Mini LED backlight intelligent detection system

Abstract

According to the Mini LED backlight intelligent detection system, an optical detection module shoots full-bright, full-black and gray-scale pictures through a high-frame-rate camera, brightness uniformity is calculated in combination with transmittance data of a diffusion plate, a color gamut coverage rate is analyzed through a spectrograph, and it is ensured that the color gamut coverage rate meets the HDR10 + standard; and the electrical test module adopts multi-power-supply voltage parallel detection, measures the voltage value of each lamp bead in real time, records voltage offset and marks a fault position. The AI defect identification module preprocesses the acquired image, extracts a defect area by using Hough circle transformation and mask processing, and generates a defect thermodynamic diagram; the dynamic performance optimization module suppresses halo and improves contrast by adjusting Local Dimming algorithm parameters; the data uploading and cloud platform integration module compresses data through edge computing equipment and uploads the data to a cloud end for statistical analysis, so that the yield of a production line is monitored in real time, and the detection efficiency, precision and display quality are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of display, in particular to a Mini LED backlight intelligent detection system for the automatic detection and quality control of the brightness, chrominance, uniformity and defects of a backlight module. BACKGROUND

[0002] In the prior art, the detection method of Mini LED backlight modules is relatively single and inefficient, which is difficult to meet the daily production needs of factories. The specific technical problems include:

[0003] Low detection efficiency: traditional manual visual inspection or single optical detection cannot meet the rapid detection needs of high-density LEDs.

[0004] Insufficient accuracy: small defects such as dead lights, weak lights, and color spots (Mura) are difficult to be recognized by conventional algorithms.

[0005] Dynamic performance is missing: there is a lack of real-time testing of backlight dynamic dimming (Local Dimming) effect.

[0006] Insufficient safety: hot plug operation is easy to damage the lamp beads, and lacks an automatic protection mechanism.

[0007] The above problems seriously restrict the production efficiency and product quality of Mini LED backlight modules, and an efficient, accurate and dynamic detection system is urgently needed to solve these problems. SUMMARY

[0008] The purpose of this section is to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0009] In view of the above problems existing in the prior art Mini LED backlight module detection method, the present application is proposed.

[0010] Therefore, the technical problem solved by the present application is to solve the problems of low detection efficiency, insufficient accuracy, missing dynamic performance and insufficient safety of the existing Mini LED backlight module detection method.

[0011] To solve the above technical problems, the present application provides the following technical solutions: a Mini LED backlight intelligent detection system, comprising the following modules: an optical detection module, using a high frame rate camera to shoot full brightness / full black / gray scale pictures, calculating brightness uniformity in combination with diffusion plate transmittance data, synchronously analyzing color gamut coverage by a spectrometer to ensure compliance with the HDR10+ standard; an electrical test module, using a multi-power voltage parallel detection system to drive the backlight module, recording voltage offset of abnormal lamp beads, and marking fault positions; an AI defect recognition module, performing Hough circle transformation and mask processing on collected images, inputting a mixed model to classify normal / abnormal areas, and outputting a defect heat map; a dynamic performance optimization module, adjusting Local Dimming algorithm parameters according to test results to suppress halos and improve contrast; a data uploading and cloud platform integration module, compressing data through an edge computing device, uploading to the cloud for statistical analysis, and realizing real-time monitoring of production line yield; wherein the modules communicate with each other through a data bus.

[0012] As a preferred scheme of the Mini LED backlight intelligent detection system described in the present application, when the optical detection module uses a high frame rate camera to shoot full brightness / full black / gray scale pictures, it is set as: high frame rate camera parameters: frame rate >= 30fps; resolution >= 1920x1080; exposure time is automatically adjusted according to LED brightness; shooting mode: all LED lamp beads are turned on when shooting full brightness pictures to detect overall brightness uniformity; all LED lamp beads are turned off when shooting full black pictures to detect light leakage; when shooting gray scale pictures, the brightness increases from 0% to 100% to detect brightness gradient uniformity.

[0013] As a preferred scheme of the Mini LED backlight intelligent detection system described in the present application, wherein: the optical detection module calculates brightness uniformity in combination with diffusion plate transmittance data, specifically comprising the following steps: S1: measuring transmittance at different positions using a spectrometer to generate a transmittance distribution map; S2: measuring actual brightness values, correcting according to the transmittance distribution, and calculating brightness uniformity; wherein the brightness uniformity is calculated according to the following formula:

[0014]

[0015] wherein L i is the corrected brightness value at the i-th position.

[0016] As a preferred scheme of the Mini LED backlight intelligent detection system described in the present application, wherein: the optical detection module analyzes color gamut coverage by a spectrometer, specifically comprising the following steps: Q1: measuring spectral distribution of different colors using a spectrometer; Q2: calculating the coverage area of the measured spectrum and the HDR10+ standard color gamut; wherein the color gamut coverage is calculated according to the following formula:

[0017]

[0018] As a preferred scheme of the Mini LED backlight intelligent detection system, when the electrical test module adopts multiple power supply voltages for parallel detection, each lamp bead is independently driven, the voltage value of each lamp bead is measured in real time, the voltage offset is recorded, the voltage offset threshold is set, and when the voltage offset threshold is exceeded, it is determined to be abnormal, and the position coordinates of the abnormal lamp bead are recorded synchronously; wherein the voltage offset is calculated according to the following formula:

[0019] Voltage offset = |V 实际 -V 理论 |

[0020] Wherein, V 实际 and V 理论 are the actual measured voltage and the theoretical voltage respectively.

[0021] As a preferred scheme of the Mini LED backlight intelligent detection system, before the image is processed by the AI defect recognition module, the image is also pre-processed; wherein the pre-processing step includes noise removal processing and contrast enhancement.

[0022] As a preferred scheme of the Mini LED backlight intelligent detection system, when the dynamic performance optimization module suppresses halo and improves contrast, the luminance difference between adjacent regions is adjusted to reduce halo phenomenon; by increasing the dark area brightness, the overexposure of the bright area is reduced, and the overall contrast is improved.

[0023] As a preferred scheme of the Mini LED backlight intelligent detection system, the statistical analysis of the data uploading and cloud platform integration module is based on the following model:

[0024]

[0025] The present application provides a Mini LED backlight intelligent detection system, which has the following beneficial effects:

[0026] 1. Multi-modal data fusion: synchronously analyze optical, electrical and dynamic light adjustment data to generate a comprehensive quality evaluation report;

[0027] 2. Dynamic light adjustment optimization: adjust backlight drive parameters through feedback mechanism to improve local light adjustment linearity;

[0028] 3. Adaptive algorithm: automatically optimize detection parameters for different Mini LED packaging processes (such as COB / POB);

[0029] 4. Cloud platform integration: connect quality management cloud service to monitor yield on production line in real time and predict process bottlenecks;

[0030] The application not only improves the detection efficiency and precision, but also realizes real-time testing and optimization of the dynamic performance of the backlight module, thereby providing reliable technical support for the wide application of Mini LED display technology. DETAILED DESCRIPTION

[0031] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present application.

[0032] In view of the low detection efficiency, insufficient precision, lack of dynamic performance and insufficient safety of the existing Mini LED backlight module detection method, the present application provides a Mini LED backlight intelligent detection system, which comprises the following modules:

[0033] The optical detection module uses a high-frame-rate camera to shoot full-brightness / full-black / gray-scale pictures, calculates the brightness uniformity in combination with the diffuser plate transmittance data, synchronously analyzes the color gamut coverage rate through a spectrometer, and ensures compliance with the HDR10+ standard;

[0034] The electrical test module drives the backlight module using a multi-power-voltage parallel detection system, records the voltage offset of abnormal lamp beads, marks the fault position, and facilitates subsequent maintenance;

[0035] The AI defect recognition module performs Hough circle transformation and mask processing on the collected images, extracts potential defect areas, inputs a hybrid model to classify normal / abnormal areas, and outputs a defect heat map, thereby realizing accurate identification of micro-defects such as dead lamps, weak brightness and color spots;

[0036] The dynamic performance optimization module adjusts the Local Dimming algorithm parameters according to the test results, suppresses halos and improves contrast;

[0037] The data uploading and cloud platform integration module compresses data through an edge computing device, uploads it to the cloud for statistical analysis, and realizes real-time monitoring of the production line yield;

[0038] Among them, each module communicates with each other through a data bus to realize fusion and analysis of multi-modal data.

[0039] Further, when the optical detection module uses a high-frame-rate camera to shoot full-brightness / full-black / gray-scale pictures, it is set as:

[0040] High-frame-rate camera parameters: frame rate ≥ 30 fps; resolution ≥ 1920x1080; exposure time is automatically adjusted according to LED brightness;

[0041] Shooting mode: all LED lamp beads are turned on when full bright picture, detecting overall brightness uniformity; all LED lamp beads are turned off when full black picture, detecting light leakage; when gray scale picture, from 0% to 100% brightness increment, detecting brightness gradient uniformity.

[0042] Further, the optical detection module combines the diffuser transmittance data to calculate the brightness uniformity, which specifically includes the following steps:

[0043] S1: using a spectrometer to measure the transmittance at different positions, generating a transmittance distribution map;

[0044] S2: measuring the actual brightness value, correcting according to the transmittance distribution, and calculating the brightness uniformity;

[0045] Wherein, the brightness uniformity is calculated according to the following formula:

[0046]

[0047] Wherein, L i is the corrected brightness value of the i-th position.

[0048] Further, the optical detection module analyzes the color gamut coverage rate by using a spectrometer, which specifically includes the following steps:

[0049] Q1: using a spectrometer to measure the spectral distribution of different colors;

[0050] Q2: calculating the coverage area of the measured spectrum and the HDR10+ standard color gamut;

[0051] Wherein, the color gamut coverage rate is calculated according to the following formula:

[0052]

[0053] Further, when the electrical test module adopts multiple power supply voltages for parallel detection, each lamp bead is independently driven, the voltage value of each lamp bead is measured in real time, the voltage offset is recorded, the voltage offset threshold (such as ±5%) is set, and when the voltage offset threshold is exceeded, it is determined as abnormal, and the position coordinates of the abnormal lamp bead are recorded synchronously, which is convenient for subsequent maintenance;

[0054] Wherein, the voltage offset is calculated according to the following formula:

[0055] Voltage offset = |V 实际 -V 理论 |

[0056] Wherein, V 实际 and V 理论 are the actual measured voltage and the theoretical voltage, respectively.

[0057] Further, the AI defect recognition module further includes pre-processing of the image before processing the image.

[0058] The pre-processing step includes denoising and contrast enhancement.

[0059] It should be noted that the AI defect recognition module includes the following steps:

[0060] 1. Image pre-processing

[0061] Objective: Denoising and contrast enhancement of the original image to improve the accuracy of subsequent processing.

[0062] 1.1 Denoising

[0063] Median filtering:

[0064] Median filtering is a nonlinear filtering method suitable for removing salt and pepper noise. The core idea is to replace the value of the center pixel with the median value of the pixels in the neighborhood, thereby eliminating isolated noise points.

[0065] Implementation steps:

[0066] Define the filter window size (e.g., 3x3).

[0067] Traverse the image and calculate the median value of the pixels in the 3x3 region around each pixel.

[0068] Replace the value of the center pixel with the median value.

[0069] Mathematical expression:

[0070] I filtered (x, y) = median{I(x-i, y-j) | i, j = 0, 1, 2}

[0071] Where I represents the original image, I represents the input image data; I filtered represents the filtered image, and x, y represents the pixel coordinates in the image, representing the position of each pixel in the image; i, j represents the offset of the filter window, representing the offset position of the filter when sliding on the image (0, 1, 2).

[0072] Gaussian filtering:

[0073] Gaussian filtering is suitable for removing Gaussian noise (such as camera sensor noise). It performs weighted averaging of the image using a Gaussian kernel to smooth the image.

[0074] Without further elaboration, Gaussian filtering directly uses the existing conventional Gaussian filtering formula, with the following steps: define the Gaussian kernel size (e.g., 5x5) and standard deviation (σ); calculate the Gaussian kernel weight; use the Gaussian kernel to perform convolution operation on the image.

[0075] 1.2 Contrast Enhancement

[0076] Histogram Equalization: Histogram equalization enhances image contrast by stretching the histogram of the image.

[0077] Implementation Steps:

[0078] Compute the grayscale histogram H(g) of the image, where g represents the grayscale level (0-255).

[0079] Compute the cumulative distribution function (CDF):

[0080]

[0081] where M and N are the height and width of the image, respectively, representing the size of the image; H(g) is the grayscale histogram, representing the number of pixels in each grayscale level (0-255) in the image; g is the grayscale level, representing the brightness value in the image; CDF(g) is the cumulative distribution function, representing the cumulative probability from grayscale level 0 to g.

[0082] Map the CDF to new grayscale levels:

[0083] I enhanced (x, y) = round(CDF(I(x, y)) * 255)

[0084] where I enhanced is the enhanced image, representing the image after histogram equalization processing.

[0085] Gamma Correction: Gamma correction adjusts image brightness through nonlinear transformation, improving visual effects.

[0086] Implementation Steps:

[0087] Define the Gamma value (usually take γ = 2.2);

[0088] Transform each pixel:

[0089]

[0090] where I gamma is the Gamma-corrected image, representing the image after Gamma transformation processing; γ is the Gamma value, usually taking 2.2, used to adjust the brightness and contrast of the image.

[0091] 2. Hough Circle Transformation

[0092] Objective: Locate the circular region of the LED lamp bead, extract the potential defect region.

[0093] 2.1 Circle Detection

[0094] Hough circle transform principle:

[0095] Hough circle transform converts the geometric properties of circles into points in parameter space through polar coordinate transformation. For each pixel (x, y), its possible circle parameters are (a, b, r), where (a, b) is the center and r is the radius. The confidence of each parameter is counted through a voting mechanism.

[0096] Implementation steps:

[0097] ① Define the search range:

[0098] Minimum radius r min : Set according to the size of the lamp bead (such as 10 pixels).

[0099] Maximum radius r max : Set according to the size of the lamp bead (such as 30 pixels).

[0100] ② Initialize the parameter space accumulator A(a, b, r).

[0101] ③ Traverse each edge point (x, y) in the image:

[0102] Calculate the possible center (a, b):

[0103] a = x - rcosθ, b = y - rsinθ

[0104] Where θ is the polar angle (0 ≤ θ < 2π); A(a, b, r) is the accumulator, representing the number of votes for each possible circle (defined by center (a, b) and radius r) in the parameter space; a, b are the center coordinates, representing the center position of the detected circular region; r is the radius, representing the radius of the detected circular region.

[0105] Vote for (a, b, r) in the accumulator.

[0106] ④ Find the (a, b, r) with the highest number of votes in the accumulator, which is the detected circular region.

[0107] Parameter settings:

[0108] Accumulation threshold: Set a threshold (such as 100), only when the accumulator value exceeds the threshold, it is considered that a circle is detected.

[0109] 3. Mask processing

[0110] Objective: Highlight potential defect areas and extract edge features.

[0111] 3.1 Binary processing

[0112] Thresholding: The image is divided into foreground (defect region) and background (normal region) by setting a threshold value.

[0113] Implementation steps:

[0114] ① Calculate the average gray value μ of the image.

[0115] ② Set the threshold value T = μ + k, where k is the adjustment coefficient (such as 50).

[0116] ③ Binaryzation for each pixel:

[0117]

[0118] Where I binary is the binary image, representing the image after thresholding processing; T is the threshold value, representing the brightness threshold of the segmented image, pixels higher than this value are set to 255, otherwise set to 0.

[0119] 3.2 Edge detection

[0120] Canny edge detection: Canny edge detection is a multi-stage algorithm, including Gaussian filtering, gradient calculation, non-maximum suppression and double threshold detection.

[0121] Implementation steps:

[0122] ① Use a Gaussian filter to smooth the image.

[0123] ② Calculate the gradient amplitude and direction:

[0124] Use the Sobel operator to calculate the horizontal gradient Gx and the vertical gradient Gy:

[0125]

[0126] Gradient amplitude:

[0127]

[0128] Gradient direction:

[0129]

[0130] Where G x , G y is the Sobel operator, representing the operator used to calculate the horizontal and vertical gradients of the image; G is the gradient amplitude, representing the gradient intensity of each pixel in the image; θ is the gradient direction, representing the gradient direction of each pixel in the image.

[0131] ③ Non-maximum suppression: Check adjacent pixels along the gradient direction and keep the pixel with the maximum gradient amplitude.

[0132] ④ Double-threshold detection:

[0133] Set high and low thresholds T low and T high ( e.g. T low = 50, T high = 100).

[0134] Mark pixels with gradient magnitude greater than T high as strong edges.

[0135] Mark pixels with gradient magnitude between T low and T high as weak edges, which are kept if they are connected to strong edges.

[0136] 4. Hybrid model classification

[0137] Objective: Classify normal and abnormal regions, generate defect heat map.

[0138] 4.1 Feature extraction

[0139] Brightness feature: Extract the average brightness value L avg and standard deviation σ L of the defect region:

[0140]

[0141] Color feature: Convert the image from RGB space to HSV space, extract hue (H), saturation (S) and lightness (V) features.

[0142] Shape feature: Extract the area A, perimeter P and circularity C of the defect region:

[0143]

[0144] Where L avg is the average brightness, representing the average brightness value of the defect region; σ L is the brightness standard deviation, representing the dispersion degree of the brightness of the defect region; C is the circularity, representing the circularity of the defect region, the value is closer to 1, the circularity is higher; A is the area of the defect region, representing the size of the defect region; P is the perimeter of the defect region, representing the boundary length of the defect region.

[0145] 4.2 Classifier training

[0146] Support Vector Machine (SVM): SVM is a supervised learning algorithm that separates two classes of samples by finding a hyperplane.

[0147] Training steps:

[0148] ① Collect the feature vectors X of normal and abnormal regions.

[0149] ② Define label y, where y=1 represents normal and y=0 represents abnormal.

[0150] ③ Choose a kernel function (e.g., radial basis function RBF):

[0151] K(x i , x j ) = exp(-γ||x i -x j ||) 2 )

[0152] Optimize the objective function:

[0153]

[0154] subject to y i (w·x i +b) ≥ 1-ξ i , ξ i ≥ 0

[0155] Convolutional Neural Network (CNN): CNN is a deep learning algorithm suitable for image classification tasks.

[0156] Training steps:

[0157] Build network architecture (e.g., LeNet-5).

[0158] Define loss function (e.g., cross-entropy loss):

[0159]

[0160] Optimize network parameters using backpropagation algorithm.

[0161] Verify model performance and adjust hyperparameters (e.g., learning rate, batch size).

[0162] where K is the kernel function, representing the function used to map input data to high-dimensional space; γ is the kernel parameter, controlling the smoothness of the kernel function; L is the loss function, representing the loss value of the classifier, used to measure the classification effect; w is the classifier weight, representing the decision boundary of the classifier; C is the penalty coefficient, representing the punishment degree of the classifier for misclassification; ξi is the relaxation variable, representing the tolerance degree of the classifier for misclassification.

[0163] 4.3 Defect heat map generation

[0164] Heat map generation: According to the classification result, generate a heat map to visually display the defect location and severity.

[0165] Implementation steps:

[0166] Initialize the heat map H with the same size as the original image.

[0167] For each pixel (x, y), assign a heat value according to the classification result:

[0168]

[0169] where p(x, y) is the defect probability, representing the probability of each pixel in the image belonging to the defect area; H is the heat map, representing the heat distribution map of the defect area.

[0170] Visualize the heat map using color mapping (such as JET colormap).

[0171] Furthermore, when the dynamic performance optimization module 400 suppresses halo and improves contrast, it adjusts the brightness difference between adjacent areas to reduce halo phenomenon; by increasing the brightness of dark areas and reducing overexposure in bright areas, the overall contrast is improved.

[0172] It should be noted that for the dynamic performance optimization module 400:

[0173] 1. Local Dimming algorithm parameter adjustment

[0174] Objective: Adjust the parameters of the Local Dimming algorithm to optimize the dynamic dimming effect of the backlight module and improve display quality.

[0175] 1.1 Parameter definition

[0176] Region brightness gain (Gr): used to adjust the brightness of a specific area to enhance or weaken the display effect of that area.

[0177] Dynamic range compression coefficient (C DR ): used to compress the dynamic range of high brightness areas to prevent overexposure.

[0178] Halo suppression coefficient (S halo ): used to reduce the halo phenomenon caused by LED array dimming.

[0179] 1.2 Feedback mechanism

[0180] Input: results of optical detection and AI defect recognition, including brightness uniformity, color gamut coverage, defect heat map, etc.

[0181] Output: optimized LocalDimming parameters (Gr, C DR , S halo ).

[0182] Implementation steps:

[0183] Collect the results of optical detection and AI defect recognition, analyze the current backlight module display effect.

[0184] According to the analysis results, calculate the parameter range that needs to be adjusted.

[0185] Use optimization algorithms (such as gradient descent or genetic algorithm) to find the optimal parameter combination.

[0186] Update the LocalDimming algorithm parameters and re-evaluate the display effect.

[0187] Repeat steps 2-4 until the preset optimization target is reached.

[0188] 1.3 Optimization algorithm

[0189] 1.3.1 Gradient descent method

[0190] Objective function: Define a loss function to measure the difference between the current display effect and the ideal effect.

[0191] L = α·L uniformity + β·L halo + γ·L contrast

[0192] Where α, β, γ are weight coefficients; L uniformity is the brightness uniformity loss; L halo is the halo loss; L contrast is the contrast loss.

[0193] Optimization steps:

[0194] Initialize parameters G r (0) , C DR (0) , S halo (0) .

[0195] Calculate the objective function L.

[0196] Calculate the gradient including the partial derivatives of G r , C DR , S halo .

[0197] Update parameters:

[0198]

[0199] Where η is the learning rate; G r is the regional brightness gain, indicating the gain value used to adjust the brightness of a specific region; C DRDynamic range compression coefficient, representing the coefficient for compressing the dynamic range of high brightness regions; S halo Halo suppression coefficient, representing the coefficient for reducing the halo phenomenon; L is the loss function, representing the loss value of the classifier, used to measure the classification effect; α, β, γ are weight coefficients, representing the weights of different loss terms; L uniformity Luminance uniformity loss, representing the loss value of luminance uniformity; L halo Halo loss, representing the loss value of the halo phenomenon; L contrast Contrast loss, representing the loss value of contrast; η is the learning rate, representing the step size of parameter update.

[0200] Repeat steps 2-4 until L converges.

[0201] 1.3.2 Genetic algorithm

[0202] Population initialization: randomly generate a set of parameter combinations (chromosomes), each containing G r , C DR , S halo .

[0203] Fitness function: define the fitness function to measure the goodness of parameter combinations.

[0204]

[0205] where Fitness is the fitness function, representing the fitness value of individuals, used to measure the goodness of individuals; L is the loss function, representing the loss value of the classifier, used to measure the classification effect.

[0206] Selection operation: select chromosomes with high fitness for reproduction according to the fitness function.

[0207] Cross operation: randomly select two chromosomes, exchange part of the parameters, and generate new offspring.

[0208] Mutation operation: randomly change the value of a parameter of the offspring to increase the diversity of the population.

[0209] Iteration: repeat the selection, cross, and mutation operations until the preset number of iterations or the fitness converges.

[0210] 2. Halo suppression and contrast enhancement

[0211] Objective: reduce the halo phenomenon by adjusting the luminance difference between adjacent regions; enhance the overall contrast by increasing the brightness of dark areas and reducing the overexposure of bright areas.

[0212] 2.1 Halo suppression

[0213] Principle: The halo phenomenon is caused by the dimming of the LED array. By adjusting the brightness difference between adjacent regions, the visual impact of the halo can be reduced.

[0214] Implementation steps:

[0215] Define the brightness difference threshold T of adjacent regions diff .

[0216] Calculate the brightness difference between the current region and its adjacent regions:

[0217] D = |L current -L neighbor |

[0218] If D > T diff , adjust the brightness gain G of the current region r so that D approaches T diff .

[0219] G r = G r + ΔG r

[0220] where ΔG r is the adjustment amount, calculated according to D and T diff ; D is the brightness difference, representing the brightness difference between the current region and its adjacent regions; L current is the current region brightness, representing the brightness value of the current region; L neighbor is the adjacent region brightness, representing the brightness value of the adjacent region; k is the adjustment coefficient, representing the step size of parameter adjustment; T diff is the brightness difference threshold, representing the maximum allowed brightness difference.

[0221] 2.2 Contrast enhancement

[0222] Objective: Enhance the contrast of the image by dynamically adjusting the brightness, and improve the display effect.

[0223] Implementation steps:

[0224] Calculate the global contrast of the image:

[0225]

[0226] where L max , L min , L avg are the maximum brightness, minimum brightness and average brightness of the image, respectively.

[0227] According to the global contrast, adjust the dynamic range compression coefficient C DR :

[0228]

[0229] wherein T low and T high are preset contrast threshold values.

[0230] Adjusting the brightness gain G r to increase the brightness of dark regions and reduce overexposure of bright regions:

[0231]

[0232] wherein L dark and L bright are brightness threshold values of dark and bright regions; C global is a global contrast, representing the overall contrast of the image; L max is a maximum brightness, representing the maximum brightness value in the image; L min is a minimum brightness, representing the minimum brightness value in the image; L avg is an average brightness, representing the average brightness value in the image; T low,Thigh is a contrast threshold value, representing the low threshold and high threshold of the contrast; L target is a target brightness, representing the desired brightness value to be achieved; and L current is a current brightness, representing the brightness value of the current region.

[0233] 3. Specific formula

[0234] 3.1 Brightness uniformity calculation

[0235]

[0236] wherein L i is the brightness value of each region, representing the brightness value of each region in the image.

[0237] 3.2 Halo suppression adjustment

[0238] G r = G r + ΔG r = G r + k·(T diff - D)

[0239] wherein k is an adjustment coefficient, representing the step size of parameter adjustment; T diff is a brightness difference threshold value, representing the maximum allowed brightness difference; and D is a current brightness difference, representing the brightness difference between the current region and its adjacent region.

[0240] 3.3 Contrast enhancement adjustment

[0241]

[0242] wherein L targetL is the target brightness; k is the adjustment coefficient, indicating the step of parameter adjustment; L target L is the target brightness, indicating the desired brightness value; L current L is the current brightness, indicating the brightness value of the current region.

[0243] Further, the data uploading and cloud platform integration module 500 specifically includes:

[0244] 1. Data acquisition and compression

[0245] Data acquisition:

[0246] Collect data from optical detection, electrical testing, AI defect recognition, and other modules.

[0247] Data compression:

[0248] Use JPEG, PNG, and other compression algorithms to reduce data volume.

[0249] Ensure the integrity of compressed data to avoid information loss.

[0250] 2. Data uploading and storage

[0251] Uploading method:

[0252] Upload to the cloud through Wi-Fi, 4G, or wired network.

[0253] Storage service:

[0254] Use cloud storage services (such as Aliyun OSS, AWS S3) to store detection data.

[0255] 3. Data analysis and visualization

[0256] Statistical analysis:

[0257] Calculate key indicators such as yield and defect rate.

[0258]

[0259] Process bottleneck prediction:

[0260] Predict possible process bottlenecks through machine learning models.

[0261] Visualization interface:

[0262] Use charts, heat maps, and other forms to display analysis results for real-time monitoring and decision-making.

[0263] The present application provides a Mini LED backlight intelligent detection system, which has the following advantages:

[0264] 1. Multi-modal data fusion: Simultaneous analysis of optical, electrical, and dynamic dimming data to generate a comprehensive quality assessment report;

[0265] 2. Dynamic dimming optimization: Adjusting backlight drive parameters through a feedback mechanism to improve local dimming linearity;

[0266] 3. Adaptive algorithm: Automatically optimizing detection parameters for different Mini LED packaging processes (such as COB / POB);

[0267] 4. Cloud platform integration: Connecting quality management cloud services to monitor production line yield in real-time and predict process bottlenecks;

[0268] The present application not only improves detection efficiency and accuracy, but also realizes real-time testing and optimization of the dynamic performance of backlight modules, providing reliable technical support for the widespread application of Mini LED display technology.

[0269] In order to verify the beneficial effects of the present application, the following simulation experiments are carried out:

[0270] 1. Verification of optical detection module

[0271] Objective: Verify the accuracy of the optical detection module in brightness uniformity and color gamut coverage detection.

[0272] Test design:

[0273] Test samples: Prepare 5 Mini LED backlight modules, 3 of which have good uniformity, and 2 of which have poor uniformity (achieved by artificially adjusting LED brightness).

[0274] Test equipment: Use a high-frame-rate camera (frame rate 30fps, resolution 1920x1080) and a spectrometer.

[0275] Test steps:

[0276] Use the system to take full-brightness, full-black, and gray-scale pictures.

[0277] The system calculates the brightness uniformity in combination with the diffuser plate transmittance data.

[0278] Use the spectrometer to manually measure the color gamut coverage and compare it with the system results.

[0279] Results and analysis:

[0280]

[0281]

[0282] Conclusion: The error between the system detection results and the manual measurement results is less than 0.3%, indicating that the optical detection module has high precision.

[0283] 2. Electrical Test Module Verification

[0284] Objective: Verify the accuracy of the electrical test module in detecting voltage offset and marking the fault location.

[0285] Test Design:

[0286] Test Samples: Prepare 10 Mini LED backlight modules, each containing 100 LED beads, with 5 beads having voltage offset exceeding the threshold (±5%).

[0287] Test Equipment: Multi-power voltage parallel detection system.

[0288] Test Steps:

[0289] System drives the backlight module and measures the voltage value of each bead in real time.

[0290] The system records the voltage offset of abnormal beads and marks the location.

[0291] Manually check abnormal beads to verify the accuracy of the system's marking.

[0292] Results and Analysis:

[0293]

[0294]

[0295] Conclusion: The system detection accuracy is 100%, indicating that the electrical test module has high reliability.

[0296] 3. AI Defect Recognition Module Verification

[0297] Objective: Verify the performance of the AI defect recognition module in identifying defects such as dead lights, weak lights, and color spots.

[0298] Test Design:

[0299] Test Samples: Prepare 20 Mini LED backlight modules, 10 of which contain dead lights, weak lights, or color spot defects.

[0300] Test Equipment: High-frame-rate camera, AI defect recognition system.

[0301] Test Steps:

[0302] System takes pictures of the backlight module and performs preprocessing (denoising, contrast enhancement).

[0303] Use Hough circle transformation and mask processing to extract the defect area.

[0304] System generates a defect heat map and marks the abnormal area.

[0305] Manual inspection of defects, verification of system recognition results.

[0306] Results and analysis:

[0307]

[0308]

[0309] Conclusion: The system recognition accuracy is 100%, indicating that the AI defect recognition module has high precision.

[0310] 4. Dynamic performance optimization module verification

[0311] Objective: To verify the effect of the dynamic performance optimization module in suppressing halos and improving contrast.

[0312] Test design:

[0313] Test samples: Prepare 5 Mini LED backlight modules for display effect testing before and after optimization.

[0314] Test equipment: High frame rate camera, spectrometer.

[0315] Test steps:

[0316] System detects brightness uniformity and contrast before optimization.

[0317] Dynamic performance optimization module adjusts Local Dimming parameters.

[0318] System detects brightness uniformity and contrast after optimization.

[0319] Compare the results before and after optimization.

[0320] Results and analysis:

[0321]

[0322] Conclusion: After optimization, the brightness uniformity is improved by an average of 3%, and the contrast is improved by an average of 10%, indicating that the dynamic performance optimization module effectively improves the display effect.

[0323] 5. Data upload and cloud platform integration module verification

[0324] Objective: To verify the performance of the data upload and cloud platform integration module in data transmission, storage and analysis.

[0325] Test design:

[0326] Test samples: Randomly select 100 Mini LED backlight module detection data.

[0327] Test equipment: edge computing device, cloud platform.

[0328] Test steps:

[0329] System collects and compresses test data.

[0330] Data is uploaded to the cloud platform via Wi-Fi.

[0331] The cloud platform performs statistical analysis and generates yield reports.

[0332] Verify data transmission speed and accuracy of analysis results.

[0333] Results and analysis:

[0334]

[0335] Conclusion: The data upload and analysis process is stable, and the data integrity reaches 100%, indicating that the cloud platform integration module is highly efficient.

[0336] Summary

[0337] Through the above test verification, the Mini LED backlight intelligent detection system proposed by the present application performs excellently in the following aspects:

[0338] Optical detection module: high-precision detection of brightness uniformity and color gamut coverage.

[0339] Electrical test module: high accuracy in detecting voltage offset and marking fault positions.

[0340] AI defect recognition module: high-precision identification of dead lights, weak lights, and color spots.

[0341] Dynamic performance optimization module: effectively improves brightness uniformity and contrast.

[0342] Data upload and cloud platform integration module: efficient data transmission and analysis.

[0343] These results fully demonstrate the excellent technical effects of the system, providing reliable technical support for efficient detection and quality control of Mini LED backlight modules.

[0344] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be included in the scope of the claims of the present application.

Claims

1. A Mini LED backlight intelligent detection system, characterized in that, The system comprises the following modules: An optical detection module uses a high-frame-rate camera to take full-brightness / full-black / gray-scale pictures, calculates brightness uniformity in combination with diffuser transmittance data, synchronously analyzes color gamut coverage by a spectrometer, and ensures compliance with the HDR10+ standard; An electrical test module drives the backlight module using a multi-power-voltage parallel detection system, records voltage deviation of abnormal lamp beads, and marks fault positions; An AI defect identification module performs Hough circle transformation and mask processing on collected images, inputs a hybrid model to classify normal / abnormal areas, and outputs a defect heat map; A dynamic performance optimization module adjusts LocalDimming algorithm parameters according to test results, suppresses halos, and improves contrast; A data uploading and cloud platform integration module compresses data by an edge computing device, uploads the data to the cloud for statistical analysis, and realizes real-time monitoring of production line yield; The modules communicate with each other through a data bus.

2. The Mini LED backlight intelligent detection system of claim 1, wherein: When the optical detection module uses a high-frame-rate camera to take full-brightness / full-black / gray-scale pictures, the following parameters are set: High-frame-rate camera parameters: frame rate ≥ 30 fps; resolution ≥ 1920x1080; exposure time is automatically adjusted according to LED brightness; Shooting mode: all LED lamp beads are turned on for full-brightness pictures to detect overall brightness uniformity; all LED lamp beads are turned off for full-black pictures to detect light leakage; for gray-scale pictures, the brightness is increased from 0% to 100% to detect brightness gradient uniformity.

3. The Mini LED backlight intelligent detection system of claim 2, wherein, The optical detection module calculates brightness uniformity in combination with diffuser transmittance data, specifically comprising the following steps: S1: Measure transmittance at different positions using a spectrometer to generate a transmittance distribution map; S2: Measure actual brightness values, correct them according to the transmittance distribution, and calculate brightness uniformity; The brightness uniformity is calculated according to the following formula: wherein L i is the corrected luminance value for the i-th position.

4. The Mini LED backlight intelligent detection system of claim 3, wherein, The optical detection module analyzes color gamut coverage by a spectrometer, specifically comprising the following steps: Q1: Measure spectral distribution of different colors using a spectrometer; Q2: Calculate the coverage area of the measured spectrum and the HDR10+ standard color gamut; The color gamut coverage is calculated according to the following formula:

5. The Mini LED backlight intelligent detection system of claim 4, wherein: When the electrical test module uses multi-power-voltage parallel detection, each lamp bead is independently driven, the voltage value of each lamp bead is measured in real time, voltage deviation is recorded, a voltage deviation threshold is set, and if the voltage deviation exceeds the threshold, the lamp bead is determined to be abnormal, and the position coordinates of the abnormal lamp bead are recorded synchronously; The voltage deviation is calculated according to the following formula: Voltage offset = |V 实际 - V 理论 | where V 实际 and V 理论 are the actual and theoretical voltages, respectively.

6. The Mini LED backlight intelligent detection system of claim 5, wherein: The AI defect identification module further comprises image preprocessing before image processing. The preprocessing steps include denoising and contrast enhancement.

7. The Mini LED backlight intelligent detection system of claim 6, wherein: When the dynamic performance optimization module suppresses halos and improves contrast, it adjusts the brightness difference between adjacent areas to reduce halos, and increases dark area brightness to reduce bright area overexposure and improve overall contrast.

8. The Mini LED backlight intelligent detection system of claim 7, wherein, The data uploading and cloud platform integration module performs statistical analysis according to the following model:

Images