transmittance difference of >15%;

[0100] Generate composite feature maps with a fusion resolution of 12000×9000;

[0101] Multi-level parameter optimization:

[0102] Brightness balance compensation:

[0103] Particle swarm algorithm parameter configuration:

[0104] Learning factors c1 = 1.52, c2 = 1.68;

[0105] Speed ​​limit v_max = 0.15 × search space;

[0106] The global optimal solution update threshold ΔE<0.004;

[0107] Brightness uniformity increased from 91.7% to 98.5%;

[0108] The maximum brightness difference dropped from 23.4% to 2.1%.

[0109] Color gamut dynamic matching:

[0110] Deep learning model training:

[0111] Batch size = 256, initial learning rate = 0.001;

[0112] Loss function = ΔEab 2 +0.3×SSIM;

[0113] Training took 8 hours (4×A100 GPUs);

[0114] Online reasoning:

[0115] Input tensor size 512×512×3;

[0116] Inference time ≤ 15ms;

[0117] Time domain response correction:

[0118] Pre-emphasis circuit design:

[0119] Establishment time constant τ = 1.8ms;

[0120] Overshoot suppression rate ≥85%;

[0121] Measured results:

[0122] Grayscale 0→255 transition time shortened from 5.2ms to 1.7ms;

[0123] The rate of eliminating smear phenomenon is 92.3%;

[0124] Real-time feedback control stage:

[0125] Deploy 256 PT1000 temperature sensors, with a distribution density of 1 per 0.5 m2; Kalman filter life prediction model:

[0126]

[0127] Where α = 0.17 / year, β = 0.023 / ℃;

[0128] Dynamic adjustment strategy:

[0129] Temperature compensation: ΔPWM = γ × (T_avg - 25°C), γ = 0.48% / °C; Current fine-tuning: recalibration is triggered when ΔI > 0.5%;

[0130] Light attenuation compensation: If the attenuation rate is greater than 15%, the module will be marked and replaced.

[0131] The comparison of key indicators before and after correction is shown in Table 2:

[0132]

[0133]

[0134] Table 2

[0135] Advantages of drone collaboration:

[0136] The system's four-machine parallel acquisition and correction efficiency is increased by 240%.

[0137] The accuracy of 3D point cloud scanning reaches ±1.8mm, which is better than the ±5mm error of traditional 2D scanning.

[0138] Multi-level optimization effect:

[0139] The convergence speed of the brightness compensation PSO algorithm is 35% faster than the traditional gradient descent method.

[0140] The ResNet-18 color gamut matching model reduces ΔEab by 42% compared to the traditional 3DLUT method.

[0141] Drive system innovation:

[0142] The GaN driver conversion efficiency reaches 98.7%, while the traditional MOSFET solution is 94.2%.

[0143] PAM4 encoding increases data transmission bandwidth from 12Gbps to 15.6Gbps.

[0144] Verification of this embodiment on curved LED displays in sports stadiums demonstrates that innovative designs, including drone swarm 3D scanning, a multi-level optimization engine, and an intelligent drive system, have resulted in comprehensive improvements in calibration accuracy, efficiency, and reliability. Compared to traditional solutions, brightness uniformity improved by 7.4%, color accuracy ΔEab decreased by 71%, and energy consumption decreased by 23.7%, breaking through the technical bottleneck of curved screen calibration. All system indicators met or exceeded design requirements, providing a complete solution for LED display quality control in ultra-large and complex scenarios.

[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. LED display screen calibration method, characterized in that, The following steps are involved: S1: Use a swarm of drones to perform 3D point cloud scanning on the display surface and construct a 3D feature matrix of brightness, chromaticity, and timing. S2: Use multi-level optimization strategy to execute in sequence: Brightness balance compensation based on particle swarm algorithm; Color gamut dynamic matching based on deep learning; Grayscale response correction driven by time domain transfer function; S3: Real-time monitoring of temperature, current, and optical attenuation parameters through a distributed fiber optic sensor network, and feedback to the correction engine.

2. The LED display screen calibration method according to claim 1, characterized in that: The step S1 comprises: S11: Use TOF synchronous trigger mechanism to control the drone imaging unit; S12: Correct lens distortion using LiDAR ranging data to generate a sub-pixel aligned display surface topology map; S13: Extract the normalized brightness distribution under the Gamma 2.2 curve and calculate the CIE 1931XYZ space color gamut coverage.

3. The LED display screen calibration method according to claim 1, wherein: The multi-level optimization strategy in step S2 includes: S21: In the brightness balance compensation stage, the particle swarm algorithm parameters c1 = 1.5, c2 = 1.7 are set, and the inertia weight ω decreases linearly from 0.9 to 0.4; S22: In the color gamut dynamic matching stage, a training set containing 100,000 color block samples is used, and the loss function is the sum of squared color differences ΔEab; S23: In the grayscale response correction stage, the pre-emphasis time constant is calculated according to the formula t_rise=K×exp(-T / τ), where τ=2ms.

4. The LED display screen calibration method according to claim 1, wherein: The feedback control in step S3 includes: S31: Use the Kalman filter to predict the LED lamp life. The calculation formula is L_t = L_0 × (1-α)^t, where α = 0.15 / year; S32: When the brightness attenuation of a single LED is detected to be more than 15%, the local recalibration process is automatically started; S33: Dynamically adjust the PWM duty cycle according to the change of the ambient temperature, with the temperature coefficient γ = 0.5% / °C.

5. An LED display screen calibration system, comprising the LED display screen calibration method according to any one of claims 1 to 4, characterized in that: include: UAV visual acquisition module, equipped with a dual-spectral imaging system and LiDAR rangefinder; Edge computing nodes, integrated with FPGA chips and configured with YUV 4:4:4 color space conversion units; Multi-stage correction engine, including a three-stage cascade processing architecture of brightness compensation channel, color gamut mapping channel and timing compensation channel; Intelligent drive unit, using GaN-based PWM driver.

6. The LED display screen calibration system according to claim 5, characterized in that: The UAV visual acquisition module includes: The spiral trajectory planning algorithm generates a 3D point cloud scanning path within the 0.5-3m airspace in front of the display screen; Adaptive exposure control unit, according to the formula E_t=K×(L_amb+σ 2 / SNR_target) dynamically adjusts imaging parameters, where SNR_target ≥ 45dB; The multispectral fusion unit superimposes visible light and near-infrared imaging data to generate a composite feature map.

7. The LED display screen calibration system according to claim 5, characterized in that: The multi-stage correction engine includes: In the brightness compensation channel, the particle swarm optimization algorithm is used to allocate the brightness weight of the LED module, and the iterative convergence threshold ε = 0.005; The color gamut mapping channel deploys a deep residual network based on the ResNet-18 architecture. The input layer receives CIE 1931XYZ colorimetric data and outputs a 3×3 colorimetric transformation matrix. The timing compensation channel constructs the time domain transfer function H(s)=K / (Ts+1), performs pre-emphasis processing on the driving signal, and compensates for the grayscale response delay.

8. The LED display screen calibration system according to claim 5, characterized in that: The intelligent driving unit includes: Fourth-order current calibration circuit, achieving 0-20mA current closed-loop control through Hall effect sensor; PAM4 encoding module, which encodes the correction data into a four-level pulse signal; The temperature compensation submodule dynamically adjusts the drive voltage according to the PT1000 sensor data, with a compensation coefficient of β = -0.02V / °C.

9. The LED display screen calibration system according to claim 5, characterized in that: Also includes: Embedded self-check module performs a quick correction verification every 24 hours and generates a CRC32 check code; The detachable heat dissipation component controls the module temperature at ΔT≤5°C through the heat pipe structure; The security encryption unit uses the AES-256 algorithm to perform end-to-end encryption transmission of calibration data.

Images