Micro-led display uniformity detection method and system

Abstract

The application discloses a Micro-LED display uniformity detection method and system, and relates to the technical field of display device detection; the photo-induced luminescence peak wavelength of a Micro-LED chip is collected pixel by pixel, is mapped to a CIEDE2000 perceived color difference space, perceived features are extracted from five dimensions of global gray co-occurrence matrix anisotropy, local texture deterioration, multi-directional discrete Fourier period energy, multi-scale local standard deviation and macro boundary energy, after a statistical benchmark is constructed by Monte Carlo random shuffling, Z-Score standardization and asymmetric truncation, multi-dimensional features are fused based on information entropy weight and combined with a root mean square of color difference, and a unified defect index UDI is calculated; the problems that a traditional single luminance standard deviation cannot comprehensively represent spatial topological defects of a Micro-LED array and ignores wavelength chrominance difference are overcome, and multi-dimensional quantitative evaluation from physical wavelength to human visual perception is realized. Experiments show that the unified defect index of the mixing scheme is reduced from 11.894 to 0.708, the uniformity improvement amplitude reaches 94.0%, and the application can be applied to Micro-LED chip mixing arrangement effect evaluation and process optimization.

Description

Technical Field

[0001] This invention relates to the field of display device testing technology, and in particular to a method and system for detecting the uniformity of Micro-LED displays based on photoluminescence wavelength and color difference perceived by CIEDE2000. Background Technology

[0002] Micro-LED display technology boasts advantages such as high brightness, high contrast, fast response speed, long lifespan, and low power consumption, and is considered a crucial development direction for next-generation display technology. However, during the fabrication of Micro-LED display panels, due to the non-uniformity of epitaxial wafer growth, positional deviations in mass transfer, and differences in bonding processes, the display panels are prone to problems such as uneven brightness, color deviation, and Mura defects after being lit, severely impacting display quality and user experience. Therefore, accurate detection of the uniformity of Micro-LED display panels is a critical step in ensuring product quality and guiding process optimization.

[0003] Existing display uniformity testing methods mainly include the nine-point method, the global standard deviation method, and the Mura index method. The nine-point method only selects nine sampling points on the panel for brightness calculation, which cannot reflect the spatial distribution characteristics of the entire area and is insufficient for detecting local periodic defects and large-scale gradient defects. Although the global standard deviation method can characterize the overall dispersion, it is insufficient for the perception of defects in the frequency bands that are sensitive to the human eye and cannot distinguish the spatial distribution pattern of defects. The Mura index method usually relies on manually set thresholds or weight parameters, resulting in highly subjective test results and poor consistency among different inspectors, making it difficult to meet the needs of automated production line testing.

[0004] Furthermore, most existing detection methods evaluate based on brightness or grayscale values, neglecting the impact of color deviation caused by differences in the peak wavelengths of photoluminescence between Micro-LED chips on human visual perception. For Micro-LED display panels using a split-mixing process, the spatial distribution of chip wavelengths directly determines color uniformity; detection methods based solely on brightness cannot effectively characterize spatial topological defects caused by wavelength distribution.

[0005] Some existing technologies use a multi-index weighted average method for comprehensive testing, but the weight of each index is usually determined by human experience and lacks objective basis. More importantly, the weighted average model treats benign features that are better than the reference baseline and malignant defects that are worse than the reference baseline equally, resulting in an inverted test result. That is, some solutions that perform well in a certain dimension may get a poor overall score due to the weighting and cancellation of other dimensions, which cannot truly reflect the uniformity level of the display panel.

[0006] Therefore, there is an urgent need to develop an objective, accurate method and system for detecting the uniformity of Micro-LED displays that conforms to the characteristics of human eye perception, in order to solve the above-mentioned technical problems. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for detecting the uniformity of Micro-LED displays based on photoluminescence wavelength and CIEDE2000 perceived color difference, so as to realize multi-dimensional quantitative evaluation from physical wavelength to human visual perception.

[0008] This invention is achieved through the following technical solution:

[0009] The method for detecting uniformity in Micro-LED displays includes the following steps:

[0010] S1. The Micro-LED display back panel is scanned pixel by pixel using a photoelectric detection device to collect the peak wavelength of photoluminescence of the Micro-LED chip at each pixel position and construct a two-dimensional wavelength matrix.

[0011] S2. Map the two-dimensional wavelength matrix to the CIE 1931 standard color system, convert it to the CIELAB uniform color space via the XYZ color space, and calculate the CIEDE2000 perceived color difference of each pixel using the array average color coordinates as a reference to construct the perceived color difference matrix.

[0012] S3. Extract multi-dimensional perceptual features from the perceptual color difference matrix. The multi-dimensional perceptual features include global gray-level co-occurrence matrix anisotropy, local texture deterioration degree, multi-directional discrete Fourier periodic energy, multi-scale local standard deviation and macroscopic boundary energy, and calculate the root mean square of the perceptual color difference matrix as the basic reference feature.

[0013] S4. Perform N Monte Carlo random shuffles on the perceived color difference matrix to generate random baseline distributions for each perceived feature, and calculate the mean and standard deviation of each random baseline distribution, where N≥30.

[0014] S5. Perform Z-Score standardization on each perceptual feature relative to the random baseline distribution to obtain the standardized deviation of each perceptual feature.

[0015] S6. Perform asymmetric truncation on the standardized deviation, truncating the standardized deviation less than zero to zero, to obtain truncated feature values; the asymmetric truncation is used to penalize only the malignant defects that are worse than the random baseline, while not adding extra gain to the benign features that are better than the random baseline.

[0016] S7. Calculate the fusion weight based on the information entropy of each perceptual feature, and after weighted fusion of the truncated feature value and the fusion weight, calculate the integration defect index in combination with the basic reference feature, and output the integration defect index.

[0017] Further, in step S2, mapping the wavelength to the CIE 1931 standard colorimetric system specifically includes: using cubic spline interpolation to interpolate the color matching function of the CIE 1931 standard observer to obtain the tristimulus values ​​X, Y, and Z corresponding to each wavelength, wherein the Y channel is uniformly normalized to 100, and the X and Z channels are taken as non-negative values ​​after interpolation; after XYZ to CIELAB conversion under the D65 standard illuminator, the average value of the CIELAB coordinates of all pixels in the array is used as a reference benchmark to calculate the CIEDE2000 color difference pixel by pixel.

[0018] Further, in step S3, the global gray-level co-occurrence matrix anisotropy is represented by calculating the contrast of the gray-level co-occurrence matrix in four directions (0°, 45°, 90°, and 135°) after percentile truncation and normalization of the perceptual color difference matrix, and the ratio of the maximum contrast difference in the orthogonal directions to the average contrast in the four directions is used; the local texture deterioration is represented by calculating the anisotropy of each sub-region through a sliding window and taking the average of the top 5% of high-value regions; the multi-directional discrete Fourier periodic energy is represented by performing one-dimensional discrete Fourier transform on the row and column mean profiles of the perceptual color difference matrix and taking the normalized spectral peak value; the multi-scale local standard deviation is represented by calculating the global standard deviation after performing uniform filtering on the perceptual color difference matrix at multiple scales and taking the multi-scale mean value; the macroscopic boundary energy is represented by calculating the average value of the Sobel gradient magnitude after Gaussian smoothing of the perceptual color difference matrix.

[0019] Furthermore, in step S5, the formula for calculating the Z-Score standardization is as follows: ;

[0020] Where Xᵢ is the measured value of the i-th perceptual feature, μᵢ is the mean of the random baseline distribution of the i-th perceptual feature, σᵢ is the standard deviation of the random baseline distribution of the i-th perceptual feature, and Zᵢ is the standardized deviation.

[0021] Furthermore, in step S6, the calculation formula for the asymmetric truncation process is as follows: ;

[0022] Where Zᵢ represents the standardized deviation. Let be the i-th truncation feature value.

[0023] Furthermore, in step S7, the formula for calculating the fusion weight is:

[0024] ;

[0025] Where Hᵢ is the information entropy of the i-th perceptual feature in the random baseline distribution, and ΣHⱼ is the sum of the information entropies of all perceptual features.

[0026] Furthermore, in step S7, the formula for calculating the integrated defect index is:

[0027] ;

[0028] Where Base_RMS is the root mean square value of the perceptual color difference matrix, i.e., the basic reference feature, wᵢ is the fusion weight of the i-th perceptual feature, and n is the number of dimensions of the perceptual feature.

[0029] A Micro-LED display uniformity detection system includes a photoelectric detection device, a data preprocessing module, a color difference conversion module, a multi-dimensional feature extraction module, a random baseline generation module, an asymmetric truncation module, an entropy-weighted fusion module, and an output module. The photoelectric detection device is used to scan the Micro-LED display backplane pixel-by-pixel and acquire the peak wavelength of photoluminescence. The data preprocessing module is used to construct a two-dimensional wavelength matrix and repair abnormal pixels. The color difference conversion module is used to convert the wavelength matrix into a CIEDE2000 perceived color difference matrix. The multi-dimensional feature extraction module is used to extract multi-dimensional perceived features and basic reference features from the perceived color difference matrix. The random baseline generation module is used to generate a random baseline distribution for each perceived feature through Monte Carlo random shuffling and perform Z-Score standardization. The asymmetric truncation module is used to perform asymmetric truncation processing on the standardized deviation. The entropy-weighted fusion module is used to calculate the fusion weight based on information entropy and calculate the integration defect index. The output module is used to output the integration defect index.

[0030] The beneficial effects of this invention are as follows:

[0031] 1. This invention uses a photoelectric detection device to collect the peak wavelength of photoluminescence of Micro-LED chips pixel by pixel. After mapping with the CIE 1931 colorimetric system and CIEDE2000 color difference conversion, the physical wavelength domain is mapped to the human-perceived color difference space. Combined with multi-dimensional feature extraction, random baseline standardization, asymmetric truncation and entropy weighted fusion, a complete detection technology chain is formed, which can realize the automated and objective detection of Micro-LED display uniformity and effectively characterize the color uniformity defects caused by wavelength spatial distribution.

[0032] 2. An objective baseline distribution is generated by Monte Carlo random shuffling. The features of each dimension are mapped to a unified deviation measurement space through Z-Score standardization, eliminating the subjectivity of manually setting thresholds and weights. The detection results have good repeatability and consistency, and are suitable for batch detection on production lines.

[0033] 3. An asymmetric truncation mechanism is introduced, which sets the feature values ​​that are better than the random baseline to zero and only punishes the malignant defects that are worse than the random baseline. This conforms to the asymmetric characteristics of human eyes’ perception of display uniformity and effectively solves the problem of inverted detection results caused by the mutual cancellation of benign features and malignant defects in the traditional weighted average model. The consistency between the detection results and subjective visual evaluation is significantly improved.

[0034] 4. The fusion weights are automatically allocated based on the information entropy of each dimension feature. The weighted Euclidean distance is used to calculate the integration defect index, and the root mean square of the perceived color difference is used as the physical baseline. This ensures the full integration of multi-dimensional information and has clear physical meaning. At the same time, the integration defect index can be directly used to evaluate the uniformity improvement effect of the mixing and arrangement scheme and guide the optimization of chip mixing and laser transfer parameters. Attached Figure Description

[0035] Figure 1 This is a schematic flowchart of the Micro-LED display uniformity detection method in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the multidimensional perception feature extraction module in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the photoelectric detection pixel-by-pixel scanning acquisition device in an embodiment of the present invention;

[0038] Figure 4 This is a comparison of the five-dimensional sensing features Z-Score radar between the non-mixed control group and the mixed experimental group in this embodiment of the invention.

[0039] Figure 5 This is a comparison chart of the integration defect index between the non-mixed control group and the mixed experimental group in an embodiment of the present invention.

[0040] In the diagram: ①-backplate, ②-Micro-LED chip array, ③-photoelectric detection probe, ④-horizontal scanning rail, ⑤-vertical bracket, ⑥-data acquisition line. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0042] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0043] like Figure 1 As shown, the Micro-LED display uniformity detection method includes the following steps:

[0044] S1. The Micro-LED display back panel is scanned pixel by pixel using a photoelectric detection device to collect the peak wavelength of photoluminescence of the Micro-LED chip at each pixel location, and a two-dimensional wavelength matrix is ​​constructed. For example... Figure 3 As shown, the photoelectric detection device includes a backplate ①, a Micro-LED chip array ② disposed above the backplate ①, a photoelectric detection probe ③ located directly above the chip array, a horizontal scanning rail ④ that moves the probe horizontally, a vertical support ⑤ that fixes the probe, and a data acquisition line ⑥. The photoelectric detection probe ③ moves pixel by pixel along the horizontal direction via the horizontal scanning rail ④ to scan and acquire data from the Micro-LED chip array ②. The acquired photoluminescence spectrum signal is transmitted to the data acquisition and processing unit via the data acquisition line ⑥, and the peak wavelength of each pixel is extracted by peak detection. The photoelectric detection probe is a hyperspectral imager or spectrometer with a spectral resolution of not less than 5 nm and a spatial resolution of not less than 10 μm. For pixels with missing wavelength data or abnormal quality markings, interpolation repair is performed using the local mean of the effective pixels in a 5×5 neighborhood.

[0045] S2. Map the two-dimensional wavelength matrix to the CIE 1931 standard colorimetric system. Use cubic spline interpolation to interpolate the color matching function of the CIE 1931 2-degree standard observer to obtain the tristimulus values ​​X, Y, and Z corresponding to each wavelength. The Y channel is uniformly normalized to 100, while the X and Z channels are taken as non-negative values. Under a D65 standard illuminator, convert XYZ to the CIELAB uniform color space. Using the mean of the CIELAB coordinates of all pixels in the array as a reference, calculate the CIEDE2000 perceptual color difference pixel by pixel to construct the perceptual color difference matrix.

[0046] S3. Extract five-dimensional perceptual features and basic reference features from the perceived color difference matrix, specifically including:

[0047] (1) Global gray-level co-occurrence matrix anisotropy (GCV-OA): The perceptual color difference matrix is ​​truncated and normalized from 1% to 99 percentile, and the contrast of the gray-level co-occurrence matrix in the four directions of 0°, 45°, 90° and 135° is calculated. The contrast is represented by the ratio of the maximum contrast difference in the orthogonal directions to the average contrast of the four directions multiplied by 100.

[0048] (2) Local texture deterioration degree: The perceptual color difference matrix is ​​traversed by a sliding window of a preset size, the anisotropy in each window is calculated, and the average value of the top 5% high value area is taken to represent the degree of local texture deterioration; the size range of the sliding window is 8×8 pixels to 15×15 pixels;

[0049] (3) Multidirectional Discrete Fourier Periodic Energy (MDFT): Perform one-dimensional discrete Fourier transform on the row mean profile and column mean profile of the perceived color difference matrix respectively, remove the DC component and take the maximum value of the normalized spectrum amplitude, and use the larger value in the row and column directions to characterize the intensity of the periodic defect.

[0050] (4) Multi-scale local standard deviation S-Value: The perceptual color difference matrix is ​​uniformly filtered using multiple scales such as 3×3, 5×5, and 10×10. The global standard deviation of the filtered data at each scale is calculated, and the multi-scale mean is used to characterize the low-frequency slowly changing cloud spot defect.

[0051] (5) Macroscopic boundary energy Edge: After Gaussian smoothing of the perceptual color difference matrix, the Sobel gradient magnitudes in the horizontal and vertical directions are calculated, and the intensity of the macroscopic abrupt gap boundary is characterized by multiplying the mean gradient magnitude by 1000.

[0052] (6) Base_RMS: Calculate the root mean square value of the perceived color difference matrix as the physical background reference quantity for the integrated defect index.

[0053] S4. Perform N Monte Carlo random shuffles on the perceptual color difference matrix, where N≥30. After each random shuffle, recalculate the above five-dimensional perceptual features to obtain N random sample values ​​for each feature. Then, calculate the mean μᵢ and standard deviation σᵢ of the random baseline distribution of each feature.

[0054] S5. Standardize each perceived feature value Xᵢ of the actual scheme with respect to the corresponding random baseline distribution using the Z-Score formula:

[0055] ;

[0056] The standardized deviation Zᵢ of each perceptual feature is obtained.

[0057] S6. The standardized deviation Zᵢ is asymmetrically truncated, and the calculation formula is as follows:

[0058] ;

[0059] The standardized deviations less than zero are truncated to zero to obtain the truncated eigenvalues. The technical principle of the asymmetric truncation process is that the human eye's perception of display uniformity is asymmetric; defects that are worse than expected are keenly perceived, while uniform areas that are better than expected do not produce additional perceptual gain. Therefore, only positive deviations are penalized.

[0060] S7. Calculate the fusion weight based on the information entropy of each perceptual feature. The calculation formula is as follows:

[0061] ;

[0062] Where Hᵢ is the information entropy of the i-th perceptual feature in N random shuffle samples. The truncated feature values ​​are weighted and fused with the fusion weights, and combined with the basic reference features to calculate the integration defect index. The calculation formula is as follows:

[0063] ;

[0064] The Integrated Defect Index (UDI) is obtained and output to a display device for display.

[0065] The Micro-LED display uniformity detection system includes a photoelectric detection device, a data preprocessing module, a color difference conversion module, a multi-dimensional feature extraction module, a random baseline generation module, an asymmetric truncation module, an entropy-weighted fusion module, and an output module, with each module electrically connected in sequence; wherein, the hardware structure of the photoelectric detection device is as follows: Figure 3 As shown.

[0066] The present invention will be further illustrated by specific embodiments below.

[0067] Example 1 (Separate Mixed Experimental Group):

[0068] The detection method of this invention is used to perform uniformity testing on a Micro-LED display backplane that has undergone a splitting and mixing process. The chips in the display backplane are derived from the same green GaN-based Micro-LED epitaxial wafer, with a chip size of 34μm × 58μm, a pixel pitch of 230.4μm, and a sorting wavelength of 523.5 ± 1.5nm. The splitting and mixing process employs a two-dimensional checkerboard topology, arranging chips in different wavelength ranges in a cross-hatching pattern according to a preset rule.

[0069] Adopting such Figure 3 The photoelectric detection device shown performs pixel-by-pixel scanning of the display backplane, acquiring the peak wavelengths of photoluminescence from each chip and constructing a wavelength matrix. This matrix is ​​then converted using CIEDE2000 to obtain the perceived color difference matrix. Five-dimensional perceptual features and basic reference features are extracted, and 50 Monte Carlo random shuffles are performed to generate a random baseline. After Z-score normalization, asymmetric truncation, and entropy-weighted fusion, the integrated defect index (UDI) is obtained. The detection results show that the UDI for this mixed experimental group is 0.708.

[0070] like Figure 4 As shown, Z-Score analysis was performed on the five-dimensional perception features (GCV-OA, Local-OA, MDFT, S-Value, Edge) of the split-mix experimental group. The results showed that the standardized deviation of each dimension was close to zero or negative. After asymmetric truncation, the truncation feature value of most dimensions was zero, indicating that the spatial topological features after the split-mix process were close to or better than the random shuffle baseline, with no significant malignant defects.

[0071] Comparative Example 1 (Unspecified control group):

[0072] Using the same chip pool as in Example 1, without any mixing process, the chips were directly arranged on the display backplane according to the original epitaxial wafer wavelength distribution order, serving as a non-mixing control group. The uniformity of the display backplane in this control group was tested using the detection method of this invention, and the detection process was completely consistent with that of Example 1.

[0073] The test results showed that the Integration Defect Index (UDI) of the unmixed control group was 11.894. Figure 4 As shown, the control group exhibits large positive values ​​for the normalized deviation in five dimensions: global gray-level co-occurrence matrix anisotropy (GCV-OA), local texture degradation (Local-OA), multi-directional discrete Fourier periodic energy (MDFT), multi-scale local standard deviation (S-Value), and macroscopic boundary energy (Edge). This indicates that the unmixed chip arrangement is significantly inferior to the random shuffle baseline in terms of spatial topology, exhibiting severe periodicity and large-scale gradient defects.

[0074] Results Comparison and Analysis:

[0075] like Figure 5 As shown, comparing the detection results of Example 1 and Comparative Example 1, the Integrated Defect Index (UDI) of the mixed experimental group was 0.708, significantly lower than the UDI value of 11.894 in the unmixed control group, representing a reduction of 94.0%. This result indicates that the mixed process can effectively improve the wavelength distribution uniformity of the Micro-LED display backplane, and the detection method of this invention can objectively and accurately reflect the improvement effect of the mixed process on uniformity.

[0076] like Figure 4 As shown in the Z-Score radar comparison chart of the five-dimensional sensing features, the non-mixing control group exhibits significant positive deviations in all dimensions, with the most significant deviation occurring in the Global Gray-Level Co-occurrence Matrix Anisotropy (GCV-OA) dimension. In contrast, the mixed-data experimental group approaches the zero axis in all dimensions, with some dimensions showing negative deviations (better than the random baseline). This radar chart visually demonstrates the improvement effect of the mixed-data process on various spatial topological defect dimensions, verifying the effectiveness of the multi-dimensional feature detection method of this invention.

[0077] To further verify the effectiveness of the method of the present invention, a correlation analysis was performed between the detection results of the method of the present invention and the subjective visual ranking of professional display inspectors. The results show that the detection results of the method of the present invention have a high degree of consistency with the subjective evaluation, which is superior to the traditional global standard deviation method and the traditional Mura index method.

[0078] Furthermore, to verify the technical effectiveness of the asymmetric truncation mechanism, the integrated defect index of the embodiment was recalculated with asymmetric truncation disabled (using the traditional weighted average model, retaining positive and negative values). The results show that after disabling asymmetric truncation, eigenvalues ​​superior to the random baseline in some dimensions cancel each other out with malignant defects, leading to biased detection results that fail to accurately reflect the true uniformity level of the array. This comparison verifies the crucial role of the asymmetric truncation mechanism in improving detection accuracy.

[0079] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the uniformity of Micro-LED displays, characterized in that, Includes the following steps: S1. Collect the peak wavelength of photoluminescence from the Micro-LED chip and construct a two-dimensional wavelength matrix; S2. Map the two-dimensional wavelength matrix to the CIE 1931 colorimetric system and convert it into a CIEDE2000 perceived color difference matrix; S3. Extract multi-dimensional perceptual features from the perceptual color difference matrix and calculate the root mean square of the color difference as the basic reference feature; S4. Perform N Monte Carlo random shuffles on the perceived color difference matrix to generate random baseline distributions for each feature, where N≥30; S5. Standardize each sensory feature relative to the random baseline distribution using Z-Score to obtain the standardized deviation of each sensory feature. S6. Perform asymmetric truncation on the standardized deviation, truncating values ​​less than zero to zero; S7. Calculate the fusion weight based on the information entropy of each perceptual feature, and after weighted fusion of the truncated feature value and the fusion weight, calculate the integrated defect index in combination with the basic reference feature and output it.

2. The Micro-LED display uniformity detection method according to claim 1, characterized in that, In S2, the CIE 1931 tristimulus values ​​are calculated using cubic spline interpolation. After XYZ to CIELAB conversion under D65 illuminator, the CIEDE2000 perceived color difference is calculated with the array average color coordinates as the reference.

3. The Micro-LED display uniformity detection method according to claim 1, characterized in that, The multidimensional perception features include global gray-level co-occurrence matrix anisotropy, local texture deterioration, multi-directional discrete Fourier periodic energy, multi-scale local standard deviation, and macroscopic boundary energy.

4. The Micro-LED display uniformity detection method according to claim 1, characterized in that, The formula for calculating the Z-Score standardization in S5 is as follows: ； Where Xᵢ is the measured value of the i-th perceptual feature, μᵢ is the mean of the random baseline distribution of the i-th perceptual feature, σᵢ is the standard deviation of the random baseline distribution of the i-th perceptual feature, and Zᵢ is the standardized deviation. The calculation formula for the asymmetric truncation in S6 is as follows: ； Where Zᵢ represents the standardized deviation, Z truncated,i Let be the i-th truncation feature value.

5. The Micro-LED display uniformity detection method according to claim 1, characterized in that, The formula for calculating the fusion weight in S7 is as follows: ； Among them, Hᵢ and Let ΣH be the information entropy of the i-th and j-th perceptual features in the random baseline distribution, respectively, and let ΣH be the sum of the information entropy of all perceptual features; The formula for calculating the integration defect index is: ； Where Base_RMS is the root mean square value of the perceptual color difference matrix, i.e., the basic reference feature, wᵢ is the fusion weight of the i-th perceptual feature, and n is the number of dimensions of the perceptual feature.

6. A Micro-LED display uniformity detection system, used to implement the Micro-LED display uniformity detection method of claims 1-5, characterized in that, It includes a photoelectric detection device, a data preprocessing module, a color difference conversion module, a multidimensional feature extraction module, a random baseline generation module, an asymmetric truncation module, an entropy weighted fusion module, and an output module, which are connected in sequence.

7. The Micro-LED display uniformity detection system according to claim 6, characterized in that, The photoelectric detection device includes a back plate, a Micro-LED chip array disposed on the back plate, a photoelectric detection probe, a horizontal scanning guide rail, a vertical support, and a data acquisition line. The photoelectric detection probe scans pixel by pixel along the horizontal scanning guide rail.

8. The Micro-LED display uniformity detection system according to claim 6, characterized in that, The data preprocessing module includes an abnormal pixel repair unit, which is used to repair abnormal pixels using the local mean of effective pixels within a 5×5 neighborhood.

Images